=Paper=
{{Paper
|id=Vol-2604/paper19
|storemode=property
|title=Going parallel: using earlier translations as background for facilitating re-translation technique
|pdfUrl=https://ceur-ws.org/Vol-2604/paper19.pdf
|volume=Vol-2604
|authors=Tetiana Anokhina,Iryna Kobyakova,Svitlana Shvachko
|dblpUrl=https://dblp.org/rec/conf/colins/AnokhinaKS20
}}
==Going parallel: using earlier translations as background for facilitating re-translation technique==
https://ceur-ws.org/Vol-2604/paper19.pdf
Going parallel: using earlier translations as background
for facilitating re-translation technique
Tetiana Anokhina1 [orcidID0000-0002-8859-5568], Iryna Kobyakova 2[0000-0002-9505-2502], Svitlana
Shvachko3[0000-0002-2119-1884]
1
National Dragomanov Pedagogical University, Kyiv, Ukraine
2
Sumy State University, Sumy, Ukraine
3
Sumy State University, Sumy
anokhina_mail@yahoo.com,kobyakova@ukr.net,shvachko.07@ukr.net
Abstract. As it goes for the re-translation technique, the modern data enable access
and analysis of the previous translations for further re-translations based upon the
previous empirical studies and earlier translations. Also, the students' re-translations
are a perfect illustrated material to compile the little educational corpora being com-
parable due to the slight differences which make up the material of translation lacunae
possible for analysis of the corpus-based studies. There are different tools for going
parallel, including ParaConc, SketchEngine, tools applicable to the students (re-
translation based on the previous translation) which illustrate their own ways to deal
with difficulties of translation and lacunae. The paper provides an overview of tools
such as AntConc for working with the re-translated words, MWUs and lacunae in
translation corpus called ReTRans, the corpus of translator's re-translations. It is sup-
posed to have texts of the original and published translation in order to build small
self-made corpus of re-translations.
Keywords: students' translations, Wordfast, SketchEngine, AntConc Tool, bitext, re-
translation technique.
1 Introduction
The corpus studies are applicable to many scientific spheres today. The linguistic
areas of applied linguistic and translation studies are actively interacting with corpus
linguistics giving new results approved statistically by the rich and overwhelming
corpus material.
Among many problems of translation studies, the problem of finding the best
matching equivalent is actual and popularized. To solve this problem and many others
the corpus material and corpus tools appear to be the helping hand for translators and
researchers.
The most popular modern tools for corpus analysis are concordancers which are ef-
fective tools for applied linguistics research and translation studies facilitating the
learning of vocabulary, learning how to use fixed and non-fixed collocations in for-
eign languages. Among popular tools that are worth mentioning for further observa-
tions and discussions is Wordfast.
Copyright © 2020 for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
The Wordfast family are great translation tools that have the corpus match func-
tioning. To prepare and use self-made parallel corpus the Wordfast aligner tool
(https://www.wordfast.net/?go=align) can assist. There is a tendency in modern lin-
guistics to have a corpus-based research. For the purpose of analysis he or she may
use some tools ready to work with raw material and other requiring preparatory work
for research and study.
If there are some txt files in the personal library, the AncConc tool
(https://www.laurenceanthony.net/software/antconc/) will be serving for the analysis.
If a researcher has at least the text of original and the text of translation the ParaConc
(https://paraconc.com/) tool can be the needful choice. The tool working with both
single units and multiword expressions might be helping to prepare the personal cor-
pus along with using the existing corpora is SketchEngine
(https://www.sketchengine.eu/).
Using the friendly interface of SketchEngine it is possible for linguists to compile
their own corpus without writing any code. This popular tool helps to extract the
searching items, one word or multiword units.
One can select well prepared annotated corpora in SketchEngine. It uses complete-
ly everything for compilation your own corpus (docs, pdfs, excel files, runs tmx files
for setting up your parallel corpus) but it doesn’t work with scans. Using SkectEngine
our lab has started to compile ReTrans corpus comprising re-translations of popular
and classical works made by students of our university.
