Intersteno 2021 Online Competition Report (Mar 27, 2021 - May 12, 2021)


  1. About Intersteno
  2. My results
  3. Prologue
  4. Strategies
    1. Reconsideration of keyboard layouts (reducing strokes of frequent diacritical marks)
    2. Reconsideration of keyboard layouts (utilizing dead keys)
    3. Intensive training of difficult languages
    4. Training of typing speed
    5. Training schedule for Intersteno 2021
    6. The sequence of 16 languages in the championship
  5. Tactics
    1. Collecting, inputting, re-inputting training texts
    2. Examining/implementing problems and diacritical marks of training texts
    3. Selecting dialects
    4. Selecting keyboard
    5. Arranging other environments
    6. Participating in the contest
  6. Tips for each language
    1. Hungarian
    2. English
    3. Russian
    4. Japanese
    5. Czech, Slovak
    6. Finnish
    7. French
    8. Turkish
    9. Italian
    10. Dutch, Germany
  7. Sentiment and reflection
  8. For the Intersteno 2022
  9. References and acknowledgements
  10. Appendices

1. About Intersteno

Intersteno is the trademark of International Federation for Information Processing, founded in 1887 for stenographers and typists. It holds on-site stenography and typing championship every two years since 2007, and internet typing contest every year since 2003. This report contains the internet contest held from March to May in 2021.

[Official ranking]

[Official ranking (Japanese competitors)]

Every competitor can join the online championship with maximum 16 languages from 17 languages. The 17 languages include Italian, English, Germany, French, Dutch, Spanish, Turkish, Portuguese, Romanian, Japanese, Finnish, Czech, Slovak, Hungarian, Russian, Polish, and Croatian/Serbian, from the sequence in the training website (Taki software).

Every competitor types first-sight texts for ten minutes for each language. Points are calculated by [total strokes - 50 * errors] for each language. Total points are calculated by sum of 16 languages. Therefore, it is important not only to increase speed but also to improve accuracy. By the way, definition of errors is different from that of Mainichi PC typing competition (hereinafter "MaiPaso"), which is one of the most popular Japanese typing contests. For example, errors in one word (e.g., "taht", "thaaat", "tt" instead of "that"), one line omission, are counted as only one error.


2. My results

[Table 1: My results of online competition 2021]

LanguagePointsStrokes/mErrorsPlaceParticipants
French6020(+254)602.0(+5.4)0(-4)3(+2)165(+22)
Italian5826(+197)582.6(-0.3)0(-4)3(+1)238(-4)
Portuguese5714(+375)571.4(+27.5)0(-2)2(+1)104(+10)
English5654(-477)590.4(-27.7)5(+4)3(-)325(-295)
Finnish5279(+277)532.9(+17.7)1(-2)1(+1)136(+3)
Turkish5276(+122)542.6(+12.2)3(-)11(+4)227(+14)
Spanish5220(-133)547.0(+1.7)5(+3)4(-1)152(-16)
Romanian5111(-32)521.1(+1.8)2(+1)2(-1)82(+12)
Dutch5052(-299)535.2(-14.9)6(+3)3(-1)202(+1)
Croatian5050(-199)520.0(-14.9)3(+1)1(+3)205(-19)
Germany4802(-423)510.2(-27.3)6(+3)7(-3)253(-12)
Slovak4738(-28)483.8(-2.8)2(-)2(-)185(-202)
Polish4723(-84)497.3(+1.6)5(+2)6(-2)99(-5)
Czech4613(-167)481.3(-1.7)4(+3)4(-2)238(-417)
Hungarian4559(+179)475.9(+12.9)4(-1)3(-)91(-20)
Japanese4507(-)460.7(-)2(-)2(-)25(+25)
Russian----45(+10)
Total82144(-207) 48(+6)1(+1)839(-677)

* Numbers in bracket are difference compared to last year.
* Numbers in bracket in "Total" row are the sum of points of 16 languages (except Japanese in 2020, except Russian in 2021).
* [+] in bracket of Points, Strokes/m, Place, and Participants means improvement, while [-] means deterioration. Those in blanket of Errors mean the opposite.
* I selected Belgian dialects of Dutch and French, while I selected original Germany, Italian, and Croatian.
* I collected Places at 0:00 (GMT) May 13. Also, I summed data of dialects (e.g., original Germany and Swiss Germany).

[Table 2: Stats of my ten-minutes training in 2021]

LanguageMax pointsAve. pointsAve. strokes/mAve. errors
French6020(+112)5745(+103)583.3(+11.3)1.75(+0.19)
Italian5826(+92)5554(+48)566.2(+5.6)2.15(+0.15)
Portuguese5714(+243)5361(+74)547.5(+5.8)2.29(-0.33)
English6302(-97)5792(-123)594.5(-9.4)3.06(+0.57)
Finnish5385(+180)5171(+87)528.3(+12.9)2.25(+0.85)
Turkish5671(+15)5422(+203)553.6(+18.8)2.29(-0.29)
Spanish5568(-7)5386(+38)556.7(+11.5)3.62(+1.54)
Romanian5389(+54)5086(+2)522.5(+3.6)2.78(+0.67)
Dutch5641(+91)5276(-28)543.8(+4.8)3.25(+1.52)
Croatian5345(+53)5063(+7)519.6(+1.1)2.67(+0.07)
Germany5406(+75)5160(+27)529.1(+1.9)2.63(-0.15)
Slovak4937(+34)4753(+85)487.5(+4.0)2.43(-0.90)
Polish4723(-84)4514(-84)480.2(+6.0)5.75(+2.89)
Czech4840(+59)4577(+31)477.0(+8.3)3.85(+1.05)
Hungarian4665(-374)4495(-112)471.2(-4.5)4.33(+1.33)
Japanese4903(-2)4599(+50)473.3(+3.3)2.67(-0.33)
Russian4153(-428)3937(-245)418.0(-14.4)4.86(+2.02)
Total90488(+16)85891(+164)  

* Results in bracket are difference compared to last year.
* The data includes results of online competitions.
* The data includes errors and points deduction due to bugs in Taki software.
* I practiced Russian, while I did not participate in Russian of the competition.


3. Prologue

Last year, I was defeated by Sean, who got brilliant 83,490 points by only 15 languages in the Intersteno online multilingual competition. In addition, my goal of last year, "85,000 points by 16 languages", remained undone. Then, I set the goal of 2021 to break my record of last year and recapture the world champion.

On the other hand, Japanese representative Mr. Sumino, who had organized Japanese typists until last year, was absent this year with some reason. Then, to participate in the online contest 2021, I decided to organize Japanese typists on behalf of him. As I started registration in March, Japanese typists successfully secured six weeks for the championship period. The period is much longer than before.


4. Strategies

I set the theme in 2021 to do my best to fight with international typists by 16 languages. On the background, Intersteno online contest 2021 were held by 16 languages in order to realize fairer championship in the situation that it is difficult for non-Japanese typists to type Japanese including lots of Kanjis (Chinese characters). By the way, Japanese (transliterated), which includes only Latin characters and had been discussed since last year, was not implemented yet. I heard that transliterated competition text was under preparation for Intersteno online competition 2022.

In fact, there is still advantage for Japanese typists. Japanese typists can select 16 languages from 17 languages. That is, Japanese typists can exclude one "difficult" language except Japanese (maybe Russian or Hungarian). On the other hand, non-Japanese typists must select 16 languages except Japanese, as very few of them can type Japanese. However, there is few advantages for me because Japanese typists must select Japanese as mother language. It is difficult to get high points in Japanese. For example, in the online competitions 2018 and 2019, my Japanese points were less than Russian which includes only Cyrillic letters, and Czech which contains lots of diacritical marks.

The basic strategy for Intersteno 2021 was to improve speed of all 16 languages. I tried to aggressively improve speed, as I wrote in For the Intersteno 2021 (Japanese only) last year. Indeed, it is desirable to improve speed with high accuracy. However, points cannot be higher without improving speed. Basically, I adopted the strategies established in 2016. I thought there was room for enhancing my typing skills through the same approach. However, I changed some contents if necessary.

(1) Reconsideration of keyboard layouts (reducing strokes of frequent diacritical marks)

I have considered and evolved original keyboard layouts (hereinafter "KL") for each language utilizing Microsoft Keyboard Layout Creator (hereinafter "MSKLC") since 2015. MSKLC is one of the major software allowing us to change KL flexibly. Then, I compared my original KL with popular KLs of each language (including, but not limited to QWERTZ and AZERTY), grasped advantages and disadvantages of them, and re-considered the optimal KLs. As a result, my KLs did not significantly change from previous ones. If I participated in only one language, it would be better to adopt the popular KL of the language in order to improve speed and accuracy. However, I was afraid of increasing errors by confusing different allocations of the same keys among plural languages when participating in multilingual contest.

Finally, I allocated diacritical marks on []\=qvwxyz keys of US international KL as table 3. Except for Japanese and Russian, I created original KLs based on US international KL, which was familiar to me. To reduce learning cost, I allocated similar characters on the same keys as far as I could. However, I gave priority to some frequent diacritical marks.

It should be noted that these KLs are not completed. I will evolve them through trial and error.

