This study aims at assessing the quality of major Korean-English patent translations, that is, Korea Patent Abstract (KPA) translated by human translators, and machine translation (MT) by K2E-PAT, an automatic Korean-English patent translation system ...
This study aims at assessing the quality of major Korean-English patent translations, that is, Korea Patent Abstract (KPA) translated by human translators, and machine translation (MT) by K2E-PAT, an automatic Korean-English patent translation system created by the Korean Intellectual Property Office (KIPO), using criteria based on the characteristics and functions of the translation, and, moreover, the norms of KPA and MT.
As technological competition has become fierce in all aspects and global economy has become more connected, intellectual property becomes increasingly significant, boosting the demand for intellectual property related translation services. Intellectual property translation, mostly referring to patent translation, is a key path for the protection of intellectual property rights at home and abroad as well as a major means to provide up-to-date information on the technologies newly invented in domestic and foreign countries. However, despite its importance, intellectual property translation has been recognized as one of the secondary services necessary for the creation and protection of intellectual property rights in Korea. Against this backdrop, in 2012, one of Korea’s media groups raised the issue of KPA’s quality in an article, and, since then, KIPO and related agencies have exerted remarkable efforts to improve the quality of KPA. However, these efforts have been made without any in-depth consideration of the definition of quality itself, and, therefore, they have not led to the establishment of practical criteria for quality assessment of patent translations.
Intellectual property translation has another notable feature, that is, the active utilization of MT. KIPO and other major patent offices abroad as well as the World Intellectual Property Organization (WIPO) have developed MT systems specialized in patent translation. Among those systems, K2E-PAT is KIPO’s Korean-English automatic translation system run by Siriussoft, an MT company that mainly provides English MT of the patent official gazette in Korea. Though Korea is among the countries leading the trend of MT in the intellectual property industry, the quality of MT also has been barely discussed at a practical level.
With the need for research in mind, this study seeks to answer the following four research questions.
First, what are the quality assessment criteria applicable to KPA and MT, respectively, given the characteristics of technical translation in which patent translation is included as one genre, as well as the respective functions and norms of KPA and MT?
Second, from the perspective of Chesterman’s (1997) professional norm, what implications do the features of the patent translation process have for the quality of Korean-English patent translation?
Third, what do professional Korean-English patent translators think of the quality assessment criteria for evaluating Korean-English patent translation, and what do they think of the overall quality of KPA and MT?
Fourth, how are the quality levels of KPA and MT assessed based on the criteria identified and confirmed through the above three research questions?
To answer the research questions, three types of research methods, that is, literature review, in-depth interview, and text analysis, were employed. By adopting these integrated methods, this study took into consideration the criteria practitioners think is important in the patent translation process as well as those identified in the related literature, to identify practical quality assessment criteria.
The literature review presents an overview of literatures on technical translation, patent translation, and MT in the patent translation area to highlight the characteristics and functions of KPA and MT. Moreover, not only guidelines from patent offices but also quality control processes from patent translation service providers are reviewed.
The in-depth interview was conducted with ten professional Korean-English patent translators in order to gain ideas about the general patent translation process, expectancy norms and professional norms in patent translation, general quality of KPA, and general quality of MT. The purpose of the interview was to shed light on the practical patent translation process and confirm the quality assessment criteria drawn from the literature review.
For the text analysis, 447 semiconductor-related KPAs and their corresponding MTs published between August 1 and October 31, 2014 were collected from KIPO’s Korea Intellectual Property Rights Information Service (KIPRIS). In addition, from KIPRIS, 447 source texts (STs) of KPAs and MTs were gathered, and 1,918 semiconductor-related patent abstracts written for the Patent Cooperation Treaty in English, the target language, were collected as parallel texts for reference.
Based on the aforementioned methods, the key findings of the study are as follows.