Other tools serve differently but they all coincide in working with machine read-
able libraries, small and huge corpus data, give statistic information and may be used
for corpus-based research and translation studies. AntConc, ParaConc, Wordfast
aligner tool and SketchEngine are corpus managers and text analysis tools are able to
work with numerous texts and corpora in various languages. They can provide collo-
cation search and generate statistics related to the co-occurrences of collocations'
constituents. A successful computational treatment of one word or multiword expres-
sions leads to solving language ambiguity of the broad context. Word clusters and
bundles tools search help in better collocations learning. The shifts of meaning of
multiword units and idioms can be effectively learned relating the corpus-based ap-
proach to the interpreted data.
The idiom search is a effective in the corpus environment. We find the corpus tools
are the needful tools today for linguistic research and translation studies. This study
presents the study of corpus tools, their functioning and ability to analyze small or
huge corpus data.
In our observation the software tools have shown its output based on their proper-
ties analysis. The above mentioned tools have been tested to have concluded the opin-
ion of highly innovative tools with friendly interface serving for multifunctional ap-
proach and all worth trying and using.
2 The Ukrainian translation studies and corpus-based
lacunology
Today translation students should not only learn the foreign language, enhancing
writing, speaking, listening skills and improving their pronunciation and grammar [7].
It is the technological time when electronic texts and comparable corpora are ac-
cessible online; using the digitization forces translation students can include some
additional skills to their CV.
As Veiga (2016) states the students should be working in the highly competitive
profile of the translation market, to work with computer tools designed for linguistic
analysis and translation [12].
It is important to analyze corpus-based technologies to develop techniques for
identifying terms from a specialized field searching for them in internet and extracting
from the comparable corpora, as our students English :: Ukrainian.
In this observation, we find the developed corpus-based tools used for students
who are aiming to relate their research to the modern tendencies in applied and corpus
linguistics for going parallel in their translation analysis and for better extracting ele-
ments with multiple realizations in contrasting languages (English and Ukrainian).
There are a variety of corpus-based studies that are applicable to many scientific
spheres to confirm theories by the data available from corpora. The linguistic areas of
applied linguistic and translation studies are actively interacting with corpus linguis-
tics giving new results approved statistically by the rich and overwhelming corpus
material. Among many problems of translation studies, the problem of finding the
best matching equivalent is actual and popularized. To solve this problem and many
others the corpus material and corpus tools appear to be the helping hand for transla-
tors and researchers.
The most popular modern tools for corpus analysis are concordances which are ef-
fective tools for applied linguistics research and translation studies facilitating the
learning of vocabulary, learning how to use fixed and non-fixed collocations in for-
eign languages.
The modern linguistics go hand in hand with translation studies and corpus-based
lacunology, the science with studies rare frequent words or non-equivalent words in
other cultures as lacunae. These lacunae can be zero places in one language or diffi-
culties of translations that can have different realization in multiple re-translations.
The phenomenon of the lack of specific elements in the culture of one ethnic group
against another in English termed as gap. Also, lacunae can be the aligned to the eth-
nographic elements, phraseologic units, multi words in the parallel translations result-
ing in either deletion of lacunar element or rendering “with the phrase, not by a word”
[10].
We find the multi-word expressions (MWEs) to be one of the most heterogeneous
phenomenona in the fields of Natural Language Processing (NLP) [2]. Today it is
possible for find huge data comprising also MWEs in different corpora due to the
electronic revolution.
Alghamdi and Atwell (2017) emphasize that the best computational practice re-
lated to MWEs processing is connected with new techniques and tools arise both with
machine translation developing and corpus studies. The both constituents help to find
better extraction models and computational ways of study [2].
2.1 Lacuna as a unique phenomenon
Thus, lacuna is a unique phenomenon which mirrors zero reflection of non-equivalent
vocabulary. Lacunae are quasi-comparable units that can refer to various referents of
ethnic cultures. The comparable corpora return the lacunar queries in multiple varia-
tions. We trace lacunae as difficulties of translation.
