Free graded reader resource

[IronMandarin]

Hello everyone,

Inspired by the Inkstone app that I found was so useful and free, I dedicated some of my time here in China to create a graded reader website, IronMandarin.

 

What you can do with this website:

• read a text
• choose a text by category or by HSK level
• see how close a text really match a given HSK level
• save words in your know list, to give you an idea of where YOU stand to a given text (this is your personal profile)
• save words in a 'to learn' list, and I plan to add a SRS (Spaced Repetition Software) functionality, or maybe just allow export in a convenient format for anki import?
• see the most frequent unknown words in a text (so once you add 的，一，个... to your personal list they don't appear in the frequency listing anymore)
• switch the character set from simplified to traditional
• analyze your own text:
  • without logging in, you can analyze a text, for example an email or an article, it will be better for you than google translation
  • logging in, you can publish a text so it is saved on the website. It can be public, or private (so no sharing say, personal emails)

 

It is not fully developed, as I also study and work, but more functionalities are planned, such as frequency list over one category or a set of articles, to be able to focus on specific vocabulary, which I find especially useful for HSK 6.

 

I publish text from different sources but I know a few Chinese teachers here in Chengdu that help me. You can also participate by publishing some texts.

 

The segmentation algorithm is automated, based on Jieba, but it makes a lot of mistakes and currently I spend quite some time reviewing the segmentation, maybe I'll have a look at the code to patch a few common mistakes (for example it doesn't split numbers, or number and measure word).

 

I had before some question about monetization. I dedicate quite some time to this project, and I need to pay the writers more if they spend some more time on the project.

I will make pretty soon a Patreon page, hopefully it will be enough to make this project sustainable.

The website also offers the possibility of tailored analysis and advice for a reader to progress, based on his current word list. This and some premium content could make for a premium membership in the future if a Patreon is not enough.

The website is not free to gain enough traction to put it full price more expensive than a real newspaper. The core functionalities intend to remain free, on a donation or freemium model.

 

Hopefully this will help all of us Chinese learners

Let me know if you have any suggestions!

[imron]

5 hours ago, IronMandarin said:

maybe I'll have a look at the code to patch a few common mistakes

Jieba is a statistics based segmenter.  It's not so much the code you need to patch but rather the probabilities used in the statistical model.  You could probably hard-code a bunch of different exceptions, but the whole point of using a statistical model is to avoid the need to hard-code exceptions in the first place.

 

For what it's worth, I'm currently working on a statistical segmenter for Chinese Text Analyser that uses the Jieba data files for probabilities (the current version of CTA uses a first longest match algorithm, which is fast but even more inaccurate than Jieba).

[IronMandarin]

Ok, thanks for the information I'll have a look into that. It was not my priority but that could be useful to dig a bit into the technique.

[imron]

Let me know if you have any questions about it, or can't figure out why it does something in a given way.  I've been going over it in detail recently so have a good idea of how most of it works.