中文文本同频词统计规律及在关键词提取中的应用

摘要

通过大量实验对中文文本中同频词的统计规律进行了研究，利用齐普夫定律推导出了适合中文文本的同频词数的数学表达式，能更准确地表示出不同长度的文本中各频次的同频词数；借助同频词数的数学表达式，重新确立了中文文本中高频词和低频词的界分公式，并通过实验验证了该公式能够更好地界分高频词和低频词。将提出的统计规律应用于中文文本关键词提取，有效提高了关键词提取效率，在文本长度不小于3010词的前提下，频次为1和频次为2的词不必参与TF-IDF值的计算，可将计算效率提高2～7倍，且没有造成关键词丢失。解决了学术界关心的如何处理中文低频词的问题，对关键词提取中如何处理低频词提供了可操作标准。%This paper presented a statistics law on the same frequency words in Chinese text based on a large number of experi-ments.It deduced the mathematical expression of the same frequency words based on Zipf’s law,which could be applied to Chinese text better.Moreover,it re-established the boundary points formula of high-frequency words and low-frequency words, and then verified its correctness.Finally,it applied the proposed statistics law to keywords extraction.Previous academic re-search on how to deal with low-frequency words was rare and nobody gave a concrete solution.This paper provided a standard method on how to deal with the low-frequency words in the application of keywords extraction.It notes that text length must be no less than 3 010 words and it can ignore the calculation of words occurring once and twice when calculating the value of TF-IDF.This method raises the efficiency by 2~7 times but no loss of keywords.