Learn Japanese with gengo.tech!

[xasdasdas]

I’ve been developing a tool that helps analyze Japanese text for personal use, and after many months of work on it, I’m finally looking forward to how other people can use it.

It’s still very much a work in progress so I’m looking forward to any kind of feedback and suggestions.

It takes text and automagically adds kana, translations, meanings, and more to it.

Click here to access the website

Main features

The main features are:

• Text can be inputted via OCR (image → text)
• Furigana
• Translation
• Text-to-speech
• Dictionary and grammar breakdown
• Chat assistant (based on GPT3.5)

Translation

Each sentence is automatically translated. You can see the translation by hovering over the translation button. If you click, the translation will be copied into your clipboard.

Text-to-speech

Each sentence has text-to-speech automatically generated. While it plays, the active words will be highlighted, so you can follow along more easily.

Logical blocks

Each sentence is divided in logical blocks, which don’t have to be exact words.
You can hover over the sentence and they’ll be highlighted, along with their meaning being shown.

Dictionary

Instead of hovering, you can also click on a logical block. When you do so, detailed information will be pulled from the dictionary and shown on the right panel.

There is also an option to select a different mode, which will instead focus on the specific grammar of each part of the sentence, inspired by Jisho:

Chat assistant

Instead of the dictionary, there is an option to instead show an AI assistant based on the language model GPT3.5. It will know be aware of the text and can answer questions about it, or the Japanese language in general.

Account

Every full analysis is saved into your account, and can be accessed from the sidebar. There’s also a full-text search function.

OCR

You can upload (or paste) images with Japanese text in it and it’ll try to parse it.

You can choose to either just receive the text…

Or do a regular, full text analysis on it for a more in-depth look.

The future

I have many ideas on how to improve the tool in the future.

• Anki / AnkiConnect integration
• WaniKani integration (levels + your vocabulary)
• Kanji
• Pitch accent
• Notes
• Shareable analysis
• Easy way to download analysis and audio you’ve generated

.. and many more.

Pricing

The tool is free while I collect user feedback and gauge the costs (server hosting, 3rd party services, development, etc)

Feedback

Apart from this thread, I also made a discord server for more direct communication about the app! It’s a platform I enjoy using.

[Kumirei]

What parser are you using under the hood? Mecab, Spacy, something else?

[yatyisam]

@xasdasdas

Apart from this thread, I also made a discord server for more direct communication about the app! It’s a platform I enjoy using.

The Discord link says that it is invalid for me.

[xasdasdas]

Heya! I’m using two parsers, which are Sudachi and an in-house one which is based on GPT3.5.

Since the latter is non-deterministic, I’ve added an option to re-run it:

[Ted]

this is really cool fren! excited to follow along :3

[whinette]

[xasdasdas]

The tool makes use of an external content moderation service to comply with certain external APIs we use for analysis.

If you feel like it was a false positive, I’d be more than happy to take a look at it

Feel free to send the text to gengotech@protonmail.com, here, or on Discord, whichever place you feel more comfortable with. Thanks!

[whinette]

Oh it sure wasn’t.

[xasdasdas]

Haha, got it.

I’ll look into modifying the behavior so that instead of flat out failing, it creates an analysis with limited functionality, by not using the APIs that specifically reject such content.

[xasdasdas]

Today’s patch

• Now you can crop images before uploading them
  

• Added WaniKani level information (and link) on dictionary entries
  

• Text inputted in the new analysis page will be saved locally

• Centered tab in new analysis page

• Improved the sorting of dictionary entries, now it takes into account the context in the sentence

• Now you can analyze text even if marked as inappropiate by OpenAI’s moderation API

• Now selected words will have a thicker border for more clarity
  

• Fixed a bug where the reading for certain dictionary entries would get cut off because of the length
  

I’ll keep an eye on the thread for feedback If anything behaves strangely, try clearing the local browser cache, sometimes it can be stubborn.

[xasdasdas]

Got a few new features during the last week of development. Most of those are work in progress but I’m fairly happy about the progress.

Conjugation info

Now verbs and adjectives show additional contextual information. Additionally, I’ve added conjugation tables.

