| With the rapid development of educational informationization, the Internet is playing an increasingly vital role of getting access to teaching resources. The amount of digital resources is experiencing a remarkable growth, which provides a broad range of options for learners, but at the same time reduces the filtering efficiency. And therefore, selecting the required one efficiently from a huge collection of resources is quite important. Automatic Abstract Technology compresses the texts, extract the core contents, and assists learners to have a better and quicker understanding of the main content of text.In the first place, this paper makes an analysis and research concerning algorithms of key words extraction and various approaches of extracting the topic sentences. The author choose the algorithm based on word frequency to get feature words of the article, and then calculate each sentences’ weight value on the basis of words’ comprehensive weight, location, structure and some other sentence’s feature. Finally, the author performed a series of experiments and then implemented the function of Automatic Abstract in Siyuan Network Teaching Platform, which can generate abstract with high performance.Firstly, the Automatic Abstract System transforms the format of received resources in order to obtain its plain text and perform a shallow parsing on the text structure for those need abstract. Because Word Segmentation is the first vital step of Automatic Abstract, the author choose ICTCLAS, whose performance is satisfactory and then extract key and effective words with artificial corpus. And subsequently the algorithm of semantic similarity measurement is raised, which is based on semantic dictionary, to weaken the negative impact on the redundancy resulted from synonymous sentences. Additionally, Association Rules is applied to the generated abstract, aiming to enhance its readability. Finally, the system has implemented the presentation of key words and automatic abstract. |
PDF Full Text Request