Where does the training data come from?

[polm]

Where does the training data come from?

[othmane-ab]

It appears he didn't add the training data in the repository

[HeroadZ]

hi, there. @polm See you again.
I'm a beginner of NLP and interested in japanese. But I can't find dataset with labels in japanese for sentiment analysis.
Could you give me some advices? 😄

[polm]

Sorry, I am not aware of any good training set, which is why I was curious what was used for this model. pymlask is rule based but seems solid.

Since you are affiliated with an academic institution you may have access to this Rakuten dataset.

https://rit.rakuten.co.jp/data_release_ja/

[HeroadZ]

Sorry, I am not aware of any good training set, which is why I was curious what was used for this model. pymlask is rule based but seems solid.

I just saw an article which compare the current japanese sentiment analysis tools. It seems like that there exist problems for the pymlask. If test sentence doesn't contain any word in dictionary, the polarity is None. If you are interested, check this.

@ikegami-yukino how do you think about the article above? Thank you for your work.

Since you are affiliated with an academic institution you may have access to this Rakuten dataset. https://rit.rakuten.co.jp/data_release_ja/

Thank you for your advices. But maybe I cannot access it because the research I'm doing now is from a company where I work as an intern. 😢

[polm]

If you want to close this that's fine, but I still don't see an explanation of the training data anywhere. Did I miss it?