Publicação
Creating a social media-based personal emotional lexicon
| Resumo: | One of the major problems when using lexicon in sentiment analysis is that they do not cover all possible words in a text and frequently they miss the more expressive to describe the emotions of the text's author efficiently. This problem occurs because people in non-official, on formal channels, communicate using slangs, neologisms, new patterns based on abbreviations (as "aka", "brb" and "asap") and the different meanings, making challenging to analyse texts using a finite subset of a language. This is a problem because some unknown words can completely change the meaning of a sentence, producing misunderstandings. In this paper we present an approach to expand an emotional lexicon for a specific author, producing a customised lexicon which represents how the author "feels" the words. In our experiments, we got an increase of 35.34% and 107.02% in the dictionary size when compared to the original lexicon using two different authors, and identifying different emotions from the same text according to each author's lexicon, i.e. interpreting the text according to the author's "point of view". |
|---|---|
| Autores principais: | Martins, Ricardo |
| Outros Autores: | Almeida, J. J.; Novais, Paulo; Henriques, Pedro Rangel |
| Assunto: | Sentiment Analysis Machine Learning Natural Processing Language Ciências Naturais::Ciências da Computação e da Informação Engenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática |
| Ano: | 2018 |
| País: | Portugal |
| Tipo de documento: | comunicação em conferência |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade do Minho |
| Idioma: | inglês |
| Origem: | RepositóriUM - Universidade do Minho |
| Resumo: | One of the major problems when using lexicon in sentiment analysis is that they do not cover all possible words in a text and frequently they miss the more expressive to describe the emotions of the text's author efficiently. This problem occurs because people in non-official, on formal channels, communicate using slangs, neologisms, new patterns based on abbreviations (as "aka", "brb" and "asap") and the different meanings, making challenging to analyse texts using a finite subset of a language. This is a problem because some unknown words can completely change the meaning of a sentence, producing misunderstandings. In this paper we present an approach to expand an emotional lexicon for a specific author, producing a customised lexicon which represents how the author "feels" the words. In our experiments, we got an increase of 35.34% and 107.02% in the dictionary size when compared to the original lexicon using two different authors, and identifying different emotions from the same text according to each author's lexicon, i.e. interpreting the text according to the author's "point of view". |
|---|