文体特征与网络中文学术文献的自动识别——学术文献文体与新闻报道文体的比较研究

摘要

学术文献具有鲜明的文体特征,且部分特征能够用于网络中文学术文献的自动识别与检索,提高学术文献的相对检准率.本文分析了学术文献的部分文体特征,并调查了检索网络中文学术文献时的主要干扰文献--新闻报道的文体特征,从特有表述、平均句长、中西文字符比例三个方面,对两类文献的文体特征进行了分析对比.最后就本文的研究结果如何用于网络中文学术文献检索系统(NSIRS)进行了探讨.%Academic documents have outstanding stylistic features that can be explored to facilitate the automatic identification and retrieval of the Chinese academic papers on the web. This paper analyses some of the stylistic features of the academic papers and of the news reports which form the main noise when searching for academic document. Three aspects of those documents are compared : typical expressions, average length of sentences and ratio of Chinese characters to Roman alphabets. Findings are applied in improving the precision of the system NSIRS designed to identify Chinese academic papers on the web.