By Bakampa Brian Baryaguma*
[Dip. Law (First
Class) – LDC; Cert. Oil & Gas – Mak; LLB (Hons) – Mak]
*Legal Assistant, Alma Associated
Advocates, P.O. Box 115280, Kampala.
Email: bakampasenior@gmail.com; Mobile:
+256753124713.
November 2025
ABSTRACT
Background and Objectives
This article
looks at the role of artificial intelligence (AI) in text translation. It studies
AI’s innovations in that realm, its problems and future expectations of it.
Part 1 of the
article is the introduction, defining key terms, indicates the relevancy and
need for translation and places AI in the scheme of things. Part 2 deals with
innovations introduced by AI in text translation. Part 3 addresses problems
caused by AI in the process of translating texts. Part 4 peeks into the future
of and with AI in text translation, analyzing industry expectations. Part 5
gives the conclusion.
Methodology/Approach
This is an analytical research involving an
examination of the role of artificial intelligence in translating texts.
The study is theoretical, relying on a qualitative approach, using a desk study
method of data collection from available literature.
Key Findings
AI
is more of an enabler than a disruptor and accordingly should be enabled – not
hobbled. AI-powered systems have flawless, despite impressive advancements. AI is creating new roles. Hence it is not necessarily
replacing human translators.
Relevance
and Implications
The
study is significant for equipping practitioners in the translation industry
and the general public with greater knowledge that helps guide and even
evaluate their decisions and actions. The knowledge attained clears or at the
very least dilutes doubts and suspicions against AI as a spoiler and disruptor
of people’s livelihoods.
Key words: Artificial intelligence; Foreign
language; Text translation; Machine learning.
____________________________________________________________
1.
Introduction
There are 7,159 languages used in the world today.[1]
Since there is no single global language, it follows by default that each of
these languages is foreign in places
or countries where it is neither native nor official. Wiktionary defines a
foreign language as, “A language that is not widely or officially spoken in a
particular place;”[2] or “A language that is not one's native tongue.”[3]
This is where the need for translation comes in handy so as to bridge the gap
between respective languages.
The author, Oni Samuel Boluwatife, observed that the ability
to bridge language gaps is critical to the functioning of today’s increasingly
interconnected world, necessitating a higher demand for translation; and that the
landscape of translation is changing rapidly from being the domain of skilled
human linguists who possess deep knowledge of languages and cultures, due to the
emergence and integration of Artificial Intelligence (AI).[4] Oni added that:
Within the realm
of translation, AI has revolutionized traditional practices by introducing
automation, machine learning, and real-time processing. The rise of AI-powered
translation tools—especially Neural Machine Translation (NMT)—has transformed
how individuals, corporations, and governments communicate across linguistic
boundaries. Services like Google Translate, Microsoft Translator, DeepL, and
Amazon Translate exemplify this shift, offering users quick and often surprisingly
accurate translations of text, speech, and even images.[5]
From Oni’s observations above, it is evident that AI enables
machine translation, by which AI
algorithms translate text and speech from one language to another, powered by
models that learn the relationships between languages. Otherwise, other than
translation, AI also generates content
like text, images or audio and human-like text for summaries, reports and
customer service responses, based on the data they were trained on.
But one may wonder or ask, what is AI? AI is a field of computer science focused on
creating machines that can perform tasks typically requiring human intelligence,
such as learning, reasoning and problem-solving.[6] It works by
analyzing vast amounts of data to recognize patterns, make decisions and adapt
to new information. AI does many things.
2.
Innovations of
AI in Text Translation
AI has revolutionized text translation primarily through the
development of neural machine
translation (NMT)[7] and large language models (LLMs).[8] These are
computer systems that understand the context of a sentence (not just individual
words), enabling more accurate,
contextually relevant and rapid translations, moving beyond older, less
sophisticated methods of word-for-word rules.
