Comparer des méthodes

	Traitement automatique du langage naturel (TALN) pour les médias sociaux	Embeddings BERT	TF-IDF
Domaine	Fouille de textes	Fouille de textes	Fouille de textes
Famille	Process / pipeline	Process / pipeline	Process / pipeline
Année d'origine	2017	2019	1988
Auteur d'origine	Community-established benchmark (SemEval shared tasks, Cardiff NLP group)	Devlin, Chang, Lee & Toutanova (Google AI)	Salton & Buckley
Type	NLP process pipeline for short, noisy social-media text	Contextual transformer text-representation method	Text vectorization / term-weighting scheme
Source fondatrice	Rosenthal, S. et al. (2017). SemEval-2017 Task 4: Sentiment Analysis in Twitter. Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017). ACL.	Devlin, J., Chang, M.-W., Lee, K. & Toutanova, K. (2019). BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. NAACL-HLT, 4171-4186.	Salton, G. & Buckley, C. (1988). Term-weighting approaches in automatic text retrieval. Information Processing & Management, 24(5), 513-523.
Alias	Sosyal Medya Metin Analizi, social media text mining, Twitter NLP, short-text NLP	contextual embeddings, transformer embeddings, BERT Tabanlı Metin Gömülmeleri	term weighting, tf-idf weighting, TF-IDF Vektörizasyonu
Apparentées	5	4	3
Résumé	Social Media NLP is a specialised natural-language-processing pipeline designed for the short, noisy, and informal text that appears on platforms such as Twitter, Reddit, and comment sections. Unlike general-purpose NLP, this pipeline accounts for platform-specific conventions — hashtags, emojis, abbreviations, and code-switching — enabling tasks such as hashtag analysis, viral content detection, and public-opinion measurement. The benchmark tradition for this approach was established through the SemEval-2017 Task 4 shared task (Rosenthal et al., 2017) and the TweetEval unified benchmark (Barbieri et al., 2020).	BERT-based text embeddings, introduced by Devlin and colleagues at Google AI in 2019, turn text into context-sensitive dense vectors using a bidirectional Transformer encoder. Because the meaning of a word shifts with its context, BERT produces richer representations than static methods such as Word2Vec or topic models like LDA.	TF-IDF, introduced by Salton and Buckley (1988), is a term-weighting scheme that scores each word in a document by how often it appears there and how rare it is across the whole collection. It turns raw text into weighted document vectors, giving high weight to terms that are frequent in one document but uncommon elsewhere.