Authors:
Nina Rizun
1
and
Wojciech Waloszek
2
Affiliations:
1
Department of Applied Informatics in Management, Gdansk University of Technology, Gdansk and Poland
;
2
Department of Software Engineering, Gdansk University of Technology, Gdansk and Poland
Keyword(s):
Textual Content Classification, Hierarchical Sentiment Dictionary, Text Tonality, Evaluation the Quality, Bigrams, Polarity Scores.
Related
Ontology
Subjects/Areas/Topics:
Artificial Intelligence
;
Knowledge Discovery and Information Retrieval
;
Knowledge-Based Systems
;
Mining Text and Semi-Structured Data
;
Symbolic Systems
Abstract:
This paper presents the methodology of Textual Content Classification, which is based on a combination of algorithms: preliminary formation of a contextual framework for the texts in particular problem area; manual creation of the Hierarchical Sentiment Dictionary (HSD) on the basis of a topically-oriented Corpus; tonality texts recognition via using HSD for analysing the documents as a collection of topically completed fragments (paragraphs). For verification of the proposed methodology a case study of Polish-language film reviews Corpora was used. The main scientific contributions of this research are: writing style of the analyzed text determines the possibility of adaptation of the Texts Classification algorithms; Hierarchically-oriented Structure of the HSD allows customizing the classification process to qualitative recognition of text tonality in the context of individual paragraphs topics; texts of Persuasive style most often are initially empowered by authors with a certain
tonality. The tone, expressed in the author's opinion, effects the qualitative indicators of sentiment recognition. Negative emotions of the author usually reduce the level of vocabulary variability as well as the variety of topics raised in the document, but simultaneously increase the level of unpredictability of words contextually used with both positive and negative emotional coloring.
(More)