20201005 trans

Cao Runxin 曹润鑫

Modern Interpreting with Digital and Technical Aids

Challenges for Interpreting in the 21st Century

Martin Woesler, Hunan Normal University/China

Abstract

The world is growing in tandem with the Internet, freedom of travel and globalization. Inevitably translation and interpreting are in greater demand, especially online during the coronavirus pandemic. With the beginning of the 21st century, interpreting faces new technical and digital challenges requiring new methods of delivery. Technical developments in the course of digitization have been on the rise and are approaching real-time use simultaneous capability. The technology supporting the interpreter is becoming more and more effective, while an increasing number of systems, such as artificially intelligent programs, are competing with the human being. The interpreter must now be a technology organizer and adroitly adapt to technologically predefined interpreting situations, such as video conferences with augmented reality, tele-interpreting, etc.

Chang Huiyue 常慧月

The movements of migrants within the EU and the influx of refugees from crisis areas outside of it render community interpreting especially of rare languages in unprecedented demand, often resulting in the unsavory use of non-professional and sub-standard interpreters. Existential problems for the entire profession become apparent. Due to networking via the Internet, unqualified interpreters pour onto the market from low-wage countries: with dumping prices and low-quality services, they discredit the profession of the professional interpreter. Meanwhile remuneration practices have been declining, with payment of interpreting services often being delayed or payment defaulted. All these lead to a devaluation of the profession of the interpreter while digital technology throws into doubt the need for the role of the human as interpreter or translator.

Chen Han 陈涵

The positive side of technological advancement is that communication (including translation and interpretation) becomes digital and therefore can be enhanced with artificial intelligence. This enhancement takes interpreting and translation to a new quality level. Translation and interpretation theory needs to adapt to translation and interpretation in the age of artificial intelligence, the focus, which has moved with the functional approaches to the translator, now moves to the target text audience. The new way of interpreting is a human, but digital-technically determined hybrid form of human-machine interactive interpretation, with due respect for the human participation expressed in the form of professional remuneration.

Chen Hui 陈惠

Keywords

interpreting; artificial intelligence; simultaneous interpreting; real-time support; interpreting technology; digitalisation; interpreting software

1.Literature Review

In 2006, Honegger notes in a survey that most prospective interpreters do not use software in the interpreting booth, but work with paper glossaries which they have created with MS Word or MS Excel. Fantinuoli (2011, 50) states five years later that most interpreters still manage their terminology traditionally with MS Word or MS Excel. In 2015 20- to 30-year-old interpreting students state in a survey by Gacek that "software solutions are not sufficiently known among prospective interpreters and therefore are not used" (Gacek 2015, 82). The purpose of the present study is therefore to draw attention to the functionality of software for interpreting booths.

Chen Jiangning 陈江宁

The conference interpreter Anja Rütten (2007) presents valuable information in her information-rich overview, Information and Knowledge Management in Conference Interpreting, and in her numerous blogs in 2013 and 2014 about experiences with hardware and software in conference interpreting. In this paper, also lists of software and technology from Drechsel's interpreting practice in 2005, 2013 are used.

Standard works on the craft of interpreting (Pöchhacker 2004; Stoll 2009) also form the basis of the present study. Only their suggestions are taken up here and they ought to be reconsidered from the perspective of support by software and technology. In addition, current individual studies are evaluated, such as on distance interpreting (Kalina 2010) and explanations on the use of office programs (Fantinuoli 2011), on web corpora (Gurevych 2013), and on Qtrans (Scholz 2008; Gacek 2015). To compare it with the analogue age, the very early Braun 1999 study on video conferencing is used.

Chen Jiaxin 陈佳欣

Since software and technology are developing rapidly, a study inevitably becomes obsolete quickly. To arrive, nevertheless, at more generally valid statements, the present study focuses on developments and functionality, explaining concrete software and hardware only by way of example, with the knowledge that products are often replaced by others and that only a few can last for a long period of time. In order to get an overview of software and hardware, the current master's thesis by Gacek (2015) is used. However, it only refers to a small selection of products and does not come to generally valid conclusions. This study also addresses the polarizing discussion on how much technology and software the interpreter needs. While Spitzer provocatively speaks of the "digital dementia" of interpreters in 2012, Conway (2014) has taken a more balanced approach to the question of the cost-benefit ratio of the computer in the interpreting booth.

Chen Jingjing 陈静静

This paper will also introduce special aspects, like problems posed by trendy community interpreting (Andres 2009) and cheap competition for interpreters, as well as placing networked communication and work on texts in the context of the development of swarm intelligence and a collective consciousness. Finally, a new type of interpreting is called for, which expects at least technical competence from the interpreter, and at most a hybrid human-machine working method. In this paper, the thesis, already passionately advocated historically, that the computer or artificial intelligence could never replace the human being in certain functions such as language (Austermühl 2004), is no longer categorically excluded.

Chen Sha 陈莎

2.Simultaneous Technology/Technology Working in Real Time

While in the past technology did not allow simultaneous work support, i.e. it was not possible for translators to work with simultaneous or consecutive interpreting, this disadvantage seems to have been largely overcome technically at the beginning of the 21st century, although there are still too few apps that make use of these new possibilities.

