Testing Distributional Hypothesis in Patent Translation

The 2014 Conference on Computational Linguistics and Speech Processing ROCLING 2014, pp. 185-192  The Association for Computational Linguistics and Chinese Language Processing Testing Distributional Hypothesis in Patent Translation Hsin-Hung Lin and Yves Lepage Abstract This paper presents a wordlist-based lexical richness approach to testing distributional hypothesis for genre analysis in translation studies. In recent years, there has been continuing interest in patent translation. However, there are only a few lay their interests on comparison between native and non-native writing. The proposed approach to terms distrubution of technical words contained in United States Patent and Trademark Office (USPTO) and Japan Patent Office (JPO) in terms of lexical variation, lexical density and lexical sophistication, in brief, highlights distributional similarity of technical genre, and in particular, distibutional difference of academic and general genres. Keywords: Patent Translation, Native Characterization, Corpus, Co-Occurrence. 1. Introduction As globalization has resulted in rapid greater economic growth, the challenges of interdisciplinary interaction in pursuit of precise patent writing have incredibly increased. In Lin and Hsieh (2010a), English patent documents were statistically extracted and computationally examined from LexisNexis Academic, a database for legal professionals. They compiled a reference corpus of independent claim texts and lay the focus on their collocation features. Mutual information is attainable with the help of selectional collocation features underlining specific clausal types represented in natural language processing of patent specification. While their work appears to fill a niche in the ESP (English for Specific Purposes) field (and particularly in the English for Occupational Legal Purposes), Lin and Hsieh (2010b) further compiled a modern patent language technical term list with statistical-retrieval methodologies as a mandatory  Graduate School of Information, Production, and Systems, Waseda University, Japan. E-mail: nobuhiro602@toki.waseda.jp; yves.lepage@waseda.jp 185 approach. The research content and statistical investigations assist patent attorneys expand the vocabulary size for the advancement of patent writing at an international level. Lin and Hsieh (2011) proposed a mixed-method approach to detecting scholarly discourse in patent technical documents. The Patent Technical Word Corpus (hereafter PTWC), containing 16 million word tokens, was compiled to elucidate the underpinning principles in identifying discourse elements, text-structure components, and the location of references. Whereas most existing IPR (intellectual property rights) databases accessible for information retrieval, the creation of PTWC, based on corpus-statistics and text-processing technology, refines more decisive characteristics of terminological knowledge as potential contribution for evaluation of technical documents. To characterize technical genre in translation studies, we use lexical richness based on technical wordlist to test distributional hypothesis. 2. Technical Terms Distribution We firstly conduct a quantitative survey based on USPTO Glossary to rank the distribution of technical terms used in United States Patent and Trademark Office (USPTO) and Japan Patent Office (JPO) within the time period from year 2010 to 2013. Table 1 below presents the statistical results. ‘Comprising’, a term of art used in claim language which means that the named elements are essential in describing the invention, ranked the first in USPTO. According to USPTO Glossary, it is a transitional phrase that is synonymous with …

Content truncated to 500 words