Sentiment analysis in Spanish-language tweets
Abstract
SENTAT: Self-Attention for Twitter Sentiment Analysis in Spanish. Increasingly, social media are used to express opinions and sentiments about companies, institutions, products, services, politics, etc. Companies and organizations in general have shown their interest in knowing the opinions and sentiments that their activities and products arouse in society. In this sense, Twitter has become an excellent tool for these purposes. The joint use of machine learning and natural language processing techniques allows us to automatically determine the opinion on a certain topic by analyzing social media. The scientific community has made great efforts to analyze, structure and process the information generated in social networks using machine learning, deep learning, and natural language processing techniques. The main characteristic of our approach is that it uses deep learning models specifically trained for Twitter in both Spanish and English languages. In addition, the models have been tested in international competitions where they have obtained the best results in analyzing opinions in Spanish from Spain as well as in various variants of Spanish from Latin America. To use SENTAT, a Docker container is provided that allows it to run on Linux, Windows, and MacOS operating systems. The classification models are accessible through a web application independent of the operating system thanks to the provided REST API. The set of libraries and third-party software are freely distributed.
Technical specifications
Type of technology
SOFTWARE