TU Darmstadt / ULB / TUbiblio

Assessing SRL Frameworks with Automatic Training Data Expansion

Hartmann, Silvana and Mújdricza-Maydt, Éva and Kuznetsov, Ilia and Gurevych, Iryna and Frank, Anette :
Assessing SRL Frameworks with Automatic Training Data Expansion.
[Online-Edition: https://aclweb.org/anthology/W/W17/W17-0814.pdf]
Proceedings of the 11th Linguistics Annotation Workshop (LAW XI) at EACL 2017 Association for Computational Linguistics
[Conference or Workshop Item] , (2017)

Official URL: https://aclweb.org/anthology/W/W17/W17-0814.pdf

Abstract

We present the first experiment-based study that explicitly contrasts the three major semantic role labeling frameworks. As a prerequisite, we create a dataset labeled with parallel FrameNet-, PropBank-, and VerbNet-style labels for German. We train a state-of-the-art SRL tool for German for the different annotation styles and provide a comparative analysis across frameworks. We further explore the behavior of the frameworks with automatic training data generation. VerbNet provides larger semantic expressivity than PropBank, and we find that its generalization capacity approaches PropBank in SRL training, but it benefits less from training data expansion than the sparse-data affected FrameNet.

Item Type: Conference or Workshop Item
Erschienen: 2017
Creators: Hartmann, Silvana and Mújdricza-Maydt, Éva and Kuznetsov, Ilia and Gurevych, Iryna and Frank, Anette
Title: Assessing SRL Frameworks with Automatic Training Data Expansion
Language: English
Abstract:

We present the first experiment-based study that explicitly contrasts the three major semantic role labeling frameworks. As a prerequisite, we create a dataset labeled with parallel FrameNet-, PropBank-, and VerbNet-style labels for German. We train a state-of-the-art SRL tool for German for the different annotation styles and provide a comparative analysis across frameworks. We further explore the behavior of the frameworks with automatic training data generation. VerbNet provides larger semantic expressivity than PropBank, and we find that its generalization capacity approaches PropBank in SRL training, but it benefits less from training data expansion than the sparse-data affected FrameNet.

Title of Book: Proceedings of the 11th Linguistics Annotation Workshop (LAW XI) at EACL 2017
Publisher: Association for Computational Linguistics
Uncontrolled Keywords: UKP_reviewed;Natural Language Processing;UKP_p_QAEduInf;UKP_p_CLARIND
Divisions: Department of Computer Science
Department of Computer Science > Ubiquitous Knowledge Processing
DFG-Graduiertenkollegs
DFG-Graduiertenkollegs > Research Training Group 1994 Adaptive Preparation of Information from Heterogeneous Sources
Event Location: Valencia, Spain
Date Deposited: 23 Feb 2017 11:35
Official URL: https://aclweb.org/anthology/W/W17/W17-0814.pdf
Identification Number: TUD-CS-2017-0045
Export:

Optionen (nur für Redakteure)

View Item View Item