PROJECT TITLE :

Context-Aware Sliding Window for Sentiment Classification

ABSTRACT:

Sentiment categorization is a hot topic in science, with applications in a variety of fields. Many researchers have presented approaches to accurately discern feelings in the past. However, the focus is on the papers' syntactic and semantic properties. These features are useful, but they disregard the user's previous feelings. We hypothesize in this study that previous attitudes aid the classifier in efficiently linking the user's history with the contents of the current tweet. As a result, learning algorithms are able to correlate past activities in order to determine current feelings. We propose three sliding window features to collect past sentiments from time series data for this purpose. On diverse Machine Learning and Deep Learning techniques, we offer seven versions of Context-aware Sliding Window (CSW) features in this study. We also present a temporal dataset of user tweets that has been manually classified by nine human annotators. The proposed dataset contains 4,557 tweets from 36 people. The results show that six state-of-the-art baseline approaches are significantly improved.