# Social media analysis

A modern bidirectional encoder-only Transformer model for Russian, pre-trained on approximately 2 trillion Russian, English, and code tokens, with a context length of up to 8,192 tokens.

This is a binary classification model trained using AutoTrain, designed to distinguish between memes and non-meme images.

A multilingual summarization model fine-tuned on the CrossSum dataset, supporting cross-lingual summarization for 45 languages

This model is a RoBERTa model pre-trained on Twitter data, fine-tuned on the tner/tweetner7 dataset for named entity recognition in tweets.

A named entity recognition model fine-tuned on the TweetNER7 dataset based on the Twitter-RoBERTa model, specifically designed for entity recognition in tweets.

Baseline Twitter POS tagging model trained on Tweebank V2 named entity recognition benchmark with 95.21% accuracy

A sentiment analysis model fine-tuned based on distilbert-base-uncased, trained on the tweet_eval dataset for tweet sentiment classification

A sarcasm text classifier based on RoBERTa-base architecture, used to distinguish whether a text is sarcastic

A model fine-tuned based on bert-base-uncased for identifying hateful meme text content

This is a sentiment analysis classifier based on SlovakBERT, specifically optimized for Slovak tweets, capable of distinguishing between negative, neutral, and positive sentiment levels.

Hate speech detection model fine-tuned on distilbert-base-uncased, trained on tweet_eval dataset

A sentiment analysis model fine-tuned based on bert-base-uncased for tweet sentiment classification

This is a binary classification model trained via AutoNLP for detecting hate speech content in text.

A text classification model trained on Russian sensitive topic datasets, capable of identifying 18 types of potentially reputation-damaging inappropriate content

This model is used to detect hate speech in Portuguese, fine-tuned from a multilingual BERT model as a monolingual model.