Site Search
Computer Science

Photos

Robin Gras, Ph.D.Dr. Robin Gras
Dr. Robin Gras
Xiaobu Yuan, Ph.D.Dr. Xiaobu Yuan
Dr. Xiaobu Yuan
Dr. Scott GoodwinDr. Scott Goodwin
Dr. Scott Goodwin
Jessica Chen, Ph.D.Dr. Jessica Chen
Dr. Jessica Chen
Windsor WaterfrontWindsor Waterfront Park
Windsor Waterfront Park
Christie Ezeife, Ph.D.Dr. Christie Ezeife
Dr. Christie Ezeife
Imran Ahmad, Ph.D.Dr. Imran Ahmad
Dr. Imran Ahmad
Dr. Robert KentDr. Robert Kent
Dr. Robert Kent
Dr. Luis RuedaDr. Luis Rueda
Dr. Luis Rueda
Alioune Ngom, Ph.D.Dr. Alioune Ngom
Dr. Alioune Ngom
Dr. Ziad Kobti lecturingDr. Ziad Kobti
Dr. Ziad Kobti
Arunita Jaekel, Ph.D.Dr. Arunita Jaekel
Dr. Arunita Jaekel
Lambton TowerLambton Tower
Lambton Tower

Thesis Proposal Announcement: Improving Document Embedding Using Retrofitting

Add this event into your calendar using the iCAL format
  • Mon, 01/14/2019 - 10:00am - 12:00pm




The School of Computer Science at the University of Windsor is pleased to present…………

 

Improving Document Embedding Using Retrofitting
MSc Thesis Proposal by:  Zeeshan Mansoor
Date:  Monday, Jan 14th, 2019
Time: 10:00 am – 12:00 pm
Location: 3105, Lambton Tower
 
 
Abstract: Data-driven learning of document vectors that capture linkage between them is of immense importance in the natural language processing (NLP). These document vectors can, in turn, beused for tasks like information retrieval, document classification, and clustering. Inherently, documents are linked together in the form of links or citations in case of web pages or academic papersrespectively. Methods like Paragraph Vector try to capture the semantic representation of the document using only the textual information. These methods ignore the network information altogether while learning the representation. Similarly, methods developed for network representation learning like Node2Vec or DeepWalk, capture the linkage information between the documents but they ignore the textual information altogether. In this thesis, we propose a method based on Retrofitting
word vectors using semantic lexicons for improving the word embeddings. The proposed method tries to incorporate both the textual and network information together by bringing citing documents closer together in the vector space.

 

 

Internal Reader: Dr. Saeed Samet
External Reader: Dr. Myron Hlynka
Advisor: Dr. Ziad Kobti
Co-advisor: Dr. Mehdi Kargar
Chair: Dr. Jianguo Lu



Christine Weisener
cweisen@uwindsor.ca
(519)253-3000 ext.3716