CC BY
76 Арабистика Евразии, № 14, Июль 2021 16
АРАБИСТИКА ЕВРАЗИИ Y • \ 1 JJIJJ И i tA^ljjVl ¡jjluiljjll
EURASIAN ARABIC STUDIES LBРАЗИЯ АРАБИСТИНАСЫ Eurasian Arabic Studies, № 14. July 2021
ФИЛОЛОГИЧЕСКИЕ НАУКИ
jJL-LJ I p j_Ls_J I
PHILOLOGY
УДК 8Г322.4
Original Paper Оригинальная статья
AN INVESTIGATION OF GOOGLE'S ENGLISH-ARABIC TRANSLATION
OF TECHNICAL TERMS
Reima S. Al-Jarf
King Sand University
reima. al.jarf@gmail. com
Received: April 5, 2021 Поступила в редакцию: 5 апреля 2021 г.
Reviewed: April 25, 2021 Одобрена рецензентами: 25 апреля 2021 г. Accepted: April 30, 2021 Принята к публикации: 30 апреля 2021 г.
Abstract
Many EFL learners resort to Google Translate (GT) to check meanings of difficult words. The aim of the present study is to check the breadth of coverage and accuracy of the Arabic translation equivalents that GT gives to technical terms. It also aims to discuss the semantic, contextual, syntactic, morphological and orthographic deficiencies in the Arabic equivalents to technical terms that GT yields. A random sample of technical terms with different Greek and Latin roots and different prefixes and class-changing suffixes was collected. The Arabic meaning of each word with its derivatives was looked up in Google Translate. It was found that GT gives Arabic equivalents to the full range of meanings of some terms such as 'mobilization' and 'technical', but it is inconsistent in translating terms with varying prefixes, roots combined with the same suffix, compounds and blends. A team of linguists or specialists is needed to revise, add and supervise scientific terms included in the
Арабистика Евразии, № 14, Июль 2021
Y • \ 1 jjIJJ И i iAjuiljj^l
Eurasian Arabic Studies, № 14. July 2021
АРАБИСТИКА ЕВРАЗИИ EURASIAN ARABIC STUDIES
LBРАЗИЯ АРАБИСТИНАСЫ
Google English-Arabic translation dictionary. Students wishing to use the Google English-Arabic translation tools need to be cautious and are advised to verify meanings obtained from GTwith their teacher or a specialist.
Keywords: Google Translate, machine translation, online translation, web-based translation, translation memory, technical terms, semantic deficiencies, contextual deficiencies, syntactic deficiencies, morphological deficiencies, orthographic deficiencies.
For citation: Al-Jarf R.S. (2021). An investigation of Google's English-Arabic translation of technical terms. Eurasian Arabic Studies, 14, 16-37.
АНГЛО-АРАБСКИЙ ПЕРЕВОД НАУЧНО-ТЕХНИЧЕСКИХ ТЕРМИНОВ
В GOOGLE ПЕРЕВОДЧИКЕ
Р. С. Ачъ-Джарф
Университет им. Короля Сауда
reima. al.jarf@gmail. com
Аннотация
Многие изучающие английский в качестве иностранного языка обращаются к сервису Google переводчик (Google Translate) для проверки значения сложных слов. Цель настоящего исследования - проверить наличие эквивалентов английских научно-технических терминов на арабском языке и точность их перевода. Также обсуждаются семантические, контекстные, синтаксические, морфологические и орфографические недостатки перевода у арабских эквивалентов, полученных при переводе в Google Translate. Материалом для исследования послужила случайная выборка технических терминов с греческими и латинскими корнями, а также с разными словообразовательными элементами - префиксами и суффиксами. Арабское значение каждого слова и его производных было проверено в Google переводчике. Было обнаружено, что сервис выдает все имеющиеся варианты перевода некоторых технических терминов, например, таких как «мобилизация» и «технический», но ошибается при переводе терминов с приставками, корнями в сочетании с одним и тем же суффиксом, сложными словами и словами-гибридами. Для решения данных проблем видится необходимым создание команды лингвистов или специалистов для
Арабистика Евразии, № 14, Июль 2021
Y • \ 1 jjIJJ И i iAjuiljj^l
Eurasian Arabic Studies, № 14. July 2021
АРАБИСТИКА ЕВРАЗИИ EURASIAN ARABIC STUDIES
LBРАЗИЯ АРАБИСТИНАСЫ
редактирования, добавления и проверки научно-технических терминов, включенных в словарь перевода Google с английского языка на арабский. Студенты, пользующиеся инструментами перевода Google с английского на арабский, должны проявлять внимательность, и им рекомендуется проверять переведенные единицы со своим учителем или специалистом. Ключевые слова: Google переводчик, машинный перевод, онлайн-перевод, веб перевод, память переводов, научно-технические термины, семантические ошибки, контекстные ошибки, синтаксические ошибки, морфологические ошибки, орфографические ошибки.
Для цитирования: Ачь-Джарф Р.С. Англо-арабский перевод научно-технических терминов в Google переводчике //Арабистика Евразии. 2021. № 14. С. 16-37. (на английском языке)
INTRODUCTION
Recent advancements in information and communication technology made the world a small village with a multitude of languages. To deal with this multilingualism, automated or machine translation (MT) tools have become an urgent need. A variety of free translation services for the automatic translation of texts and Webpages between a variety of languages are now available online such as Yahoo's Babel fish, Babylon, Bing Translator, dictionary.com Translator and Google Translate (GT). GT, in particular, supports translation between 64 languages1.
A review of the literature has shown that online MT systems have numerous advantages. They constitute an easier and faster way of translation between different languages as opposed to slow, expensive human translators (Bakhshaei, Khadivi, Riahi and Sameti, 2010). They make it possible for layman Internet users to perform cross-language searches (Wu and He, 2010), help remote groups of stakeholders to translate synchronous text-based chats in order to overcome language barriers that might exist among them (Calefato, Lanubile and Minervini, 2010), and are considered an effective tool for translating Web queries (Wu and He, 2010). Furthermore, they are used as a language learning tool by advanced learners (Garcia and Pena, 2011.
Some other studies that focused on specific MT systems such as GT showed that GT is more effective than Apertium-service in translating synchronous text chat
1 http://support.google.com/traiislate/
Арабистика Евразии, № 14, Июль 2021
Y • \ 1 jjIJJ И i iAjuiljj^l
Eurasian Arabic Studies, № 14. July 2021
АРАБИСТИКА ЕВРАЗИИ EURASIAN ARABIC STUDIES
LBРАЗИЯ АРАБИСТИНАСЫ
messages exchanged during four distributed engineering workshops in which groups of stakeholders were remotely negotiating software requirements. GT produced significantly more adequate English-Italian translations than Apertium and helped stakeholders overcome language barriers (C ale fa to, Lanubile and Minervini, 2010). In addition, MT systems, such as GT, were found to be effective in translating PubMed titles for patients. Analysis of statistical machine translation (SMT) output for English and six foreign language translation pairs within the biomedical domain showed a high performance for German, French and Spanish-English bi-directional translation pairs for both GT and the researchers' system (Wu, Xia, Deleger and Solti (2011).