As for the first research question, the quality assessment criteria for KPA and MT, respectively, were identified based on the characteristics and functions of each corpus. Specifically, for KPA, four criteria were identified: accurate understanding of the technical idea communicated through the ST; accurate translation of technical terms; correct style required for patent translation, including correct grammar and characteristics of the genre of patent translation; and strict literal translation, i.e. no addition or omission from the ST. The first three criteria are derived from the characteristics of technical translation, which includes patent translation as a genre, and the last criterion is based on the function of KPA, which is to give an accurate delivery of the patent information. For MT, the study came up with two criteria - searchability and readability - based on its key function of gist translation. In this study, searchability was assessed in terms of whether technical terms were correctly translated; while readability was assessed by whether readers can find the information they need from the sentences in the MT corpus. In addition, it was noted how the quality of the ST itself might affect the quality of KPA and MT alike for rather different reasons. This is because patent translation, particularly KPA, is a kind of documentary translation, so translators cannot modify words, phrases, or sentences in the ST. On the other hand, in MT, machines do the translation. Accordingly, whether the ST is simple and clear enough for machines to understand easily is a key for readable and understandable translation. In MT, there are established guidelines for the use of simple, clear language, called controlled language. After identifying the respective criteria for assessing the quality of KPA and MT, this study confirmed the criteria from the perspective of Chesterman’s (1997) expectancy norms and professional norms. Specifically, regarding expectancy norms in patent translation, the study reviewed quality assessment criteria set out by “norm authorities,” that is, major patent offices and WIPO. In addition, it looked into quality assessment criteria suggested by companies that offer professional patent human translation and MT services to identify professional norms in patent translation.
Regarding the second research question, a noteworthy fact is that most professional patent translators are engineers or college graduates who majored in science or engineering. They tend to translate texts in the domain of their own major. This demonstrates that domain-specific knowledge is very important in patent translation. In addition, the factors that can affect the quality of patent translation include the quality of the ST, cheap translation rates, and non-qualified translators and reviewers.
KPA has more issues that could influence the quality in comparison to general patent translation. First of all, KPA is produced by translators at the Korea Institute of Patent Information (KIPI) as well as other translation service providers (through outsourcing), making it more complex to streamline and manage the quality control process. Further, compared with general patent translation providers, KIPI has more non-majors working on translations, who are assumed to lack the specific domain knowledge required to accurately understand the ideas of the ST. Moreover, lower translation rates and non-major reviewers/editors can be factors that affect the quality of KPA.
Concerning the third research question, all of the interviewees agreed with the two expectancy norms laid out by the patent offices and WIPO: accuracy and readability. However, most of them noted that the two norms have highly different weights. That is, in the interviewees’ opinions, accuracy is much more important in the quality of patent translation, as most patent documents are written for audiences with “ordinary skill in the art.” If readers have no difficulty in reading and understanding the ideas communicated through the translated texts, the readability of the translation does not matter.
Regarding professional norms, the ten translators confirmed the criteria drawn from the related literature. They suggested accurate usage of technical terms, no addition/omission, accurate understanding of the ST, and style as the factors they take into consideration for quality control in the patent translation process to meet the expectancy norms recognized by norm authorities. These factors are the same as the criteria identified through the literature review.
As professionals, the ten interviewees agreed the overall quality of KPA has greatly improved, though it is not good enough. Some respondents mentioned that external factors such as the quality assessment process, evaluators/reviewers, and bidding process could materially affect the quality of KPA. On the other hand, as potential readers, the ten translators indicated that the overall quality of Korean-to-English MT results is so poor that most readers would not understand even the gist of the translation. Despite this evaluation, some interviewees used patent-specialized MT to search for equivalents for technical terms or to understand patent documents written in a third language, e.g. German. These usages somewhat prove the function of MT, that is, gist translation.
