search

teinhonglo / Information-retrieval

52 stars 33 forks source link

Relevance-based Word Embedding #1

Closed hamed-zamani closed 6 years ago

hamed-zamani commented 7 years ago

Hi,

I've noticed that you have an implementation of relevance-based word embedding in this repo. Is this the implementation of the following paper? https://dl.acm.org/citation.cfm?doid=3077136.3080831

Best, Hamed Zamani zamani@cs.umass.edu

teinhonglo commented 7 years ago

This is refer the first method (Relevance Likelihood Maximization Model) in the paper.

In my understanding: Input is original BOW representation, which feed into feedforward neural network, then output is reformulated representation with other Pseudo Relevance feedback.

Best, Tien-Hong, Lo

hamed-zamani commented 7 years ago

Thanks for the response. Have you tested it? Are the quality of embeddings better than word2vec? Are you sure that the implementation is correct?

I am asking these questions, because people reach me regarding this paper and I would like to point them to your repo, if you have tested it and you're sure that the implementation is correct.

Thanks, Hamed