Despite the numerous advantages of MT systems, they have a number of shortcomings as revealed by findings of some studies. In a study by Sheppard (2011), French researchers who struggled with English found GT highly useful in translating technical documents. However, Sheppard found that GT has significant limitations: It yields faulty translations that users should be aware of when they use it; it tends to translate word for word; it cannot select the appropriate equivalent according to context; it is insensitive to syntax and idiomatic expressions; in addition to confidentiality issues. Geer (2005) added that statistical translation technology must clear many hurdles such as inadequate accuracy and problems recognizing idioms before it can be useful for mission-critical tasks. MT systems typically give the general meaning of the words, but sometimes provide gibberish. They cannot accurately deal with complicated syntax, grammar, figures of speech, and jargon that native speakers take for granted (Goldsborough, 2009).
Furthermore, GT has limitations when translating compound and proper nouns encountered in general information articles and patent texts. Stefansson (2011) found that neither text type is better suited for Swedish-English MT than the other, and neither had an error rate below 20%. Words were erroneously omitted in the English output or were incorrectly translated in relation to context. However, in the general information articles, the most prominent errors were related to the Swedish version not being maintained in the English output, such as translating Abrahamsberg as Abraham rock. Compound and proper noun errors had varying impact on the meaning of the texts. Some distorted the meaning of the word completely; others were of minor importance.
To overcome MT problems and improve its quality, some researchers proposed alternative software and paradigms. For example, Komeili, Hendavalan and Ali Rahimi (2011) introduced MT software consisting of Padideh, Pars, Google and
Арабистика Евразии, № 14, Июль 2021
Y • \ 1 jjIJJ И i iAjuiljj^l
Eurasian Arabic Studies, № 14. July 2021
АРАБИСТИКА ЕВРАЗИИ EURASIAN ARABIC STUDIES
LBРАЗИЯ АРАБИСТИНАСЫ
Hezareh bilingual dictionary to overcome irregularities and problems in English-Persian MT software. In another study, Bakhshaei, Khadivi, Riahi and Sameti (2010) ran a statistical phrase-based system on Farsi-English language pairs using BLEU, an algorithm for evaluating the quality of translation accuracy. The proposed system achieved an improvement of 1.84%, relative to baseline accuracy, which was an increment from 16.97% to 18.81% in the best case.
As for Arabic, Aljlayl, Frieder and Grossman (2002) empirically evaluated the use of an MT-based approach, using the TREC-7 and TREC-9 topics and collections to improve query translation in an Arabic-English cross-language information retrieval (CLIR) system, called ALKAFI. In another study, Kholidy and Chatterjee (2010) identified errors by examining the outputs of some aligner annotation tools that already exist, and studied the main rules and guidelines required for building an aligner tool for correcting most alignment errors in Arabic. To overcome problems in the MT of Arabic relative clauses, Pedersen, Eades, Amin, and Prakash (2004) tried out an approach for parsing Arabic relative clauses in the tradition of the computational Paninian grammar framework, which led to deriving a common logical form for equivalent sentences. Their analysis was based on the development of a lexicalised dependency grammar for Arabic with applications for MT. They paid attention to resumptive pronouns in relative clauses that refer to the antecedent of the main clause, in the retrieval of syntactico-semantic relationships. In Chinese, the translation of derived words by MT systems received special attention. Wang and Harkness (2010) utilized a head transduction model for automatically translating derived words from English to Chinese, in order to solve the problem of derived words that are usually treated as unknown words by MT systems, and building a bilingual lexicon with a richer content.
There is no doubt that studies, such as the ones reported above and others, have lead to the improvement in MT in some language pairs, yet there are still areas that need to be improved at the word, sentence and text levels in other languages such as Arabic.
Geer (2005) notes that few researchers have worked on approaches that compare and analyze documents and their already-available translations to statistically determine the likely meanings of phrases, to help statistical systems use this information to translate new documents.
The area of terminology is of special importance in MT, due to the influx of English technical terms and productivity of new inflected and derived forms in this age of great advancements in science and technology. There is a need to find out whether
Арабистика Евразии, № 14, Июль 2021
Y • \ 1 jjIJJ И i iAjuiljj^l
Eurasian Arabic Studies, № 14. July 2021
АРАБИСТИКА ЕВРАЗИИ EURASIAN ARABIC STUDIES
LBРАЗИЯ АРАБИСТИНАСЫ
MT translation systems such as GT are capable of providing accurate Arabic equivalents to English technical terms, especially emerging ones. Therefore the present study aims of investigate the English-Arabic translation of technical terms by GT. Specifically, the study aims to examine the following:
(i) the breadth of coverage of technical terms, i.e. whether GT gives Arabic equivalents to the most recent English technical terms (whether single terms, compounds, blends or acronyms), cover all derivatives, all senses, all combinations of a particular affix with a variety of roots and all possible combination of a variety of affixes with a particular root;
(ii) (ii) the semantic, syntactic, morphological, contextual and orthographic accuracy of Arabic equivalents given by GT to English technical terms, whether GT is consistent in translating the same affix attached to different roots, the different affixes attached to the same root, and irregularities in translating compounds, blends and acronyms.
Although the Internet has greatly increased students' access to many language and communication tools, web-based machine translation (WBMT) is still ineffective for translating text into another language, especially when the user of the software is not able to make gramma tic ality and acceptability judgments in the target language (Williams, 2006). In this respect, results of the present study shed light on the usefulness of GT as a tool for translating English technical terms into Arabic . The paper describes specific limitations of WBMT with examples from English-Arabic GT translation of technical terms and areas that still need to be improved and give some recommendations to computational linguists for enhancing the quality of English-Arabic GT translation and to teachers when presenting WBMT such as GT to students in order to promote their language awareness and electronic literacy, which could help reduce their misuse of this tool.
Results of the present study are also useful to translation students at the college of Languages and Translation (COLT), King Saud University, Riyadh, Saudi Arabia who take 18 specialized translation courses in 18 subject areas including engineering, physical science, medicine, petroleum, computer science, sociology, and others, with English and Arabic as their working language pairs and who frequently check GT for Arabic equivalents of technical terms in those subject areas.
Results will help raise their awareness of the shortcomings of the Google translation of scientific terms. Students wishing to use the GT in English-Arabic translation need to be cautious and are advised to verify Arabic equivalents obtained from GT with their instructor or a specialist.