[Table 3: Allocation of diacritical marks]

Keyboard layout[]\=qvwxyz
Japanese@[]^      
Russianпчьщфжшхуй
Englishéç       
DutchëèDK       
Italian and FrenchéDK'DK  à   
Finnish and Germanyäöüß      
PortugueseãõçDK  à í 
Spanisháéíü  ó   
Polishąęłóść ż  
Czecháéížů ěč  
Slovakáéížú ľč  
Hungarianáéíóü öő  
Turkishışçİü öğ  
Croatianćščw  đž  
Romanianăşţwâ î   

* DK means dead key.
* In Czech, šňřťý are allocated on 34567. In Slovak, šňŕťý are allocated on 34567.
* Blanks mean that there are no changes from US international KL. For example, "q" in English is allocated on "q" in the US international KL.

[The process I reached the current KL]

On the competition 2015, I had too short time to prepare for all 17 languages. So, I immediately decided to adopt "general expanded KL for Intersteno" created by Curry-san (hereinafter "Curry KL"). Curry KL is the most suitable KL for short-term preparation as it can be easily introduced and memorized. For example, it is not necessary to install KL for each language. On the other hand, it has disadvantage that the more frequent diacritical marks appear, the more typing speed decreases. Especially, the speed significantly decreased in Czech, Slovak, Polish, Hungarian, and Turkish. Furthermore, I wasted time to type umlauts by three strokes (e.g., I typed ä by [Shift + : -> a]) in Finnish. Although only two diacritical marks are included in Finnish, they frequently appear.

I was a pity that I could not reach the supreme to learn popular KLs for each language as Junior-san did. It was because the location of some characters and marks differs for each language. It is obvious that QWERTZ KL or AZERTY KL adversely affects QWERTY KL (e.g., English). Furthermore, I intuitively knew that allocations of marks such as [,.;:'"] should be common for all languages. Otherwise, they would definitely lead to lots of errors.

Thus, I considered original KLs to type languages which I could not type by US international KL (e.g., Croatian, Romanian, Turkish, Hungarian, Polish, Czech, Slovak, and Russian). First, I learned how to use MSKLC and create original KLs based on US international KL. Second, I investigated frequency of characters for each language. I not only referred to Multilingual Typing Wiki (Japanese only) but also collected data from training texts in Taki software as shown in Tactics 1.

Based on the research results, I created KLs in which each frequent diacritical mark can be typed by one stroke. For example, äö in Finnish are allocated on []. On the other hand, Hungarian has so many diacritical marks that there are not enough keys to type each diacritical mark by one stroke. I would not like to allocate diacritical marks on the number keys on the top row or on [,./] except some difficult languages. So, I searched rare characters in 26 alphabet letters. Consequently, I allocated üöő on qwx. As Czech and Slovak have more diacritical marks than Hungarian, it is obvious that there are not enough keys to type each diacritical mark by one stroke. Then, I decided to allocate some keys on the number keys on the top row. Referring to popular KLs and frequency, I allocated šňřťý on 34567 in Czech KL, while šňŕťý on 34567 in Slovak KL.

Of course, the method above can be also applied to the seven languages which I typed by US international KL in 2015 (e.g., Germany, French, Italian, Dutch, Spanish, Portuguese, and Finnish). For example, I created French/Italian KL so that I could type frequent ' (apostrophe) by one stroke. I made Portuguese KL so that I could type çãõ by one stroke for each. As to English, I allocated ’éç on []\ to type them by one stroke for each, because they rarely appears in training texts.

At first, I thought that it is inefficient to use different KL for each language. Then, I used "Dutch/Spanish/Portuguese common KL" for example. However, the KLs of these three languages were separated through optimizing each KL. Finally, only "Italian/French KL" and "Finnish/Germany KL" remained as common KLs.

(2) Reconsideration of keyboard layouts (utilizing dead keys)

For some language, there are not sufficient keys by the methods above. In such case, US international KL adopts concept of "dead key". For instance, á is input by two strokes [:->a] using dead key [:]. Curry KL covers almost all Intersteno languages by only one KL by utilizing plural dead keys.

Also, If I allocated diacritical marks on [qwvx], I cannot input qwvxQWVX instead. Then, I allocated them on [:->qwvxQWVX] applying dead keys. Similarly, I allocated [34567#$%^&] on [`->34567#$%^&] in Czech and Slovak. Also, I cannot type []{}\|=+ which are allocated on []\=. So, I allocated [=] on [`->-] as it sometimes appears.

Dead keys I have used since Intersteno online contest 2018 are shown as table 4.

[Table 4: Allocations of diacritical marks (dead keys)]

Keyboard layoutDead keys
'`others
Common àèìòù 
RussianIVXёэъ“”«»—–= 
English “”£€‘…–¬ 
Dutcháéíóú äïöü (\)
Italian and Frenchèìòùçæœw“”€’«»ºª–…=âêîôû (]) ëïöÿ (=)
Finnish and Germanyå픀= 
Portugueseáéóúwy“”ºª°=âêôïöü (=)
Spanishúñwºª= 
Polishéńźqvx= 
Czechóúďqwx34567#$%^&= 
Slovakóďäôqwx“”‘’–ĺ34567#$%^&= 
Hungarianäúűqwx“”„–àõû[]= 
Turkishâîþqwx“”’= 
Croatianðæx«»–è= 
Romanianqǎ“”„’ºª–ã…= 

* These allocations contain characters appeared in texts in Taki software, even if they are not originally used for each language. For instance, ä in Hungarian, ð in Croatian, etc.

I deleted almost all dead keys from US international KL, but kept [`] (àèìòù), [Shift + `] (ãñõũ) and [Shift + 6] (âêîôû). I thought these dead keys would be useful in some proprietary nouns of foreign countries.

In addition, I moved dead keys if necessary. For example, I allocated dead key on [`] instead of ['] in English KL. This is because I would like to type frequent ['] with one stroke. US international KL allocates ['] on ['] key. However, ['] is also used as a dead key. Then, ['] is typed by ['->Space]. It is very inefficient. Similarly, I moved tréma dead key from [;] to [=] in order to type frequent [;] by one stroke in Italian and French KL.

(3) Intensive training of difficult languages

To overcome difficult languages, I set one language for each year and intensively practiced it.

[Hungarian in 2016]

In the Intersteno online contest 2015, my speed of Hungarian was the least and my point of Polish was the worst. In addition, I could get only around 3,000 points in Turkish, Czech, and Slovak. Conversely, I could expect significant increase of points by practicing these difficult languages intensively. It was much more possible for me to raise 1,000 points (nearly equal to increase speed by 100 strokes per minute) on Hungarian and Polish rather than to realize it on English and Dutch, which can be typed relatively easily.

First, I investigated and considered the fundamental cause of losing speed in Hungarian. Then, I judged that it would be significantly effective to reduce strokes by optimizing KL. Then, I considered original Hungarian KL as I wrote before. Moreover, I gradually evolved the original KL through practicing one-minute texts on Taki software ten times a day for three months.

[Russian in 2017]

In the Intersteno online contest 2016, I got the worst point in Russian. Furthermore, I could get only around 4,000 points in Czech, Polish, Hungarian, Romanian, and Slovak. Conversely, I could expect significant increase of points by practicing these difficult languages intensively. It was much more realistic for me to raise 500 points on Russian and Czech rather than to raise 500 points on English and Dutch.

First, I investigated and considered the root cause of losing speed in Russian. Consequently, I found that I did not get used to Cyrillic characters. Then, I judged that it is quite effective to get used to Cyrillic characters by increasing quantity of practice. Then, I kept practicing one-minute training on Taki software ten times a day for four months. In parallel, I kept inputting and re-inputting ten-minutes texts.

My typing speed of Russian did not reach 400 strokes per minute (hereinafter "spm") in one-minute training. But I steadily got speed by intensive training. I reached 450 spm on Dec. 20, and 500 spm on Jan. 15. On ten-minutes training, I reached plateau before 4,000 points and 4,150 points. But I kept intensive training steadily so as to stimulate potential. Then, I got over 4,270 points at the competition 2017 and training in advance.

[Turkish in 2018]

In the Intersteno online contest 2017, my point in Turkish decreased by 189, while that in Russian increased by 787. I thought that it was because I did much fewer training of Turkish than Russian.

First, I investigate and consider the fundamental cause of losing speed in Turkish. Then, I found that I did not get used to diacritical marks as well as frequent words and sequences. Thus, I judged that it is extremely effective to get used to typing Turkish by increasing quantity of practice. Then, I kept intensive Turkish training for four months. As I knew that there were over 100 ten-minutes texts, I mainly practiced ten-minutes training from the early stage. In parallel, I kept re-inputting ten-minutes texts.

Thanks to inputting and re-inputting ten-minutes texts, I found clear growth in the early stage. I reached 5,000 points in ten-minutes training in the first weekend. After practicing a few weeks, I acquired constant speed of 500 spm. Finally, I reached 5,500 points in ten-minutes training. But I kept intensive training steadily so as to stimulate potential. Then, I barely got 4,773 points at the competition although I was so nervous.

...

