Researched
Distributional Semantics (Context)
In the 1950s linguists such as Harris and Firth proposed that words used in similar contexts have similar meanings; turned into word vectors, this idea is the root of word2vec and language models.
Open in the interactive tree →Martin Joos (1950) defined a word's meaning through the probabilities of the words around it, Zellig Harris (1954) argued that similar words occur in similar surroundings, and J. R. Firth (1957) summed it up as knowing a word by the company it keeps. The psychologist Charles Osgood (1957) placed word meanings as points in a space, and information-retrieval researchers (Salton, 1971) described words and documents as vectors. Combined, these ideas became the word vectors of word2vec (2013) and the embeddings at the heart of large language models.
Prerequisites
- Probability Theory1654Joos (1950) defined meaning as the probabilities of words in context.
- Comparative Linguistics1786
Unlocks
- Word Embeddings (word2vec)2013Word2vec learns word vectors from the contexts in which words occur.