基于逆序文本对齐的缩写词识别算法研究

Abbreviation Recognition Algorithm Based on Reverse Text Alignment

针对生物术语的缩写词识别问题,提出了一种基于逆序文本对齐的搜索算法,它实现简单,不需要大量训练数据。该算法在对Medstract标准语料库测试中,准确率和召回率分别为91%和93%;在对包含128篇全文文本的大测试集SBQTL测试中,准确率和召回率分别为96%和84%。在详细分析实验结果后,提出了将文本预处理以及语法规则等自然语言处理技术融入搜索算法作为未来工作的方向。

The identification of abbreviations in biomedical literatures is important for all text mining tools .An abbreviation recognition algorithm based on reverse text alignment was proposed and it is easy to be implemented and need not training data . The algorithm achieves 91%precision and 93%recall on the gold standard corpus ＆quot;Medstract＆quot;and 96%precision and 84%re-call on the larger test data that includes 128 full text literatures .After analyzing the errors produced by the approach , the further improvement work of the approach was discussed .A recognition tendency was proposed that the natural language treatment tech -nology of text pretreatment and grammatical rules should be combined into text alignment algorithm .