Arguments against the positive effects of advanced technology include references to over-coding, the abundance of information, etc. These phenomena appear to be a hindrance, especially in connection with the interpreting profession, which requires the highest level of concentration.

Chen Sunfu 谌孙福

Spitzer (2012) for example, poses the question of whether computer work causes a mental deterioration of society. Rütten (2014a), on the other hand, refers to Spitzer, pointing out the advantages of the computer, arguing that it has voice output, retrievability through full-text search, spell checking, sorting, categorization and transmission functions. She suggests as a compromise: "Talking about research, challenging one's own memory and always asking about what is meant and the context - if we take this to heart, we have a good chance of making the computer a valuable training tool and an excellent assistant.”

Chen Yongxiang 陈永相

2.1 Technology to Support the Interpreter

2.1.1 Preparation

Even if a consecutive interpreter is standing next to the speaker, armed with a stenographer's notepad and pencil, and interprets him or her in an apparently analogous manner, even he or she cannot do without technology these days.

Even during the preparation stage, his duty of care requires him to take into account the accessible sources, some of which are electronic.

Even more technical possibilities are offered if the interpreter asked to use his own initiative receives speech manuscripts beforehand. These can be scanned, transcribed and prepared, for example by automatic color coding of verbs and realities, which Stoll (2009, 84f.) also recommends for analogue preparation, as it improves anticipation and simplifies syntactic planning.

Cheng Yusi 成于思

In addition to the concrete preparation of the text to be interpreted, more general preparation can also be supported technically.

Drechsel (2005, 16f.) lists, for "preparation for the conference topic and the creation of glossaries, electronic tools such as search engines, web catalogues, topic portals, scientific websites, company or customer websites, library catalogues, online libraries, specialised services for e-publications, online magazines, newsgroups and others are used.”

Online comparison with existing acronyms/abbreviations (Stoll 2009, 85) is also helpful.

Deng Jinxia 邓锦霞

Of course, ad hoc knowledge acquisition (Gile 2009) prior to the conference also counts as preparation, such as requesting speech manuscripts and presentations, often in file form. Preparation does not only take place in the weeks before the assignment ("advance preparation", Gile 2009), but also on site a few minutes before the assignments ("last minute preparation", Gile 2009) and during the interpreting breaks ("in-conference preparation", Gile 2009). Stoll categorizes the areas of interpreting preparation as "general technical", "terminological" and "interpreting strategy" (Stoll 2009, 86).

Ding Daifeng 丁代凤

Here, search engines or electronic dictionaries are often faster than paper dictionaries. Moreover, if, for example, the conference program/list of speakers is updated on a website, the interpreter can adjust his or her planning by accessing the website without further consultation with the client.

In principle, all technical aids should meet the requirements of being user-friendly, having a fast or real-time response time and being manageable.

Fang Jieling 方洁玲

2.1.2 Speech to Text in Real-time with Low Error Rate

Every smartphone today has a mode or apps with which it records speech and converts it into text in real time with a now justifiably low number of errors. This simple function alone is valuable for the apparently analogue consecutive interpreter: He can put his shorthand pad in a folder in which, for example, the smartphone is inserted on the left and has written down the spoken text as text.

Gan Fengyu 甘奉玉

While he works with his notes in interpreting notation on the right, as he used to do on the right, and can insert the pen into a holder on the folder if necessary, the transcription on the left enables him to find his way back into the context at a glance or, at most, with a wiping movement in the event of a pause in his hesitation (colloquially known as a 'hang up') or even a 'blackout'.

Drechsel (2013) demonstrates the use of the following programs/websites on the Ipad (which helps him to concentrate on individual processes): Documents (manages documents and allows editing), Interplex, LookUp (terminology databases), Wikipedia, Google.

Gao Mingzhu 高明珠

Tablets like the Ipad have the following advantages over the conventional laptop, notebook or netbook in the cubicle: They are lighter, smaller, handier, the battery lasts longer, typing is silent, you can use apps, your work is not interrupted by updates or pop-ups, you can take notes by hand and record things in the background for archiving or follow-up. Of course, due to the data protection regulation, the consent of the client must be obtained before recording.

Gong Yumian 龚钰冕

The disadvantage of not being able to see several windows at the same time appears to be an advantage to Drechsel (2013) from his personal experience, because it allows him to concentrate better on one thing. However, he neither has all the Office applications with him nor his domestic terminology databases. A possible compromise could be a laptop with a touchscreen.

Gu Dongfang 顾东方

2.1.3 From Text in Source Language to Text in Target Language

Not yet very professional but available as prototypes are transcription systems that offer a translation in a second screen (e.g. translator.google.com, fanyi.baidu.com, for European languages DeepL.com) in addition to the real-time transcription of the source language presented in 1.1.1. On the smart device, for example, an upright screen can be divided into an upper transcription area and a lower translation area.

Guan Qinqing 管钦清

Here, too, constant trial and error is required to improve the quality of the interpretation. It is also individually different how much different information the individual interpreter is able or used to process. If the flow of data is too large, there is a risk that concentration or the flow of production will be impaired.

2.1.4 Keyword Cloud

Helpful for the interpreter's work is the insertion of key terms and their translation into the target language in real time in the form of dynamic clouds, i.e. in real time and the longer a pause before the interpretation of the relevant term lasts.

