Fine Grained Insincere Questions Classification using Ensembles of Bidirectional LSTM-GRU Model

Published in Forum for Information Retrieval Evaluation, Kolkata, 2019

Recommended citation: S.D. Das, A. Basak and S. Mandal. Fine Grained Insincere Questions Classification using Ensembles of Bidirectional LSTM-GRU Model . Forum for Information Retrieval Evaluation, Kolkata (2019). http://ceur-ws.org/Vol-2517/T5-5.pdf

abstact
In this paper, we have described our deep learning based system for fine-grained insincere questions classification, which is the CIQ track in FIRE 2019. Our pipeline uses ensembles of bidirectional LSTMGRU model with different word embedding techniques namely Glove, FastText, and Paragram. We have also used the checkpoint ensemble method to enhance performance alongside a combination of two different embeddings per-ensemble. Our pipeline has secured the first position in this track with an F1 score of 67.32%.

download paper here

@article{dasfine, title={Fine Grained Insincere Questions Classification using Ensembles of Bidirectional LSTM-GRU Model}, author={Das, Sourya Dipta and Basak, Ayan and Mandal, Soumil} }