Please use this identifier to cite or link to this item: http://scholarbank.nus.edu.sg/handle/10635/147849
Title: IMPROVING INFORMATION EXTRACTION WITH WEIGHTED MULTI-TASK LEARNING
Authors: BENJAMIN YAP YAN HAN
Keywords: Natural Language Processing, Information Extraction, Multi-task Learning, Deep learning
Issue Date: 11-May-2018
Citation: BENJAMIN YAP YAN HAN (2018-05-11). IMPROVING INFORMATION EXTRACTION WITH WEIGHTED MULTI-TASK LEARNING. ScholarBank@NUS Repository.
Abstract: Information extraction is a challenging task that extracts structured information from unstructured text. Traditional machine learning techniques for information extraction require task-specific training data and manually engineered features, many of which require a pipeline of external sub-systems to extract named entities, etc. On the other hand, deep learning systems do not require any feature engineering and therefore only require task-specific training data. However, the performance of deep learning systems often suffers when the amount of training data is limited. In this work, we propose the use of multi-task learning to leverage training data from related auxiliary tasks to augment limited task-specific data for training a deep learning model in an end-to-end manner. We also use a loss function which automatically weights the importance of the different tasks. Our experiments on two datasets show that our approach outperforms traditional machine learning systems and deep learning systems trained only on task-specific data of limited quantity.
URI: http://scholarbank.nus.edu.sg/handle/10635/147849
Appears in Collections:Master's Theses (Open)

Show full item record
Files in This Item:
File Description SizeFormatAccess SettingsVersion 
YapYHB.pdf616.91 kBAdobe PDF

OPEN

NoneView/Download

Page view(s)

7
checked on Oct 11, 2018

Download(s)

1
checked on Oct 11, 2018

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.