Gui Yizhi 桂一枝

In my opinion, the development of a corresponding app would be a desideratum. It would not be used intensively, but would be conceived as an additional screen for the corner of the eye, which is preferably perceived subconsciously and which one can turn to when one has a hesitation pause and is looking for stimulation. Here, terms that have been interpreted according to the proposal should turn green and others grey. It should be possible to turn back the displayed cloud by fractions of a second in the timeline by pointing and wiping. Of course, it should be capable of learning, i.e. it should be able to memorize frequent non-standard translations and offer them itself.

In the post-processing phase, the program should offer a list of terms to be practised and has the non-standard translations approved manually by the interpreter.

Guo Lu 郭露

2.1.5 Effects of Technology

The interpreting profession has changed due to hardware and software development, especially since the 1990s, which has essentially made interpreting easier and better. Whether a technology or an app makes it into the consecutive or simultaneous interpreting situation and even into the interpreting booth is subject to the highly subjective decision-making power of the individual interpreter, just like the interpreting notation, which varies from person to person. New techniques should only be tried out if they bring about a real improvement in use.

Han Haiyang 韩海洋

It is not in anyone's interest to use the latest technology for its own sake if this means a loss of quality in the interpretation. Conversely, an interpreter should be open to accept new techniques and software if they can improve his work.

2.1.5.1 The Presence of the Interpreter at the Place of Assignment

The interpreter can influence physical and spatial matters by being present at the place of assignment:

a) He/she can acquire vocabulary in advance in exchange with the client, compile vocabulary lists, practise them and also physically take them into the booth as paper printouts or on the screen.

Han Wanzhen 韩宛真

b) He can request speech manuscripts in advance or speak directly to speakers on site and ask for a copy of their speech manuscripts.

c) In case of early contact with the Interpreter Equipment Company, he may be able to influence positioning the interpreting booth so that he can see the speaker and the PowerPoint presentation. In the case of video conferences (e.g. by zoom, voom/Tencent Meeting, BigBlueButton, Skype, Tencent Classroom, MS Teams etc.), he can make sure that he can see all the speakers, including those who appear on screens, so that he can also recognize non-verbal signals and incorporate them into the interpretation. In principle, it is recommended that the interpreter is also shown on screens or in video conferences on a small side window in order to better understand the interpretation.

He Changqi 何长琦

d) Before the event, the interpreter can be introduced to the interpreting equipment by the technician and test the audio system, determine and announce the channel-language assignment (experienced clients post these and instruct the staff who issues the receivers to inform the participants of the assignment), adjust the volume of the headphones and arrange the aids on the work table.

Hu Baihui 胡百辉

e) He/she can coordinate with his/her interpreting colleague regarding, e.g. glossaries, agenda, and exchange documents, then both can determine the approximate intervals between changes or simply notice during the assignment if the colleague is at a loss and give a helping hand with writing a missing term on a piece of paper. He/she can also nudge him/her to draw his/her attention to something. He/she can share glossaries and documents electronically, e.g. with the Interplex program simply with a swipe to the left. He/she can also take over the interpretation prematurely in case of exhaustion of his/her colleague.

Hu Huifang 胡慧芳

The number of distance interpreting assignments has already been increasing before the coronavirus pandemic, but with Corona, it exploded. Before Corona, a combination of telephone and web communication was used. With the coronavirus pandemic, Zoom was used more frequently and the Business version comes with a conference interpreter menu.

AIIC, BDÜ, European institutions and other international organizations agreed (Kalina 2010) on the following conditions as minimal professional standards for distance interpreting:

Hu Jin 胡瑾

- As a rule, a direct view of the room should be possible.

- To avoid multilingualism, distance interpreting should only be used in exceptional cases when more than six active languages are spoken. But even then, as many booths as possible should be available in the room and the booths of the distance interpreters should not be too far away.

- High-definition monitors should be installed in front of and not in the booths.

- The remote interpreter should be able to communicate directly with the client and the cameramen.

- Teams should regularly consist of a minimum of three remote interpreters".

Ji Tiantian 纪甜甜

2.1.5.2 Higher Validity through Explanation and Correction Functions

In the university lecture hall, students critically accompany, verify and check for errors in the statements of the lecturer by simultaneously reading in Wikipedia, Baike Baidu, etc., and in search engines such as Google, and Baidu and, if necessary, manually validate or question the statements in accompanying chats (e.g. in the WhatsApp or WeChat group of the course). In a technically mediated interpreting situation, it is now possible to show alternative translation suggestions to participants. This is useful in reducing misunderstanding when a) politically/religiously/culturally sensitive terms appear, b) different groups of recipients may have different understanding of the same term, for instance, a socialist-authoritarian country and a liberal-democracy] state perceive political ideologies, systems and concepts through a spectrum of shades and hues. Recognition of the various nuances, the multifaceted implications and the degree of value-laden presumptions for each individual recipient serve to ease tension and foster communication.

Jiang Fengyi 蒋凤仪

2.1.6 Special Features in the Booth

Computers or smart devices allow you to view presentations, documents and notes. However, reference by typing during the interpreting process is hardly possible due to time constraint. Therefore, text to speech and speech to text modules become more important.