In addition, I intensively practiced Romanian and Hungarian for five weeks aiming for TypeRacer events year-end and new year holidays held by Tanon-san. It was a particularly good opportunity for me because I felt stuck of both languages in the Intersteno online contest 2017. Also, I intensively practiced Russian for four weeks immediately before Intersteno online contest 2017. It was because I worried about decline as a reaction to intensive training last year.

(4) Training of typing speed

In 2015-2017, I practiced English ten-minutes training every day to strengthen base speed and endurance. In 2018, I adopted TypeWell English for increasing base speed. Since 2019, I have used TypeWell Japanese R.

I kept practicing TypeWell Japanese R from last June to the fast half of March. It is considered that base speed training through TypeWell Japanese R was helpful for increasing speed of French, Italian, and Japanese. On the other hand, the quantity of English typing training clearly decreased. Deterioration of English typing skill affected almost all fast languages. I felt that something was wrong when practicing Portuguese ten-minutes texts in March. Moreover, I found my fingers did not move smoothly when typing Spanish, Germany, and Dutch. Such uncomfortable feeling continues to the online contest in April.

From December to February, I kept practicing one-minute trainings of Turkish, all French, original Italian ten times a day. I called this training type-all-untyped-texts project. Although I collected hundreds of one-minute texts respectively, I found that I had typed few of them yet. In the result, I covered about 95% of Turkish and original Italian, about 75% of all French texts. I moved on to ten-minutes trainings in the early stage for these languages. Moreover, I collected and input texts of Swiss Italian, Serbian (Cyrillic), Serbian (Latin) immediately after those languages were added in February and March. By such daily training, I could limit deterioration of Turkish, French, Italian, and Croatian typing skills as less as possible.

I continued re-inputting Taki training texts and detecting errors on every weekday. Consequently, I newly detected over 200 errors from approx. 1.5 million characters texts and improved accuracy. I will mention the details later.

(5) Training schedule for Intersteno 2021

As I felt base-speed training was enough, I started training in middle March. I practiced four languages a week considering the schedule for online contest. The concrete training schedule is shown as table 5. I intended to practice 16 languages except Japanese in four weeks.

[Table 5: Training schedule]

PeriodLanguages
3/15-21Czech, Slovak, Portuguese, Spanish
3/22-28Russian, Hungarian, Turkish, English
3/29-4/11Croatian, Romanian, Italian, French
4/12-18Polish, Finnish, Germany, Dutch

I have worked from home since last April due to Covid-19. Then, I was able to start training immediately after the work on weekdays. Thus, I started from rehabilitation by one-minute trainings. After that, I gradually introduced ten-minutes trainings assuming the competition. Just like last year, I thoroughly eliminated fruitless trainings such as simple ten-minutes typing for meeting the quota. Thus, average points of ten-minutes trainings are almost the same as those of last year. However, one-minute or ten-minutes training texts of Spanish, Finnish, Slovak, Polish, Hungarian are extremely limited. Then, I cannot practice first-sight texts as I can easily memorize part of texts. Then, I sometimes joined TypeRacer races for first-sight texts.

Table 6 shows the comparison of training quantity for Intersteno online championship 2015-2021. The training quantity in 2021 is slightly more than that in 2019, but much less than that in 2016, 2017, 2018. However, I could practice necessary and sufficient adjustment for most of languages as I have two months for training.

[Table 6: Number of ten-minutes trainings]

Language2015201620172018201920202021
French74414939(+6)12(+3)
Italian6111611510(+5)13(+3)
Portuguese62114151413(-1)14(+1)
English1124586151751(+44)17(-34)
Finnish1786855(-)4(-1)
Turkish6179155657(+51)7(-50)
Spanish12149142013(-7)13(-)
Romanian811184079(+2)9(+3)
Dutch16121611611(+5)8(-3)
Croatian72516151215(+3)9(+3)
Germany162016969(+3)8(-1)
Slovak1281214106(-4)7(+1)
Polish911101367(+1)4(-3)
Czech121316351110(-1)13(+3)
Hungarian6331318626(+20)9(-17)
Japanese924103695(-4)9(+4)
Russian03594441467(+53)7(-60)
total160552375598147323(+176)163(-160)

* The number above includes championship and speculation investigation.
* The number of each year includes all ten-minutes trainings from the next day of previous online championship to the last day of online championship this year.

(6) The sequence of 16 languages in the championship

It is clearly inefficient to type all 16 languages competition texts in one or two days. Repeated ten-minutes concentrations result in severe physical and mental fatigue. Moreover, less quality and quantity of training would lead to confusion of diacritical marks and frequent sequences among different languages.

For the Intersteno online championship 2021, I secured 42 days, from April 1st to May 12th. In fact, I started contest on March 28th because I got contest ID earlier than I thought. So, I successfully adopted the strategy to postpone not-well-prepared languages to national holidays in May. Since late April, I slightly reduced the load to three languages per week. Then, I secured quality and quantity of training for each language, and consequently avoided deterioration of points as less as possible.

The training schedule for online championship is shown as table 7.

[Table 7: Training schedule for 2021 online championship]

PeriodSlow languagesMiddle languagesFast languages
3/22-28(Russian, Hungarian) Turkish, (English)
3/29-4/11 Croatian, RomanianFrench, Italian
4/12-18PolishFinnish, GermanyDutch
4/19-24PolishFinnish, Portuguese 
4/25Czech, SlovakPortugueseSpanish
4/26-30Czech, Slovak Spanish
5/1-2Hungarian English, Japanese, Spanish
5/3-4Hungarian Japanese

* Hungarian and Turkish KLs are similar (öü are allocated on wq).
* The KLs of French and Italian are the same. Also, I have abundant experience of them.
* Czech and Slovak KLs are similar (some diacritical marks are allocated on 34567).
* Spanish, Czech, Slovak, and Hungarian KLs are similar (áéí are allocated on []\).
* Germany and Dutch are included in Germanic.
* The KLs of Germany and Finnish are the same.
* Polish and Croatian are included in Slavic.

Basically, I have used almost the same schedule since 2016. I dispersed fast languages (English, French, Italian, Turkish, Spanish, and Dutch) and slows language (Russian, Hungarian, Polish, Czech, and Slovak) as far as I could. I intended not only to keep speed of fast languages but also to raise the base speed of slow languages.

In addition, it is efficient to practice languages with high affinity or languages using the common KLs in the same week. On the other hand, I took care that I did not type confusing languages (e.g., French and Spanish, Hungarian and Germany, Polish and Czech) in the same week. By the way, Czech and Slovak also have confusing part. However, I thought that I could have more advantage by the synergy of similar KLs. Therefore, I decided to practice Czech and Slovak in the same week.

……

The final schedule for Intersteno online championship 2021 is shown as table 8. As I have work on weekdays, I participated in the championship on holidays.

[Table 8: Schedule for 2021 online championship]

Date / timeLanguageTotal pointsDifficultyNotes
3/28(Sun) eveningTurkish5276normalToo nervous. Wrong line break position in middle phase
4/10(Sat) eveningFrench11296normalCareful enough to reduce errors and loss
4/11(Sun) eveningItalian17122normalToo many mistyping in first half, ideal acceleration in second half
4/11(Sun) nightCroatian22172hardToo many mistyping in first half, wrong line break position in second half
4/11(Sun) nightRomanian27283normalLuck of training. Losing focus in the final stage
4/18(Sun) eveningGermany32085hardFailed as I reached peak at the training just before
4/18(Sun) eveningDutch37137hardToo nervous. Too many mistyping, wrong line break position two times
4/24(Sat) eveningFinnish42416normalBest condition. Felt collapsed at the middle but recovered
4/24(Sat) eveningPolish47139easyToo exhausted in the final stage
4/25(Sun) eveningPortuguese52853easyIncreased speed while emphasizing accuracy
4/30(Fri) pmSlovak57591hardEndured by training effects enough
4/30(Fri) nightCzech62204hardSo nervous that fingers often stopped
5/2(Sun) amSpanish67424normalSo nervous that fingers were cold, at last mistyped space key (-100 points)
5/2(Sun) pmEnglish73078normalSo nervous that left hand made lots of mistyping
5/4(Tue) amJapanese77585easyNot so nervous. But left hand was teased by Roman input method
5/4(Tue) pmHungarian82144normalAccelerated on frequent [egészség...] (health)


5. Tactics

(1) Collecting, inputting, re-inputting training texts

I collected one-minute and ten-minutes training texts from the Taki software. The results are uploaded as
IS_research.zip. I have done this in order to not only investigate frequency of diacritical marks, but also grasp cause and tendency of errors. In the training for Intersteno online championship 2015, all I could do was to guess the cause of errors from result screen. Sometimes I could find possible answers by inputting words including errors into Google translation or Google search. However, I could not find the correct words in all cases. Then, the training becomes too inefficient.

Then, I collected 1,961 one-minute texts and 535 ten-minutes texts (added 584 one-minute texts and 22 ten-minutes texts in this year). Moreover, I typed all collect training texts and organized them using spreadsheet so that I could accurately grasp the cause of errors. The detailed method is shown as table 9.