Арабистика Евразии, № 14, Июль 2021
Y • \ 1 jjIJJ И i iAjuiljj^l
Eurasian Arabic Studies, № 14. July 2021
АРАБИСТИКА ЕВРАЗИИ EURASIAN ARABIC STUDIES
LBРАЗИЯ АРАБИСТИНАСЫ
MATERIALS AND METHODS Data Collection
A random sample of 200 technical terms covering different disciplines such as medicine, technology, politics, economics, psychology, computer science, linguistics, literature, astronomy and others was collected from Webster's Online Dictionary and news websites such as BBC. The sample covered the following: (i) 90 Terms with different Greek and Latin roots and affixes: electroencephalograph, hydrology', internationalization, biometeorology, biometeorologist,
counterproductive, instrumentalization, macroeconomics, paralympic, psychometric, interoperability, reflexology, hermeneutics, bolemia, phlebostasis, psychometrist, arteriofibrosis, arteriorrhagia, endocentric, exocentric. (ii) 60 compounds such as: bypass surgery, extended family, facial expressions, handling services, video message, workstation, workstation, workstation, Euro games, Bionic bodies, arctic archaeology'). (iii) 10 Blends such as flexicurity, socioeconomic, workaholic, radiotherapy, aromatherapy, counter-productive, videoconferencing, (iv) 10 acronyms such as CIA, HTML & PM\ and (v) 10 miscellaneous terms as in: utilities, utility, tweet. Thus, the sample covers various noun and adjective forms and terms with correct and incorrect Arabic equivalents for comparison purposes. Data analysis
The author looked up each term in isolation, i.e., without a context, in GT. The percentages of terms with available Arabic equivalents and those with faulty Arabic translation equivalents were calculated. Faulty Arabic translations were classified into those with coverage and accuracy shortcomings. Coverage refers to whether a term and/or its Arabic meaning exist in the GT corpus, whereas accuracy refers to whether Arabic equivalents obtained match the English source terms semantically, contextually, syntactically, morphologically, and orthographic ally. In addition, the author classified each faulty Arabic translation given by GT into a semantic, contextual, syntactic, morphological, or orthographic deficiencies. Each of which is defined below.
• Semantic deficiencies refer to Arabic equivalents with a faulty meaning such as: (a) giving far-fetched meanings; (b) giving inaccurate meanings to single terms with multiple meanings; (c) giving the literal meaning of a component part rather than the specialized meaning of the whole term as a block sequence; (d) when new meanings are missing; (e) selecting a meaning of one element of a compound that does not collocate with the other element with which it is used; (f) treating compounds as
Арабистика Евразии, № 14, Июль 2021
Y • \ 1 jjIJJ И i iAjuiljj^l
Eurasian Arabic Studies, № 14. July 2021
АРАБИСТИКА ЕВРАЗИИ EURASIAN ARABIC STUDIES
LBРАЗИЯ АРАБИСТИНАСЫ
single isolated parts that are not related to each other, not as block sequences with a fixed meaning; (g) inconsistency in translating terms containing different affixes when they combine with the same root, and in translating the same affix when it combines with different roots; (h) translating part of a term (a root or affix), compound or blend and ignoring the other; and (i) confusing terms that are orthographic ally similar; and (J) giving unfamiliar Arabic equivalents.
• Contextual deficiencies refer to: (a) giving an Arabic meaning that does not fit the context in which the term is used; (b) treating a compound or blend as consisting of single unrelated words rather than a block sequence with a fixed specialized meaning; and (c) giving a faulty meaning of an element that does not collocate with the complementary element or does not match the domain in which a particular sense is used when translating compounds with elements with multiple meaning.
• Syntactic deficiencies refer to: (a) a faulty word order when translating compounds, blends and terms containing prefixes, i.e. giving a left-to-right rather than right-to-left word order; (b) inconsistency in translating compounds sharing an element; and (c) giving definite Arabic equivalents to single terms, acronyms, compounds and blends that are definite.
• Morphological deficiencies refer to: (a) lack of agreement in number, gender and part of speech between the English source term and its Arabic equivalent; (b) inability to distinguish between different nouns or adjectives derived from the same base form but ending in different noun or adjective suffixes; and (c) inability to capture the change in meaning when a particular noun is singular, and when it has a plural form.
• Orthographic deficiencies refer to: (a) giving faulty meanings when the elements of the same compound are hyphenated, separated by a blank or agglutinated; (b) giving faulty meanings when the same term is spelled with upper and lower cases; and (c)spelling errors in the Arabic equivalents.
Some faulty Arabic equivalents in the error corpus had more than one error. In this case, the equivalent was classified several times under the relevant error categories. For example, river bed jJj^1 has three translation errors: Semantic (the bed we sleep on of the river), contextual (bed is not used in geography) and syntactic (reversed word order).
Errors in each category (semantic, contextual, syntactic, morphological, and orthographical) were computed in percentages.
АРАБИСТИКА ЕВРАЗИИ EURASIAN ARABIC STUDIES
LBРАЗИЯ АРАБИСТИНАСЫ
Арабистика Евразии, № 14, Июль 2021
Y • \ 1 jjIJJ И i tAjuiljj^l ^JJ^I illluiljjll
Eurasian Arabic Studies, № 14, July 2021
Validity and reliability
To verify the GT translation error corpus, the list of terms and their Arabic equivalents given by GT were given to two colleagues with a Ph.D. degree in linguistics and two colleagues with a Ph.D. degree in translation. They were asked to judge the correctness/incorrectness of the Arabic equivalents given by GT. To check the consistency and accuracy of the author's classification of the translation errors, they were also asked to classify the GT translation errors into semantic, contextual, syntactic, morphological, and orthographic errors using the operational definitions given above. Their analyses were compared with those of the author and disagreements were resolved by discussion. RESULTS
Coverage issues
Results of the data analysis showed that GT covers many newly coined terms such as: flexicurity, Euro zone, proxy war, hermeneutics, surreptitious, Zeitgeist, kinesophobia, claustrophobia, xenophobia, Schadenfreude. A large number of medical terms are covered by GT as well as political terms widely used in the media. It gives Arabic equivalents to the full range of meanings of some terms and sometimes explanatory equivalents according to the Part of Speech of the English terms in cases of terms with multiple parts of speech as in the following examples:
• Mobilization: (noun) ^Jj^-^ ^ t.lj*i->*i t^Lj^j *■ ajS j ^jh'i'i ^-iAjJj Ai^a.
• Claustrophobia: o-^j* '^laLJ1 j^LaVl j-11 i jUJLiVl ■m^Vj
ilhJall
• Xenophobia: 'mjU-V1 ^^j
When it lists the different Arabic equivalents, English synonyms matching each Arabic meaning are displayed on the right (See Figure 1).
