Abstract
This paper presents a method to predicate news reader emotions. News headlines supply core information of articles, thus they can serve as key information for reader emotion predication. However, headlines are always short which leads to obvious data sparseness if only lexical forms are used. To address this problem, words in their lexical forms in a headline are transferred to their concepts and concept sequence features of words in headlines based on a semantic knowledge base, namely HowNet for Chinese. These features are expected to represent the major elements which can evoke reader's emotional reactions. These transferred concepts are used with lexical features in headlines for predicating the reader's emotion. Evaluations on dataset of Sina Social News with user emotion votes show that the proposed approach which do not use any news content, achieves a comparable performance to Bag-Of-Word model using both the headlines and the news contents, making our method more efficient in reader emotion prediction.
Original language | English |
---|---|
Title of host publication | Computational Linguistics and Intelligent Text Processing - 15th International Conference, CICLing 2014, Proceedings |
Publisher | Springer Verlag |
Pages | 73-84 |
Number of pages | 12 |
Edition | PART 2 |
ISBN (Print) | 9783642549021 |
DOIs | |
Publication status | Published - 1 Jan 2014 |
Event | 15th International Conference on Computational Linguistics and Intelligent Text Processing, CICLing 2014 - Kathmandu, Nepal Duration: 6 Apr 2014 → 12 Apr 2014 |
Publication series
Name | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
---|---|
Number | PART 2 |
Volume | 8404 LNCS |
ISSN (Print) | 0302-9743 |
ISSN (Electronic) | 1611-3349 |
Conference
Conference | 15th International Conference on Computational Linguistics and Intelligent Text Processing, CICLing 2014 |
---|---|
Country | Nepal |
City | Kathmandu |
Period | 6/04/14 → 12/04/14 |
Keywords
- Concept Feature
- Concept Sequence Feature
- Emotion Prediction
- HowNet
ASJC Scopus subject areas
- Theoretical Computer Science
- Computer Science(all)