There are many lacunae in the text of the original. To find some of them students
can analyze the footnotes which eliminate lacunae by explanations given in the end of
the text or in the mode “footnote at the page”. Nowadays the footnotes are electronic
which is very convenient for reading the electronic texts.
In order to work effectively with corpus material AntConc, ParaConc, Wordfast
aligner tool and SketchEngine can be used. These are corpus managers and text
analysis tools helping to work with numerous texts and corpora, for our purposes with
parallel corpora in English – Ukrainian pair. Using these tools students can find the
needful collocation searches and look for translation provided in the self-named cor-
pus.
The extracted translation equivalents alone and translation memory file obtained
after aligning translation pairs (bitext) can be used for re-translation purposes when
students make their translation based on the re-translated text which they add as trans-
lation memory files (using Wordfast software).
Successful re-translation files are making the small corpus of students' re-
translations (reTRans). This kind of activity is effective for learning unique transla-
tion pairs, helping students to find better results comparing their results with previ-
ously translated text and other variants in the reTRans corpus (the small self-made
corpus of students translation) [6] which contains the original and different translation
English – Ukrainian pairs. It is supposed to use SketchEngine tool in order to preload
small or bigger corpora and use for analysis in the SketcthEngine environment.
As SkecthEngine is the most powerful tool it is highly recommended to use it for
automatic extraction of needful words, lemmas and multiword units ("terms" in
SketchEngine). It is also has POS (part of speech) search which enables to navigate
and extract the needful data from other accessible corpora.
As students papers are corpus-oriented but deal with such linguistic material as ad-
verbs/nouns/verbs, it is possible to add their data into ReTrans project based on Harry
Potter series, Nebo, The Da Vinchi Code and other re-translations compiled from
their smaller corpora into bigger one.
The special attention has been paid to corpus based search of translation variants.
Due to the broad context search students can interpret the data. Some scholars claim
[9] the lexical unit is very often longer than a single word [4]. As McEnery and Wil-
son (2001) state that one of the major trends in corpus linguistics over the past few
years is the increased interest in very small, highly specialized corpora. Small corpora
can be used for a great many different purposes [9].
The date of translation difficulties comprise single words (glossaries), multiword units
(terms) and ideas hidden implicitly. To follow the latter task the translation corpus has
been compiled to see how translation tasks are performed at the level of word, word
unit (N-grams) and at the semantic level (meaning that the secondary text feels the
most successful translation. Also, the needful phrase for translation may be found
online or in the downloaded corpus by the idiom search.
As Pierre Colson (2017) believes that automatic extraction is possible from large
web corpora in different languages. The use of a new algorithm based on metric clus-
tering techniques made it possible to find long N-grams from the freely accessible
web [1].
We find the translation corpora are giving more information that corpus-based dic-
tionaries or glossaries of terms. If we are talking about re-translation corpus it is rich
in contextual variants and can be effectively used for learning EFL purposes and for
learning how to translate better, how to find more translation variants. For semi-
automatic analysis the translation pairs are downloaded into separate files containing
the unique translation making the subcorpus of the larger corpus of students' re-
translations (ReTrans) (Figure 1).
Fig. 1 The SketchEngine interface of the translation corpora used for re-translation
We use the corpus tools in order to work more effectively with translation corpora
(parallel corpora). The bellow mentioned needful tools serve the crucial role in trans-
lation studies. Our observation aims at describing the corpus tools for “going paral-
lel”, their functioning and ability to analyze small or huge corpus data. In our article
we have mentioned the most popular software tools along with their output based on
properties analysis.
There are plenty of tools to be tested with highly friendly interface serving for mul-
tifunctional purposes and worth trying and using. Among them there are some free
and prepaid tools. Firstly, we regard free tools available for students with any paying
abilities. The highly innovative but prepaid tools such as TRADOS and SkecthEngine
are regarded also for being currently the best as the market place.