User dictionary

Now each user can add additional dictionary entries. They’ll be bound to your account and be available whenever you analyze a text.

Radical search

I added a new page to find Kanji by selecting the radicals that compose them.

Emojis

This is a bit of a fun one. Now every analysis will display an emoji that matches the text.

[xasdasdas]

Just a small update. Working on counters & numbers, the idea is to make it be able to interpret pretty much anything. Still a work in progress as there are many counters and it’s a fairly complex system.

Changes are available on the website if you’d like to try it out! Also, you can now login using your Discord account.

[xasdasdas]

Daily dev update with everything i’ve been working on today!

• Example sentences: I’ve added about 10,000 example sentences, covering most types of common verbs.

• Mobile UI: Working on making the app more accessible for mobile devices.

• Counters: Still working on counters, added support for special formats such as this one:

[xasdasdas]

Big thanks to everyone that has given the app a try so far!

I’ve been working on new features since the last update. I’m going to list the main ones here.

Improved dictionary

Added a section that lists all Kanji characters in the word. You can click on them for more detail.

Added a pitch accent and alternative word forms sections.

Added more detailed word frequency data.

Added audio transcription

Now you can either upload an audio file or speak into the microphone, and the audio will be transcribed into Japanese and analyzed.

Added info tab

The idea was to have an explanation of the text in English which provided the gist/main ideas and the most important words or set phrases.

Within the explanation, you can click on Japanese words to open the dictionary on the right side.

Added suggestions for AI assistant

I’ve added some pre-made questions that the AI assistant should be able to answer for most texts.

Smaller changes

• Improved the quality of example sentences
• Changed translation engine, now it should be more context aware and natural sounding.
• Most of the website icons are now emojis.
• I’ve labelled features which make use of generative AI
• Many other improvements to the Japanese parser

[xasdasdas]

Thanks to all the people that tried the website out so far!!

I still post daily updates and changelogs on the Discord server!

It also has cool educational minigames like shiritori (w/ a server leaderboard ), JLPT quizzes, prefecture map quizzes, hiragana/katakana quizzes, etc.

Plus, it has other utilities like a translate command, a question command to ask questions to the AI assistant, attaching an image to extract the text, vocabulary/kanji lookups, etc!

So, looking forward for more people to try it! I’ve been working on it for a while.

[yatyisam]

This website is amazing and I can see how it could be very useful. I am just wondering though, is there any way to control the speed of the audio of the TTS?

[xasdasdas]

@yatyisam

This website is amazing and I can see how it could be very useful.

Thank you so much!

@yatyisam

I am just wondering though, is there any way to control the speed of the audio of the TTS?

Not at the moment, but it’s definitely doable! I’ll start working on it.

[xasdasdas]

Update! Added a new settings panel where you can configure the voice and speed of the TTS.

[yatyisam]

Wow, that was fast. Thanks @xasdasdas !

[xasdasdas]

Hey all, going to share some new features that I’ve been working on during the past weeks!

gengo.tech - Japanese Text Analysis Tool

New OCR functionality

I improved the OCR section! Now it includes an automatic translation, and it’s more accurate on text that includes furigana or is vertical (like in manga)

Dedicated page for example sentences

Demo

Now when you click on an example sentence, it will open in a separate page!

Example sentences in the analysis

Demo

Now whenever you click on a word, all the example sentences that contain it will be displayed. You can click on the button to get a new set!

Example sentences for every dictionary entry

Now when you navigate to the page for a dictionary entry, you’ll see an example sentences section. You can search for specific terms in either the original Japanese or English translation text.

Improved info tab

When you create a stand alone text analysis, the info tab now allows up to 10 summaries to be generated. As always, you can click on the words inside the summary to open them on the right panel!

More resources

I’m constantly adding links to useful resources, so hope you find those useful.

Romaji support

If you’re into that, readings can be shown in Romaji instead of Kana! You can change that in the settings. Someone requested this and thought, why not.

Thank you!

Thank you all for using my tool, it really means a lot! If you have any requests or feedback, let me know!