Research on LAK Algorithm for Short Text Topic Detection

Aijun Li, Tong Chang


In the process of topic detection for short text, the accuracy of topic detection is affected because of the limited length of short text, sparse features and lack of semantics. In this paper, we propose a LAK algorithm for short text topic detection. In the algorithm, the probability distribution of topic-words is obtained by the Latent Dirichlet Allocation model. Moreover, LAK algorithm which combining AGNES (Agglomerative Nesting) and K-means algorithm is chose to cluster the data, in order to compensate for the subjective randomness of K-means algorithm for K-value selection and the ignorance of semantics when calculating feature weights with TF-IDF. The effectiveness of LAK algorithm in topic detection is proved by experiments.


Full Text:



  • There are currently no refbacks.