3 Corpus Tools observation
It is highly important for the Ukrainian students to learn how they can make auto-
matically the alignment of parallel texts (English and Ukrainian).
As a rule, in a translation class they make parallel texts when translating from
Ukrainian and into Ukrainian. The problem is that there is only OPUS corpus avail-
able for parallel comparison which is not enough for educational purposes, so new
translation corpora should arise.
We find it is to use comparable corpora from other languages into Ukrainian and
English. This application area (the alignment of parallel texts from multilingual cor-
pora) is highly important today. That is why we are making the small corpus of stu-
dents translations based on the existing translations.
It may be considered helpful to use not only aligned 'translation pairs' from English
– Ukrainian but also from other languages with parallel component (English or
Ukrainian). Students had to find a txt file in English and in Ukrainian or scan and
send the language pair to txt files (a very time consuming task).
If a researcher has at least the text of original and the text of translation the Para-
Conc tool can be the needful choice. Other software tools have the option to include
the parallel corpora, such as SketchEngine. Wordfast tool generates translation mem-
ory files that can be used in other tools supporting the format.
Lacunae as difficult places to render in translation can be vividly observed in mul-
tiple re-translations performed by students. Lacunae thus will be found as aligned to
the expanding explanation, footnote or omission or contextual change. We find all
rare frequent units in the contrasting languages to be considered as lacunae. If the unit
is very rare (less than 1 lemma entry per corpus) it is considered to be a hapax (law
frequent unit/lacuna).
Lacunae, which are difficult to translate, are found in the contrast analysis of artis-
tic translations and their re-translated versions, since they provide more complete
information about the removal of lacunar elements in one or more versions of the
translation (lacunar units) and contain extra text, preface, afterword, etc.
There are some resources available on-line, including free corpora of translated
works are not systematized by the degree of extra-linguistic additions, commentary,
or preface (Paraconc concordance; Ukrainian Language Corps; National corps of the
Russian language).
Linguists and programmers are involved in creating independent databases, scan-
ning and sorting data according to their tasks. There are open source platforms avail-
able for development [http://opus.nlpl.eu/].
Hapaxes (rare units or lacunae) are not usually included in traditional enclosures,
as these bases are more useful for improving the work of electronic translators, for
whom it is important to have contextually dynamic correspondences.
The minimum retrains corpus must contain n-versions including the original, its
translation and re-translations (Table 1).
Table 1.
Original Translation Re-translation
Theft: a love story Translation from English Ukrainian
Harry Potter series Translation from English Ukrainian
The Da Vinci Code Translation from English Ukrainian
Alice in Wonderland Translation from English Ukrainian
Surely You're Joking, Mr. Translation from English Ukrainian
Feynman!
3.1 AntConc as concordancer tool and text analyzer for the ReTRans corpus
We find the AntConc tool efficient for multiple re-translations versions analysis, as
we had put the texts of translations in the manner text 1-n to analyse and compare re-
translation results compiling the small comparable corpus of the printed and students'
re-translations ReTRans.
The central tool used in most corpus analysis and translation studies we strongly
recommend AntConc as your first concordancer. The texts of re-translations (even not
annotated) give the information about difficulties of translation. The texts of re-
translations can be downloaded to AntConc tool (each file named 1…n) and the diffi-
cult place (usually has a footnote in translation) can be quickly found in the other text
translated by another student. It is the way how is the best variant is found.
This method can be used for effective assessment of students' translation (the fig-
ures1…n are used instead naming the file by the students' name).
AntConc can be used for this purpose. It is simple to use, it has a friendly interface
and it is free of charge. All these make it to be the best choice corpus manager tool to
work with raw texts in your compiled corpus of texts. Being used for research in
translation and computational linguistics AntConc concordancer as Sun and Wang
(2003) describe is a tool for learning vocabulary, collocations, and multiword units
[4].
Also, if there is some or files in the personal library, the AntConc tool will be serv-
ing for the analysis working directly on the raw texts of the corpus. The idea to work
with separate words is not working any more in linguistics and translation studies
[13]. Word for word translation is the past of machine translation which is relying on
editing and corpus data extraction (as Wordfast tools).
