rubert base cased sentiment new

Introduction

The rubert-base-cased-sentiment-new is a text classification model designed for sentiment analysis of Russian texts. Developed by Tatyana Voloshina, it is based on the BERT architecture, focusing on identifying sentiment in Russian language inputs.

Architecture

The model is derived from the BERT architecture, specifically tailored for sentiment analysis. It categorizes input text into three sentiment labels: neutral, positive, and negative. The model is trained using the Tatyana/ru_sentiment_dataset.

Training

Training Data

The model utilizes the Tatyana/ru_sentiment_dataset for training, which contains labeled Russian text data for sentiment analysis.

Training Procedure

Details on preprocessing, speeds, sizes, and times are not provided.

Guide: Running Locally

To run the rubert-base-cased-sentiment-new model locally, follow these steps:

1. Install Required Packages

   !pip install tensorflow-gpu
   !pip install deeppavlov
   !pip install fasttext
   !pip install transformers
   

2. Set Up the Model Download the necessary PyTorch-trained model files from the provided Google Drive folder and place model.pth.tar in the same folder as the other model files.

3. Install Additional Components

   !python -m deeppavlov install squad_bert
   !python -m deeppavlov install bert_sentence_embedder
   

4. Load and Use the Model

   from deeppavlov import build_model
   
   model = build_model('path_to_model/rubert_sentiment.json')
   model(["Сегодня хорошая погода", "Я счастлив проводить с тобою время", "Мне нравится эта музыкальная композиция"])
   

Suggested Cloud GPUs

For improved performance, consider using cloud GPUs from providers like AWS, Google Cloud, or Azure.

License

The specific licensing details for the rubert-base-cased-sentiment-new model are not provided. Users are advised to consult the model repository or contact the authors for more information.