Authors:
Sanae Yamashita
1
;
Yasushi Kami
1
and
Noriyuki Okumura
2
Affiliations:
1
National Institute of Technology, Akashi College, 679–3 Nishioka, Uozumi-cho, Akashi-shi, Hyogo, 674–8501 and Japan
;
2
Otemae University, 6–42 Ochayasho-cho, Nishinomiya-shi, Hyogo, 662–8552 and Japan
Keyword(s):
Emotion Extraction, Emotion Estimation, Response Time, Annotation.
Related
Ontology
Subjects/Areas/Topics:
Artificial Intelligence
;
Communication, Collaboration and Information Sharing
;
Information Security
;
Knowledge Management and Information Sharing
;
Knowledge-Based Systems
;
Social Networks and the Psychological Dimension
;
Symbolic Systems
Abstract:
The existing systems to estimate emotions extract some emotions from the given sentences in any and all circumstances. However, there are many sentences whoever cannot estimate emotional features. It follows that the systems should not extract some emotions all the time. Systems should return "It is difficult to estimate" as we humans do so. This paper proposes a method to classify Japanese sentences based on the difficulty level of emotion estimation. Proposed system judges the difficulty level to estimate emotions using three conditions (negative expressions, emotive expression, and machine-learned classifications). As a result, proposed system achieved 0.8 of F1 score based on mechanical evaluation.