A linguistic model identifies feelings in texts written on social networks

Words that we use every day in any field such as “fast”, “slow”, beautiful”, “hot” or “normal” implicitly carry a load of information that is increasingly important to companies and organizations. Evaluative language, the one used every day, is one of the productions of the language that makes communicative interactions interesting, since they provide essential information.

Sentiment analysis is a field that has received great attention in recent years due to the massive use of social networks. The use of these virtual communities has generated large amounts of evaluative text produced by users around all kinds of products and services.

Given the interest that the analysis of these texts has aroused, a research team from the Department of Romance Philology of the Rovira i Virgili University (URV) in Tarragona has developed a technique that includes different mathematical and linguistic methods and that manages to formally model the statements. evaluative and capture or extract the feeling (or evaluation) behind these linguistic expressions of a diffuse nature. The result of their research, which has been carried out in collaboration with the IRAFM center of excellence in the Czech Republic, has been published in the academic journal Mathematics.

To analyze sentiment, computational tools are used that detect and evaluate evaluative language, in terms of polarity, that is, they automatically classify texts based on the positive or negative connotation of the language used. With this analysis, an attempt is made to determine the attitude of a person with respect to an issue. The attitude can be a judgment or evaluation, its affective state (emotional state of the author when writing), or the emotional communicative intention (the emotional effect that the author tries to cause in the reader). The development of these sentiment analysis tools requires formal models that can describe the evaluative language in terms that a machine is capable of processing.

Adrià Torrens and María Dolores Jiménez. (Photo: URV)

Evaluative language is said to be diffuse or vague, since it is very difficult to delimit its meaning from everyday words such as good, bad, big, small, love, hate, etc. For example, a 5-year-old boy can be “tall” if he is 130 cm tall, and an adult basketball player, on the other hand, is “tall” if he is 220 cm tall. This variability can also be found between cultures: for example, the final meaning of the adjective “tall” is surely different in the American and Japanese conceptions. Although the final meaning is different, everyone can understand that “high” means high value on a height scale. A model to characterize this “fuzziness” in meaning is a diffuse model, and this is the basis of the proposal of this research, headed by Adrià Torrens and María Dolores Jiménez, from the Research Group in Mathematical Linguistics of the Department of Romance Philologies of the URV, together with Vilém Novák, from the University of Ostrava, in the Czech Republic. (Source: URV)