In exceptional situations, such as a 'blackout' or a hesitation break, the technique of stalling, i.e. inserting neutral expressions without new information together with deceleration of the target text production, could be useful strategies (Pöchhacker 2004, 132ff).

Jiang Hao 姜好

2.1.7 Software Examples

When naming selected software, I follow the list and evaluation of Janovska (2011).

To this day, interpreters continue to manage their terminology with Office programs such as Word or Excel (Fantinuoli 2011, 50). However, searching is cumbersome with Ctrl+F, and only the first entry found is displayed. Better suited are Interplex simple, LookUp (search without Enter key, which also has a separate module for Word, Janovska (2011, 80); it also has input fields such as customer, topic and project for additional information, terminology extraction, vocabulary trainer, semantics filter and sorting functions, Honegger (2006, 2) and Rütten (2014f), Terminus (a) good multilingual terminology management system and TermDB developed by conference interpreter Christian Vogeler but discontinued in 2012. But here, too, Fantinuoli (2011, 51) complains that the hit rate cannot be reduced by stop words and that the errors in the glossary cannot be corrected.

Jiang Qiwei 蒋淇玮

Gacek (2015) presents numerous programs: Interpreters' Help (Gacek 2015, 55), a new browser-based web application from 2014 for all browser-based operating systems and for Boothmate for Mac OS X. It offers the possibility of synchronization offers with fast search function. The currently available beta version and the license for students are free of charge.

Gacek (2015, 56) also introduces the Glossary Assistant program, which is still in development. It presents glossaries clearly on tablets, - the focus is not on mobile phones, - and makes them editable, see also Rütten (2014g) and Martin (2014). The program is currently only available for Android tablets and is free of charge.

Kang Haoyu 康浩宇

The Intragloss program, which Gacek (2015, 57) presents, was "developed by AIIC conference interpreter Dan Kenig and software developer Daniel Pohoryles". "The aim of the program is to facilitate the creation and use of even extensive multilingual glossaries". "Glossaries can be created directly within a preparation document or within a website.” Gacek values this as “very useful” and gives the example “making a list of problematic terms after receiving a last-minute document”. He also points to the fact that “Appropriate filter and sorting functions alphabetical, by website, by chronological order are also provided. A glossary created from a web page is automatically updated, which is an advantage for future. Like InterpretBank software, Intragloss also offers the possibility of automatic translation of terms, with the search for translation suggestions in large translation portals (IATE, GDT, Linguee, Termium, WordReference, etc.). Other useful functions include the clear comparison of different language versions of a document". A free trial version is available for Macintosh devices, see also Kenig (2014).

Kang Lingfeng 康灵凤

Gacek (2015, 61-63) tests and describes in detail the InterpretBank program in version 3 2014, which aims to "provide interpreters with a sophisticated corpus-linguistic tool". "InterpretBank was developed by Claudio Fantinuoli, a graduate interpreter working in the Department of Translation, Linguistics and Cultural Studies at the Johannes Gutenberg University of Mainz" (Gacek 2015, 62). "InterpretBank is a modular tool that provides interpreters with computer support in the field of knowledge and terminology management before, during and after a simultaneous interpreting assignment" (Gacek 2015, 63). Gacek lists the three modes offered by the program: "TermMode: Module for creating and managing glossaries; in addition, various functions can be used, such as automatic translation of terms and searching for definitions from the Internet. MemoryMode: Module for visual memorization of bilingual glossaries. ConferenceMode: Module for cabin-friendly reference during the interpretation." (Gacek 2015, 63).

Kong Xianghui 孔祥慧

Rütten (2014) judges: "Very user-friendly, many nice functions; organised by glossaries (which, technically speaking, are subject areas tagged to each entry), has all the essential data categories (customer, project etc.) and a very nice flashcard-like memorising function. The quick-search function ignores accents. It is limited to five languages and you cannot add endless numbers of individual data fields. Intrepretbank is platform independent, works on Windows, Mac and Android (Gacek 2015, 63), costs 69 Euro, for students 39 Euro, university teachers (and their students) get a free demo license.

The interpreters also use very simple websites and programs such as the search engine Google, the Internet encyclopaedia Wikipedia, dictionary programs such as Langenscheidt and systems for teamwork such as Google Docs. Typical concordance programs are WebSleuth, WebResearch, Black Widow, Site Ripper.

Kong Yanan 孔亚楠

Terminology extraction enables Easy Concordance, and Concordance. Text to speech can handle most operating systems (Microsoft Windows, Macintosh Operating System), there are also special programs for it like Ivona Reader (from 49 €) with the free MiniReader version, TextAloud (19.99 €). There are programs for fast reading (so-called "Improved Reading") like "A Faster Reader" (for Android smart devices) and programs for managing documents like the app "Documents" (for Apple iOS), in which among other things search functions can be used and a web browser can also be opened, or OneNote. The website www.wortwarte.de is supposed to present neologisms, but it is not up to date (visited September 11, 2020 by M.W.).

