Every app has a dictionary now. The Aedict suite.

In my recent forays into searching for better solutions to working with Japanese on Android I became aware of a new addition by the developer of Aedict, called (unflatteringly, but descriptively) "Aedict Reader Popup".

I think I've mentioned the following already, but the existence of a dictionary that uses the EDICT Japanese dictionary, i.e. Aedict, was the reason I originally bought a smartphone. Thus allowing me to hold a book in one hand and dictionary in the other while on the bus. Although I was almost ready to jump ship when Aedict turned paid, Aedict3 remained head-and-shoulders above the competition.

Looking through the various Japanese dictionaries in development, there are a number of promising (even free) alternatives available. Indeed, I have Jsho installed on my phone just to see how it develops from here (however, its current lack of technical dictionaries is a bit of a turnoff at the moment*).

It was while looking at these apps that I (quite accidentally) came across the aforementioned Aedict Reader Popup. Despite having my eye on Aedict's competition, I thought I would at least buy it and try it.

I was first presented with the screen to the right. This is AedictReader, a web browser that automatically translates Japanese to English. By accidentally skipping the next dialogue, I thought for a moment this was the entire app. But the settings allow it to be used pretty much anywhere (I've yet to find anywhere it doesn't run; hell, it will even display over World of Goo if I tell it to**). The reader does, however, serve as a useful introduction to Aedict Reader Popup's functions.

Once enabled to work across all apps (I think that is the default setting), copying text will result in a popup, as shown to the left. Any dictionaries you have enabled in Aedict3*** will be consulted for a match. The picture shows an example of the app being used on a biosciences document. Since I have the life sciences dictionary installed, "transcriptional activity" is correctly displayed when "転写活性" is selected and copied in my document viewer (I've blurred the rest of the writing in the GIMP*4 but you get the idea).

Tapping on the popup gives access to more advanced functions, including the ability to quickly switch to Aedict3 to have a closer look at the word and scrutinize it further. There is also the option to temporarily remove the popup or minimize it (down to about the size of a fingerprin; easy to push, but unobtrusive on my largish AQUOS Crystal).

You can set the popup app to remain in the notification bar; from there you can activate the popup even if using a program from which you cannot copy/paste. If you have the paid Aedict3 Kanjipad plugin, this lets you copy the character in-situ*5. You can also do a normal search, though with the app-of-interest, the popup, and the keyboard on screen, it does get crowded. On tablets; however, this is very manageable.

When reading books on the Kindle, Aedict's new popup app even seems to outperform Kindle's own dictionaries (see below), especially considering Aedict is aware of verb tense, while Kindle's is not!

The only thing I can imagine that would make this better is integration of the OCR package, so that I can hover the popup over some uncopyable text and have the popup translate it. A tall order, but given the completeness and robustness of the app, I wouldn't be surprised if such a feature does appear in the future.

Conclusion

Today, I expect almost every learner of Japanese, everyone with Japanese as a second language, absorbs the language in some way via their phones. Whether it is film, books, instant messaging, news, websites, or even games, any tool that could be used to take the chore out of looking up a new word, or a forgotten word, makes these activities even more rewarding than before. Thus, it is easier to spend more time getting better at Japanese.

The popup app extends the features of the Aedict suite into almost any/all app(s) and does so without compromising the ability to search multiple dictionaries. It works and responds well on phones and tablets, even where the most basic system functions (such as copy/paste) fail. From my own experiences, I surmise that there has never been a more robust, accurate, and flexible Japanese dictionary tool in history*6, besides -perhaps- the human mind.

The ol' Nexus 7 running Kindle. Left: Kindle's own Japanese-English dictionary shows the entry, and I can get a full definition from there (but that opens up a new window, so I cannot see the word with its definition in the context of the book). Right: pressing the copy icon brings up Aedict's definition, complete with example sentences, all in one place so that I can still see the word in context. Moreover, having the popup installed doesn't interfere with the apps existing functions, so I (of course) still have Kindle's Wikipedia links etc.

*It is also light on other features, such as different methods for kanji lookup, though handwriting recognition is probably the way forward. Given the dictionary aims to be "simple, lightweight, fast and accurate. No unnecessary splash screens and no need to download any additional files", I'm not sure we'll see Jsho becoming the powerhouse that Aedict is.

**Rest assured, this is optional.

***And usefully, ONLY those dictionaries you have enabled in Aedict3's word search.

*4 "GNU Image Manipulation Program", before you get any odd ideas...

