word2vec prototype

[githubmanticore]

word2vec may be useful for

• phrase completion suggestion
• synonyms suggestion
• text similarity calculation (e.g. finding related posts)

Some articles:

• https://medium.freecodecamp.org/how-to-get-started-with-word2vec-and-then-how-to-make-it-work-d0a2fca9dad3
• https://datascience.stackexchange.com/questions/9785/predicting-a-word-using-word2vec-model
• https://stackoverflow.com/questions/44951605/word2vec-guesing-word-embeddings

Let's make a prototype of this by building word2vec vectors during indexation and then making it possible to use the vectors. The questions to answer are:

• resources consumption: how much more RAM/disk will be needed
• indexation latency: how much higher it will be
• functionality: what we can and what we can't do with that
• search performance overhead: how enabling word2vec in an index will affect normal searches (that don't have anything to do with word2vec)
• word2vec search performance: how fast is searching through the word2vec built structures