darinochka / news_predictions

News Predictions

Flask application with models for classification and summarization of news in Russian.

Description

The application contains two handlers with calls to the text classification model (/predict) and the summarization model (/summary_v2)

• /predict: Uses trained fasttext model for classification
• /summary: We use a model pre-trained on the dataset of russian news mbart_ru_sum_gazeta

The application is used to operate another application - a telegram bot and placed in a separate repository

About the code:

This project is a modular bot, made using Python 3 and the following:

• FastText
• Pre-trained model mbart_ru_sum_gazeta
• Flask

Setting up

Before all, clone this repository.

You need to have a trained fasttext classification model and filling congig.yml file:

CLASSIFICATOR_MODEL_RU: "fasstext_model.bin"
SUMMARIZE_MODEL_RU: "IlyaGusev/mbart_ru_sum_gazeta"

Using Docker

Simply, run the following command:

docker-compose up --build -d

Without Docker

Run main file news_predictions/app/src/wsgi.py => python3 main.py

Using handlers

It is possible to use only Russian text for models.

• classification

  curl -i -H "Content-Type: application/json" -X POST -d '{"news": ["какая-то первая новость", "какая-то вторая новость"]}' http://localhost:5000/predict
  

• summarization

  curl -i -H "Content-Type: application/json" -X POST -d ‘{“text: “текст для суммаризации”}’ http://localhost:5000/summary_v2
  

Authors

Contributors names and contact info

ex. Artem Ponomarev ex. @aspnmrv