The invention puts forward an abstract sample information retrieval system based on a context, and an abstract sample characterization representing method of the abstract sample information retrieval system. In the system, the abstract sample characterization representing method comprises the following steps: utilizing a word2vecor to extract semantic characteristics, and obtaining the word2vecor of an abstract word; then, carrying out ''optimal fitness division'' clustering on the word2vecor of the abstract word, and replacing the abstract word with clustering centroid according to a clustering result; and finally, according to the centroid and the word frequency of the abstract word represented by the centroid, forming a word2vecor clustering centroid frequency model (ST-IDF (Inverse Document Frequency)) used for carrying out characterization representing on the abstract sample. By use of the abstract sample information retrieval system, the execution frequencies of clustering and fitness calculation can be lowered, the performance of the similarity analysis of the abstract sample is improved, and sample classification accuracy is improved.