On the other hand, GT provides no translation or a partial translation to 11% of the terms in the sample, whether single terms, compounds, blends, or acronyms as in "bolemia, endocentric, exocentric, phlebostasis, psychometrist, arteriofibrosis, arteriorrhagia, thermite bomb aromatherapy ^jJl telecommuting ^ i>". 70% of the common acronyms in the samples such as "PDF, FBI, CIA, G4S, DNA, FIFA, HTML " are not translated, although the Arabic meaning of the full form is given such as "Hypertext Markup Language lk^I i ij^jj
There is a lack of balance in GT's coverage of terms containing medical roots. For example, the author found 17 terms containing the root arterio-, 11 terms containing the root phlebo-, and 6 terms containing the root veno- only.
Al-Jarf, R.S. (2021). Ail investigation of Google's English-Arabic translation of technical terms.
Eurasian Arabic Studies, 14. 16-37.
Арабистика Евразии, № 14, Июль 2021 25
АРАБИСТИКА ЕВРАЗИИ Y • \ 1 jjIjj И i (AjuiIjj^I ¡jjluiljjll
EURASIAN ARABIC STUDIES LBРАЗИЯ АРАБИСТИНАСЫ Eurasian Arabic Studies, № 14. July 2021
Go gle retma.al.jarf@gmail eom -
Translate From: English - ^ To: Arabic ▼
English Arabic Spanish Detect language Arabic English Spanish
class X
p a = m Ф) у
Newt Click the words above to edit and view alternate translations Dismiss
noun
class, category, group, denomination, grade, division
degree, grade, point, class, stage, stair
class, type, brand, species, kind, classification
separation, claBS, season, decisiveness
t>> type, kind, sort, species, form, class
batch, payment, impulse, class, installment, pay
share, portion proportion, ration... lot class
style, type, class, make, stripe, mold
liji. status, standing, rank, stature, dignity, class
L ll class, grade, schoolroom, form
Ojij superiority, excellence, supremacy, preponderance, pre-eminence
^S .LH -1 U.ll class
* y~ ' -w j ¿1 JJj consequence, class., sphere, station
i-lJ class
fJiiiSyitiUa class
.wljU jJi^ class
verb
class, grade, classify, sort, rate, group
ii^«»ftB class
MijS class
class
Figure 1: Source text and target text windows, list of translation equivalents, Part of speech and English synonyms for which an Arabic meaning is given by GT
In addition, Arabic equivalents given by GT do not cover all forms, i.e., parts of speech derived from the same term, especially nouns and adjectives. Although an equivalent is given to 'psychometrics, psychometric; grammar, grammarian & grammatical', no equivalents are given to 'psychometrist & grammatically'. GT gives no equivalents to exocentric & endocentric' either.
Accuracy Issues
Data analysis revealed that GT gives correct Arabic equivalents to some new terms such as Nanotechnology, Muon, hermeneutics, surreptitious, Zeitgeist, kinesophobia, claustrophobia, xenophobia, Schadenfreude, war proxy', smart phone, supernova, broadband internet, militant nexus, servicewomen, peri-mortem, Euro zone, plug-ins, painkillers and radiotherapy. It also gives correct Arabic equivalents to 88% of the medical terms in the data containing affixes and/or roots and widely used political terms. However, Arabic equivalents given by GT have several semantic, contextual.
АРАБИСТИКА ЕВРАЗИИ EURASIAN ARABIC STUDIES
LBРАЗИЯ АРАБИСТИНАСЫ
Арабистика Евразии, № 14, Июль 2021
Y • \ 1 jjIJJ И i tAjuiljj^l ^JJ^I illluiljjll
Eurasian Arabic Studies, № 14, July 2021
syntactic, morphological and orthographic deficiencies, each of which is described in detail below.
Semantic Deficiencies GT gives mistranslations with far-fetched equivalents as in examples 1-14:
1. reflexology jt&ill 9. physiotherapy sljlj^ -inJJl Z^JLlJI
2. macroeconomics 4j*nhH
3. consortium ^L^YI 10. socioeconomic status
4. hydrometer Jh^Jt inL^syij y^Li^yi
5. phlebangioma y^j/jMf-dl11. social system ¿^L^YI ¡»Lki
6. Flexicurity 12. smart board ¿¿¿i//
7. bypass surgery ¿¡^/j^. i±LJ^ 13. medical conditions iuiJ/ t-ij
8. counterproductive^^ 14. heartburn lJ^HMj^
In translating terms with multiple meanings, GT gives meaning that are inaccurate as in "logistics: cjblj^Jl iAjJ-j Ajflkla tijajiii; Technical 'lW lk^s technocrat
ts^p^VI (jjLui ^Jc ^jLoj t>; gives the literal meaning of the component
part rather than the specialized meaning of the term as a block sequence as in 'flexicurity physiotherapy ^»jiMl sIjIa^. In other cases, new meanings are
missing as in 'tweet S**^ *j^1 ^Ul There is no mention a
'tweet' written on Twitter.
In compounds containing elements that have multiple meanings such as bodies, board, GT selected a meaning of one element that does not collocate with the other element with which it is used as in bionic bodies '-A^1 jjj; smart board
macroeconomic jL-ajSVt. This is probably because GT does not treat those
compounds as block sequences with a fixed meaning, but as single isolated parts that are not related to each other.
GT is also inconsistent in translating terms containing different affixes when they combine with the same root, and in translating the same affix when it combines with different roots. For example, GT makes no distinction in meaning between intercellular and intracellular and both are translated as Ij^kk ¿31 j. In some cases, '-gram' & '-graph' are given the same equivalent when combined with the root arterio- as in arteriogram & arteriograph AAj"1 / but translated differently when combined with the root phlebo- as in phlebogram j & phlebograph jJj-^j. A completely different equivalent is given to -gram in electroencephalogram ^bj^ll ^jll and no translation is given to -gram when it combines with the root veno- as in venogram or when -graph is used in
Арабистика Евразии, № 14, Июль 2021
Y • \ 1 jjIJJ И i iAjuiljj^l
Eurasian Arabic Studies, № 14. July 2021
АРАБИСТИКА ЕВРАЗИИ EURASIAN ARABIC STUDIES
LBРАЗИЯ АРАБИСТИНАСЫ
electroencephalograph Similarly, GT is inconsistent in translating '-algia
when it combines with different roots such as neuro- & arthr-. It gives an equivalent to 'neuralgia ' ¡JVl but not arthalgia.
