作者: Vanni Zavarella , Alexandra Balahur , Jenya Belyaeva , Mijail A. Kabadjov , Ralf Steinberger
DOI:
关键词: Sentiment analysis 、 Information retrieval 、 Context (language use) 、 Interpretation (philosophy) 、 Computer science 、 Artificial intelligence 、 Newspaper 、 Natural language processing 、 News values 、 Annotation
摘要: Recent years have brought a significant growth in the volume of research sentiment analysis, mostly on highly subjective text types (movie or product reviews). The main difference these texts with news articles is that their target clearly defined and unique across text. Following different annotation efforts analysis issues encountered, we realised opinion mining from other types. We identified three subtasks need to be addressed: definition target; separation good bad content expressed marked explicitly, not needing interpretation use world knowledge. Furthermore, distinguish possible views newspaper ― author, reader text, which addressed differently at time analysing sentiment. Given definitions, present work opinions about entities English language news, apply concepts. Results showed this idea more appropriate context approaches taking into consideration produce better performance.