A topical hierarchy at different levels of granularity from short texts has many valuable applications in the areas of opinion analysis

vision detection

semantics mining and graph construction.Aiming at how to mine the hierarchy of topics from short texts

a topical hierarchy mining framework integrated anchor extraction is proposed based on the modification of the tradition phrase definition.Firstly

the topic inference and the anchor extraction are conducted in the proposed framework.Secondly

frequent anchor phrases are found by applying associate rule mining.Finally

a kind of rank method is used to quantity the criterion of anchor phrases in order to find the most representative topical phrase by ranking.Compared to the topic analysis method of the word co-occurrence graph

the word co-occurrence integrated into anchor is more beneficial to build the higher level of topics.Experiments with datasets from the two Chinese short texts are performed

and the results show that the proposed method can generate interpretably phrases and be used for classification prediction.