This repository contains the implementation of three different models for document retrieval;
- MP_original: an implementation of the MatchPyramid approach presented in https://arxiv.org/abs/1606.04648
- Prob_WE: a modification of the MatchPyramid model using the Probabilistic Word Embeddings proposed in: https://arxiv.org/abs/1805.07594;
- MP_WN_WE: an implementation of a new MatchPyramid model using pre-trained WordNet (http://www.di.fc.ul.pt/~ahb/pubs/2018SaediBrancoRodriguesEtAL.pdf) and FastText word embeddings to perform retrieval.