In other instances, GT translates part of the term (a root or affix), compound or blend and ignores the other root, base or element as shown in the underlined parts of the following examples: telecommuting i>, electroencephalograph lt^js^I phlebitis ajjj Ulia^ ajJj phleboasthenia phlebo i> j^1, thermite bomb aromatherapy GT confuses terms that are orthographic ally similar. For instance, 'neutrons & neutrinos' are given the same meaning '<—^jjJj^and 'experiential & experimental' are given the same Arabic equivalent
GT gives unfamiliar Arabic equivalents that are rarely used such as psychometric j^j^J1"A cyber sarcoma hydraulics hydrology dAjjUl,
hydrometer J^^1, biometeorology <—radiotherapy ^U^Wtfl , although
such terms have more commonly used Arabic terms.
In some cases, GT transliterates some source term although an Arabic equivalent exists as in examples 15-22:
15. videoconference j^jtP^ 19. hydrologist W-Ajj-^'
16. technological 20. hydraulic y^Asj-^
17. romanticism ^^Lajj 21. Paralympic ^JOM1
18. Psychometric lsS^j^*^ 22. neuralgia L^ljy^
Results of the data analysis in the present study have revealed that in 16% of the total errors and 30% of faulty compounds, the faulty Arabic meaning provided by GT does not fit the context in which the term is used in Arabic. In such contextual errors, GT treats a compound or blend as consisting of single unrelated words rather than a block sequence with a fixed specialized meaning used in a special domain. Although, 'radiotherapy ^JU^ smart phone ^^ were translated correctly;
'physiotheraphy, aromatherapy, and smart board were not. It seems that GT does not recognize the domain in which such terms are used and the exact Arabic terms used in those domains in cases such as: flexicurity aj^1 irritable colon J^VI ¡y^ aAM extended family 1 ^^ liaison interpretingj^j , microeconomics
^»«T'ti ojsYI and reflexology' . Compounds and blends that refer to fields of
study (liaison interpreting, microeconomics), diseases (irritable bowel syndrome, bypass surgery, bolemia) and specific concepts used in sociology (extended family, socioeconomic status), technology (smart board), geography (solar system, river
Contextual Deficiencies
Арабистика Евразии, № 14, Июль 2021
Y • \ 1 jjIJJ И i iAjuiljj^l
Eurasian Arabic Studies, № 14. July 2021
АРАБИСТИКА ЕВРАЗИИ EURASIAN ARABIC STUDIES
LBРАЗИЯ АРАБИСТИНАСЫ
bed), medicine (medical conditions, serum iron), and linguistics (underlying meaning) should be translated as block sequences with a fixed meaning. In compounds in which an element such as (bodies, conditions, bed, core, hoard or return) has multiple meanings, a faulty meaning that does not collocate with the complementary element together with the domain in which a particular sense is used need to be taken into consideration.
Data analysis has shown that compounds constitute a major weakness in English-Arabic translation by GT and the main deficiency in translating English compounds into Arabic is word order. In 60% of the compounds in the sample, GT gives Arabic equivalent with a reversed word order, i.e. a left-to-right rather than the Arabic right-to-left word order as in Burmese dissident ¿j'^1'1^1 arctic archaeology'm^1
jliiVl JUAI^ lexical transference J^ aj^sj-Ji ; Martian air lSj^1 plastic
money lM1 security/ system ^Lk^ in such examples, GT follows a linear
left-to-right direction (sequence) in translating an English compound which is contrary to the natural Arabic word order. No rules seem to be provided in the GT dictionary for the correct re-ordering of elements of an equivalent Arabic compound. The same reversed word order (linear left-to-right direction) is followed in the translation of English blends and terms containing prefixes such as flexicurity i>Vi ujA counter evidence physiotherapy (UJl ^JUJl. In those
examples, GT seems to break such blends into their component parts and handle those parts or elements as independent unrelated elements rather than treating them as block sequences with fixed Arabic meanings.
In translating compounds sharing an element, GT is also inconsistent. For example, smart phone y^l nuclear fusion iSjy^ jV-^l Euro zone jjj^1 are
translated correctly, but Smart board ^^ <ji1VA nuclear sabotage "Mj^and
Euro games lA^VI jjj^1 are not. Political system gr^^1 f1-^1 is translated correctly, whereas Solar system Aalkll ^U^ social system ^^VVl f^ and security system
^Lkill ¿>1 are not. Pine wood pinewood jiy^ pinewoods
jjji~=^are correctly translated, however river bed J^ is not It seems that GT
does not follow any Arabic-language rules in translating such compounds and that correct Arabic equivalents given depends on the frequency of use of the compound, i.e., pairs that are commonly used in the GT corpus are correctly translated. Another major syntactic deficiency in English-Arabic translation by GT is giving definite Arabic equivalents. It was found that in 55% of the data, GT gives Arabic equivalents to single terms, acronyms, compounds and blends that are definite, i.e.
Syntactic Deficiencies
Арабистика Евразии, № 14, Июль 2021 29
АРАБИСТИКА ЕВРАЗИИ Y • \ 1 jjIjj И i (AjuiIjj^I ¡jjluiljjll
EURASIAN ARABIC STUDIES LBРАЗИЯ АРАБИСТИНАСЫ Eurasian Arabic Studies, № 14. July 2021
containing the Arabic definite article /—al/ as in the following examples: Circumnavigation <—sljUl, proxy J^j^1 , neutrinos ^jjJj^1, neuralgia ^j"-»»^ facial expressions H^jM fossiltSjjh^1 Jj^^l fossilization
gastrointestinal jW^l globalized tweet ^¿Jl, utilitarian
understandability touristic Physiotherapy In translating
single terms, the Arabic equivalent, in most cases, should be indefinite, i.e. without the Arabic definite article —al.
Morphological Deficiencies Findings of the present study have shown that the most common morphological problem in English-Arabic translation by GT is lack of agreement in number, gender and part of speech between the English source term and its Arabic equivalent. Agreement errors constitute 14% of the total errors. In some cases, GT gives single Arabic equivalents to plural English source terms as in radii »ji*^, utilities ^lil^ and gives a plural Arabic equivalent to a singular English source term as in smart phone Ajiill t-fljljfr!!, Burmese dissident i-^jjA hydraulics hydrology ¿jLuUI ,
logistics cjblj^Jl j instrumentalization cjljjli In translating romanticism some of the equivalents are masculine 'ur^j' and others '^Ap^jj & AA^ ' are feminine. In addition, some Arabic equivalents do not match the English (source) term in part of speech. GT does not seem to distinguish between English source terms that are nouns and those that are adjectives based on their final suffix. English adjectives are given an Arabic equivalent that is a noun as in contextual meteorological
4jj=Jl ±^»jVI, globalized ^Technical touristic and gastrointestinal
jWA1. English nouns are given Arabic equivalents that are adjectives as in hermeneutics an English noun that refers to a dealer or doer is given an
equivalent that refers to the field of study as in hydrologist The compound
noun 'bypass surgery' is given the Arabic equivalent ¿'^j'"' that is a
verb phrase. Some nouns are given Arabic equivalents that are prepositional phrases as in instrumentalization and fax Furthermore, GT does not distinguish
between different nouns or adjectives derived from the same base form but ending in different noun or adjective suffixes. Both understandability & understanding are given the same Arabic equivalent '
Overall, GT does not seem to be able to capture the change in meanings when a particular noun is singular and when it has a plural form. GT gives the same Arabic equivalent S-^UJlfor both 'utility' and 'iutilities, although the former has a completely different meaning from the latter.
