seara/rubert-tiny2-russian-sentiment

This is RuBERT-tiny2 model fine-tuned for _sentiment classification of short Russian texts. The task is a multi-class classification_ with the following labels:

0: neutral
1: positive
2: negative

Label to Russian label:

neutral: нейтральный
positive: позитивный
negative: негативный

Usage

from transformers import pipeline
model = pipeline(model="seara/rubert-tiny2-russian-sentiment")
model("Привет, ты мне нравишься!")
# [{'label': 'positive', 'score': 0.9398769736289978}]

Dataset

This model was trained on the union of the following datasets:

• —Kaggle Russian News Dataset
• —Linis Crowd 2015
• —Linis Crowd 2016
• —RuReviews
• —RuSentiment

An overview of the training data can be found on S. Smetanin Github repository.

_Download links for all Russian sentiment datasets collected by Smetanin can be found in this [repository](https://github.com/searayeah/russian-sentiment-emotion-datasets)._

Training

Training were done in this project with this parameters:

tokenizer.max_length: 512
batch_size: 64
optimizer: adam
lr: 0.00001
weight_decay: 0
epochs: 5

Train/validation/test splits are 80%/10%/10%.

Eval results (on test split)