Syntactically Enriched Multilingual Sentiment Analysis
-
-
[1] Universidade da Coruña
Universidade da Coruña
A Coruña, España
-
- Localización: Proceedings of the Workshop on Hybrid Intelligence for Natural Language Processing Tasks (HI4NLP 2020) co-located with 24th European Conference on Artificial Intelligence (ECAI 2020): Santiago de Compostela, Spain, August 29, 2020 / coord. por Pablo Gamallo Otero
, Marcos García González, Patricia Martin-Rodilla, Martín Pereira Fariña, 2020, págs. 5-6 - Idioma: inglés
- Enlaces
-
Resumen
Sentiment analysis of natural language texts needs to deal with linguis- tic phenomena like negation, intensification or adversative clauses. In this talk, I present an approach to tackle such phenomena by means of syntactic information. Our approach combines machine learning and symbolic processing: the former is used to obtain dependency trees for input sentences, and the latter to obtain the sentiment polarity for each sentence using handwritten rules that traverse the tree. Thanks to uni- versal guidelines for syntactic annotation, our approach is applicable to multiple languages without rewriting the rules. Additionally, very accurate parsing is not needed for our approach to be helpful: fast and simple parsers will do, even if they lag behind state-of-the-art accuracy.