In AntConc, multiword units can be investigated using the Word Clusters Tool. As
Lawrence (2004) states:
This tool displays clusters of words that surround a search frequency. The
search term can be specified as a substring, word, phrase or regular expres-
sion as in the Concordancer, Plot and View File tools, and the number of
additional words to the left and right of the search term can also be speci-
fied. It is also possible to set a minimum frequency threshold for the clusters
generated [13].
An alternative way to sear bundles [3], which are equivalent to N-grams, where n
can vary usually between two and five words. Few corpus analysis programs offer
this feature [5], but AntConc includes lexical bundle searches as an option in the
Word Clusters Tool [13].
Lonfils and Vanparys (2001) explain how the AntConc works like a little but effec-
tive 'ant' as its logo shows, with most essential features of standard software applica-
tions to work with words in the contexts and set expressions, stable or free multiword
units (MWEs) [8].
The needful function is bundles tool effective for translation studies of the un-
known MWEs or word clusters, foreign collocations difficult for learners to acquire.
Having one text or a book, the smallest corpus, the researcher can easily extract the
searched item in the nearest context with enable its proper usage.
The program let us see the bundles two and five words) equivalents [3].
AntConc performs all operations directly on the raw texts of the corpus. This is
useful in that the user is often switching or modifying the target corpus for a particular
need, as the program does not need to do any pre-processing of the data, for example,
creating an index. On the other hand, because AntConc does not use an index, it can
only work effectively with small scale corpora [13].
Most corpus analysis programs offer users the ability to see collocates of a search
term in a table, where the frequency of the most common words to the left or right of
the search term is indicated. Learners often find such tables difficult to interpret and
so the current version of AntConc offers no implementation of this feature. However,
for advanced learners this can be a severe disadvantage that will be addressed in the
next release of the program.
Students may use AntConc program [13] with statistics to run in the selected cor-
pus, working both wit annotated and non-annotated data search results. Young learn-
ers have the helping advice using simple keyboard shortcuts which they can follow,
namely: working with statistics just to copy and paste your results into a spreadsheet
program for further analysis and research. The program enables to see the bundles as
two, three, four or five words equivalents also called N-grams. Also, this free tool has
offline concordancing feature.
AntConc supports different formats of annotated data and non-annotated data. It
runs raw text, also it works with text in HTML/XML format with possibility to view
or hide embedded tags used in HTML/XML [13] Student do not use all the possibili-
ties at one time but as they get interested in what they can do with annotated corpora,
they are trying to use their knowledge in their Bachelor' and Master' papers.
4 Translators' engines and CATs
Some free tools as ZExtractor are used for the automatic extraction of term candidates
in a text or set of texts. The paid for Ukraine is a perfect program SketchEngine which
is a powerful linguistic data analysis system that has several features. Among them
there is the Keywords feature to identify words in a text or set of texts.
In order to covert files into text formats, Notepad for Windows can be used. Also,
in order to compile data we Microsoft Excel allowing filtering, calculate, compare
and analyze data.
All these programs are helping young researchers to compile their own corpus
(small or huge) for research or translation purposes, for instance for compiling the
specialized dictionaries to use them by machine translation tools such as TRADOS,
WORDFAST or other tools [12].
The popular tools worth mentioning and further observations and discussions are
Wordfast family. The Wordfast tools are great translation tools that have the corpus
match functioning. To prepare and use self-made parallel corpus the Wordfast aligner
tool can assist. There is a tendency in modem linguistics to have a corpus-based re-
search. It is the tool we are using to align our ReTRans corpus of students re-
translations of the popular and classical masterpieces.
For the purpose of analysis translation students may use some tools ready to work
with raw material and other requiring preparatory work for research and study. Short
pieces of text (2 to 160 pages) are aligned.
