Pleco's Text Import and Flashcard Issues

hoshi

Member
I purchased the flashcard option, and while it supports importing entries from a text file, the import process has some serious shortcomings. The biggest problem is that it doesn't faithfully preserve the original dictionary data. Punctuation such as commas, middle dots in names, exclamation marks, ellipses, and other characters are stripped out or ignored. This is a major issue for a dictionary app, and it is disappointing that it has remained unaddressed for years, with no apparent sign of improvement.

Here are some examples of entries that are incorrectly treated as duplicates and skipped during import:

鞋好不好只有脚知道[鞋好不好只有腳知道] xie2 hao3 bu4 hao3 zhi3 you3 jiao3 zhi1 dao4 You only know whether a shoe fits by wearing it.
鞋好不好,只有脚知道

风里来雨里去[風裡來雨裡去] feng1 li3 lai2 yu3 li3 qu4 To work hard through all kinds of hardships.
风里来,雨里去

食不言寝不语[食不言寢不語] shi2 bu4 yan2 qin3 bu4 yu3 Don't talk while eating or in bed.
食不言,寝不语

马照跑舞照跳
马照跑,舞照跳

Pleco apparently considers these pairs duplicates, even though the punctuation changes their meaning, usage, or presentation. The same problem occurs with names, titles, and other entries where punctuation is significant.

Punctuation in names and titles is not optional

For example, the entries below are distinct, but Pleco seems to merge them or strip the punctuation, resulting in incorrect or lost data.

达·芬奇 → 达芬奇
The middle dot is removed, even though it is part of the original name format.

雅诗·兰黛 vs. 雅诗兰黛

雅诗·兰黛 — Estée Lauder, the person (American businesswoman).

雅诗兰黛 — Estée Lauder, the company (cosmetics and skincare manufacturer).

These need to remain separate entries.

艾德·伍德 vs. 艾德伍德

艾德·伍德 — Ed Wood, the person (American film director), as well as the film Ed Wood.

艾德伍德 — Ed Wood, the film title.

Again, these are distinct dictionary entries.

破·地狱 vs. 破地狱

破·地狱 — The Last Dance, the movie.

破地狱 — A Taoist ritual in which written names are burned and tiles are broken with an iron rod as part of a ceremony intended to guide the deceased to heaven.

东京 vs. 东京!

东京 — Tokyo, Japan; also historical names for Luoyang during the Han dynasty and Bianliang during the Northern Song, as well as Tonkin, a region of Vietnam bordering China.

东京! — TOKYO!, the movie.

人·鬼·情 vs. 人鬼情

人·鬼·情 — Woman-Demon-Human, the movie.

人鬼情 — A romantic relationship between a human and a ghost.

就这样 vs. 就这样…

就这样 — A literary expression meaning "That's it" or "So it ends"; also the title of a novel by Marguerite Duras.

就这样… — And Just Like That..., the American TV series.

德·桑蒂斯 vs. 德桑蒂斯

德·桑蒂斯 — De Santis.

德桑蒂斯 — DeSantis.

These are not interchangeable. For example:

朱塞佩·德·桑蒂斯 — Giuseppe De Santis, Italian film director.

罗恩·德桑蒂斯 — Ron DeSantis, American politician.

约翰·内森 vs. 约翰内森

约翰·内森 — John Nathan, American scholar and translator of Japanese literature.

约翰内森 — Johansen, a surname in its own right.

One is a specific person's name, while the other is a surname in its own right.

These are not merely cosmetic differences. Punctuation, spacing, and character formatting can distinguish people, organizations, titles, expressions, and entirely different meanings. Removing them or treating the resulting entries as duplicates can corrupt dictionary data.

Full-width alphabet conversion is another problem

Chinese text increasingly uses English abbreviations and acronyms directly, such as WTO, NBA, and similar terms. However, Pleco seems to convert Latin letters to full-width characters in some cases. For example, even if the original dictionary entry contains WTO, it may end up being converted into WTO (full-width characters).

This creates a mismatch when using Clip Reader, because the text itself contains ordinary WTO, not WTO. As a result, the entry is no longer matched.

The same issue occurs with mixed Chinese-English entries:

WTO[WTO] wto World Trade Organization
穿PRADA的恶魔[穿PRADA的惡魔] chuan1 PRADA de e4 mo2 The Devil Wears Prada

In the second example, PRADA is converted into full-width characters, so it fails to match the original text in Clip Reader as well.

Other examples include:

伊隆·马斯克[伊隆·馬斯克] yi1 long2 ·ma3 si1 ke4 Elon Musk

The middle dot is removed, resulting in 伊隆马斯克, which again fails to match the original text in Clip Reader.

The conversion process is painfully slow

The most frustrating part of creating a user dictionary from a text file is how long the process takes. With a large number of entries, importing can take close to an hour.

A self-made converter could reduce this process to a matter of seconds. However, even with such a converter, entries consisting entirely of Roman letters, such as WTO, cannot be handled properly and still have to be added through Pleco's built-in text import.

In practice, this means maintaining two user dictionaries on the phone: one created using my own converter, and another containing entries added through Pleco's built-in text import to handle the entries the converter cannot process.

For an app as mature and widely used as Pleco, these limitations are frustrating. Text import should preserve the source data faithfully, handle punctuation correctly, avoid false duplicate detection, and provide a much faster way to import large dictionaries.
 
Sorry, would you like us to refund your purchase?

Honestly, the reason we haven't changed this is not because we can't, but because people seem fine with it - you are as far as I know the first person to ever complain about most of these issues - and for all of your frustrations there are also substantial benefits, particularly e.g. given the ambiguity in how a lot of dictionaries treat punctuation. Changing this behavior at this point would mean breaking behaviors that a lot of other people are relying on and have not expressed any objections to. (you mention "for years" - had you yourself been waiting for that to be resolved? if so then I'm sorry you didn't write to us earlier)

Part of maintaining a mature and widely-used app is not springing changes on people unexpectedly, a lesson reiterated for me recently when we tried a more accurate but less reliably positioning OCR algorithm on Android and people revolted, so on iOS we rolled it out quietly for new users and did not touch a thing about old users' experience unless they went in and turned on the new system.

That being said, it seems like this basically boils down to four problems:
  1. Duplicate handling of punctuation. We don't currently offer a way around this - as I said, not something people have asked for - but now that you've raised an issue it's something we could consider making optional. This would basically just need to apply to duplicate checking and identical entry merging? Would you be OK with - as a consequence of this - a dictionary that did not differentiate between 雅诗·兰黛 and 雅诗兰黛 merging both meanings, including the wrong one, with your entry for 雅诗兰黛 and not with 雅诗·兰黛? (you could always delete or disable that dictionary for imports)

  2. Alterations to full-width characters; this should be fixed in our next major update.

  3. Reader searches of A-Z characters; the basic issue is that we don't currently make regular letters tappable in the reader because we offer a lot of mixed Chinese/English texts where that could get annoying. This could however be easily addressed by a setting if we can't figure out a way to intelligently limit it to actual Chinese texts.

  4. Slow user dictionary imports. This is not so much that the system is inherently slow as that it wasn't designed to scale well to large dictionaries (we thought people would be using it for word lists of a few thousand entries tops); that issue is already is fixed for our next major update. In the meantime honestly the format isn't that complicated, if you have an account with a decent LLM service I suspect you could hand it a backup of one of our user dictionary databases (they're regular boring SQLite files, nothing hidden) and it could write you an efficient desktop converter in minutes.
 
Back
Top