User talk:Maximilian Janisch

See also User talk:Maximilian Janisch/latexlist.

Manual texification of articles

Hello, I am a new user at EoM. I noticed that you and User:Ulf Rehmann were in the process of "texifying" articles that had only image-based formulae. I just wanted to ask what method is preferred to manually convert images that cannot be processed with MathPix to LaTeX, and to offer my help in the process of texifying articles that need it. Thanks. 07:25, 10 December 2020 (UTC)

Hello, sorry for the late reply! I can happily tell you that all formulas except for the articles listed here have been retexified successfully. So there is not much work to do (if you want, feel free to manually retexify one of these 68 articles, it is not very fun work though). Feel free to make changes to the content too, the EoM is in need of that :)

PS: If you want faster replies, it is safer to send me an e-mail at mail@maximilianjanisch.com :) --Maximilian Janisch (talk) 22:10, 16 December 2020 (UTC)

Nice to meet you. This is no urgent message, so I will leave it here. I will periodically remove entries from the "still to texify" list and add them to the "texified" list as I complete their texification. 06:26, 21 December 2020 (UTC)

Amazing and merry christmas! --Maximilian Janisch (talk) 19:28, 23 December 2020 (UTC)

Thank you for your correction efforts

Moved from User_talk:Liuyao.

Hi Liuyao,

thank you for your recent efforts in correcting mistakes made by the automatic Texification.

There were two types of Texification performed: The fully automatic type and the semi-automatic type. The semi-automatic type of correction was manually checked by humans, so it should be fine. There seems to be a systematic set of errors in the automatic correction though; Feel free to let me know about them and maybe we can fix those automatically across the whole encyclopedia.

Best, --Maximilian Janisch (talk) 23:18, 17 January 2022 (UTC)

Thank YOU for the great work! Happy to fix what I see, but if you can automate the process, the majority of the typos that I'm fixing are the following:

* subscripts (and superscripts) closing } misplaced, such as \sum_{n=1}^3 becoming \sum_{n}=1^3

* Some (but not all) appearances of \dots need , before and after

* linebreak after inline TeX, introducing unwanted spaces (maybe should remove all linebreaks unless for new paragraph or display TeX?)

** "hyphenated" terms such as $n$-dimensional appearing as $n$- dimensional

** misplaced space when inline TeX followed by an opening (

* \overline XXX gets a \; which I'm not sure should be fixed or not.

Some commutative diagrams are harder to fix, I'm afraid. See class field theory for an example. Liuyao (talk) 05:51, 18 January 2022 (UTC)

This is a good list: For background, I will note that there were 270000 formulas with no TeX code initially. For about four fifths of these formulas, the editor-in-chief of the EoM, User:Ulf_Rehmann, found the original Nroff codes. These are markup codes which can be translated to LaTeX automatically. All articles that were translated by him this way can be found at Category:TeX_auto. Apparently the translation contains some systematic mistakes, as for instance the one you mention above. I will try to see if they can easily be corrected automatically. For instance, it would be easy for me to adapt my previous programs in order to automatically replace all occurences of "\sum_{X}=Y^Z" by "\sum_{X=Y}^Z". (But some encoding mistakes may be hard to correct automatically.)

--Maximilian Janisch (talk) 15:40, 20 January 2022 (UTC)

Thank you for the explanation! I remember visiting EOM a few years back, and it's much improved now! I think it may be time for some automation on (internal) hyperlinks, which sounds like a natural task for Natural Language Processing. Liuyao (talk) 16:30, 20 January 2022 (UTC)

One "naïve" approach would be to collect a list of all names of articles in the EoM (this is very easy, the list already exists, cf. Special:AllPages) and to look in each article whether a word which is the name of another article appears there and if yes, to link it. This would not be more than a few days of work for me to go through the whole encyclopedia.

More sophisticated approaches (synonyms, NLP, etc.) would take longer --Maximilian Janisch (talk) 11:24, 21 January 2022 (UTC)