Lei Fangyuan 雷方圆

The program BootCaT creates topic-specific web corpora: "The users first define the search terms. Then, web pages are collected, which contain the combination of these search terms", see Gurevych (2013, 546). At the end a text corpus is created, “with which one can quickly get an overview of the content and terminology, for example, of the keywords ‘energía solar’ in Spanish or ‘solar energy’ or ‘solar power’ in German”, see Rütten (2008).

Lei Kuangxi 雷旷溪

In addition, there are Termprofile, Endnotes, Qtrans-Search Bar. The latter is rated by Gacek (2015), 50 as "no pop-up windows, faulty queries or other inconveniences". Also Scholz (2008) presents Qtrans: "One of the great advantages of the tool, however, is its low threshold: the software can be used without training, is system-independent, requires no installation and is immediately ready for use.” Scholz also explains, that it is not based on its own search technology, but passes parameters to other services; “therefore it is easily adaptable and can integrate any internal and external data sources via HTTP” (Scholz 2008).

Li Haiquan 李海泉

For preparation and simultaneous use, there is the glossary software Interplex, which is capable of handling multitasking. It can import glossaries from Word or Excel and is available for Windows, iPhone and iPad. In 2020, it costs US$ 75 and there is a free demo version available. More features for conference interpreters are offered by Term LookUp, which cost US$99 in 2019. In 2020 the author of this paper only found the free version and IntelliWebSearch (free).

Li Lili 李丽丽

2.2 Simultaneous Interpretation with Video Conference Systems

The current video conferencing systems all allow simultaneous interpretation. For this purpose, every participant simply needs two devices and joins two meetings at the same time. For more languages, more meetings are offered. Common solutions are BigBlueButton (open source, free), Zoom, Skype for Business, Microsoft Teams, WebEx, Zoom, GoToMeeting and in China voom (alias WeMeet, Tencent Meeting), TencentClassroom, WeChat.

Li Lingyue 李凌月

Every participant should mute his/her microphone (except when speaking) and can listen to the original language on the first device and use one ear plug from the 2nd device to listen to the target language. If the original language does not need to be heard, the participant can also turn off the sound of the 1st device and plug in both ear plugs from the 2nd device or listen to the sound from the loudspeakers of the 2nd device. It is not recommended to have both devices on loudspeaker at the same time to avoid acoustic feedback. Sometimes, if the web speed is too slow, turning off the camera (not the loudspeaker) may help.

Li Liqin 李丽琴

Every participant should mute his/her microphone (except when speaking) and can listen to the original language on the first device and use one ear plug from the 2nd device to listen to the target language. If the original language does not need to be heard, the participant can also turn off the sound of the 1st device and plug in both ear plugs from the 2nd device or listen to the sound from the loudspeakers of the 2nd device. It is not recommended to have both devices on loudspeaker at the same time to avoid acoustic feedback. Sometimes, if the web speed is too slow, turning off the camera (not the loudspeaker) may help.

Li Luyi 李璐伊

With the business version of Zoom, you can simply choose the language channel by clicking on a flag. As the host of the meeting, one must activate the interpretation function and invite the respective interpreters.

Li Meng 李梦

So far, conference interpreters often sat in the back of the conference in their boxes and where not visible to the audience. With video conferencing systems like Zoom it is now technically possible to blend in a video of the interpreter, which enhances understanding. However, so far the video conference communication situation still feels artificial, much different from the analogue situation, with only faces, voices and shared screens. The next generation of digital communication and interpretation is the Virtual Reality Room (e.g. using the hardware Oculus Quest VR glass and the software Spatial), in which participants can upload a photo to create realistic avatars and then ‘look around’ to see the speaker and the interpreter. The participants can even ‘walk up to’ the speaker to sit in the first row or they can ‘place’ speaker and interpreter next to each other.

Li Yongshan 李泳珊

2.3 Artificially Intelligent Programs as Competition for Interpreters

The debate about the extent to which technology can replace humans is as old as the first fantasies of artificial humans. It is still a polarizing discussion today and is conducted in a highly emotional manner.

Austermühl (2004) explains: "We believe that even the latest technology and up-to-date machines cannot replace the human brain when it comes to language transfer. Many times, the concepts are too complex to reduce them to the level of machine-readable language."

Li Yu 李玉

2.3.1 Machine Translation vs. human translation

For this section 2.3.1 I am indebted to the final exam paper of my student Jia Liwen 贺丽文 from my 2019/2020 Master course on Translation Studies at Hunan Normal University, Foreign Studies College. Although I come to slightly different conclusions, the paper sums up some discussions I had with the students in class and provides literature review and a field study worth quoting extensively here.

Lin Min 林敏

Machine translation, commonly known as MT, can be defined as “the application of computers to the translation of texts from one natural language into another” (Huchins 1986, 1). The term “machine” is outdated, since we refer to computers today or to digital/eletronic instead of analogue translation, the term “machine translation” is mainly understood in contrast to human translation and therefore has potential to sustain.

Lin Xin 林鑫

Machine translation pioneers were the United States, the Soviet Union and European countries. The initial stage can be dated 1947-1954: In 1946, the world’s first electronic computer was born. Soon, in the second year, 1947, an American scientist Warren Weaver and a British engineer A. D. Booth firstly proposed to translate languages by modern electronic computers. It was in 1954 that Georgetown University cooperated with International Machine Cooperation (IBM) on a project, which created the world’s first machine translation system breaking the restriction of word-to-word translation. It was recognized as a breakthrough of machine translation and demonstrated to the public and the scientific community the feasibility of machine translation for the first time.

