[Purpose/Significance] Aiming at the problem of insufficient information analysis ability in the current public security business about identification and screening of crime clues in hotline texts, a method of hotline text data crime clue screening based on keyword mining is proposed to help business departments improve relevant intelligence and judgment [Method/Process] Considering that algorithms such as automatic text classification are subject to the problem of sample size, this paper firstly identified the key information of the known attribute data and established a seed lexicon, and then used Word2Vec to expand the seed vocabulary from the perspectives of similar words and synonym words to form a professional thesaurus, and finally used a semantics-based integral screening model to screen criminal clues in the hotline text data. [Result/Conclusion] This paper conducted a crime clue screening experiment on 1 050 priori hotline text data in Jinan City. After actual comparison and index analysis, the recall rate reached 86%. The specific identification of crime information in the text data of the city hotline achieved the expected effect and realized the effective screening of crime clues.