Please use this identifier to cite or link to this item:https://hdl.handle.net/20.500.12259/57570
Type of publication: Konferencijų tezės nerecenzuojamuose leidiniuose / Conference theses in non-peer-reviewed publications (T2)
Field of Science: Informatika / Informatics (N009)
Author(s): Markievicz, Irena;Tamošiūnaitė, Minija;Vitkutė-Adžgauskienė, Daiva;Krilavičius, Tomas
Title: Text mining for robotic action ontology engineering
Is part of: Data analysis methods for software systems : 8th international workshop, Druskininkai, Lithuania, December 1-3, 2016 : [abstracts book]. Vilnius : Vilnius University Institute of Data Science and Digital Technologies, 2016
Extent: p. 36-37
Date: 2016
ISBN: 9789986680611
Abstract: The development of the action ontology (ACAT) from domain specific texts allows to discover previously unknown dependencies between robotic actions and their environment objects. This study explains the conceptual model of the ontology actions and environment objects and relations between them. Two main ACAT ontology classes determine the hierarchical structure of action and object hyponymy/hypernymy, troponyny: „action“ and „object“. Each action and object synset contains a subset of synonymous entities. All synsets from the ontology are described by the semantic roles, used in action execution by robots: main action, main object, primary object and secondary object. Study also explores various text mining methods for action ontology learning: collocation extraction, frequency lists, bag-of-words, word space model and Heart’s hyponomy recognition patterns. Verbnet thematic roles and frames are used to identify text syntactic and semantic structure – in this way recognized new text patterns allow to define dependencies between ontology synsets. Robotic action classes are identified by text classification with SVM machine learning method, where action categories are treated as classes, and appropriate verb context – as classification instances. The action ontology completeness and utility is evaluated empirically, by running as additional source of knowledge base in instruction processing system. This study introduces the preliminary testing results of the ACAT ontology usage in instruction processing to sequence of robotic execution tasks (chosen rotor assembly and biotechnology laboratory scenarios). While the explicit knowledge is parsed directly from the instructions, reasoning on queried ACAT ontology data allows to cover instruction tacit knowledge. It helps to execute human readable instructions with polysemous information, omissions, too general or non-robotic actions and not relevant texts
Internet: https://hdl.handle.net/20.500.12259/57570
Affiliation(s): Informatikos fakultetas
Sistemų analizės katedra
Taikomosios informatikos katedra
Vytauto Didžiojo universitetas
Appears in Collections:Universiteto mokslo publikacijos / University Research Publications

Files in This Item:
marc.xml7.21 kBXMLView/Open

MARC21 XML metadata

Show full item record
Export via OAI-PMH Interface in XML Formats
Export to Other Non-XML Formats

Page view(s)

158
checked on Mar 30, 2020

Download(s)

14
checked on Mar 30, 2020

Google ScholarTM

Check

Altmetric


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