Topic recognition of message threads in social networking

Abstract

This project presents the topic recognition of message threads in English using TF-IDF based on the distribution of similar words. First, a program is designed to extract the words from a sequence of sentences which taken from news articles. Next, a program is designed using TF-IDF with C programming language to compare words similarity between the words around the data and the words around the example sentences from Facebook, Twitter and blogs using TF-IDF coefficient. The similarity measures the occurrence of words around the word with all example sentences from collected. Finally, the performance of this proposed similarity measurement method is evaluated by measuring the precision, recall, and f-measure of the word identification. Furthermore, the test results presented the advantage and disadvantages of the proposed similarity measurement method that applied to classify the English word based on the distribution of similar words. Overall, the performance of our proposed method is good for word classification with three word extraction strategy.