I need to specify semantically close words in a corpus. For instance newspaper/magazine in the following:

I read [newspaper/magazine] yesterday.

In this example both words are close in meaning. I need to find these kind of words.

How can I find those words?


1 Answer 1


Add a regularization term for synonymity to the objective function (to be maximized):

$\log P(w_i|h) \rightarrow \log P(w_i|h) + \lambda \frac{1}{|S|} \sum_{(i,j) \in S} sim(w_i, w_j)$ where $S$ is the set of synonyms, and $sim$ is your word similarity function (e.g., cosine similarity).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.