Item – Theses Canada

​
OCLC number
1122757083
Link(s) to full text
LAC copy
Author
Yang, Wei,1995-
Title
End-to-end neural information retrieval
Degree
Thesis (M.Math)--University of Waterloo, 2019.
Publisher
Waterloo, Ontario, Canada : University of Waterloo, 2019.
Description
1 online resource (x, 54 pages) :illustrations (chiefly colour)
Notes
"A thesis presented to the University of Waterloo in fulfillment of the thesis requirement for the degree of Master in Computer Science."
Includes bibliographical references (pages 45-54).
Abstract
In recent years we have witnessed many successes of neural networks in the information retrieval community with lots of labeled data. Yet it remains unknown whether the same techniques can be easily adapted to search social media posts where the text is much shorter. In addition, we find that most neural information retrieval models are compared against weak baselines. In this thesis, we build an end-to-end neural information retrieval system using two toolkits: Anserini and MatchZoo. In addition, we also propose a novel neural model to capture the relevance of short and varied tweet text, named MP-HCNN. With the information retrieval toolkit Anserini, we build a reranking architecture based on various traditional information retrieval models (QL, QL+RM3, BM25, BM25+RM3), including a strong pseudo-relevance feedback baseline: RM3. With the neural network toolkit MatchZoo, we offer an empirical study of a number of popular neural network ranking models (DSSM, CDSSM, KNRM, DUET, DRMM). Experiments on datasets from the TREC Microblog Tracks and the TREC Robust Retrieval Track show that most existing neural network models cannot beat a simple language model baseline. How- ever, DRMM provides a significant improvement over the pseudo-relevance feedback baseline (BM25+RM3) on the Robust04 dataset and DUET, DRMM and MP-HCNN can provide significant improvements over the baseline (QL+RM3) on the microblog datasets. Further detailed analyses suggest that searching social media and searching news articles exhibit several different characteristics that require customized model design, shedding light on future directions.
Other link(s)
hdl.handle.net
uwspace.uwaterloo.ca
Subject
Natural language processing (Computer science)
Text processing (Computer science)
Information storage and retrieval systems.
Information retrieval.
Neural networks (Computer science)
Word processing.
Word processing operations.
Traitement automatique des langues naturelles.
Traitement de texte.
Systèmes d'information.
Recherche de l'information.
Réseaux neuronaux (Informatique)
information retrieval, neural network, text matching