NLP - What is the best/most accurate method of detecting the meaning of a word?

Question

I am looking for an accurate method of detecting the meaning of a word based on a context. Eg the word 'base' could have several meanings: bottom, military, main part etc.

I know that SpaCy can detect difference in meaning based on Part Of Speech (POS) tags, eg duck noun vs verb, but in the example above this technique does not work. Other method eg Word2Vec works on single word embedding (i think) so its not good neither.

Any ideas how to tackle this using Python?

Answer 1

(1) Train an LSTM to predict the next word in the sequence. The state of the RNN is then a representation of the sentence so far. (2) The predicted word will take into account the context, so will be a disambiguated version of the actual word found. (3) To improve the predictions use a BiLSTM, so it also takes into account the following context.