JeSemE models for lexical semantic change

Models for diachronic lexical semantics used by the Jena Semantic Explorer (JeSemE) web site described in our COLING 2018 paper "JeSemE: A Website for Exploring Diachronic Changes in Word Meaning and Emotion".

Also described and applied in Johannes Hellrich's Ph.D. thesis "Word Embeddings: Reliability & Semantic Change" who was funded by the Deutsche Forschungsgemeinschaft (DFG) within the graduate school "The Romantic Model" (GRK 2041/1).

One ZIP file per corpus, each containing several CSV files:

• CHI.csv with χ² word association values (structure: word-id, word-id, time, value)
• EMBEDDING.csv with SVD-PPMI word embeddings (aligned; structure: word-id, time, values)
• EMOTION.csv with VAD word emotion values (structure: word-id, time, values)
• FREQUENCY.csv with relative word frequency values (structure: word-id, time, value)
• PPMI.csv with PPMI word association values (structure: word-id, word-id, time, value)
• SIMILARITY.csv with word embedding derived word similarity values (structure: word-id, word-id, time, value)
• WORDIDS.csv mapping words to their corpus specific IDs

Corpora are:

• coha: Corpus of Historical American English

• dta: Deutsches Textarchiv 'German Text Archive'

• google_fiction: Google Books N-Gram corpus, English fiction subcorpus

• google_german: Google Books N-Gram corpus, German subcorpus

• rsc: Royal Society Corpus