Introducing tenjinreader.com

Posted on February 27, 2018

My thoughts on the motivation and design of tenjinreader.com

Background

There can be many different ways to learn a language. Talking (speaking and listening) in your day to day life is the best way.

The second best is studying. Studying happens in a calm, relaxed, and distraction free environment. It takes a lot of effort / discipline to study, but it does give better results.

I am right now focusing my efforts more on studying/reading because it is easier than talking (for me, I dont talk a lot). I learned English by reading dozens of books, and I think I can learn Japanese also.

Motivation behind the SRS

I have been doing SRS reviews from the very start of Japanese learning. I find them very helpful to retain the vocabulary.

I wanted to try speech recognition for doing my reviews, so this was the main motivation behind developing a new browser based SRS app from scratch.

There were a few shortcomings in the existing tools, which led me to the current design.

Wanikani and HouHou SRS

I used Wanikani for an year, but I realised I dont want to spend so much time learning Kanji. Also I wanted a customised study, the flixibility to chose which words I learn first, and for which words I learn the Kanji form.

Then I moved on to HouHou SRS, it was a much better experience, as I could suspend review items which were nor relevant or difficult, and instead focus on the vocabulary which I can understand.

I used it for almost an year, it was going good but then I installed Linux as a primary OS, and now I could not longer do my reviews.

Anki

Also during this time I started using Anki for doing the production reviews, for much of the same vocabulary as HouHou SRS. I found it odd to use two different tools, one for recognition reviews and other one for production.

Nevertheless these are the good things about Anki

In my opinion the repetition part of the SRS should be emphasised over the game/quiz part. Its more important to have a look at the vocabulary again and again, than to answer it correctly.

SRS in tenjinreader

The ability to answer the reviews using voice was a really the driving factor behind the design of SRS in tenjinreader

For this feature I surveyed and tried out the existing speech to text technologies for a few months, like Kaldi and Julius.

I even got the Juluis speech recognition engine to work with a browser frontend. But the quality of recognition was unacceptable. It would be very frustating for a user if she has to repeat the answers multiple times. The language learning itself is a difficult process, and giving an answer should not be this difficult.

So I decided to use the Google’s speech recognition (via Chrome or Chromium browser).

Motivation behind the design of Reader

After coming to Japan I immediately realised my studies were incomplete. I had never studied the grammar and had no clue how to use the words in sentences.

I went through the Tae Kim’s awesome grammar guide (and still refer it). But what I really needed now was a way to not just learn useful words, but also see them used in sentences.

So soon I started with a school textbook of 5th grade containing short stories. The content was interesting but it was quite difficult for me to read it properly, as I had to constantly refer to dictionary or online resources, and sometimes use Google translate’ Camera feature just to figure out the start and end of words.

When I tried to look for computer based tools for reading Japanese, I found them to be much worse than SRS. The reading of Japanese material is especially complicated because of the Kanji and lack of spacing between words.

I soon found japanese.io, and it was a big improvement over the earlier tools. After using it for few months I found these problems.

Overall it was manageable but not a good experience. So this was the major motivation to create this reader.

With tenjin reader