Clasificación de comentarios hacia futbolistas en Instagram

Abstract

[en] Currently, multiple machine learning models have been implemented for sentiment analysis that have the ability to classify text according to whether it is positive or negative, both individual words and complex sentences. However, the models with the highest hit rates have required high computational power to classify the text in question and also to be constantly updated with more examples. In this case, the aim is to classify the polarity of offensive comments on social networks, specifically on Instagram and directed towards professional footballers. Therefore, the objectives of this study have been defined firstly as the autonomous collection of data and the creation of a dataset to then train models. Following this thread, the next objectives are to investigate the different methodologies, technologies and models of the Python machine learning library, Scikit-learn. Finally, after making a comparison between the 5 selected models, one of these models will be chosen to determine the polarity of the comments previously extracted by sentiment classification (“sentiment analysis”). Despite the low level of personal knowledge available in the field of Natural Language Processing at the beginning, and the lack of computational capacity, the results of the model can be considered satisfactory, since a coherent classification based on a well-founded justification is being obtained. However, if the initial planning had been more accurate, the results could have been improved and if these data are intended to be used in another project, the model should be trained on a machine with higher computational capacity by which the model can be trained for a longer time with more advanced methods, such as some of those that are nowadays considered as part of the state of the art in this field.