4.1 The translation experiment
The mentioned above tools serve an educational purpose. The re-translation data is
easily accessible for research and analysis. In class students are to compare what was
done in translations and using the experience of many other translators in order to
make the re-translation copy readable. Their practice includes editing the segments of
the text, using machine translation, and creating the memory based on the existing
translation. For educational purposes they translate the previously translated pieces.
Wordfast anywhere tool that can be used freely gives students to work with one
memory at the time. It is important for students to find the text file in English and
Ukrainian to align (usually it is classical masterpiece freely downloadable) or to buy
e-text and its translation for the compilation translation work.
When this material is ready students are making their translation try (from one
page to chapter) to compare and analyze with the original. The next stage is compiling
MWUs to MWUs corpus [11].
5 Final remarks
In our studies we rely on the new parallel corpora based mostly on classical prose and
best-sellers translated into English.
Step by step students are translating masterpieces into Ukrainian, making the small
corpora of their secondary translations with the possibility to compare with the previ-
ous translations from the languages they can comprehend into Ukrainian.
The translations available from English and into English are making the huge cor-
pora potential as they include worldwide classic and modern popular prose translated
into English.
The idea is to facilitate their translation using the previous translation. Using the
technological tools such as AntConc, SketchEngine as corpus tool and translation
software Wordfast/Trados they align the translation pair into the translation memory
before they translate the text making the new re-translation with their own footnotes
and translation lacunae elimination.
References
1. Colson, J.-P.: The IdiomSearch Experiment: Extracting Phraseology from a Probabilistic
Network of Constructions. Europhras, LNAI 10596, 16-28 (2017)
2. Alghamdi, A., Atwell, E.: Towards Comprehensive Computational Representations of
Arabic Multiword Expressions. Europhras, LNAI 10596, 415-431(2017)
3. Biber, D., Johansson, S., Leech, G., Conrad, S., & Finegan, E.: Longman grammar of spo-
ken and written English. Longman, London (1999)
4. Bowker, L., Pearson, J.:Working with Specialized Language: A Practical Guide to Using
Corpora.Routledge, London / New York (2002)
5. Coniam, D.: Concordancing oneself: Constructing individual textual profiles. International
Journal of Corpus Linguistics (2), 271-298 (2004)
6. Krajka, J.: Language Teachers Working with Text: Increasing Target Language Awareness
of Student Teachers with Do-It-Yourself Corpus Research. In: Working with Text and
Around Text in Foreign Language Environments. Springer, Cham (2016)
7. Lee, D., Swales, J.: A corpus-based EAP course for NNS doctoral students: Moving from
available specialized corpora to self-compiled corpora. English for Specific Purposes
\textbf{25}, 56-75 (2006)
8. Lonfils, C. and Vanparys, J.: How to design user-friendly CALL interfaces. Computer As-
sisted Language Learning (5), 405-417 (2001)
9. Mcenery, T. and Wilson, A.: Corpus Linguistics. An Introduction. Second edition. Edin-
burgh Press, Edinburgh (2001)
10. Markovina, I., Sorokin, Y. Kul'tura y tekst. Vvedenye v lakunolohyyu: ucheb. posobye
[Culture and text: the textbook]. GEOTAR-Media, Moscow (2010)
11. Osenova, P., Simov, K.: Modelling multiword expressions in a parallel Bulgarian-English
newsmedia corpus. In: Multiword expressions. Insights from a multi-lingual perspective
(Phraseology and Multiword Expressions 1).Manfred Sailer & Stella Markantonatou
(eds.), 247-271 (2018)
12. Veiga, A. Using corpus linguistics tools to help translation students create technical glos-
saries. In: ICICTE Proceedings, 341-347 (2016)
13. Lawrence, A. AntConc: Design and Development of a Freeware Corpus Analysis Toolkit
for the Technical Writing Classroom. In: IEEE International Professional Communication
Conference Proceedings, 729-737 (2003)
14. Sun, Y.-C., Wang, L.-Y. (2003). Concordancers in the EFL classroom: Cognitive ap-
proaches and collocation difficulty. In: Computer Assisted Language Learning, 16, 83-9