DLR-Logo -> http://www.dlr.de
DLR Portal Home | Imprint | Privacy Policy | Contact | Deutsch
Fontsize: [-] Text [+]

One Class Text Classification using an Ensemble of Classifiers

Mayaluru, Hemanth Kumar Reddy (2020) One Class Text Classification using an Ensemble of Classifiers. Master's, Rheinische Friedrich-Wilhelms-Universität Bonn.

[img] PDF


Traditional classification algorithms work in a closed-world scenario where the training data contains all existing classes. In contrast, open set classifiers can handle new input that does not belong to any of the classes seen during training. Open set classification has been studied intensively in the computer vision domain, primarily in handwriting recognition, face recognition, object classification and computer forensics. Here we are interested in open set classification in natural language processing in one class document classification. We propose a new system based on autoencoder for one class classification of documents leveraging the full text. Extending further, we propose a novel ensemble based classifier model, a combination of several basic classifiers, to detect if an incoming document belongs to the class known from training or an unknown class. We compare and evaluate our methods on existing one class classification datasets for NLP - 20 Newsgroups, reuters and webkb. We also extract and use a new full-text dataset from arxiv.org. Our methods significantly outperforms the current state-of-the-art approaches for one class document classification.

Item URL in elib:https://elib.dlr.de/141147/
Document Type:Thesis (Master's)
Title:One Class Text Classification using an Ensemble of Classifiers
AuthorsInstitution or Email of AuthorsAuthor's ORCID iDORCID Put Code
Date:January 2020
Refereed publication:Yes
Open Access:Yes
Number of Pages:61
Keywords:Natural language processing, text classification, autoencoders, ensemble methods
Institution:Rheinische Friedrich-Wilhelms-Universität Bonn
Department:Institut für Informatik
HGF - Research field:other
HGF - Program:other
HGF - Program Themes:other
DLR - Research area:no assignment
DLR - Program:no assignment
DLR - Research theme (Project):no assignment
Location: Köln-Porz
Institutes and Institutions:Think Tank
Deposited By: Hamm, Dr. Andreas
Deposited On:26 Feb 2021 15:11
Last Modified:29 Mar 2021 12:37

Repository Staff Only: item control page

Help & Contact
electronic library is running on EPrints 3.3.12
Website and database design: Copyright © German Aerospace Center (DLR). All rights reserved.