[Table 9: How to collect and analyze training texts on Taki software]

(1) Register a very long word in the Japanese word conversion dictionary.
- For example, register the word "あいうえお...だぢづで" (including 60 Japanese Hiraganas) on "p" (double-byte p).
- It is because Japanese Input Method Editor (hereinafter "IME") in Windows 7 and Windows 10 can preserve words which contains maximum 60 double-byte characters.
- Use double-byte characters because they can fill the text field twice as fast as single-byte characters do.

(2) Capture all texts.
- First, capture the screen soon after the training text appears on Taki software.
- Input "p→Space" by Roman input method using Japanese IME. Then, the registered word above will appear on the text field.
- Repeat inputting the word seven times, and stroke Enter key twice so that the input result will be settled.
- Then, the training text in the upper screen will be scrolled about one page.
- If you do not have Japanese IME, repeat typing "asdfjkl;" as fast as possible in order to scroll the text.
- Capture the screen again.
- Repeat the same works until text scrolling on the upper screen stops.

(3) Concatenate captured images into one image using Microsoft Paint.
- On average, one-minute text includes three images, ten-minutes text includes 25 images.
- Delete unnecessary parts.

(4) Type training texts into Notepad watching the image.
- Set font Courier New (MS Gothic in Japanese), font-size 12pt.
- Adjust the width of Notepad so that line break position is the same as that of Taki screen.
- To preserve diacritical marks, set character encoding "Unicode".

(5) Paste input results to spreadsheet.
- When using Windows 7, delete blank lines which was made by pasting.
- When using Windows 10, enter line breaks before pasting.

(6) Paste input results from spreadsheet to Google translation screen.
- Paste the text every 15 lines.
- If there are strange translation results, there may be some input errors. Then, compare the text with images and correct errors if necessary.
- Re-input texts of difficult languages, and compare the result with previous results using EXACT function to detect errors.

(7) Search first three lines by Google to find the source.
- Search each line using quotation marks.
- Definite sources are useful to detect errors and to anticipate competition texts.

(8) Paste all texts from spreadsheet to Microsoft Word and count characters by tis function.
(9) Investigate frequency of diacritical marks and rare characters such as qwvx using replace function of spreadsheet.
- Macro is also useful.
- Frequency analyses tool on web recommended by Tsukasa@dvorak-san is fantastic.

Moreover, I re-input almost all texts including but not limited to difficult languages. I have re-input 1,709 one-minute texts and 528 ten-minutes texts, including about 7.0 million characters in total. In this year I added 414 one-minute texts and 102 ten-minutes texts, including about 1.5 million characters. In addition to that, I utilized EXACT function to compare re-input results with input results in the past. Then, I detected over 2,200 errors in total (increased about 200 in this year). As the result of such re-inputting, it is possible that residual error rate decreased from about 0.05% to less than 0.001%.

(2) Examining/implementing problems and diacritical marks of training texts

Training texts in Taki software contains problems shown as
table 10. I reported over 200 errors in total to Intersteno person in charge soon after I found them. I intended to improve accuracy of training texts and efficiency of training. Also, they contain diacritical marks and special marks shown as table 11. I examined such problems and diacritical marks and reflected them to my original KLs before 2018. I have utilized these KLs this year.

[Table 10: Problems in Taki software]

No.ExplanationLanguages
1including few ten-minutes texts,
so it is hard to practice first-sight texts
Finnish (2), Polish (3), Slovak (3),
Hungarian (4), Spanish (5)
2spelling errorsEnglish (e.g., "competitition", "artisric")
3including diacritical marks
unfamiliar with the language
English (éçàèìò), Hungarian (äõû), Turkish (îûþ),
Croatian (ðæè), Romanian (ãǎ),
Italian (î), Portuguese (ïö)
4sometimes no texts shown in upper screenEnglish, Croatian, Italian,
original French, Belgian French
5including mahjong-tile-like mark (Win7/Vista),
or blank rectangle (Win10)
all Germany, Italian, all French,
English, Belgian Dutch
6including black diamond mark
contains white [?]
Belgian French, Italian
7apostrophes become errors in some textsEnglish, French, Swiss French,
Italian, Turkish, Romanian, Slovak
8hyphens become errors in some textsJapanese, English, Hungarian, Croatian,
Romanian, French, Italian, Russian, Slovak
9including other special markssee table 13
10some one-minute texts are too shortTurkish
11some ten-minutes texts are too shortall French, Turkish
12in some error cases,
the end of each line is hidden in result screen
English, Italian, Polish, Turkish,
Finnish
13unconfirmed (under-conversion) texts
remained when finishing become errors
Japanese
14including one-minute texts of other languageRomanian, Swiss Germany
15at the start, [`] is input automaticallyEnglish, Hungarian, Czech,
Dutch, Slovak

As to No.7, by typing [Shift + 7] on Japanese KL or ['->Space] on US international KL, ' (apostrophe) is displayed. However, some texts contain ’ (single quote). Single quote can be barely distinguished from apostrophe because single quote has slight angle in Courier New font.

As to No.8, by typing [-] on Japanese KL or US international KL, [-] (hyphen/minus) is displayed. However, some texts include [–] (en dash) as well as hyphen. Some Japanese texts have [―] (two-byte dash). It is almost impossible to distinguish these marks in Courier New font. By the way, some Russian texts include [—] (em dash). The mark can be distinguished from other similar marks because it is slightly long.

As to No.9, most of special marks can be input by Curry KL shown as table 13. My original KL is also useful. On the other hand, competition texts do not contain these special marks in principle (exceptions: In 2016, all French competition texts included two unfamiliar quotation marks that are not included in the table below. In 2018, Turkish text included [¬].). Therefore, some people may ignore irrational deduction in the trainings and assume that there are no such bugs in competition texts.

No.15 occurred in English in the online contest 2017. It seems that I unconsciously typed [Alt + `] in order to change input method from Japanese two-byte input method. Then, [`->It] becomes [Ìt], for example. As my KL is based on US international KL, vowels will have grave accent, while consonants are preceded by [`]. It is important to take a few seconds for seeing input texts before going ahead.

[Table 11: Special marks in training texts of Taki software]

MarksCurry KLDq KLLanguages
Shift + ' -> 1' -> 1Turkish, English, Portuguese,
Slovak, Russian
Shift + ' -> 2' -> 2Turkish, English, Portuguese, Slovak,
Russian, Romanian, Finnish, Hungarian
(unallocated)' -> 3Hungarian, Romanian
£Alt Gr + $' -> 4English
Alt Gr + 5' -> 5English, Italian, Finnish
Alt Gr + 9' -> 6English, Slovak
Alt Gr + 0' -> 7English, Slovak, Turkish, Italian, French,
Belgian French, Romanian
«Alt Gr + [' -> 8all French, Croatian, Russian
»Alt Gr + ]' -> 9all French, Croatian, Russian
ºAlt Gr + :' -> 0Italian, Spanish, Portuguese, Romanian
ª(unallocated)' -> -Spanish, Portuguese
ǎ(unallocated)' -> aRomanian
left Alt +
0151 on numeric keypad
' -> mRussian
left Alt +
0150 on numeric keypad
' -> nEnglish, Hungarian, Croatian, French,
Romanian, Italian, Russian, Slovak, Finnish
þAlt Gr + t' -> pTurkish
§Alt Gr + S' -> sPortuguese
Tab(unallocated)' -> tTurkish
œAlt Gr + x' -> xall French
æAlt Gr + z' -> zCroatian
¿Alt Gr + /' -> /Spanish
°(unallocated)` -> 0Portuguese
(unallocated)` -> .English, Belgian French, Italian, Romanian
¬Alt Gr + \` -> \English, Turkish
½Alt Gr + 7Alt Gr + 7Romanian

* Curry KL is the KL created by Curry-san in May 2015. It is also based on US international KL.
* Dq KL is my original KL. Only English KL uses dead key [`] instead of ['].

(3) Selecting dialects

When selecting dialects in Germany, French, Dutch, Italian and Croatian, I utilized my training texts collection. In concrete, I intended to improve efficiency in trainings and earn more points in competitions by selecting dialects with less bugs and diacritical marks.

Germany includes original Germany and Swiss dialect. I have adopted original Germany since 2019. The main difference is Swiss dialect uses [ss] instead of [ß] (Eszett) in principle. Then it takes shorter time to learn how to type Swiss dialect, which contains fewer diacritical marks. However, in the long run, physical burden will be less by memorizing the position of [ß] rather than enduring [ss], repeating the same key. In fact, there are few differences, because the frequency of [ß] is low. By the way, Swiss Germany competition texts in 2016 and 2018 included plural [ß]s. Then, it is necessary to prepare the KL including [ß] and grasp its allocation.

French includes original French, Swiss dialect, and Belgian dialect. I have adopted Belgian dialect including the least bugs since 2016. But I could not find essential differences among dialects in my investigation. After that, Intersteno technician taught me that original French and Swiss dialect often uses a space before marks such as [;:!?], while Belgian dialect does not. It was good for me to select Belgian dialect, which has familiar speculation to me.

