TAL pour Sciences Humaines et Sociales

Statistiques Lexicales

• Transparents en ligne (Google Slides)

Programme

Day 1: (6h)

• Zipf law, transition point technique
• ZipfR http://zipfr.r-forge.r-project.org/
• http://zipfr.r-forge.r-project.org/materials/ESSLLI/01_introduction.slides.pdf
• Pointwise Mutual Information, Kullback-leibler distance

Day 2: (6h)

• Perplexity, n-grams based language models
  • SRILM http://www.speech.sri.com/projects/srilm/
  • https://web.stanford.edu/class/cs124/lec/languagemodeling.pdf
• Anomaly detection (static and time-dynamic)
• How to do a cloud tag?
  • Example with http://www.wordle.net/

Sémantique Distributionnelle

• Transparents en ligne (Google Slides)

Programme + materiel pour le cours:

Latent Semantic Analysis

• http://www.datascienceassn.org/sites/default/files/users/user1/lsa_presentation_final.pdf
• http://blog.josephwilk.net/projects/latent-semantic-analysis-in-python.html
• Demo: http://lsa.colorado.edu/

Explicit Semantic Analysis

• Esalib: https://github.com/ticcky/esalib/tree/master/src

Latent Dirichlet Allocation

• Tutorial LDA with Python by J.Barber
• An example of LDA for topic modeling: the Sarah Palin emails

Semantic Vectors

• Google word2vec pre-trained Vectors
• FastText pre-trained vectors