1. Go to data directory (https://github.com/zhanymkanov/russian_reviews_dataset/tree/master/data)
2. Download raw/cleaned data

or

git clone https://github.com/zhanymkanov/russian_reviews_dataset
cd data

https://github.com/zhanymkanov/reviews_parser

https://github.com/zhanymkanov/reviews_tazalau

1. 143k Russian reviews
2. 8.7k Kazakh reviews
3. 256 undetected language reviews (numbers like "10/10", English or Kazakh in latin)

• Review text might be in text and/or in plus and/or in minus
• At least one of the columns is filled

Number of categories: 17

• computers
• smartphones
• perfumes
• watches
• wearables
• etc.

• Integers from 1 to 5

• russian
• kazakh
• other

• Stopwords are removed (find them in reviews_tazalau/code/constants.py)
• Tags are removed (e.g. <br>, </br>, <p>)
• Russian words are lemmatized (via pymystem3)
• text, plus and minus columns are concatenated into one combined_text column