Dutch includes original Dutch and Belgian dialect. But I could not find essential differences among dialects in my investigation. The reason was that some texts of both dialects were exactly the same. However, I found that some diacritical marks have different points among dialects. Especially, [.] in Belgian dialect has two points. So, I can earn totally about 1% more points. I have adopted Belgian dialect since 2019, as I found this speculation.

Italian includes original Italian and Swiss dialect, which was added in February 2021. There are no essential differences in both dialects. Although some diacritical marks have different points between dialects, they are rare. Therefore, the impact to total points is limited.

Non-Croatian typists can select Serbian (Latin) or Serbian (Cyrillic) instead of Croatian. Serbian (Latin) is almost the same as Croatian. Serbian (Cyrillic) is similar to Russian, but more difficult because it includes additional diacritical marks.

(4) Selecting keyboard

In Intersteno online contest 2017-2019, I mainly used Topre REALFORCE106 (All 30g). except for Japanese Kana input method. It was because I was confident of increasing speed with fewer errors by All 30g. In fact, total errors in the online contests 2016-2019 were 47->28->35->34. The averages per language were 2.76->1.65->2.06->2.00. Errors clearly decreased compared to 2016.

I started experiment of applying All 30g on Dec. 19, 2016. In 2003, I could not handle this keyboard efficiently. It was because very slight touches on adjacent keys easily become errors. However, I have gradually decreased such errors by All 30g through improving accuracy using Topre REALFORCE89U (45g/30g) over ten years. Then, All 30g is better as it helps me increase speed. In addition to that, the response of space key of 45g/30g deteriorated through overuse of many years. That was the reason I adopted All 30g in Internet online contest 2017.

However, I used 45g/30g again in Japanese contest on May 3, 2019. This is because All 30g was also broken through overuse of many years.

In October 2019, before increasing Japanese income tax (from 8% to 10%), I procured new keyboard, Topre REALFORCE (R2TLSA-JP3-IV). I selected this keyboard on the criteria of All 30g and having no numeric keypads. This keyboard was useful for making new record in TypeWell Japanese R. However, it has significant disadvantage of increasing errors due to slight touch of adjacent keys. Thus, I was afraid that it is not useful for Intersteno. However, I resolved this problem by setting APC (Actuation Point Changer) "2.2mm for frequently using keys, 3.0mm for other keys".

(5) Arranging other environments

I intended to systematically eliminate human errors by creating and following
operation manual for preparing and conducting championship.

There is a fatal bug in Taki software. The shortcut key [Ctrl + z] on Firefox deletes all input results. Then, it is indispensable to disable this shortcut key or to use another browser. I installed Google Chrome. By the way, [Ctrl + z] on Google Chrome deletes only one character typed recently. If you dare to use Firefox, you had better temporarily disable Ctrl keys using ChangeKey or KeySwap.

Even if you use Google Chrome, Ctrl keys should be disabled. It is because there are too much harmful shortcut keys shown as table 14. Especially, [Ctrl + w] (close the tab or window) and [Ctrl + r] (refresh the tab or window) are fatal. For example, when typing a capital letter, the little finger of left hand may hit left Ctrl key quite easily if it slips downward from left Shift key.

[Table 14: Harmful Ctrl shortcut keys on Google Chrome]

Shortcut keysAdverse effects
Ctrl + wClose the tab or window (FATAL)
Ctrl + rRefresh the tab or window (FATAL)
Ctrl + tOpen a new tab
Ctrl + uOpen a new tab and show the source
Ctrl + oOpen a file
Ctrl + pOpen the print screen
Ctrl + sOpen the save screen
Ctrl + dAdd a bookmark
Ctrl + f/gOpen the search screen
Ctrl + hOpen a new tab and show the history
Ctrl + jOpen a new tab and search downloads
Ctrl + e/k/lMove to search window
Ctrl + ;Enlarge font size
Ctrl + nOpen a new window

In addition, if you use different KL for each language, it is indispensable to disable shortcut key [Shift + left Alt] which changes KL (in case of using Firefox, Windows Vista). Otherwise, KL may be changed into unintended one at unintended timing due to errors of capital letters or marks using Shift key. The detailed settings are found by Google search. For example, I referred to How to disable the Windows 10 language shortcut?.

Furthermore, font size can be enlarged by [Ctrl + ;]. This shortcut key is almost indispensable when typing Italian, Spanish, Portuguese, Hungarian, Turkish, Czech, and Croatian, which contain two or more type of i, í, ì, ı. By the way, text screen size is small in www.intersteno.org. So, it is necessary to adjust it using mouse immediately after starting. On the other hand, such adjustment is not necessary in www.intersteno.it. I recommend the original Italian website because it saves a few seconds at the beginning.

Before starting the competition or training, applications and browser tabs not in use must be finished. In my case, regular reception of e-mails by Outlook sometimes froze my browser when typing. In addition, it took over ten seconds to display Japanese conversion candidates because Firefox plug-in went out of control. In case of such accidents, there are no relief measures such as retry. The points may be calculated at the place of freezing, or zero, which means disqualification. Furthermore, it is recommended to stop background processes such as Windows Update, Live Update, Java Update, and regular scan of security software in the ten minutes of the competition.

Finally, it is necessary to reload the Taki software tab or window immediately before participating in the contest. This is because the contest session will reach timeout after some time and be changed into practice mode.

……

In addition, I have been annoyed with automatic loss of focus from the input field on Windows 10 computer while typing since 2019. The trouble occurred in the contest in Finnish and Japanese in 2019 as well as Romanian in 2021. At first, I considered that it results from some shortcut keys including left Alt key, Alt Gr key, or left Ctrl key. Then, I tried to resolve the trouble by removing keytop of these keys, or temporarily disabling these keys by software. Moreover, it seems that there are other causes such as interruption by security software or printer driver. So, I introduced My Window Logger to find the causes but have not succeeded yet.

On April 7, 2020, I raised home electricity contract from 30A to 40A. If the circuit breaker trips, my computer keeps working with its battery. But the router is turned off, which leads to lose internet connection. It is dangerous enough to lose points on internet contest.

(6) Participating in the contest

Needless to say, the competition should be conducted with best physical and mental condition, at the best time. In my case this year, holiday evening and after dinner until 8:00 p.m. were the best time. This is because there are almost no works, raising children, and housework. Also, my physical condition is the best. On the other hand, the competition should not be conducted after work weekdays, after mountaineering, with not enough rehabilitation, with accumulated eye strain or fatigue of brain, arms or fingers.

In addition, it is recommended that the number of competitions per one set (about an hour) be one, the number of competitions per one day be at least two. However, I conducted two competitions per one set and three competitions per one day if I was well-prepared and in good shape.

Moreover, it is important to exercise ten-minutes training immediately before competition. There are two reasons for this. First, the KL is certainly set for the language. Second, it is useful to avoid confusion with other languages. It is essential to adequately get used to frequent words or sequences for the language. However, it is also indispensable to avoid fatigue in fast languages such as English and Japanese. So, do not type ten minutes at maximum speed, but save the energy for the competition. For instance, stop at five minutes or just type five one-minute trainings. Since 2018, I prepared and properly used a few policies such as "emphasize speed in first half, emphasize accuracy with less speed in second half", or "emphasize speed in second half only if the brain and brawn do not wake up in first half". In conclusion, type near the best speed and accuracy, but save the power in order to reach the peak at the competition.


6. Tips for each language

(1) Hungarian

For the Intersteno online contest 2016, I practiced Hungarian almost every day for about six months as I wrote in the strategy
(3) Intensive training of difficult languages. Here I show the process of evolving KL as table 12. Major breakthroughs are highlighted by blue bold characters.

[Table 12: Version up history of my Hungarian keyboard layout]

DateVersionEvolution
2015/10/18commonAllocated öü on [無変換 + py] so that I could type them by two strokes
(but not suitable for me)
2015/10/20commonChanged Curry KL.
allocated umlauts on dead keys ];
2015/10/21commonChanged Curry KL.
allocated umlaut and ñ on dead key [;], circumflex on dead key []]
2015/10/223Newly created original KL.
allocated őű on ]\ so that I could type them by one stroke
2015/10/234allocated áöü on []\
2015/10/246allocated áé on [] (based on investigation results of diacritical marks frequency)
allocated öü on dead key [;], allocated őű on dead key [\]
2015/10/277or8allocated öüő on wqx, allocated qwx on `=\
2015/11/49?
2015/11/1010allocated „ on ['->3]
deleted öü from dead key [;] (so that I could type ; by one stroke)
deleted ćĺńŕśýź, which are not used in Hungarian
2015/11/2712allocated ű on \, allocated x on [Alt Gr + v]
2015/11/2713allocated íó on ]^
allocated űqwx on ['->ysdv]
(At that time I did not know how to allocate qwx on ['->qwx])
2015/11/2714allocated űqwx on ['->jfgv]
2016/2/315allocated qwx on ['->qwx]
2016/3/2816allocated em dash on ['->m] (but it was unnecessary)
2016/4/1118allocated en dash on ['->n]
2017/3/219allocated õû on [`->ou]

(2) English

English is the most familiar language for me except Japanese. Also, I have typed English for many years. However, such conceits led to a humiliating defeat in the online contest 2015. I did not practice English intensively. Then, I got only 4897 points (539.7 strokes per minute, ten errors). Both speed and accuracy were unbelievable catastrophe for me.

