(2016-09-23, 4:56 pm)Xavier22 Wrote:
(2016-09-23, 9:49 am)ZorgBlub Wrote: Sorry for the delay.
I just published a new version to fix some bugs namely, the problem with the settings and the immersive mode.
It's absolutely great!
Thank you for the update. The koujien and daijirin J-J EPWING dictionaries work almost just how I want them to now. There are still a few issues but now I can use this program to do some reading so this will help a lot. Much appreciated.
If I were to suggest a few additional features, they would be in the following order, from most to least important, at least just for me personally. Apologies if any of these are ridiculous requests, I am not a dev so I don't know how difficult any of these would be to implement.
- Deinflection support for EPWING dictionaries. If I hover over "死んだ" in Koujien I only get results for "死" while Edict and Wordnet properly deinflect to "死ぬ".
- Ability to default to the dictionary that has the longest count of highlighted characters. For example, say I have two dictionaries active: EPWING Koujien, and Edict. If I tap on "異世界", all I get on Koujien is "異". I'd like to have it be able to auto-switch to Edict because that actually has an entry for "異世界". This would be very helpful for multi-kanji terminology as well as idioms and phrases that Edict has whereas other dictionaries do not.
- There are a few things I think would be good additions to the "saving words" function:
Finally, two very minor issues. For some reason Kenkyusha J-E dictionary seems to be broken. Other EPWING dictionaries work just fine though. See https://i.imgur.com/XsgTWXW.jpg
- You can press and hold on an Edict definition to save the word and sentence to Ankidroid. Unfortunately you can't do the same with the other dictionaries; if you could do this with EPWING dictionaries as well it would be great.
- Even when I tried it with Edict, for some reason every Anki card had the text "No context sentence". The rest of the info (kanji, reading, definition) worked fine but it would not mine the source sentence as well.
- This might be asking a lot, but it would be cool if there were an option to have pressing and holding on a word share it and its sentence to applications besides just Ankidroid. Typhon's Ankidroid import function currently does not have all the options that desktop firefox Rikaisama does (pitch accent, frequency info, word audio, etc.), and this is fine. But what might be neat is if I could just press and hold on a word to send it and its source sentence to a note program such as Evernote or Google Keep. That way, after I finish reading on the Android device, I can just go to the note program on my computer and mine the sentences into Anki using Rikaisama. However, this feature is not extremely important since I can already do this by manually highlighting the whole sentence and then sharing it to Evernote/Google Keep. It would just make it more efficient if it could be done automatically via a function of the application.
Also, http://typhon-reader.org/ seems to be offline so I can't get a key or whatever. I don't really care about online sync so this is not a big deal for me but I thought you should know.
And that is everything! Thank you for the update, I will be doing a lot more reading on the go from now on.
Ok, I am gonna reply to the different points but long story short, the first two are trickier than it seems. The last one is doable but not a priority I agree.
1. The deinflection is based on the availabilty of the part of speech information of a dictionary entry. The reason for that is that the deinflector works by mapping of the ending of the word to the stemmed version of the ending. Now, the advantage of that method is that it is fast. The inconvenient is that it can lead to false positives. And the part of speech is used to avoid stemming those that are not inflectable.
This information is not available from an EPWING dictionary which is presentation-based and not structure-based.
That said, what I can propose is to add an option to add a deinflector on all dictionaries which is not too hard but with the warning that it can lead to (more) false positives.
A more complicated solution would be to use a more powerful stemmer but that can potentially slow down the lookups, to be analysed.
2. This is a bit tougher because the way the dictionary lookup works is that it looks up only for the current dictionary and then the next one in the direction of the last swipe to be prepared in case the user swipes. It does not lookup the rest of the dictionaries and that's important to keep the lookup super fast. Maybe in the future, it could lookup the other dictionaries in another thread and switch if the result arrives in time but that might create an unpleasant delay that switches while you read the first entry.
3. The EPWING to anki features requires a more advanced parsing that is dictionary-dependent of the entries which is what cb4960 implemented. So that means a parser per version of dictionary like Kenkuysha 5th, Daijisen, Daijisen 2th, etc… And for different reasons, I cannot reuse directly cb4960's code.
While possible to adapt, the main problem for me is to be able to test those parsers while not possessing the dictionaries (I have only Kenkuysha 5th).
But that could potentially bring some good things along like coloring of the EPWING entries.
I will check the context sentence, there might be a bug.
I read the code from cb4960 and the pitch accent and frequency info could be added pretty easily but I don't know about the audio. But I was thinking of implementing a history information based on two files:
- The matches to have a history of the searches
- A possibility to long click (thus overriding the anki creation) on an entry to save its content as a csv to import it elsewhere, potentially a better app to generate flashcards.
Sorry if somethings are not making sense, I did not review my post :p