Skip to content
Context Attentive Document Ranking and Query Suggestion
Python
Branch: master
Clone or download
Latest commit 5bd241f Apr 14, 2018
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
multi_task_models initial commit Mar 8, 2018
ranking_baselines initial commit Mar 8, 2018
suggestion_baselines initial commit Mar 8, 2018
LICENSE initial commit Mar 8, 2018
README.md initial commit Mar 8, 2018

README.md

Multitask Learning for Document Ranking and Query Suggestion

This is a PyTorch implementation of the proposed framework and baseline models described in the ICLR 2018 paper Multi-task Learning for Document Ranking and Query Suggestion.

Multi-task Neural Session Relevance Framework

We propose to use a multitask learning framework to jointly learn document ranking and query suggestion, where improved performance on both tasks can be achieved. We propose a multitask neural session relevance framework (M-NSRF) that utilizes previous queries and click-through information from the same search session. M-NSRF is composed of two major components, document ranker and query recommender. Document ranker combines current queryand previous session information and compares the combined representation with the document representation to rank the documents. Query recommender tracks users’ query reformulation sequence considering all previous in-session queries using a sequence to sequence approach. The proposed model is depicted in the following figure.


Figure: Architecture of the Multitask Neural Session Relevance Model (M-NSRM). M-NSRM uses bi-LSTM with max pooling to form query and document representations and use LSTM to gather session-level information. These recurrent states (current query representation and session-level recurrent state, which summarizes all previous queries) are used by query decoder and document ranker for predicting next query and computing relevance scores.

You can’t perform that action at this time.