The published scientific and technical abstracts provide semantic fact data of problems, methods, and results in scientific research activities, as well as reliable factual data for the use of artificial intelligence in knowledge services. If they can be accurately identified and separated, an intelligent and innovative knowledge question and answer system can be realized. This study proposed a semantic recognition and classification method for scientific and technical abstracts. The method first sorted the scientific and technical abstracts according to the syntactic and semantic functions and then performed statistical analysis of the number distribution of class and sentence positions, sentence type, and semantic structure of the sentence. The semantic word order characteristics of sentences were finally classified and merged based on this analysis, and the classification and experiment of the problems of scientific and technical abstracts were realized. The accuracy of the classification was 99%. The effectiveness of semantic recognition and classification algorithm was verified manually. The experimental results showed that this method had a simple algorithm and displayed high classification accuracy and good universality.