ABSTRACT

This research analyzes the impact of the filtering of word n-grams, using their grammatical category, on the identification of sentiment based on text comments from social networks in Spanish. The impact of filtering n-grams containing adjectives, adverbs and interjections was investigated. It was determined that it is possible to reduce the volume of processed data and at the same time achieve an improvement of up to 30 % in the accuracy when classifying an annotated corpus of test comments separating those that contain sentiment from those that do not.

Key Words: Natural Language Processing; Part of Speech; Sentiment Analysis