Makine öğrenmesi ve ses işleme kullanılarak sosyal ağlar üzerinde duygu analizi

Polarity classification is one of the most fundamental problems in sentiment analysis. Our study strives to develop a new definition, extraction technique and utilization of features based on the audio data for polarity classification on Twitter messages. The background of work relies on a recent study which suggests that brain uses sound as a part of language generation and words are comprehended as they are converted into sound. Using sound is effective especially for social media messages which are likely to contain misspelled or shortened words, where the sound is similar to the actual word (e.g., thank u, b4). Our results show that one of our proposed feature set definitions demonstrate an improvement in accuracy in comparison to existing studies.