These systems/models enable computers to understand, interpret and generate
human language by using machine learning and deep learning to process text and
speech. This allows for applications (like language translation, sentiment
analysis, chatbots and virtual assistants), which automate tasks, extract
insights from unstructured data and create more intuitive user experiences. Oni Samuel Boluwatife stated that,
AI has also
opened the door to new modes of communication, including multimodal translation
(combining text, speech, and images) and real-time voice interpretation. These
capabilities are changing how people interact across languages, whether through
augmented reality applications, travel tools, or AI-powered assistants. In
professional settings, this means faster negotiations, better customer
interactions, and more agile international operations.[9]
The following are examples of key innovations brought by AI
in text translation:
A.
Improved Accuracy and Fluency
Because deep learning algorithms in AI
models use embeddings (i.e. sentences
each converted into numbers that represent meaning and context) that look at
entire phrases and the broader meaning,[10] they produce
translations with significantly higher accuracy, better handling of complex
sentence structures, idioms and nuances than previous technologies that often
translated word-for-word. Consider, for example, challenges one translator
faced in translating Nicolo Machiavelli’s book, The Prince, from Italian to English:[11]
In
translating ‘The Prince’ my aim has been to achieve at all costs an exact
literal rendering of the original, rather than
a fluent paraphrase adapted to the modern notions of style and expression.
Machiavelli was no facile phrasemonger; the conditions under which he wrote
obliged him to weigh every word; his themes were lofty, his substance grave,
his manner nobly plain and serious. ‘Quis eo fuit unquam in partiundis rebus,
in definiendis, in explanandis pressior?’ In ‘The Prince,’ it may be truly
said, there is reason assignable, not only for every word, but for the position
of every word. To an Englishman of Shakespeare’s time the translation of such a
treatise was in some ways a comparatively easy task, for in those times the
genius of the English more nearly resembled that of the Italian language; to the
Englishman of to-day it is not so simple. To take a single example: the word
‘intrattenere,’ employed by Machiavelli to indicate the policy adopted by the
Roman Senate towards the weaker states of Greece, would by an Elizabethan be
correctly rendered ‘entertain,’ and every contemporary reader would understand
what was meant by saying that ‘Rome entertained the Aetolians and the Achaeans
without augmenting their power.’ But to-day such a phrase would seem obsolete
and ambiguous, if not unmeaning: we are compelled to say that ‘Rome maintained
friendly relations with the Aetolians,’ etc., using four words to do the work
of one. I have tried to preserve
the pithy brevity of the Italian so far as was consistent with an absolute
fidelity to the sense. If the result be an occasional asperity I can only hope
that the reader, in his eagerness to reach the author’s meaning, may overlook
the roughness of the road that leads him to it.
In “using four words
to do the work of one” this translator was aspiring for a more accurate and fluent translation which AI technologies
seamlessly bring to the table today – compared to older rule-based
word-for-word or phrase-based translation methods that typically provide stiff,
literal or just weird translations that don’t make much sense.[12]
B.
Contextual Understanding and
Processing
AI-powered machine learning algorithms and systems think in
concepts instead of just words.[13] They are able to process and
analyze vast amounts of text and speech data to find patterns and meaning. In so doing, they analyze entire sentences, paragraphs and
even documents to capture the broader context, unlike older rule-based or
statistical methods that often translated word-for-word and were thus very
problematic. AI’s contextual understanding boosts and amplifies the efforts of
recent translators in overcoming word-for-word translation problems to get
translations that are more natural, fluent and coherent.
C.
Real-Time Translation
AI has made instantaneous
translation for text and speech widely accessible through various apps and
integrated platforms (like Google Translate and Microsoft Translator). This
facilitates seamless global communication in personal and professional
settings, including live captions for virtual meetings. According to Oni
Samuel Boluwatife, “Integration [of AI] with cloud platforms, mobile apps, and
communication tools like Zoom or Microsoft Teams has made AI translation more
accessible than ever, supporting real-time subtitling, live interpretation, and
on-the-fly document translation.[14]
D.
