Word2vec

Word2vec is a technique in <a href="/facts/Natural_language_processing/1hjMKsSN">natural language processing</a> (NLP) for obtaining <a href="/facts/Vector_space/M1pxMLx2">vector</a> representations of words. These vectors capture information about the meaning of the word based on the surrounding words. The word2vec algorithm estimates these representations by modeling text in a large <a href="/facts/Corpus_of_text/bYeEQQdh">corpus</a>. Once trained, such a model can detect <a href="/facts/Synonym/a3BgMcdu">synonymous</a> words or suggest additional words for a partial sentence. Word2vec was developed by <a href="/facts/Tom%25C3%25A1%25C5%25A1_Mikolov/3WlvSdBT">Tomáš Mikolov</a>, Kai Chen, Greg Corrado, <a href="/facts/Ilya_Sutskever/dBXbUQ39">Ilya Sutskever</a> and <a href="/facts/Jeff_Dean/YIEWyTM4">Jeff Dean</a> at Google, and published in 2013.
Word2vec represents a word as a high-dimension <a href="/facts/Vector_(geometry)/bCLrdksy">vector</a> of numbers which capture relationships between words. In particular, words which appear in similar contexts are mapped to vectors which are nearby as measured by <a href="/facts/Cosine_similarity/vrbNb6wH">cosine similarity</a>. This indicates the level of <a href="/facts/Semantic_similarity/wuInaoaB">semantic similarity</a> between the words, so for example the vectors for walk and ran are nearby, as are those for "but" and "however", and "Berlin" and "Germany".

Word2vec open-in-new

Word2vec