Арабистика Евразии, № 14, Июль 2021
Y • \ 1 jjIJJ И i iAjuiljj^l
Eurasian Arabic Studies, № 14. July 2021
АРАБИСТИКА ЕВРАЗИИ EURASIAN ARABIC STUDIES
LBРАЗИЯ АРАБИСТИНАСЫ
Orthographic deficiencies
An interesting finding in the present study is that GT gives different Arabic meanings when the elements of the same compound are hyphenated, separated by a blank or agglutinated. Here again, some inconsistencies exist. For example, a translation is given to peri-mortem sli^ll but not perimortem. The same Arabic equivalent is given to Eurozone and Euro zone. Videoconferencing is translated as ¡io^3 Cf- Jj^ jjAiil^ whereas 'video conferencing'' is translated as o^'j^A»^- Workstation is translated as J^1 but both workstations and workstation are translated as J^ 4 in which the word order is the reverse of lU^' Score card ialkj and
heart burn m^1 ^are translated word for word (literally translated); but scorecard *bVI J?-* and heartburn s^Jl ^ are correctly translated. Interestingly, pine wood jj^i^all M^-and pinewood jjji-^i are correctly translated.
Letter case affects English-Arabic translation by GT as well. Entering the same terms in upper- or lower-case results in different Arabic translation equivalents. 'Neutrino halo' is translated as jAJjijkP when entered in lower case; but 'NUTRINO HALO' is translated as '¿jJjJj^1 Halo', when entered in upper case. The equivalent udLPj^ Halo' is meaningless as GT does not recognize HALO and hence it yields it as it is in the Arabic equivalent.
In very few cases, there are minor spelling errors in the Arabic equivalents such as Mobilization^.'¿3 which should be 'LjaU "and LASER j which should be 'jj^1' as it is the commonly used spelling. Such spelling errors seem to exist in the Arabic corpus of GT.
DISCUSSION, CONCLUSION AND RECOMMENDATIONS
Despite the fact the GT is capable of providing accurate Arabic equivalents to a wide range of English technical terms, findings of the present study have shown that GT still provides Arabic equivalents to English terms that have semantic, syntactical, morphological and orthographic limitations especially in the case of compounds. GT seems to follow a linear left-to-right word-for-word order in translating compound terms from English to Arabic, seems to treat compounds as consisting of unrelated constituents and translates each independently of the other.
Findings of this study are consistent with findings of other studies in the literature such as Sheppard (2011), Stefansson (2011), Wang and Harkness (2010), Geer (2005) and others that revealed some weaknesses, irregularities, and limitations in MT in general, between English and other languages such as French, Swedish, Farsi and Chinese, in particular.
Арабистика Евразии, № 14, Июль 2021
Y • \ 1 jjIJJ И i iAjuiljj^l
Eurasian Arabic Studies, № 14. July 2021
АРАБИСТИКА ЕВРАЗИИ EURASIAN ARABIC STUDIES
LBРАЗИЯ АРАБИСТИНАСЫ
One possible explanation for the semantic, syntactic, morphological, contextual, and orthographic limitations in the GT English-Arabic translation is that GT uses an SMT (statistical machine translation) paradigm where translations are generated on the basis of statistical models whose parameters are derived from the analysis of bilingual text corpora2. When GT generates a translation, it looks for patterns in hundreds of millions of documents to help decide on the best translation for a particular user. By detecting patterns in documents that have already been translated by human translators, GT makes intelligent guesses as to what an appropriate translation should be. Since the translations are generated by machines, not all translation will be perfect. The more human-translated documents that GT can analyze in a specific language, the better the translation quality will be. This is why translation varies across languages3 and translations between English, German, Spanish and French are more accurate than other languages.
Another possible explanation for the irregularities in GT English-Arabic translation is the poor Arabic content on the Internet compared to English, Spanish, German and others, i.e. there is a gap between the number of people who speak Arabic and the amount of information available online in Arabic. This means that the inadequate amount of Arabic-English documents in the corpus results in gaps, irregularities and translation inaccuracies.
A good MT system requires a good English-Arabic parallel corpus, i.e., a collection of original texts in English and their translations in Arabic. Internet archives contain a lot of parallel documents. To construct a good parallel corpus from Internet archives, we must have a good English-Arabic bilingual dictionary (a dictionary that translates words and phrases from English into Arabic). Fattah, Ren, Shingo, Atlam, (2003) recommended using an algorithm to automatically extract an English/Arabic bilingual dictionary from parallel texts that exist in those Internet archives. Deng, Kumar and Byrne (2007) also recommended that bilingual chunk pairs be extracted from parallel texts to create training sets for SMT. This kind of chunk alignment can significantly reduce word alignment error rate. Based on this, currently available dictionaries such as the BASM, Ajeeb, Arabization Center, Arabic Language Academy and other dictionaries may be added to enrich the Google English-Arabic Dictionary. To help improve the quality of English-Arabic translations provided by GT, the present study recommends that a team of computational linguists or subject-area specialists revise, add and supervise scientific terms included in the Google English-Arabic Dictionary.
2 http://en.wiMpedia.arg/wiki/Statistical_niachine_translation
3 http://trauslate.google.coni/about/uitl/en_ALL/
Al-Jarf, R.S. (2021). Ail investigation of Google's English-Arabic translation of technical terms.
Eurasian Arabic Studies, 14, 16-37.
Арабистика Евразии, № 14, Июль 2021
Y • \ 1 jjIJJ И i iAjuiljj^l
Eurasian Arabic Studies, № 14. July 2021
АРАБИСТИКА ЕВРАЗИИ EURASIAN ARABIC STUDIES
LBРАЗИЯ АРАБИСТИНАСЫ
Another approach was proposed by Al-Khalifa and Al-Khalifa (2011) which is based on the power of user generated content to drive Arabic translations of English words. According to Vangie Beal4 and Rachel Lebeaux5 user-generated content refers to any published information such as video, wikis, blogs, discussion form posts, digital images, audio files that an unpaid contributor, consumer or end-user of an online system or service has provided to a website and is publically available to other endusers. Al-Khalifa and Al-Khalifa's pilot experiment revealed the potential of this approach which can act as an add-on to improve the quality of Google Translate. Since GT is an SMT system, the quality of English-Arabic translations provided by GT can be improved based on the findings of prior studies and approaches proposed by English-Arabic translation researchers. Alsharaf, Cardey, Greenfield and Shen (2004) indicated that if we wish to have an MT of high quality, two issues must be resolved: The semantic aspect, i.e. meaning of the linguistic units and adapting the translation calculus to the Arabic language.
