espanolEn este articulo, se describe la estrategia que subyace al sistema presentado por nuestro grupo para la tarea de analisis de sentimiento en el TASS 2013. El sistema se basa principalmente en un clasificador Naive-Bayes orientado a la deteccion de la polaridad en tweets escritos en espanol. Los experimentos realizados han mostrado que los mejores resultados se han alcanzado utilizando clasificadores binarios que distinguen apenas entre dos categorias de polaridad: positivo y negativo. Para poder identificar mas niveles de subjetividad, hemos incorporado al sistema umbrales de separacion con los que distinguir valores de polaridad fuertes, medios y debiles o neutros. Ademas, para poder detectar si un tweet tiene o no tiene polaridad, el sistema incorpora tambien una regla basica basada en la busqueda de palabras con polaridad dentro del texto analizado. Los resultados de la evaluacion muestran valores razonablemente altos (cerca del 67% de precision) cuando el sistema se aplica para detectar cuatro categorias de sentimiento. EnglishThis article describes the strategy underlying the system presented by our team for the sentiment analysis task at TASS 2013. The system is mainly based on a naive-bayes classifier for detecting the polarity of Spanish tweets. The experiments have shown that the best performance is achieved by using a binary classifier distinguishing between just two sharp polarity categories: positive and negative. To identify more polarity levels, the system is provided with experimentally set thresholds for detecting strong, average, and weak (or neutral) values. In addition, in order to detect tweets with and without polarity, the system makes use of a very basic rule that searchs for polarity words within the analysed text. Evaluation results show a good performance of the system (about 67% accuracy) when it is used to detect four sentiment categories.
[1]
Santiago Fernández Lanza,et al.
Introducing FDSA (Fuzzy Dictionary of Synonyms and Antonyms): applications on information retrieval and stand-alone use
,
2003
.
[2]
Antonio Fernández,et al.
Sentiment Analysis and Topic Detection of Spanish Tweets: A Comparative Study of of NLP Techniques
,
2013,
Proces. del Leng. Natural.
[3]
José Carlos González Cristóbal,et al.
TASS - Workshop on Sentiment Analysis at SEPLN
,
2013
.
[4]
Patrick Paroubek,et al.
Twitter as a Corpus for Sentiment Analysis and Opinion Mining
,
2010,
LREC.
[5]
Kiduk Yang.
WIDIT in TREC 2008 Blog Track: Leveraging Multiple Sources of Opinion Evidence
,
2008,
TREC.
[6]
Pablo Gamallo Otero,et al.
A Corpus and Lexical Resources for Multi-word Terminology Extraction in the Field of Economy in a Minority Language Fco
,
2007
.