Ling Zijin 凌子瑾

From 1954 to 1975, we may define a second stage, with climax and stagnation. Due to the success of Georgetown-IBM system and the potential social, economic and intelligence benefits, quite a few countries including the United State, Soviet Union, and Japan invested heavily in the research and development of machine translation. Then, there was an upsurge of machine translation research all over the world.

Liu Bo 刘博

In 1956, the Chinese government released its “1956-1967 Prospective Plan of Science and Technology Development”, in which “automatic translation” was listed as an important task in item 41.

The poor translation caused by the rough design of the first generation machine translation system and the exaggeration of the computer capability eventually led for some people to lose confidence in machine translation. In 1964, the US Automatic Language Processing Advisory Committee (ALPAC) established by the National Academy of Sciences (NAS) carried out an investigation on machine translation including its speed, quality, costs and the demand of it and had a sound and comprehensive test, count and analysis.

Liu Jinxingqi 刘金惺琦

Later, in 1966, ALPAC published the results of the survey and its main conclusion was that no further research should be undertaken on machine translation considering its low speed, low accuracy, higher costs than human translation and inability to overcome semantic barriers. Affected by this report, researches on machine translation declined sharply and even led to a 10-year slump worldwide.

Liu Liu 刘柳

After a quiet decade of 1970s, thanks to the development of computer technology, linguistic theory and artificial intelligence research, the increasing demand for translation, and the unremitting efforts of some machine translation researchers, research on machine translation revived in the early 1980s, therefore we can define the recovery stage from 1975 to 1987. During this period, machine translation researchers no longer were blindly optimistic, instead, they paid more attention to the basic aspects of machine translation research, thus studies on machine translation systems and their development were more practical and rational, and eventually the second generation of machine translation systems emerged.

Liu Ou 刘欧

In 1975, the European Atomic Energy Agency (EURATOM) began to install SYSTRAN. In 1976, the University of Montreal, Canada, and the Translation Bureau of the Federal Government of Canada jointly developed a practical machine translation system (TAUM-METEO), officially providing weather forecasts in May 1977. This was the only machine translation system that directly published translations without post-translation editing. It marked the practical application of machine translation in technical languages and the maturity of machine translation technology in technical languages. Many of the methods and technologies used by the second generation machine translation were relatively mature and some of them are still used these days.

Liu Yangnuo 刘洋诺

Since 1987 until now, we are in the prosperity stage. Many institutes and universities began their researchers in machine translation. Since 1989, the appearance of the third generation machine translation method based on corpus has changed the vision of the whole machine translation research, marking a new period for machine translation. Many famous machine translation systems were released at this period, such as the KY-1, being developed by the Chinese Academy of Military Sciences, MT-IR-EC, an English-Chinese title and catalog translation system, developed by the Research Institute of Post and Telecommunication Science, and Huajian Chinese-English machine translation system, developed by Huajian Co. Ltd.

Liu Yi 刘艺

Due to the rapid development of computer software and hardware technology, large-scale corpora which can be read by computer can be widely used in machine translation. The larger the corpus, the richer knowledge of human linguistics it contains, and the better the quality of machine translation will be. As long as the corpus was large enough, it was expected to cover all language phenomena. In this way, the key question here is, how to automatically or semi-automatically mine relevant translation knowledge from the corpus and effectively organize the knowledge base.

Liu Yiyu 刘怡瑜

The Translation Memory (TM) technology in Trados’ Translation Workbench and the latest near-human translation machine translation systems introduced by Google, Language Weaver, Meaningful Machines and other companies since the new century and the development of artifical intelligence are all the results of the successful application of corpus technology in machine translation research.

Liu Zhiwei 刘智伟

There are two recent Chinese books on MT, by Liu Miqing and Xu Jin. Both Liu Miqing and Xujun uphold the dogma that human translation cannot be replaced by machine translation, although they don’t reason their opinions. Liu Miqing comments on the so-called misunderstanding “If machine translation succeed, then translators will lose their jobs”: “This kind of worry is unnecessary, and human translation will be in demand at any time” (Liu 2010, 402). Xu Jun states: “On the basis of the existing linguistic level and the research level of computer artificial intelligence, it is impossible to develop a machine translation system that completely replaces human translation” (Xu et al. 2009, 339-340).

Lou Cancan 娄灿灿

Both of them use the same categorization of machine translation development stages as mentioned above. Both divide traditional machine translation methods into two categories: Rule-Based and Corpus-Based. The former builds the translation knowledge base on dictionaries and grammar rule bases, while the latter builds it by making full use of the corpus. Corpus-based methods can be subdivided into statistics-based machine translation and example-based machine translation.

Luo Weijia 罗维嘉

Since language has the characteristics of flexibility and openness, the development of a machine translation system based on grammar rules is greatly limited due to the lack of human thinking and the lack of the ability to identify the text or become aware of it. On the contrary, example-based machine translation has enjoyed unprecedented development since the 1980s. With the revival of statistical methods and corpus methods, corpus-based machine translation systems based on statistics and examples are beginning to be used for large-scale processing of language materials and real texts.