In addition, word alignment (translation relationships among the words in a bitext) is an important supporting task in different Natural language processing (NLP) applications like training of MT systems, translation lexicon induction, word sense discovery, word sense disambiguation, information extraction and cross-lingual projection of linguistic information (Kholidy and Chatterjee (2010). To solve alignment problems, Deng, Kumar and Byrne (2007) derived two different alignment procedures based on the underlying alignment model. The first procedure was used to generate an initial coarse alignment of the parallel text and the second was a divisive clustering parallel text alignment procedure used to refine the first-pass alignments. It permits the segmentation of the parallel text into sub-sentence units which are allowed to be reordered to improve chunk alignment. In another study. In a third study, Dekai, Carpuat and Yihai (2006) presented a direct measurement of word alignment coverage on an Arabic-English parallel corpus using inversion transduction grammar (ITG) constraints. Their results provided evidence that ITG expressiveness appears to be largely sufficient for core MT models.
To solve agreement and word order errors in Arabic translations, Abu Shquier and Sembok (2008) recommended the use of Tarjim software by Sakhr Company which utilizes a new generation of Arabic Natural Language Processing (NLP) technologies. The approach adopted is a rule-based approach that can be used as a stand-alone tool or can be integrated in a general MT system for English sentences.
4 littp: //www. webop edia. c oni/T ERM/U/UGC. litml
5 http: //se arc he io. techtarget. c oin/defmi tion/user- genera ted-c ontent-UGC
Al-Jarf, R.S. (2021). An investigation of Google's English-Arabic translation of technical terms.
Eurasian Arabic Studies, 14, 16-37.
Арабистика Евразии, № 14, Июль 2021
Y • \ 1 jjIJJ И i tÄjiüljjVl
Eurasian Arabic Studies, № 14. July 2021
АРАБИСТИКА ЕВРАЗИИ EURASIAN ARABIC STUDIES
IEРАЗИЯ АРАБИСТИНАСЫ
A Phrase-Based Statistical Machine Translation (PBSMT) approach can be also used based on the findings of a study by Oransa, Kouta, and Sakre (2010) that utilized the Injected Tags (ITs) approach, which was applied to English-Arabic translation, and which has shown considerable improvement in the translation quality with an increase of at least 13% of BLEU score. Based on the results of another study that focused on PBSMT, Turchi, Marco, De Bie, Tijl, Goutte and Cristianini, (2012) proposed the integration of linguistic rules.
To solve the semantic problems in English-Arabic translation by GT, Jusoh and Alfawareh (2011) proposed a framework for a semantic-based translation in English-Arabic MT. Ahmed and Nürnberger (2008) also presented a word sense disambiguation approach consisting of a natural language processing method that deals with the rich morphology of the Arabic language and word sense disambiguation. This approach adapts the Naive Bayesian approach with new features that consider the Arabic language properties and the exploitation of a large parallel corpus to find the correct sense based on its cohesion with words in the training corpus. Tillmann and Zhang (2008) presented a new online relevant set algorithm for a linearly scored block sequence translation model. This online algorithm introduces "seed" block sequences which enable the training to be carried out without a gold standard block translation.
Ellison (2011) also proposed AutoLex, an application, for translating dynamic content on websites that use the Django web framework. AutoLex retrieves translations from the GT service., stores them in a database using a single table, and serves them via a user-defined accessor. In doing so, AutoLex offers a fast, cheap way to translate large amounts of content and to enable multilingual communication among users.
In addition, this study recommends that Arabic equivalents cover all parts of speech related to the same term, all parts of the word, all derived forms, all terms derived by adding the same affix to different roots, and different affixes to the same root and commonly used acronyms. Arabic equivalents to English compounds should follow the Arabic word order. Translation equivalents should make sense and should agree in number, gender and part of speech with the English source term. Some compounds also need to be translated as a block sequence.
Finally, to advance the GT bidirectional English-Arabic translation of technical terms and enhance its quality; the quality and coverage of Arabic-English translation of technical terms by GT; comparisons of English-Arabic translation by GT and other MT systems such as Babel Fish, and surveys of translation; and exploration of EFL
WW&ß (Шл Арабистика Евразии, № 14, Июль 2021 34
АРАБИСТИКА ЕВРАЗИИ Y • \ 1 jJjj И i tÄjiüljjVi
EURASIAN ARABIC STUDIES IEРАЗИЯ АРАБИСТИНАСЫ Eurasian Arabic Studies, № 14. July 2021
student-users' views of GT and the strategies they use in verifying the Arabic meanings they obtain from GT are still open for further investigation by future researchers.
BIBLIOGRAPHIC REFERENCES
1. Abu Shquier, M. & Sembok, T. (2008). Word agreement and ordering in English-Arabic machine translation. Proceedings of the International Symposium on Information Technology, 1, 1-10.
2. Ahmed, F. & Nürnberger, A. (2008). Arabic/English word translation disambiguation approach based on naive bayesian classifier. Proceedings of the International Multi-conference on Computer Science and Information Technology, 331-338.
3. Al-Jarf, R. (2018). Multiple Arabic equivalents to English medical terms: Translation issues. Internationa! Linguistics Research, 1{ 1); 102-110. doi.org/10.305 60/ilr.v lnlp 102
4. Al-Jarf, R. (2010). Translation students' difficulties with English Neologisms. Analele Universitäjii "Dunärea De Jos" Din Galap Fascicula XXIV ANUL III (2). 431-437. Romania.
5. Al-Jarf, R. (1994). A Contrastive analysis of English and Arabic morphology. Retrieved from https://www.researchgate.net/publication/31 2193999 A Contrastive Analysis of English Mid Arabic Morphology
6. Al-Jurf, R. (1995). A Contrastive analysis for translation students. Riyadh: King Saud University. Retrieved from https://www.researchgate.net/publication/268274970_Contrastive_Analy sis_for_Translation_Students
7. Aljlayl, M.; Frieder, O. & Grossman, D. (2002). On Arabic-English cross-language information retrieval: a machine translation approach. Proceedings of the International Conference on Information Technology: Coding and Computing, 2-7.
8. Al-Khalifa, A. & Al-Khalifa, H. (2011). Towards improving machine translation using user generated content. Proceedings of the International Conference on Communications and Information Technology (ICCIT), 11-14.