*5 previously you'd have to copy the character onto paper and then use the drawing to rewrite the character into the dictionary, or use another device with a Japanese dictionary. Either way, the new solution is much more elegant.

*6 good enough that I took the time to upload pictures, which I don't often do. Also its difficult to explain just how elegant the app is without them.

Handwritten Japanese Input by Google

Summary:
+Google's handwriting engine is free and highly accurate for writing Japanese characters
-If you already have good handwriting recognition as part of your dictionary, you likely wont need this

Having rediscovered that I own a blog, I thought I'd have another delve into the experience of using Japanese with Android, and quickly came upon the existence of Google's very own handwriting recognition (apparently its been around for months now, though).

I was very happy to see this for 3 reasons: 1) Google provide their input methods for free, 2) Google's products tend to work very well with each other

Reason 1 doesn't really require introduction, everyone loves a freebie, and anyone who has read any of my previous rants will know I am a techno-miser.

For point 2, I'd like to point out how all of Google's input methods* have a shortcut where you long-press the space bar to switch input methods. Thus you can switch between English, Japanese, and handwriting method fairly simply. (Many of the iWnn methods require that you pull down the notification menu and select the input method dialog from there).

So, lets do a little test. I randomly took whatever Japanese happened to be on my desk, in this case "電気料金等払込取扱票"**. Not particularly challenging, but a nice string of characters. So, I'll use a few handwriting recognition apps/input methods to see how good they are at recognizing my numpty attempts at copying these words by hand. The method counts as correctly predicting if it shows the character I want in the suggestions it displays.

Google handwriting recognition 100% correct, of which 100% were the 1st suggestion
Kanji Draw 60% correct, of which 50% were the 1st suggestion
Swype (In handwriting mode) 100% correct, of which 100% were the 1st suggestion
Aedict3 KanjiPad Extension 100% correct, of which 70% were the 1st suggestion
MyScript Stylus Beta - 90% correct, of which ~70% were the 1st suggestion

Similarly, the results for "納付書兼領収剤通知書"*** were:

Google - 90% correct, of which 100% were the 1st suggestion
Kanji Draw - 50% correct, of which 80% were the 1st suggestion
Swype - 90% correct, of which 100% were the 1st suggestion
MyScript Stylus Beta - 100% correct, of which 90% were the 1st suggestion
Aedict3 KanjiPad Extension - 100% correct, of which 90% were the 1st suggestion.

Evidently, the methods are neck-and-neck.

I'll keep Swype installed and hope that one day they'll release an update with prediction on par with Google's (Swype's language switching is faster and easier than Google's, but I've grown dependent upon how Google correctly interprets my input in the easier-to-use 12-key format).

However, I am thinking of switching from Google to Swiftkey. Why? Swiftkey's Japanese beta offers: one-tap switching between English/Japanese with very good prediction*4 in both languages and swipe input for English*5.




The other reason for this is that handwriting is only really useful when you need to input a character you cannot copy/paste and do not know the pronunciation of: ie. a random word you find in a book*6. The only reason I ever have for doing this is to look something up in a dictionary. Consequently, having bought into*7 the whole Aedict bundle (which I still think is the best dictionary around, despite previous misgivings*8) it makes sense to use the dictionary's kanjipad.


Also, note that I changed the settings of the Aedict handwriting recognition to search for kanji with the wrong number of strokes because the added accuracy saves time in the long run (if you're as sloppy a writer as me).


*at least, the ones I've used
** Yeah, I also find it depressing that the most prominent Japanese on my desk happens to be my electricity bill...
*** So many bills :(
*4 Google Japanese Input's English prediction is, for some reason, awful. Also, by very good Japanese prediction, I mean I can type "かつこうにいくとおもつた" and have it converted into "学校に行くと思った". However, Swiftkey didn't recognise the word 電磁気学 like Google does.
*5 Swype offers swipe input for both languages, but I never found it particularly reliable for Japanese, which I presume is because of the short consonant-vowel pattern of the language. Swiftkey seem to realise this and offer swipe for English and a 12-key layout for Japanese as default.
*6 or bill! (-_-;)
*7 technically, I only bought the Aedict KanjiPad Extension for the test above...
*8 And I still wouldn't recommend the Aedict OCR app.