I learned that English typing needs intensive practice. English is the most precious points source. However, in the practice for Intersteno online contest 2016, I was keenly aware that too much multilingual typing training adversely affects English typing skill. There are two reasons. First, training time for English decreases. Second, slower languages make English typing slow. Thus, I had practiced English ten-minutes training almost every day since January. I intended to keep the best feeling of typing English texts.

Moreover, I customized English KL. In 2015, I typed English using Japanese KL as I traditionally did. However, the allocations of some special marks in Japanese KL are different from those in US international KL as shown table 13. As I type almost all other languages by the KL based on US international KL, I have the risk of confusion when typing English. Therefore, I created English KL based on US international KL. However, I deleted dead keys ['] and [Shift + '], so that I could type ['] by one stroke. In addition, I allocated ’éç, which were sometimes found in training texts, on []\ so that I could type them by one stroke.

[Table 13: The difference of how to type special marks between keyboard layouts]

Special marksUS internationalJapanese
:Shift + ;'
''Shift + 7
"Shift + 'Shift + 2
@Shift + 2[
&Shift + 7Shift + 6
(Shift + 9Shift + 8
)Shift + 0Shift + 9
==Shift + -
_Shift + -Shift + \

* There are some special marks which were typed differently from Japanese KL. However, they were not included in Intersteno texts.

Furthermore, I gradually adopted optimizations of fingering as shown in table 14. I adopted only optimizations by which I could keep nearly-100% accuracy and increase speed.

[Table 14: Optimizations of fingering in English]

SequenceFingersObjectiveExamples
un, nu, huindex + middleimprove accuracyfundamental, under, tunnel, nuclear, number
po, opmiddle + ringimprove accuracyopportunity, power, potential, polluting, support, possible, policy
[exception] development, developing (ring + little). shopping (ring)
loring + middleimprove speedallow, follow, lower, ecological

(3) Russian

I created the original Russian KL based on US international KL. Although it referred to the Russian KL in Paraph KLs for Intersteno created by Paraphrohn-san, it was customized so that I could easily learn and adapt. For example, it was not comfortable for me to stroke some keys with holding down Alt Gr key, even if Alt Gr key was swapped to other key). Its reason is the same as Shift key decreases the speed of Japanese Kana input method. Therefore, I allocated dead key [:] so as not to use Alt Gr key. Also, I changed allocation of some one-stroke keys.

After getting used to Cyrillic characters and reduced recognition errors among з/э, л/п/и/ц, ъ/ь/ы, and Б/В, I started optimizing the KL. First, I allocated ьчйщ on \]z^ so that I could type each of them by one stroke. In my first Russian KL, I typed them using dead keys. But I found that their frequency was relatively high. On the other hand, as ъ is less frequent, I moved it to ['->z], using dead key ['].

This KL has two problems. First, э, which is next frequent after ьчйщ, need still two strokes. In early stage, I considered to allocate it on `. But such allocation makes time loss and fatigue as ` is far from home position. Then, I judged that I could prevent time loss and fatigue by inputting э with ['->s]. Second, п is allocated on [. Then, it is uncomfortable for me to type the frequent sequence "про" without exception. However, it was not easy to conclude which key is the most suitable for frequent п. Actually, there were the same problem in Finnish. Then, I decided to continue the consideration.

For the Internet online contest 2017, I practiced almost every day for about four month as I wrote in the strategy (3) Intensive training of difficult languages.

(4) Japanese

I heard that Japanese training texts would be renewed after the competition 2015. So, I intended to prepare for the context 2016 after the renewal of the texts. However, the text did not be renewed. Then, I prepared for the competition using the same texts as those of competition 2015.

Before 2017, I decided to only use Roman input method at an early stage. The reason is that Kana input method by Topre REALFORCE 89U/106 was not suitable for first-sight texts. Indeed, Kana input method gave me more speed than Roman input method. But I could not reduce errors in Kana input method. Especially, I could not eliminate the pattern in which the slight touch on "わ" key when typing "゛" is recognized and judged as an error. For example, "自動車(じどうしゃ; which means automobile)" becomes "じわどわうしゃ"). "゛" and "わ" are allocated on [[] and [0] respectively. So, if I type "゛" by little finger, my ring finger will slightly touch "わ". Since "゛" is frequently appeared and required high speed in Japanese Kana input method, it causes frequent errors.

By the way, [[] is frequently appeared in my Finnish, Romanian, and Russian KL. [[] is used to type [ä] in Finnish, [ă] in Romanian, and [п] in Russian. It would be a threat if it required the speed as Japanese.

It is possible to exterminate this error if I practice the same text over 300 times like the preliminary online contest of MaiPaso. However, I concluded in 2017 that it is impossible to exterminate this error when typing first-sight texts.

The main difference from 2016 was the new version of Japanese IME by procuring my 8th computer (Windows 10). The new IME could convert "安倍晋三" (the name of previous prime minister) at one time. It is very useful as the proprietary noun is frequently appeared in training texts. On the other hand, the number of conversions for other words differed slightly. So, I had to memorize the numbers again by training repeatedly.

In 2018, I aimed at 5,000 points by practicing Kana input method again using All 30g. I gradually increased points and finally exceeded 5,000 points which was not possible by Roman input method. This is because I gradually memorized frequent errors and how to prevent them by practicing the same text repeatedly. On the other hand, I could not resolve the problem of slight touch on adjacent keys, including but not limited to "わ", at first-sight or near first-sight. As a result, my speed and accuracy did not exceed those of Roman input method. Then, I experienced a catastrophic defeat at the competition.

(5) Czech, Slovak

Referring to popular KLs of both languages, I created original KL imagining Japanese JIS-Kana KL. That is, I allocated frequent diacritical marks on right end and the top row keys in order to type each of them by one stroke. The disadvantage of the KL is to deteriorate speed when two-strokes characters such as [wx34567] appear frequently. However, I judged that the advantage of typing frequent diacritical marks by one stroke would increase speed more.

The numbers of enhancing Czech and Slovak KLs using MSKLC were over ten times respectively, although they are less than those of Hungarian and English KLs. First, I allocated frequent acute accents so that I could type each of them by one stroke. Second, I allocated diacritical marks on [wx]. Third, I allocated some diacritical marks on the top row keys. Third, I allocated rare special marks such as single quote and en dash. I could not realize such evolution at one time. I had improved my KL utilizing knowledge and experience from practices of other languages. Also, I had adjusted the KL to my brain and fingers for half years.

(6) Finnish

Finnish in Intersteno is very difficult because Taki software has few ten-minutes texts. Surprisingly, there are only two texts. Then, you will gradually memorize frequent words and frequent errors through several ten-minutes training. Thus, ten-minutes texts are not suitable for preparing the competition, which uses first-site text.

Therefore, I got used to first-sight texts by typing 22 one-minute texts. I did not finish by only typing one minute (about 550 strokes). I re-input each text which contains about 1,000 characters. Moreover, I utilized 2017 and 2018 competition texts I collected. TypeRacer is also useful. In order to get used to Finnish KL, it is helpful to practice Swedish which can be typed by the same KL.

In 2019, I found that typing Roman characters in "Kanji" of TypeWell Japanese R with limiting errors was helpful for reading Finnish. By practicing "Kanji", I improved my ability to distinguish double strokes of the same key from single stroke, and to prevent frequent misrecognition between [k] and [t]. Then, I adopted "Kanji" as part of basic training of Finnish.

(7) French

I collected and input 271 one-minute texts and 98 ten-minutes texts of Belgian dialect in 2017. Total strokes are over 1.2 million including original French and Swiss dialect. Total strokes correspond to approx. 240 ten-minutes training (assuming 5,000 strokes per a training). Such abundant training significantly helped me increase points since 2017.

Furthermore, I intended to improve speed by making the most of characteristics that French has lots of English-like words. I made a list of French words whose spelling is similar to, but different from English and decreased errors with perseverance. Some of such words are shown as table 15.

[Table 15: French words which is confusing with English]

FrenchEnglish
histoirehistory
territoireterritory
mémoirememory
respiratoiresrespiratory
nécessairenecessary
militairesmilitary
environnementenvironment
gouvernementgovernment
mouvementsmovements

FrenchEnglish
rythmerhythm
algorithmealgorithm
humainehumane
organisationorganization
particulierparticular
conflitsconflicts
membresmembers
européenneEuropean
recommandationsrecommendations
jugementjudgment
comparaisoncomparison

(8) Turkish

For Intersteno online contest 2018, I decided to intensively strengthen Turkish, and kept practicing it almost every day for about five months as I wrote in the strategy (3) Intensive training of difficult languages. I put emphasis on accuracy, but improved speed gradually.

I collected and input 97 one-minute texts and 108 ten-minutes texts. total strokes are over 1.2 million. Total strokes are almost the same as those of French in 2017 and correspond to approx. 240 ten-minutes trainings (assuming 5,000 strokes per a training). Abundant training significantly helped me increase points.

As to KL, I allocated [İ] on [=] to type it by one stroke. Although it is far from home position, there is no problem as it rarely appears. I resolved confusion between i and ı by considering the characteristics that the same vowels tend to be repeatedly used in a word.

