TU Darmstadt / ULB / TUbiblio

Discovery and Analysis of Public Opinions on Controversial Topics in the Educational Domain

Vovk, Artem (2013):
Discovery and Analysis of Public Opinions on Controversial Topics in the Educational Domain.
Technische Universität Darmstadt, [Online-Edition: https://download.hrz.tu-darmstadt.de/media/FB20/Dekanat/Publ...],
[Master Thesis]

Abstract

Argumentation is used by everybody in their daily lives as well as work. People frequently need to identify arguments in favor or against a specific topic in order to present some information or make a decision. The educational domain serves as good example. Bachelor graduates often find themselves wondering if they should pursue a Master’s degree or start working in the industry. Finding pros and cons of each possibility is crucial for them in order to make up their mind. The Web is overloaded with data and it is growing constantly. It includes many arguments for topics in various fields but people are not satisfied anymore with traditional search engines that are supposed to find these arguments. Therefore, they look for more intelligent solutions and this is where argumentation mining comes in play. In this work we present a conceptual design of a system with the task to simplify the access to argumentation information concerning a specific topic. We propose to implement such a system as a search engine which looks for the arguments in the Web given a topic as a query. Because of the computation limitations we decide to concentrate only on topics from the educational domain and arguments in German language. We also implement and evaluate the critical parts of the system such as: a focused crawler, argument extraction and classification module as well as the front-end interface. For the extraction and classification part we decide to use supervised machine learning techniques. Therefore, first we collect the documents which contain the arguments. Secondly, we define the annotation scheme and perform the annotation study. As a result we create a labeled corpus, which is used for training models for the argument extraction and classification experiments. Finally, we evaluate the influence of different classification algorithms as well as the combination of different features and perform the error analysis.

Item Type: Master Thesis
Erschienen: 2013
Creators: Vovk, Artem
Title: Discovery and Analysis of Public Opinions on Controversial Topics in the Educational Domain
Language: English
Abstract:

Argumentation is used by everybody in their daily lives as well as work. People frequently need to identify arguments in favor or against a specific topic in order to present some information or make a decision. The educational domain serves as good example. Bachelor graduates often find themselves wondering if they should pursue a Master’s degree or start working in the industry. Finding pros and cons of each possibility is crucial for them in order to make up their mind. The Web is overloaded with data and it is growing constantly. It includes many arguments for topics in various fields but people are not satisfied anymore with traditional search engines that are supposed to find these arguments. Therefore, they look for more intelligent solutions and this is where argumentation mining comes in play. In this work we present a conceptual design of a system with the task to simplify the access to argumentation information concerning a specific topic. We propose to implement such a system as a search engine which looks for the arguments in the Web given a topic as a query. Because of the computation limitations we decide to concentrate only on topics from the educational domain and arguments in German language. We also implement and evaluate the critical parts of the system such as: a focused crawler, argument extraction and classification module as well as the front-end interface. For the extraction and classification part we decide to use supervised machine learning techniques. Therefore, first we collect the documents which contain the arguments. Secondly, we define the annotation scheme and perform the annotation study. As a result we create a labeled corpus, which is used for training models for the argument extraction and classification experiments. Finally, we evaluate the influence of different classification algorithms as well as the combination of different features and perform the error analysis.

Uncontrolled Keywords: UKP_a_ENLP
Divisions: 20 Department of Computer Science
20 Department of Computer Science > Ubiquitous Knowledge Processing
Event Location: Darmstadt
Date Deposited: 31 Dec 2016 14:29
Official URL: https://download.hrz.tu-darmstadt.de/media/FB20/Dekanat/Publ...
Identification Number: TUD-CS-2013-0440
Referees: Ferschke, Oliver and Gurevych, Iryna
Export:
Suche nach Titel in: TUfind oder in Google

Optionen (nur für Redakteure)

View Item View Item