Japanese and the Samsung Galaxy Fame S6810

So, I decided that, nice though the tablet/keitai phone setup was, having a working smartphone is much more handy when moving around. As much as it excels at documents and surfing, I don't terribly like using my old Nexus 7 tablet as a mp3 player. Also, its not as convenient a dictionary as a dinky smartphone you can quickly whip out whenever you spot an offending word*.

For my last trip to Tokyo, I resurrected my old HTC Tattoo by I rooting it, and swapping in my Softbank SIM card (Rooting it was a complete pain, as all the rooting instructions I could find (supposedly specifically for the HTC tattoo) didn't work**). The advantage was that I gained the ability to run programs like LINE and aedict3. However, the Tattoo simply doesn't have the required oomph to operate newer versions of Android, or the apps that run on it. Both the aforementioned apps ran terribly slowly.

So, I spent a couple of days mulling it over before deciding on a Samsung Galaxy Fame S6810 *3*. My primary concerns on purchasing were: Price, a version of Android where I could use the aforementioned apps, and a decent-ish camera. Despite numerous reviews complaining about the sluggishness of the Fame, I anticipated that compared to the 2009 smartphone technology that I was used to, that the Fame would feel nippy enough for me.

Having dabbled with the Fame for a day, I think my anticipations were on the mark. The device feels very powerful in comparison to the Tattoo, and my bread-and-butter apps (currently aedict3, ankidroid, and LINE) run very smoothly, almost as smooth as on my Nexus 7. The budget price only really became much of an issue for me when switching between applications, but the lag-times there are only on the same order as the Tattoo, despite using more intensive applications.

The disparity between the cameras is considerable, as one would hope. This makes optical character recognition very viable on the Fame, which is something that didn't really cross my mind before (I've not been using it up 'till now because the Tattoo sports a terrible camera, and the old Nexus 7 doesn't have the forward facing camera). Google Translate appears particularly competent here, and while I don't go much on some of the machine translations, you can select the source text and paste it into aedict3, or ankidroid, so it remains very useful. Google Translate's handwriting input also seems fairly friendly to slap-dash kanji stroke orders, and again can be copied. I somehow overlooked this feature previously, but may be using it a lot in future.

So in summary, the Fame excels as a handy pocket dictionary, while seeming able to cope with my other modest demands. Certainly a phone I would recommend to anyone after a cheap device that gets the job done.



*ie any word that I don't happen to know..

** that said, anyone with a working knowledge of the linux command line can solve the conundrum with a little thinking. Unfortunately I didn't document the efforts, but I doubt anyone still uses the Tattoo now.

*3* For the record, I purchased from Eternal Communications Ltd. via Amazon.co.uk. Unlike most of the suppliers in England their prices and delivery rates were both reasonable.

Aedict3, all grown up

I know I've been going on somewhat about aedict 3 recently, but it has been developing at a considerable rate lately, and after giving it some more thorough use, I thought it worth a little note here.

I'm not sure exactly when it was brought in, but aedict now has a "live search" function. When I read that in the changelog, I wasn't particularly impressed; I figured how much time does it save having to not press a button? However, it is surprising how convenient this little feature is!

Firstly, I found the time savings are actually fairly considerable in most cases (qualitatively speaking). However, what I love most about this feature is that it provides a text glossing functionality akin to the Electronic Dictionary Research and Development group's very own site*.

What is the fuss about text glossing? Well, a lot of well-known apps like "Kingsoft Office" and "Adobe Reader" are terrible at selecting Japanese text. The problem appears to be that they are used to separating words by spaces and punctuation, as we do in English. However, Japanese doesn't use whitespace characters, and so when you hold your finger on a word in either of the above programs (and I'm sure they're not the only ones**), it selects between the two nearest punctuation, which might contain several words.

After selecting Japanese text in such a program, you then have 2 options: Firstly, you can use the fiddly text selection controls to highlight just the word you want to copy into the dictionary. Secondly, you can copy the whole sentence, then edit it once you've pasted into the dictionary.

The advantage in having text glossing is that you can just dump the whole sentence into the dictionary's input box, and you get a translation of all the words in the sentence. No fiddling with text selections, no editing of the sentence; a huge time saver, especially considering the text glossing is included in the live search, so you just paste, sit back and watch the magic.

Ultimately, my praise here is as much a reflection of my frustration with other programs (such as Kingsoft Office and Adobe Acrobat) that do the bare minimum to support Japanese text. Nonetheless, if you do a lot of work with Japanese documents (I do some freelance translation, and read the odd scientific article**** in Japanese), then you will frequently need to work around this problem. Aedict3 has made this process about as painless as I expect things will get, and for those who do a lot of reading of Japanese on their android devices, I highly recommend dishing out for Aedict3.