Continuous Learning and Adaptability
AI models analyze vast amounts of
data to learn and refine their translations over time, adapting to new
vocabulary, industry-specific terminology and evolving language trends. As said
by Oni
Samuel Boluwatife,
Many
[AI-powered tools and] systems now use adaptive learning
techniques that tailor translations based on user behavior or domain-specific
content. This means that translations become more accurate, relevant, and
personalized the more they are used. In addition, collaborative efforts—such as
community-contributed corrections and post-editing—feed into the learning
process, further enhancing the quality of machine-generated translations.[15]
E.
Customization and Specialization
Translation platforms now offer
customization options, allowing models to be trained on industry-specific data
(for example legal, medical, technical) to ensure high accuracy for specialized
content. Oni
Samuel Boluwatife says that modern platforms like DeepL and Amazon Translate
have capitalized on improvements made by AI in auxiliary language tools such as
spell checkers, grammar correctors, speech recognition and optical character
recognition to offer high-quality translations tailored to business, medical
and legal domains.[16]
F.
Enhanced Efficiency and Scalability
AI tools can process and translate
large volumes of text at speeds unmatched by human translators, significantly
reducing turnaround times and costs. AI translation is remarkably speedy,
able to translate thousands of words in a matter of seconds, unlike traditional
human translation that can be a time-consuming process, especially for lengthy
or complex texts.[17] This
scalability allows people to localize content for a global audience more
efficiently.
G.
Hybrid Models and Human Augmentation
AI is not replacing human
translators per se but rather changing their role. The most effective approach
now is a hybrid model where AI provides a rapid first draft and human
post-editors then refine the translation for cultural nuances, tone and
critical accuracy. This combined approach leverages the speed of AI with the
contextual intelligence of humans. Oni Samuel succinctly captures the point thus–
Despite
fears that AI may render human translators obsolete, many professionals are
finding new opportunities by adapting to the changing landscape. They are now
working alongside AI systems, enhancing the quality of machine-generated
translations through post-editing, curating translation memories, and ensuring
that cultural and linguistic integrity is maintained. This shift represents not
just a technological evolution, but also a professional and philosophical one,
in which the translator becomes a mediator between machines and meaning.[18]
H.
Support for Low-Resource Languages
Low-resource languages are those
with limited digital content or fewer native speakers.[19] Advanced
AI models, such as Meta's SeamlessM4T, are bridging the gap for languages with
limited digital content, for example Kiswahili, by leveraging large multilingual
datasets, making translation available for a broader range of the world's
languages.
3.
Problems of AI
in Text Translation
Despite
impressive advancements, AI-powered systems are not flawless.[20] AI does not think or reason (as humans
do), but simply makes predictions, which is fraught with many problems as
discussed below:
A.
Lack of Context and Cultural Understanding
AI struggles to grasp nuances like irony, metaphors and
cultural references, which can lead to awkward or completely wrong
translations. As Oni
Samuel Boluwatife noted, “They can translate literal
meaning effectively, but they struggle with subtext, irony, humor, sarcasm,
idioms, and metaphors.”[21] This “underscores the growing complexity
of language itself and the realization that perfect, fully autonomous
translation may remain an aspirational goal rather than a present reality.
Human language is rich with ambiguity, culture, and emotion—qualities
that are difficult to encode into algorithms, no matter how advanced.”[22]
B.
Data Bias
Since AI models are only as good as the data they are
trained on, they can reproduce and amplify biases and
stereotypes present in their training data, resulting in gender, linguistic, cultural
or other forms of insensitivity in translations.[23]
C.
Inaccuracy and "Hallucination"
Translations can contain subtle, costly errors even if they
sound correct and sometimes AI generates factually incorrect output that is
still grammatically sound. AI sounds confident even when its translations that
are just wrong.[24]
D.
Data Privacy and Security Concerns
Using cloud-based AI services can expose sensitive or
personal information to third parties without explicit consent, a risk that is
exacerbated by a lack of sufficient regulation or simply none at all.
E.
Difficulty with Specialized and Complex Texts
AI often causes errors in complex texts because it fails to
accurately translate documents with highly specialized terminology (like
medical or legal texts) or complex sentence structures, thus requiring human
review. AI
may mistranslate such terms or apply general vocabulary that alters the intended
meaning. An incorrect translation of a medical instruction could put a patient
at risk, while a misinterpreted legal clause could affect the outcome of a
court case.[25]
F.