Furthermore, I memorized some frequent words, including difficult ones such as [çeşitli] (various). By the way, Turkish training texts tend to contain similar words repeatedly. If you get used to such words, you can smoothly accelerate typing in second half.

(9) Italian

At the time of Intersteno online contest 2017, I did not finish improving speed of typing Italian. However, I got used to my original KL through intensive training of French, as the KL is common with that of French. Moreover, I collected and input 55 ten-minutes texts in June - July 2017, and February 2018. Total strokes are over 0.84 million including one-minute texts. Total strokes correspond to approx. 170 ten-minutes trainings (assuming 5,000 strokes per a training). Abundant training significantly helped me increase points. Especially, I got used to frequent sequence such as [-zione], [-zza].

(10) Dutch, Germany

I wrote how to select dialects in selecting dialects. In addition, I made a list of words whose spelling is similar to, but different from other languages. Some of such words are shown in table 16.

[Table 16: Words which is confusing with other languages]

DutchGermanyEnglishFrench
ArbeidsArbeits(labor)(travail)
hebbenhabenhave(avoir)
organisatieorganisationorganizationorganisation
marktmarktmarketmarché
economieökonomieeconomyéconomie
ParlementParlamentparliamentparlament
transparanttransparenttransparenttransparent


7. Sentiment and reflection

I participated in 16 languages including plural difficult languages in limited term. As a result, I got total 82,144 points (minus 207 compared to last year). The total score was the lowest in recent five years. However, it was the 2nd place next to last year's score in 16 languages.

Of course, I do not have optimistic prospect that I can reach the same place by these points in the Intersteno online contest 2022. I do not think international typists, who have seen remarkable progress by Japanese typists since 2015, just look on idly. Especially, I anticipate more severe competition with Sean and Andrea. Absolutely, I do not think I can win the competition by keeping only accuracy without speed. Moreover, new talented rivals may appear one after another. To reach further frontier, I have to analyze failures and successes of this year.

(1) Deterioration of English typing skill

I consider that the biggest cause of failure in 2021 was deterioration of English typing skill. This was really severe strategic mistake. It is nearly impossible to make up for the failure even if I utilize some tactics. For ten months, I kept imbalanced training in which I stroke Esc key immediately after I found left-exhausting words in TypeWell Japanese R. Then, I lost muscle and endurance to keep speed on capital letters as well as left-exhausting words such as "were", "was", and "decrease". As a result, it seriously impacted not only on English but also on Spanish, Germany, and Dutch.

Moreover, I was annoyed with the situation in which my recognition of words and my neural circuit stop immediately after I perceived an error. Then, I could not prevent errors continuing, which led to time loss for about ten seconds. The reason is the imprinting of repeated serious failure in the last line of TypeWell Japanese R due to lack of endurance. I emphasized only speed and time, hitting Esc key so much. Thus, I lost endurance. Although I also emphasized speed in Intersteno, I should not have typed in the same manner of TypeWell Japanese R. I cannot use Esc key to restart in Intersteno.

Furthermore, I cannot overlook adverse effect on Japanese. Indeed, my typing speed of fast words in TypeWell Japanese R grew slightly. On the other hand, my typing speed of slow words degenerated severely. Then, I struggled with left-exhausting words such as "tikara", "kuwawaru", and "kumiawase" at the competition.

(2) Lack in practice of first-sight texts

I got used to some frequent words in training texts (e.g., "International Labor Organization") too much, while not getting used to other words. Such imbalance was significant in especially Germany, Dutch and Czech.

I utilized TypeRacer to compensate for deficiency of first-sight texts training in Finnish, Polish, and Slovak, which have less ten-minutes training texts. It is true that TypeRacer has disadvantages such as "including many too short texts that can be finished within 30 seconds", and "hard to distinguish ./,, l/I, i/j due to indistinct font". However, it was useful for practicing first-sight texts, which had been totally lack before.

On the other hand, I was too confident to think I need not practice first-sight texts in Germany, Dutch, and Czech in TypeRacer. This was because there were several practice texts in these languages in Taki software. In this meaning, do not be too confident after getting high points in trainings. It is obvious that I can get more points after getting used to frequent words. At the competition, I cannot get such effect from first-sight text.

By the way, I succeeded in French and Italian. This was because French and Italian had abundant training texts, and I got used to lots of words and sequences.

(3) Mental weakness

The higher typing competition level is, the more mental factor increases. There are some common points in defeated languages. When I experienced huge time loss in the early stage, I anticipated severe defeat and lost my fighting spirit. In the middle stage, I tried to recover by emphasizing speed, but I was so upset that mistouch increased. In the last stage, I lacked concentration and repeated adding errors in the last one second again and again. It was exactly the same pattern as crushing defeat in MaiPaso final on
Nov 23, 2007. I am really stunned by the fact that there has been no progress for 14 years.

(4) Adaptation and peaking for the competition

The quantity of training in 2021 was as less as that in 2019. Then, I had to make efforts to reach the peak at the competition. However, I failed in Germany. made private best record at the previous training of the competition. The point of the competition was 599 less than that of previous training. It was the deepest regret to repeat the same failure as Hungarian and Turkish last year.

After that, I made efforts to reach the peak at the competition, making the most of this reflection. Firstly, I made sure not to type competition forcefully. For example, in case that results of one-minute training are unstable, and that I felt some slumps in previous ten-minutes training. Secondly, I made sure not to type with full speed at previous practice in order to avoid fatigue.

In the competition, I emphasized speed as last year. For ten minutes, I kept aggressively increasing speed in the part which I found I could type fast. Furthermore, I enjoyed the process of typing each language, feeling nostalgia that it is last time I type this language in this year. Although I was nervous in several languages, I could utilize such tension for decreasing error. As a result, I made private bast record in French, Italian, Portuguese, and private second record in Finnish in the competition.

(5) The result of strengthening base speed

Improving base speed via TypeWell Japanese R significantly helped me improve base speed of Japanese texts. One reason is that some competition texts was easier to type than before. However, I think my base speed was improved to type such easy texts faster. By the way, I adopt some optimizations in TypeWell Japanese R so that I can maximize speed within only 30 seconds. For example, I type "koku" by "coku". However, I understood that "cocu" reduces more fatigue in ten-minutes typing such as Intersteno.

Deterioration of English texts typing skill affected French and Italian. But type-all-untyped-texts project, I mentioned in Strategies (4), helped me a lot.

(6) Increase in errors

Total errors were 48, increased by six from last year, and became the worst in recent six years. It was because I aggressively increased speed for all languages without fearing errors. As I anticipated more severe competition in future, I could not be satisfied with last year's speed. However, I could not compensate for 300 points deducted by errors by increasing speed.

In the competition 2021, there were three languages with zero error, which were private best record. In the three languages, I succeeded in reaching the peak at the competition. In the training, there were only 13 times in seven languages with zero error. In the 163 overall ten-minutes trainings (including competition), there were 31 one-error, 34 two-errors, 26 three-errors, and 26 four-errors. In comparison to last year's trainings, the possibility within four errors decreased from 85.4% to 79.8%, within two errors decreased from 50.2% to 47.9%, with zero error decreased from 12.4% to 8.0%. Errors were increased because I concentrated on increasing speed without fearing errors.

I succeeded in suppressing errors to some extent because I improved how to correct an error of line break position. Since 2018, when I found an error of line break position, I delete over one line and type the text again. Since 2021, I could significantly suppress time loss and decrease in points by such corrections because I could utilize mouse, arrow keys, Home/End keys. Fortunately, I did not experience huge time loss by an error of line break position in the competition 2021. Furthermore, I did not look back the previous line unless I found obvious errors in a word or a line. I put an emphasis on preventing decrease in speed by line break position check rather than reducing errors by such check.

……

All increase in points were not due to my typing skill. Especially, significant growth such as +375 points in Portuguese, +277 in Finnish, and +254 in French was mainly due to luck. Portuguese text was much easier to type than before. In French, I was not so nervous as last year and performed my full potential via daily re-inputting texts. In Finnish, I luckily finished typing with only one error. At the championship, we type first-sight text for ten minutes. Then, there are some factors of luck. As good luck does not come every time, it is necessary to steadily improve basic typing skill, that is, speed, accuracy, language skill, KL proficiency.

……

Finally, I had suffered from temptation to conduct cheating this year. When I was both an administrator and a competitor, I could conduct cheating as I like. For instance, I could register dummy competitors, collect competition texts in advance, repeat practices, and re-participate in the competition using other ID. Especially in Japanese, if I had executed these tasks, I would have got 1,000 more points. Also, if I was defeated miserably in early stage, I could delete my ID and re-register it. In the researcher's viewpoint, there is a temptation to collect and analyze competition texts for each language. To resolve this problem, I would like to improve management system next year and to prove my typing skill by keeping more points in future.


8. For the Intersteno 2022

I achieved the second part of my goal, "recapture the world champion". But I could not achieve the first part, "to break my record of last year". Also, it was far from my future goal, "85,000 points by 16 languages except Japanese, 90,000 points by all 17 languages". Therefore, I keep this goal again for the next year.

