Long-Term Goal Discovery in the Twitter Posts through the Word-Pair LDA Model

Abstract

We used the twitter posts about New Year’s resolutions as data source to capture users’ long-term goals. New Year’s resolutions are the commitments that people set for their personal goals, and generally, people plan to fulfill them for the whole following year. Therefore, we can think of such tweets as data source to explore people’s possible long-term goals. The key words in each tweet were extracted for clustering. Considering the form of word-pairs led by verbs is a more intuitive and clearer way to express people’s intentions than the one of separate words, we propose a generative model that incorporates word connections into the smoothed LDA to cluster the key words of long-term goals. The experiments demonstrate the proposed model is capable of clustering the word-pairs with better intuitive character, and clearly dividing people’s long-term goals.