9. Alsharaf, H., Cardey, S., Greenfield, P. & Shen, Y. (2004). Problems and solutions in machine translation involving Arabic, Chinese and French.
Арабистика Евразии, № 14, Июль 2021 35
АРАБИСТИКА ЕВРАЗИИ Y • \ 1 jjIjj И i (AjuiIjj^I ¡jjluiljjll
EURASIAN ARABIC STUDIES LBРАЗИЯ АРАБИСТИНАСЫ Eurasian Arabic Studies, № 14. July 2021
Proceedings of the International Conference on Information Technology: Coding and Computing, 2, 293-297.
lO.Bakhshaei, S., Khadivi, S., Riahi, N. & Sameti, H. (2010). A study to find influential parameters on a Farsi-English statistical machine translation system. Proceedings of the 5th International Symposium on Telecommunications (1ST), 985-991
1 l.Calefato, F.; Lanubile, F. & Minervini, P. (2010). Can real-time machine translation overcome language barriers in distributed requirements engineering? Proceedings of the 5th IEEE International Conference on Global Software Engineering (ICGSE), 257-264.
12.Dekai, Wu, Carpuat, M. & Yihai Shen (2006). Inversion transduction grammar coverage of Arabic-English word alignment for tree-structured statistical machine translation. Spoken Language Technology Workshop. IEEE, 234-237.
13.Deng, Y., Kumar, S. & Byrne, W. (2007). Segmentation and alignment of parallel text for statistical machine translation. Natural Language Engineering 13(3), 235-260.
14.Ellison, K. (2011). An application for automatically translating dynamic web content. M.A. Thesis. Massachusetts Institute of Technology. Dept. of Mechanical Engineering.
15.Fattah, M.A., Ren, F., Shmgo, K. & Atlam, A. (2003). English/Arabic bilingual dictionary construction using parallel texts from the internet archive. Proceedings of the IEEE 46th Midwest Symposium on Circuits and Systems, 2, 978-981.
16.Garcia, I. & Pena, M. (2011). Machine translation-assisted language learning: writing for beginners. Computer Assisted Language Learning, 24(5), 471-487.
17.Geer, D. (2005). Statistical machine translation gains respect. Computer, 38{ 10),18-21.
18.Goldsborough, R. (2009). The Si's and Nein's of computer-aided language translation. Education Digest: Essential Readings Condensed for Quick Review, 75(3), 62-63.
19.Jusoh, S. & Alfawareh, H. (2011). Automated translation machines: Challenges and a proposed solution. Proceedings of the 2tlcl International Conference on Intelligent Systems, Modelling and Simulation (ISMS), 77-82.
Арабистика Евразии, № 14, Июль 2021 36
АРАБИСТИКА ЕВРАЗИИ Y • У 1 jJjj И i tA^ljjVi ¡jjluiijjli
EURASIAN ARABIC STUDIES LBРАЗИЯ АРАБИСТИНАСЫ Eurasian Arabic Studies, № 14. July 2021
20.Kholidy, H. & Chatterjee, N. (2010). Towards developing an Arabic word alignment annotation tool with some Arabic alignment guidelines. Proceedings of the 10th International Conference on Intelligent Systems Design and Applications (ISDA), 778 -783.
21.Komeili, Z., Hendavalan, J. & Ali Rahimi, A. (2011). An investigation of the translation problems incurred by English-to-Persian machine translations: "Padideh, Pars, and Google softwares. Procedia - Social and Behavioral Sciences, 28, 1079-1082.
22.0ransa, W., Kouta, M. & Sakre, M. (2010). Injected linguistic tags to improve phrase-based SMT. The 2nd International Conference on Computer and Automation Engineering (ICCAE), 4, 329-333.
23.Pedersen, M.; Eades, D.; Amin, S. & Prakash, L. (2004). Parsing Arabic relative clauses: a Paninian dependency grammar approach. Proceedings of the 8th International Multitopic Conference (INMIC), 573-578.
24.Sheppard, F. (2011). Medical writing in English: The problem with Google Translate. La Presse medicals, 40(6), 565-566.
25.Stefansson, I. (2011). Oversdtt den hdr sidan: The advancement of Google Translate and how it performs in the online translation of compound and proper nouns from Swedish into English. Karlstad University. Retrieved from:
http://kau.divaportal.org/smash/record.jsf?pid=diva2%3A440909&dswid =-5735
26.Tillmann, C. & Zhang, T. (2008). An Online Relevant Set Algorithm for Statistical Machine Translation. IEEE Transactions on Audio, Speech, and Language Processing, 16(7), 1274- 1286.
27.Turchi, M., De Bie, T., Goutte, C. & Cristianini, N. (2012). Learning to translate: A statistical and computational analysis. Advances in Artificial Intelligence, 1-15.
28.Wang, L., Chang, B. & Harkness, J. (2010). A Method of automatic translation of words of multiple affixes in scientific literature. Proceedings of the International Conference on Asian Language Processing (IALP).241-244.
29.Williams, L. (2006). Web-Based machine translation as a tool for promoting electronic literacy and language awareness. Foreign Language Annals, 39{4), 565-578.
(Шл Арабистика Евразии, № 14, Июль 2021 37
АРАБИСТИКА ЕВРАЗИИ Y • \ 1 jjIjj И î AjjjiJl tllLjiljjIl
EURASIAN ARABIC STUDIES LBРАЗИЯ АРАБИСТИНАСЫ Eurasian Arabic Studies, № 14. July 2021
30.Wu C, Xia F, Deleger L & Solti I. (2011). Statistical machine translation for biomedical text: are we there yet? Proceedings of the AMIA Annual Symposium, 1290-9.
31.Wu, D. & He, D. (2010). A Study of query translation using Google machine translation system. Proceedings of the International Conference on Computational Intelligence and Software Engineering (CiSE), 1-4.
Information about the author
Ph.D., Professor Reima Saado Al-Jarf King Sand University Riyadh, Kingdom of Saudi Arabia reima.al.jarf@gmail.com ORCID ID: 0000-0002-6255-1305 Scopus Author ID: 14119546100 Web of Science ID AAL-3 778-2021
Информация об авторе
Ph.D., профессор Рейма Саадо Алъ-Джарф Университет Короля Сауда Эр-Рияд, Королевство Саудовская Аравия
reima.al.jarf@gmail.com ORCIDID: 0000-0002-6255-1305 Scopus Author ID: 14119546100 Web of Science ID AAL-3 778-2021
Conflicts of Interest Disclosure: The author declares Conflicts of Interest Disclosure.
Раскрытие информации о конфликте интересов: Автор заявляет об отсутствии конфликта интересов.
