基于平行语料库的专利摘要术语和词汇量化分析

Quantitative Analysis of Patent Abstract Terms and Vocabulary Based on Parallel Corpus

谷歌专利(Google Patent)为广大用户提供了便利的专利检索和翻译服务,但其(Google Patent)翻译文本仍然存在着一定的不足。本文基于谷歌专利(Google Patent)的检索文本,探讨美国专利局专利摘要在专利词汇层面的差异,进而探讨计算机通信领域专利摘要中存在的翻译共性。本文从谷歌专利(Google Patent)中随机选取100篇专利摘要文本,借助Antconc文本分析工具和微型平行语料库探究其词汇层面的翻译共性。研究表明,谷歌专利(Google Patent)在术语的层面,虽然大多数核心词能做到相对精确的对应,但仍存在着部分名词的指代偏差、修饰词搭配偏差和术语结构顺序偏差。在词汇密度和类符比方面,美国专利局的专利摘要文本和谷歌专利摘体现着一定程度的翻译共性。

Google Patent provides a convenient patent search and translation service for users, but there are still some shortcomings in its translation. Based on the search text of Google Patent, this paper explores the differences in patent vocabulary between US Patent Office patent abstracts, and then explores the commonalities of translation in patent abstracts in computer communications. This paper randomly selects 100 patent abstract texts from Google Patent, and explores the commonality of translation at the lexical level by means of Antconc text analysis tools and micro-parallel corpus. Studies have shown that Google Patent at the level of terminology, although most of the core words can achieve a relatively accurate correspondence,there are still some nouns of the deviation of the reference, modifier collocation deviation and term structure order deviation.In terms of vocabulary density and class proportions, the US Patent Office＇s patent abstract text and Google patent abstracts reflect a certain degree of translation commonality.