ReviewsScoring is a system that can score and filter reviews from booking-hotels/booking-flats services based on ML and DL approaches.
Well, in this repository I have three types of models. First one is a Spam Filter that based on TfIdf and Logistic Regression. It's can help to choose the most relevant reviews. Second one is reviews scorer that based on Convolutional 1D and Bidirectional GRU Neural Networks. And the third one is a full system combined together - scorer and filter. Before processing by models texts must be preprocessed. For this task we use NLTK and spacy. Via NLTK we just get stopwords and via spacy we do main preprocessing (lemmatization and so on).
After you clone this repository onto your PC go to directory of repository and run following command in terminal:
pip install -r requirements.txt
After all packages have installed you have to install language model for spaCy by the following command:
python -m spacy download en_core_web_sm
In repository you can find usage examples and models training. If you want to see how models was trained look at reviews_scoring.ipynb and spam-filter-fitting.ipynb, but if you want only look at usage examples see review-example.ipynb and system_test.ipynb.