In regard of the fourth research question, the study considered the quality of the ST before assessing the quality of KPA and MT. For KPA, STs were evaluated based on the four criteria of spelling, grammar, consistency in terminology, and contents (i.e. whether the main idea is omitted). For MT, STs were evaluated based on the criteria for controlled language, that is, sentence length, usage of passive voice, the amount of content a sentence includes, and punctuation. The analysis showed the STs have errors and problems that may affect the quality of both KPA and MT. KPA, aimed at delivering patent information accurately, was assessed on the basis of four criteria: accurate understanding of the technical idea communicated through the ST; accurate translation of technical terms; correct style required for patent translation, including correct grammar and characteristics of the genre of patent translation; and strict literal translation, i.e. no addition or omission from the ST. The fourth criterion, strict literal translation, was replaced with adequate addition/omission/substitution in the assessment, as the guidelines set out by KIPI in charge of the KPA business recommend KPA translators reconstruct the contents of the ST in case the ST does not include all the main ideas that are in the patent abstract or the ST is too long for all its contents to be translated. It was found that the 447 KPAs show major errors in the understanding of ideas expressed in the STs and in the translations of semiconductor terms. Above all, the errors related with the understanding of STs are partly due to complex syntax and unclear wordings in the STs. On the other hand, in terms of style, the STs overall meet the criteria well, except for the criterion of avoiding pronouns in patent documents. For the criterion of addition/omission/substitution, the changes observed in KPA complement the STs that lack key contents required for patent abstracts. In general, KPA is assessed to fulfill its function well.
MT, aimed at delivering the gist of the ST, was evaluated on the basis of two criteria: searchability and readability. Specifically, searchability was assessed in terms of the semiconductor terms used, and readability was evaluated based on the syntax of MT sentences. The results showed that the MT corpus has critical errors in terminologies, though some terms related to semiconductor processes and equipment are correct. Given that such terms are usually chosen as key words for searches, the mistranslations could disrupt searchability. In addition, regarding readability, each of the MT sentences has one or more syntactic errors, severely hampering readability. As a result, most of the MT sentences are neither readable nor understandable. According to the assessment, MT does not serve its function successfully.
Based on the findings, conclusions were drawn as follows: KPA and MT have distinctive characteristics and functions, and, therefore, the two main Korean-English patent translation outcomes need to be assessed based on different criteria. In patent translation, both accuracy and readability are important norms, though accuracy has much greater weight than readability. In addition, the patent translation process, more specifically, the KPA production process, has a number of factors that could affect the quality of translation outcomes. The assessment of the quality of KPA and MT, respectively, showed KPA overall performs well its function of delivering patent information accurately, while MT does not meet the requirements for fulfilling its function.
Given these conclusions, this study has met its goal of identifying practical quality assessment criteria based on the characteristics and functions of KPA and MT. This is significant in that it lays a foundation for the study of patent translation, which has been underexplored in translation studies. However, this is just a small beginning in research on patent translation. The following are possible future research topics in patent translation.
First, although patent translation is a part of technical translation, it has distinguishing features that set it apart from other genre translations in technical translation. These features are key factors affecting the quality of patent translation. As such, there is a need to investigate them in relation with the translation quality.
Second, industry-led efforts to enhance the quality of patent translation up until now have been focused on streamlining the translation system. For example, Certification for Intellectual Property Translators is one of the important moves. However, such efforts need to be based on a practical analysis of patent translation. Practical and textual analysis could lay a solid basis for setting up effective guidelines to assess the quality of patent translation for the training as well as the certification for professional patent translators. Especially, patent documents tend to be made up of a number of collocations. Accordingly, based on the characteristics, corpus-based research could establish concrete norms. The outcomes will surely be of substantial help in improving the overall quality of patent translation.
Third, MT plays an important role in the intellectual property industry. However, despite this significance, the quality assessment suggested MT does not well perform its basic function of delivering the gist of STs. To improve the quality of MT, there is a need to expand its boundaries to embrace tasks such as pre-editing and post-editing as well as concepts including translation memory and computer-aided translation, as discussed in translation studies.
As mentioned above, despite the importance and growing demand, patent translation has been underexplored in translation studies. Given the potential and significance of patent translation, this study is expected to pave the way to expanding and enriching this research area.
|
