Topic Discovery Using Frequent Subgraph Mining Approach

Abstract

The topic modeling has long been used to check and explore the content of a document in dataset based on the search for hidden topics within the document. Over the years, many algorithms have evolved based on this model, with major approaches such as “bag-of-words” and vector spaces. These approaches mainly fulfill the search, statistics the frequency of occurrences of words related to the topic of the document, thereby extracting the topic model. However, with these approaches the structure of the sentence, namely the order of words, affects the meaning of the document is often ignored. In this paper, we propose a new approach to exploring the hidden topic of document in dataset using a co-occurrence graph. After that, the frequent subgraph mining algorithm is applied to model the topic. Our goal is to overcome the word order problem from affecting the meaning and topic of the document. Furthermore, we also implemented this model on a large distributed data processing system to speed up the processing of complex mathematical problems in graph, which required many of times to execute.