* For the unaware, the EDRDG is the home of edict and JMdict, which is integral to aedict. I'm not sure about the specifics, but without it, and the work of Jim Breen, we'd probably be a lot poorer in the dictionary department.

** Because how do you write a program to detect word edges in Japanese? It would be very difficult, I think. A work around would be to only select contiguous kanji characters. This would make selecting hiragana words more difficult***; however, because hiragana is phonetic, it is easy to reproduce by typing elsewhere, if needs be.

*** you'd have to drag the selection over the characters you wanted, which people do when they want to select more than one word of English anyway. So not spectacularly more difficult, really.

**** scientific articles are commonly distributed as pdf files.

Aedict3 Slow and steady wins the race

A while back I posted about the glaring deficiencies between the paid-for Aedict3 and the freely available Aedict 2.

This week, Aedict 3 was updated and with it, so far as I can see, all of the features from aedict 2.

Its reassuring that the author is still working on the project, and I feel that my meagre investment in buying the paid-for version is starting to go the distance. That said, I'd still say that aedict3 has done little more than catch up with its predecessor. There are some nice changes, but these are entirely cosmetic so far as I can see. Nonetheless, it feels like there is momentum in the project, and perhaps aedict3 will soon surpass version 2.

The bottom line at the moment is that I still recommend version 2 if you're starting to seriously read in Japanese.

EDIT: Aedict 3 has finally overtaken it's older sibling, and I think people should now start seriously considering getting hold of it, especially if they read a lot of Japanese on their device see this blog post

Aedict 3: big on version numbers, lean on features

So, as I stated previously, here, new versions of aedict are now paid-for. Now the big question, was it worth £3.50 to get it?

For the impatient, the answer is no. For any readers I haven't lost by now (that's assuming anyone even made it as far as "no"), I will elaborate. Aedict3, while being flashier and maybe even easier to navigate, is decidedly less functional than the free version. There are no more "advanced" options; you cannot restrict your search to "starts with", "ends with", or "exact match". To me, this is a big setback: if you're reading through a book and there is a word you've never seen before, there is still a very good chance you'll know what either the starting or the ending character is, and not being able to search on that basis is going to make reading books somewhat more arduous.

Additionally, the search is as accurate as before, typing 問 as a search term brings up words like 受け止める, even though the latter does not contain the former. I'm not sure whether that is an error of aedict or with the dictionary file itself, but either way it doesn't lend itself to quickly finding out what a word is, or to finding out about related words/phrases.

So, as suggested in the title, aedict 3 is a significant step backwards. But I'm not terribly worried about this, and I said before, I owe the author at least a beer for his previous efforts, the quality (or lack thereof) of aedict 3 wasn't a factor in my purchase.

I hope that future versions of aedict 3 will become as feature-rich as the freely available aedict 2, and while I wouldn't recommend it just now, I would say it could be worth signing onto aedict's google group, and keeping an eye on it.

EDIT: Aedict3 has continued to improve over the months following this blog article, and I'm quite happy to endorse it fully now. See this post.

Aedict goes paid/unpaid

Recently the author of the superb android Japanese dictionary app announced that future versions of Aedict would require a one-off payment of $5 (The existing version will be available free-of charge).

To be honest, while I was disappointed, the idea of buying a beer for the author isn't one I find objectionable, after all, I've used the software a great deal in the last 4 or so years. Having a family, I can also sympathise with juggling commitments, and I've laid down a number of my own hobbies (including programming, funnily enough) too, for the same reason.

What annoys me is that the payment is under the pretence that his family situation is affecting his ability to work on the project:

"Unfortunately, I have a family now and I can no longer afford to develop Aedict for free"

And yet, later in the announcement:

"I am currently preparing a new version of Aedict"

Finally the announcement terminated with the line:

"but the only other option I had was to abandon Aedict entirely"

Gods no! What about passing it on to someone? It's already open source, and is the most useful Japanese dictionary on the Android platform, someone will pick it up.

That said, if the features are good, will I purchase aedict 3? Yes, I think so. His app is so integral to how I've learned and enjoyed Japanese the last few years that I genuinely feel I owe him it. So sure, Martin, have a drink on me, but drop the pretence, eh?