JOURNAL OF NEW CENTURY INNOVATIONS
Volume–80_Issue-1_July-2025
228
228
ADVANCED TRANSLATION OF SCIENTIFIC TERMINOLOGY
USING COMPUTATIONAL LINGUISTICS: THEORY, PRACTICE,
AND LIMITATIONS
G‘afurova Nazokat Bakhriddin’s daughter
student of Tashkent State of transport university
Annotation:
This academic paper examines the translation of scientific
terminology through computational linguistic tools. Emphasizing theoretical insights
and practical comparisons, it evaluates the capabilities and limitations of machine
translation (MT), computer-assisted translation (CAT) tools, and terminological
databases in handling domain-specific vocabulary. Drawing on translation theory
(Newmark, 1988; Vinay & Darbelnet, 1958; Vermeer, 1989) and empirical examples,
it argues that while technology improves efficiency, human expertise remains essential
for semantic and contextual accuracy.
Keywords
: Computational linguistics, machine translation, CAT tools, scientific
terminology, semantic accuracy, translation theory.
Annotatsiya:
Ushbu maqolada ilmiy atamalarni tarjima qilishda kompyuter
lingvistikasi vositalaridan foydalanishning nazariy va amaliy jihatlari o‘rganiladi.
Mashinaviy tarjima (MT), tarjima xotirasi tizimlari (CAT), hamda terminologik
ma’lumotlar bazalarining imkoniyatlari va chegaralari tahlil qilinadi. Newmark, Vinay
va Darbelnet, hamda Vermeer nazariyalari asosida ilmiy atamalarni kontekstda to‘g‘ri
va adekvat tarjima qilish uchun zamonaviy yondashuvlar taklif qilinadi. Amaliy
misollar orqali tarjima vositalari o‘rtasidagi farqlar ko‘rsatilib, yuqori sifatli tarjima
uchun gibrid yondashuvlar tavsiya etiladi.
Kalit so‘zlar
: kompyuter lingvistikasi, ilmiy atamalar, mashinaviy tarjima,
tarjima nazariyasi, kontekstual muammolar, tarjima xotirasi.
Аннотация:
В данной статье рассматриваются теоретические и
практические аспекты использования средств компьютерной лингвистики при
переводе научной терминологии. Анализируются возможности и ограничения
машинного перевода (MT), инструментов компьютерной поддержки перевода
(CAT) и терминологических баз данных. На основе теорий Ньюмарка, Вине и
Дарбельне, а также Вермеера предлагаются современные подходы к адекватному
переводу терминов в контексте. Сравнительный анализ демонстрирует различия
между инструментами перевода, предлагая гибридные стратегии для повышения
качества перевода.
Ключевые слова
: компьютерная лингвистика, научная терминология,
машинный перевод, теория перевода, контекстуальные проблемы, память
перевода.
JOURNAL OF NEW CENTURY INNOVATIONS
Volume–80_Issue-1_July-2025
229
229
The globalization of science necessitates accurate and context-sensitive
translation of scientific terminology across languages. The complexity of domain-
specific vocabulary often presents challenges for both human and machine translators.
Computational linguistics has introduced powerful tools, but the quality of output
remains uneven, especially in fields requiring precision such as medicine, law, and
engineering (
Newmark, 1988
). This paper investigates the current capabilities of such
tools and offers a theoretically grounded approach to their integration into professional
workflows.
Translation theory has evolved to address not only linguistic transfer but also
contextual, pragmatic, and cultural dimensions.
Newmark (1988)
differentiates
between semantic translation (focused on meaning) and communicative translation
(focused on effect).
Vinay and Darbelnet (1958)
identify translation techniques such
as borrowing, calque, and modulation, essential for adapting technical terminology.
Vermeer’s Skopos theory (1989)
emphasizes the purpose of translation as central to
determining strategy, particularly in scientific communication where accuracy
outweighs stylistic preferences.
Classification of Computational Linguistic Tools
Scientific terminology translation increasingly depends on three categories of
tools:
Machine Translation (MT)
: e.g., Google Translate, DeepL
Computer-Assisted Translation (CAT)
: e.g., SDL Trados, MemoQ, Smartcat
Terminological Databases
: e.g., IATE, Termium, UNTerm
MT systems like Google Translate provide rapid outputs, but as
Hutchins (2005)
notes, they lack contextual nuance. CAT tools enhance consistency using translation
memory (TM) and terminology management. Databases such as IATE provide
validated equivalents to ensure terminological uniformity (
Bowker & Fisher, 2010
).
Main Challenges in Terminology Translation
Despite advancements, several challenges persist:
Polysemy
: Terms such as
“network”
vary in meaning across disciplines (e.g.,
IT vs. biology).
Cultural absence
: Concepts like
“peer review”
may lack direct equivalents in
certain cultures.
Semantic drift
: MT systems may dilute precise meaning (e.g.,
“antibiotic
resistance”
translated as
“qarshilik”
instead of
“rezistentlik”
).
Inconsistency
: Different tools yield varied outputs for identical terms (
Koehn,
2020
).
