RuBERT for Sentiment Analysis

Short Russian texts sentiment classification

This is a DeepPavlov/rubert-base-cased-conversational model trained on aggregated corpus of 351.797 texts.

Labels

0: NEUTRAL
1: POSITIVE
2: NEGATIVE

How to use


import torch
from transformers import AutoModelForSequenceClassification
from transformers import BertTokenizerFast

tokenizer = BertTokenizerFast.from_pretrained('blanchefort/rubert-base-cased-sentiment')
model = AutoModelForSequenceClassification.from_pretrained('blanchefort/rubert-base-cased-sentiment', return_dict=True)

@torch.no_grad()
def predict(text):
    inputs = tokenizer(text, max_length=512, padding=True, truncation=True, return_tensors='pt')
    outputs = model(**inputs)
    predicted = torch.nn.functional.softmax(outputs.logits, dim=1)
    predicted = torch.argmax(predicted, dim=1).numpy()
    return predicted

Datasets used for model training

RuTweetCorp

Рубцова Ю. Автоматическое построение и анализ корпуса коротких текстов (постов микроблогов) для задачи разработки и тренировки тонового классификатора //Инженерия знаний и технологии семантического веба. – 2012. – Т. 1. – С. 109-116.

RuReviews

RuReviews: An Automatically Annotated Sentiment Analysis Dataset for Product Reviews in Russian.

RuSentiment

A. Rogers A. Romanov A. Rumshisky S. Volkova M. Gronas A. Gribov RuSentiment: An Enriched Sentiment Analysis Dataset for Social Media in Russian. Proceedings of COLING 2018.

Отзывы о медучреждениях

Датасет содержит пользовательские отзывы о медицинских учреждениях. Датасет собран в мае 2019 года с сайта prodoctorov.ru

Downloads last month
25,560
Safetensors
Model size
178M params
Tensor type
I64
·
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Spaces using blanchefort/rubert-base-cased-sentiment 5