Luo Yuqing 罗雨晴

Both authors think that translation memories are very useful. Because unlike machine translation, the translation memory technology cannot translate the entire text, but can offer previously translated phrases from documents stored in its database. Through comparing, retrieving and reusing previously translated texts, translators can promptly determine how to accomplish the translation in a more efficient way. That is to say, the help of a translation memory makes translating more convenient and it makes it more effective for translators to deal with some complicated source texts through translation memory administration.

Ma Juan 马娟

2.3.1.1 Advantages of Machine Translation

1. Among the advantages of MT are its speed and availability. If one visits a foreign country, a smart phone may make up for lack of language skills. Several free applications will help them translate texts, images, voices almost immediately, anytime and anywhere. One can take a picture of a sign and read it in one’s own language. Special apps help during conversations with foreigners. The accuracy is not high, but is also not needed in these basic conversations, which are low-end texts.

Ma Shuya 马淑雅

2. Another major advantage of machine translation are the low costs, some services even come for free. For some large enterprises and professional translators, there is an initial financial investment to buy translation software like Trados, which pays off in the mid-run.

3. Another merit of MT is the sheer volume of translation. A small application can conduct large-scale translation work in a short amount of time, impossible for humans.

4. Also for human translators, MT is a helpful tool, professional translators and interpreters can save energy and improve their efficiency. In many disciplines, there is a huge amount of specialized vocabulary, which also changes fast. MT can make sure that all special terms are translated consistently in the same way. Most human translators prefer electronic dictionaries over paper dictionaries, e.g. because they have search functions and are updated online.

Ma Zhixing 马智星

5. MT tools also enable teams of human translators to increase their efficiency of cooperation, including joint databases and standards of translation. The trend towards MT is irreversible.

2.3.1.2 Disadvantages of Machine Translation

1. A MT can only be as good as its input, processes and self-learning algorithms. In the field of speech translation, MT depends heavily on speech recognition technology. In 2020, most speech recognition systems require speakers of the standard language (e.g. Mandarin instead of Cantonese). People’s accents, dialects and other influences like a noisy background affect the accuracy of speech recognition.

2. MT depends on connectivity and electricity. Many users become aware of this dependency when they use a translation application on their smart phone and suddenly the network connection is broken or the battery is empty.

Meng Ying 孟莹

3. The establishment of the corpus is the foundation of modern MT, but the corpus itself has its limitations. Modern MT, especially in many small languages, has a small corpus. For example, under the same circumstances, sometimes the accuracy of Lao translation is not as good as that of English. Also commercial reasons may play into this.

4. Almost all MT requires human post-editing. With the progress and development of science and technology, the accuracy of MT software is getting better but so far does not match human translation quality. The text after machine translation requires proofreaders to make final modifications to the translation to ensure that it is correct and has an appropriate style.

Mo Ling 莫玲

5. MT still has deficits to understand in which context a word is used. In many languages, the same word may have several completely unrelated meanings, such as the word “spring” in English, which most commonly means “春天” in Chinese, but also means “弹簧” , “泉水” and “活力”. Another example is the word “门槛” in Chinese, which can refer to the “threshold on the door”, but the most common meaning is “the difficulty of a thing” or “the conditions for doing something”. In these cases, the context has a great influence on the meaning of words, and the understanding of the meaning depends largely on the clues one can get from the context. So far, the main advantage of human translators is the true understanding of a sentence, which is connected to relating words with their context. Another advantage is that a human translator can creatively polish the language to obtain not just a complete and accurate translation, but an appropriate one. This is undoubtedly still a big challenge for MT.

Mo Nan 莫南

6. MT so far does not possess cultural sensitivity. Human translators constantly study the relevant cultures, expand their knowledge and are able to understand specific situations. Human interactions and emotions are complex and machines lack initiative and the ability to understand or recognize slang, jargon, puns and idioms, so that the resulting MT may not conform to the values and norms of the culture of the source language and/or of the target language.

2.3.1.3 Comparison of Machine Translation with Human Translation

Machine Translation can be compared with human translation in different areas and under different aspects.

Nie Xiaolou 聂晓楼

1. In the film and television industry, there is a large demand for translation of quotes (subtitling), and the difficulty lies in the individuality of each speaker which should correspond to a characteristic style depending on the speaker. The screen format requires short and pithy translations. What’s more, in the current movies and TV plays, there are a large number of terms fashionable in the internet. The film title itself has to take all possible connotations as well as marketing aspects into account, so a human translator will think it over and over again.

Ou Rong 欧蓉

2. In the political and the diplomatic field as well as in international negotiations between countries or institutions: In these fields, human interpretation and translation is still widely used, since translation mistakes may have severe consequences for the relation of countries. When country leaders meet, the cultural accumulation of the translators can enable them to identify which content may not offend both sides and then pick the best translation.

Ouyang Jinglan 欧阳静兰

When a country leader tells a joke which is not funny in the target culture, the human translator may improvise with telling a different joke, ensuring that the visitor will laught too and the whole atmosphere stays relaxed as intended. With MT, diplomatic accidents or cultural conflicts might happen. In the first half of 2018, AI simultaneous translation was applied for the first time at the Boao Asia Forum. However, the system broke down, resulting in low-level mistakes such as inaccurate vocabulary translation and repetition. Mistakes like these are avoided by professional (human) interpreters.