JOURNAL OF NEW CENTURY INNOVATIONS
Volume–80_Issue-1_July-2025
230
230
Practical Comparison of Tools
Term
Google translate
DeepL
SDL Trados
Cloud computing
Bulutli hisoblash
Bulut asosida
hisoblash
Bulutli
texnologiyalar
asosidagi hisoblash
Antibiotic
resistance
Antibiotik
qarshiligi
Antibiotikga
chidamlilik
Antibiotikga
rezistentlik
Emission control
system
Emissiya nazorati
tizimi
Emissiya
boshqaruv tizimi
Chiqarilayotgan
gazlarni boshqarish
tizimi
As shown in the table, SDL Trados—especially when supported by subject-
specific TMs—offers more nuanced and contextually accurate translations. The phrase
“Emission control system”
is particularly well-rendered in Trados, as it reflects the
terminology used in environmental and automotive documentation (
Bowker & Fisher,
2010
).
Analysis of Comparative Results
As demonstrated in the table, automated translation systems—particularly Google
Translate and DeepL—offer rapid and generally intelligible outputs; however, they fall
short in domains that demand terminological precision and contextual accuracy. This
observation aligns with
Newmark's (1988)
assertion that translation is not merely
linguistic substitution but involves complex semantic and pragmatic transfer. For
instance, the term
"antibiotic resistance"
, when rendered as
"qarshilik"
(resistance),
may suffice in general discourse but fails to convey the technical nuance intended by
"rezistentlik"
in Uzbek medical terminology.
Moreover, while DeepL occasionally offers context-sensitive suggestions
superior to Google Translate, it often exhibits a tendency toward
stylistic
simplification
or
semantic generalization
. A good example is the translation of
"cloud computing"
as
"bulut asosida hisoblash"
(computing based on clouds), which
is grammatically correct but not widely used or recognized in Uzbek technical
discourse. In contrast, SDL Trados translates the same term as
"bulutli texnologiyalar
asosidagi hisoblash"
, aligning with established professional usage. This reflects
Vinay
and Darbelnet’s (1958)
concepts of modulation and calque when appropriately applied
through domain-aware memory banks.
Notably, SDL Trados, supported by specialized translation memory (TM) and
integrated terminology management systems, produces highly
domain-specific and
semantically appropriate translations
. For example, the phrase
“emission control
system”
is rendered in Trados as
"chiqarilayotgan gazlarni boshqarish tizimi"
, which
JOURNAL OF NEW CENTURY INNOVATIONS
Volume–80_Issue-1_July-2025
231
231
fully captures the environmental and automotive regulatory context (
Bowker & Fisher,
2010
).
Another critical issue pertains to
cultural connotation
and the translation of
realia
—terms deeply embedded in the cultural and institutional fabric of the source
language. Phrases such as
"peer review"
or
"tenure track"
often lack direct equivalents
in Uzbek and must be handled via
descriptive translation
or
conceptual
reformulation
. According to
Vermeer’s Skopos Theory (1989)
, such cases demand a
purpose-driven strategy where the translator adapts the form to match the
communicative function in the target culture.
Furthermore, modern neural MT systems, including DeepL, employ
corpus-
based probabilistic algorithms
that may generate fluent yet
falsely confident
translations
—outputs that are grammatically accurate but semantically misleading.
This phenomenon reinforces
Koehn’s (2020)
argument that human post-editing
remains indispensable, especially in specialized scientific and technical
communication.
In summary, while computational tools are indispensable for scaling translation
productivity, their effectiveness is maximized only when used in conjunction with
human expertise, contextual sensitivity, and domain-specific knowledge.
Recommendations
to improve translation quality in scientific domains:
Incorporate
domain-specific translation memory
into CAT tools.
Cross-check terms using validated
terminological databases
like IATE or
Termium.
Combine
machine translation
for initial drafts with
human post-editing
for
accuracy (
Koehn, 2020
).
Encourage
translator training
in specialized domains to ensure contextual
awareness.
Computational tools significantly improve translation productivity. However,
they are not substitutes for the nuanced understanding of human translators. Translation
is not merely about word-for-word substitution but requires sensitivity to semantic,
pragmatic, and cultural context. A hybrid model—technology-assisted, human-
reviewed—emerges as the most effective strategy for translating scientific
terminology.
References:
Newmark, P. (1988).
A Textbook of Translation
. Prentice Hall. → cited in
Sections 1, 2, 4, and 6
Vinay, J.P. & Darbelnet, J. (1958).
Comparative Stylistics of French and
English
. → cited in Section 2
Vermeer, H.J. (1989).
Skopos and Commission in Translational Action
. → cited
in Section 2
JOURNAL OF NEW CENTURY INNOVATIONS
Volume–80_Issue-1_July-2025
232
232
Hutchins, J. (2005).
Computer-Assisted Translation: The State of the Art
. →
cited in Section 3
Bowker, L. & Fisher, D. (2010).
Computer-aided Translation Technology
.
University of Ottawa Press. → cited in Sections 3 and 5
Koehn, P. (2020).
Neural Machine Translation
. Cambridge University Press. →
cited in Sections 4 and 7
