提出一种基于N元语法的英文学术文献聚类标签抽取算法,该算法利用N元语法在大规模语料库上进行先期学习生成领域短语词表,再通过K-means算法进行聚类,从聚簇中抽取N元语法项计算TFIDF值,对出现在词表中的特征项赋以更高的权值,以得分最高的特征项作为聚类标签。实验结果表明,该算法能获得更好的实验效果。同时,在抽取聚类标签时提出一种改进的TFIDF权重计算,在评价标签质量时提出一种新的标签评价方法R@N方法。 This paper proposes an algorithm for clustering label extraction based on N-gram in English academic literature. The algorithm uses N-gram to conduct the pre-learning on a large-scale corpus to generate the domain phrase list and then uses the K-means algorithm to cluster. In the extraction of the N grammatical term to calculate the TFIDF value, append the feature items appearing in the vocabulary with higher weights, and use the feature score with the highest score as the cluster label. Experimental results show that the algorithm can get better experimental results. At the same time, an improved TFIDF weight calculation is proposed when extracting the clustering labels, and a new label evaluation method R @ N method is proposed when evaluating the label quality.