I do not know when my strategies bear fruit. I found that it is not enough to conduct intensive training of slow languages in order to reach over 90,000 points. It is necessary to improve speed and keep 6,000 points or more in fast languages.

But fortunately there are some rooms to earn more total points by evolving strategies and tactics. The limitation is much more than I thought before. The key point is to aggressively improve speed. There are still some rooms to strengthen the speed with less errors. Then, what should I do specifically to strengthen my typing skill for the next year?

(1) Learning each language

I believe this approach is quite useful ultimately. In my case, English proficiency positively affects English typing skill. I have immeasurable advantage because I can understand contents and guess contexts while typing English texts. As I recognize English texts not as list of random words but as meaningful sentence, I can get more speed with less errors. Although I am not so good at short-term typing such as TypeWell English, TypeRacer, and 10FastFingers even in Japanese typists, I am good at middle or long-term typing such as Intersteno and MaiPaso. In addition, my English proficiency positively affects typing French, Italian, Spanish, Portuguese, and Romanian, as those languages contain lots of English-like words.

Moreover, I felt the advantage that I could recognize where to split very long words in Germany, which I had studied as the second foreign language. I enjoyed similar advantage in Dutch, which I did not learn but is classified as Germanic.

On the other hand, I did not feel the advantage in French, which I had selected as the third foreign language but had not studied well. The main factor of growth of French points since 2017 was to get used to diacritical marks and some frequent words. It is not relevant with French fluency. Furthermore, it may take several years to newly learn the fourth or more foreign languages so that I can feel some advantages for typing. Then, learning new languages would be challenges with long-term view.

(2) Learning frequent words and sequences for each language

This is the most effective method in the short-term. If you practice ten-minutes training 100 times, you will adequately get used to frequent words and sequences of the language. For example, my Hungarian typing speed was clearly strengthened in the intensive training for the Intersteno online contest 2016. This was because I memorized some frequent words, which includes not only easy words such as "azonban" (but), "napjainkban" (today), but also difficult words such as "egészség" (health). Then, I finally exceeded 5,000 points in 2020. In the Intersteno online competition 2021, "egészség" appeared frequently. So, I typed them with top speed feeling nostalgic.

I practiced Russian intensively for the Intersteno online championship 2017. Although I could not reach enough speed, I memorized some frequent words. Training texts in Taki software contain frequent words such as "жизни" (life), "подход" (approach), "Москве" (Moscow), "экономических" (economic), "воспроизводство" (reproduction), "распределения" (distribution), производство (production). In 2020, I finally reached 4,500 points. I believe I can reach 4,500 points in the competition if I practice Russian further. There are still rooms to get used to Cyrillic letters, frequent words and sequences.

Other difficult languages are no exception. As a result, I made a record of 4,970 points in Slovak in 2018. Since 2020, I have broken private best record in Polish and Czech. I am confident in reaching 5,000 points in these languages.

As to middle languages, I have sometimes adopted intensive training. As a result, I have felt the speed and accuracy of Portuguese since 2019. Then, I exceeded 5,700 points in the contest 2021. In addition, I broke the barrier in Germany and Finnish. Both my private best record and average points of Croatian and Romanian were improved. It is expected to reach 5,500 points after a few years.

As to fast languages, I finally exceeded 6,000 points in French at the Intersteno online contest 2021. My points of Italian and Turkish also grow for the next step. Of course, it is not the limitation. I often get over 6,200 points in English training. Indeed, diacritical marks reduce speed in other languages. However, it is unnatural that the speed is reduced over 10% because Italian, Spanish, Dutch and Turkish do not contain so frequent diacritical marks. The root cause is that I have not got used to words and sequences of the languages yet. I get more points in French, which contains frequent diacritical marks, than other fast languages. It is because I got used to French because it contains many words which has similar spelling with English.

(3) Language portfolio next year

As shown in table 17, there are still rooms to grow for all 17 languages. However, I cannot anymore expect significant growth I experienced since 2019, because I have already collected and input almost all training texts. Then, all I can do is to gradually strengthen speed and accuracy through steady training.

As to Japanese, I would like to practice Roman input method continuously and intensively, considering speculation of transliterated version. On the other hand, if I type traditional Kanji version for some reason, I will study Kana-Roman hybrid typing with trial and error. For example, how about using Roman input method in the first half, while switching to Kana input method after getting used to frequent words? This is because Roman input method is suitable for first-sight texts, while Kana input method is advantageous to increase speed.

Except for Japanese, there are some rooms to grow in the languages which I made private best record or nearly private best record.

On the other hand, the growth of the point will slow down or decrease if the training quantity is decreased. For example, Romanian points decreased by 263 in 2020 because I practiced only nine ten-minutes trainings. There are rooms to consider which language should be trained intensively, and how many ten-minutes trainings I need to minimize decline of typing skills. I would like to strategically practice each language from the perspective of overall optimization.

[Table 17: Classification of languages from the viewpoint of rooms to grow]

CategoryLanguages
Too nervousTurkish, Romanian, Czech
Not enough speedEnglish, Dutch, Spanish, Croatian
Not enough accuracyDutch, Polish, Hungarian
Not enough consideration of KLJapanese, Germany, Russian
Private best at the competitionFrench, Italian, Portuguese
Near private best at the competitionFinnish

(4) Contribution to Intersteno

It is not only to improve private total points, but also to help maintain fair and equitable competition environment. I think it will contribute to the development of Intersteno in future.

First, there still exists problems in training texts. Although I reported lots of problems in 2015 and 2016, some of them are not corrected yet. In addition, I found more bugs after that. After putting them together to some extent, I would like to report such problems to Intersteno again and request correction persistently. At the same time, I think it is a good idea to enlighten how to input marks familiar to European typists but unfamiliar to Japanese typists, such as € and £.

Second, I would like to make a proposal for speculation of Japanese. It is extremely hard for non-Japanese typists to participate in Japanese. I do not think it is a good situation. In the contest 2020, I was defeated by Sean in 16 languages except Japanese. If I had typed Japanese (including Kanjis), I would have "won in 17 languages, but lost in 16 languages". Although the limitation of maximum 16 languages was set in 2021, Japanese typists are still slightly advantageous. The similar situation may also occur next year. On the other hand, I heard that "Japanese (transliterated)", Roman characters version (transliterated), were under testing and would be introduced in 2020. Consequently, the implementation of the transliterated version was put off after 2022. I think it needs trial and error and would like to help test it. Additionally, it is necessary to discuss the coefficient of Japanese points. I would like to make a proposal so that I help resolve the current inequality. Such trial and error will be helpful for introducing new two-byte languages such as Korean in future.


9. References and acknowledgements

Pocari-san: He led and realized introduction of Japanese to Intersteno online championship 2015. Furthermore, he burdened registration fee for 120 Japanese typists at his own expense. I heard that he had proactively negotiated with representatives of international countries and resolved twists and turns in order to continue including Japanese in the contest since 2016. Without his incredible effort, I would not have known Intersteno and participated in it. Thank you very much. I worry about him as I have not heard from him since May 2020.

Junior-san: He started multilingual typing training at the earliest stage in Japan to participate in all 17 languages including Russian, the most difficult language for Japanese typists. I suppose he experienced extraordinary hardship as a front runner - for example, learning popular KLs for all languages, and investigating how to input diacritical marks and special marks from scratch.

Curry-san: Since 2015, he has proactively offered information via Multilingual Typing Wiki, so that many beginners including me could easily participate in multilingual championship. Moreover, he held weekly TypeRacer events and 10FastFingers championship for each language to help beginners prepare for the Intersteno online championship 2017.

Tanon-san: For the Intersteno online championship 2017-2020, he proactively led Japanese typists by holding and participating in weekly TypeRacer events for each language. His significant growth since Oct. 2017 encouraged me enough.

Paraphrohn-san: For the Intersteno online championship 2016, he led Japanese typists by disclosing the aggressive target, "to be the world champion by getting total 80,000 points" in the early stage and achieving it. In concrete, he created and distributed paraph KLs for Intersteno. Also, he held weekly TypeRacer events for each language. Thanks to him, the level of participants including me improved much more than previous year.

Yada-san: taught me that the error "This page and application is only for use on computers, not available for mobile devices." in Taki software can be solved by expanding screen size or reducing font size.

Drizzt-san: taught me that the problem above did not occur on www.intersteno.it. I have typed trainings and competitions only on this website since Intersteno online championship 2016.

Tsukasa@dvorak-san: introduced frequency analyses tool on web, which is quite useful for frequency analyses of specific characters in the text. I have utilized this website by analyzing mainly Russian texts.

My dear family: Thanks to them, I used most of holidays and national holidays from April to May for Intersteno multilingual championship. These holidays are originally precious time for helping raising children, housework, and family events. Thank you! ... By the way, I am not a husband/father. If I keep joining online championship in future, I will have dilemma whether to use almost all national holidays in May every year.


10. Appendices

I disclose materials related Intersteno 2021. Please utilize these materials as needed at your own risk. Also, I will delete these materials if any problem such as intellectual property is occurred.


back to dqmaniac's homepage
back to typing website
back to the last typing diary