Ouyang Ling 欧阳玲

3. Legal and technical communication. Legal translation, as well as medical, pharmaceutical, chemical etc. must be accurate, because a translation mistake may have severe consequences. The human translator spends additional time to make repeated efforts to avoid ambiguity and to improve the accuracy of the words used. Many specific terms in professional sectors have a broader or a different meaning in the standard language. For example, prejudice refers to damage, counterpart refers to a copy with the same effect, more complicated, for example, dominion refers to full ownership in civil law and sovereignty in public international law; Estoppels means that one cannot go back on the word in contract law, while in criminal procedure law, it means “forbidden to reverse confession”. Secondly, a large number of legal terms, such as “defendant”, “cause of action” and “damages”, usually do not appear in the common language. These characteristics of legal terms require people to carefully weigh and compare when translating, and give appropriate translations in specific situations.

Peng Dan 彭丹

4. Literary translation. MT in general cannot compete with human translation in the field of literature, since these kinds of translations are more complex. A typical characteristic of Western literature is to avoid repetitions. If, for example, the source is a Chinese work of literature, repetitions are more common. MT would translate these repetitions repetitively, while humans would be creative to find synonyms and variations. A good translation of literature should enable target readers to understand the world created by the source culture author and properly realize his beliefs, ideas or other things the author wants to convey through his work. Also, subtle references to other works of literature are harder to grasp by MT than by a human translator. There are often a lot of images (comparisons, illustrative expressions, motifs, metaphors, allegories) in literary works, and images have vague characteristics. When translating and dealing with these images, even experienced translators carefully consider and repeatedly weigh them. Literary works express the rich emotions of humans, it may be happy or sad, and half sad and half happy. In order to understand the subtlety of this, the translator needs to read the text carefully and weighs it over and over again. Only after careful reading and repeated deliberation the translator can really understand them and thus produce a good translation. Literary works are often historic.

Peng Juan 彭娟

In different periods, literary works created by different writers have their own imprint of their times. When people look at past literature, they cannot only translate it from the contemporary viewpoint. Therefore, when reading the original text, the translator should figure out the author’s writing intention and the emotion to be conveyed according to the background of the times, the writer’s experience, the writer’s style, etc. in order to better understand the original text and in order to better carry out the translation. Obviously, MT systems are not yet able to deal with these complicated processes. Last but not least, literary works are often fictional, and the fictional world is often beyond the imagination of the real world. Even if the machine can input all the literary works and their corresponding translations in different languages into it to build a huge corpus, literary works stay always full of infinite creativity and imagination. The current MT systems may be able to give a proper translation of some sentences of literary works, but from the perspective of development, the premise of machine translation is to establish a corpus first, thus it is always lagging behind and can never keep up with the pace of literary creation and innovation.

Peng Ruihong 彭锐宏

2.3.1.3 Field Study

In November 2019, we conducted a simple field study. We selected an original text (https://b23.tv/av9604542) among the quotes of the American movie The Pursuit of Happiness 《当幸福来敲门》: “People can’t do something themselves, they wanna tell you you can’t do it. If you want something, go get it. Period.”

Here are five MT versions from Sogou, Baidu, Netease Youdao, DeepL and Google respectively:

Sogou translation (http://bit.ly/trans_ex_1): 人们自己做不到，他们想告诉你你做不到。如果你想要什么，去拿吧。句号。

Baidu translation (http://bit.ly/trans_ex_2): 人们自己做不到，他们想告诉你你做不到。如果你想要什么，就去拿。周期。

Youdao translation (http://bit.ly/trans_ex_3): 当人们做不到一些事情的时候，他们就会对你说你也同样不能。如果你想要什么，就去争取。时期。

Google translation (http://bit.ly/trans_ex_4): 人们自己无法做某事，他们想告诉您您做不到。 如果您想要一些东西，那就去买。 期。

DeepL translation (http://bit.ly/trans_ex_5): 人们自己做不到的事情，他们就会告诉你，你做不到。如果你想要的东西，去得到它。句号。

The human translation (https://b23.tv/av9604542) is:

有些事人们自己办不到，他们就刚跟你说你也办不到。如果你想获得什么，就去争取。就这么简单。

Peng Xiaoling 彭小玲

The original text is relatively colloquial, so the overall difficulty for translation is not so high, but still the five versions of machine translation are not ideal, only the versions translated by Netease Youdao and DeepL are acceptable, but also unsatisfactory in comparison to the human translation.

Peng Yongliang 彭永亮

In the first sentence of Sogou and Baidu translation, the word “something” is ignored and the overall coherence of the translation is not high. It is also not consistent with Chinese habits. In the first sentence of the Netease Youdao translation, “当……的时候” is added, which is feasible, but compared with human translation, it is not concise enough. The DeepL translation starts strong, but does not persuade with the arbitrary addition of “的”, which destroys the grammar. The Google translation “无法做某事” reads awkward in Chinese, the reader rather would expect something like “办不到的事情”, also the auxiliary verb “想” is not appropriate.