Inconsistent Output
AI can struggle with words that have multiple meanings,
selecting the wrong one based on context and may not maintain the original tone
and style of the text.
G.
Dependence on Data Quality
The accuracy of AI translation is limited by the quality,
availability and recency of its training data, which can be scarce or outdated.
In such cases, machine translation systems simply do not have enough
information to go on to be accurate.[26]
For example AI
struggles with a lack of proficiency in less common languages.
H.
Need for Human Review
Despite advancements, AI-generated translations often
require human post-editing to ensure accuracy, quality and appropriateness.[27]
Thus,
while AI can enhance efficiency, it is not a complete substitute for
professional linguistic judgment.
4.
Prospects of AI
in Text Translation
AI presents transformative prospects in text translation by
offering unprecedented speed, efficiency and accessibility, primarily through
advanced NMT and LLM models, which tools are reshaping the field into a hybrid
model that leverages AI for initial drafts and large-scale content,
while human expertise remains crucial for nuanced, context-dependent and
sensitive materials. Current and future trends of AI in translation point
towards the following key prospects:
A.
Enhanced Speed and Efficiency
AI systems can translate vast volumes of text in seconds, a
task that would take human translators considerably longer. This speed is
invaluable for industries that require rapid communication, such as international
business, customer support and news reporting.
B.
Cost Reduction and Scalability
Automating the initial translation process significantly
reduces the cost per word compared to traditional human translation. AI tools reduce
the need for large teams of linguists by performing bulk translations
automatically and significantly cut down on labor time, thereby reducing
overall costs.[28] Cost
reduction and scalability makes multilingual content localization more
accessible for small businesses, non-profits and individuals and also enables
large corporations to scale their operations across dozens of languages
simultaneously.
C.
Improved Accuracy and Fluency (for general texts)
Modern AI, particularly NMT and transformer models like
those used by Google Translate and
DeepL, has dramatically improved translation quality, producing more natural
and contextually aware results than older methods.
D.
Real-time Communication
AI drives real-time translation tools for text, speech and
images, facilitating seamless communication during live events, virtual
meetings and international travel through apps and devices.
E.
Increased Accessibility
AI translation tools help bridge language gaps in education,
healthcare and government services, making information and resources available
to a broader multilingual audience. As Oni Samuel Boluwatife said, “AI has
democratized access to translation by making it available to a wider audience.
Free and low-cost tools like Google Translate, Microsoft Translator, and DeepL
are accessible via smartphones, browsers, and desktop applications, allowing
users to translate text, speech, and even images in real-time.”[29]
F.
New Professional Roles
Rather than replacing human translators, AI is creating new
roles, such as post-editors, AI model trainers and quality assurance experts,
who refine machine-generated content and manage complex, AI-assisted workflows.
G.
Deeper Contextual Understanding
Ongoing research aims to improve AI's ability to interpret
and integrate cultural and emotional nuances, idiomatic expressions and
figurative language, which remain current challenges.
H.
Multimodal Translation
Advancements will enable AI to translate not just text, but
also seamlessly integrate with augmented reality, virtual reality and advanced
speech-to-speech systems, potentially including lip-sync modifications for
video content.
I.
Specialized Domain Expertise
AI models will likely become more proficient in handling domain-specific
terminology (for example legal, medical, technical), requiring human experts to
specialize in these complex areas.
J.
Ethical Frameworks
The AI industry will need to address ethical concerns
regarding data privacy, potential biases in training data and accountability
for errors in critical translations.[30] Also, concerns about job displacement among professional
translators
sparked by automation of translation tasks have to be resolved. Ultimately, the most effective
approach is a human-in-the-loop model,
where AI serves as a powerful assistant to enhance productivity and human
translators provide the essential cultural sensitivity, critical thinking and
final quality control.
5.
Conclusion
It is said that,
“… the emergence of AI in translation is not merely a technological shift, but
a transformation … stand[ing] at the intersection of language and technology …”.[31]
I agree. Contrary to common perception as being a disruptor, AI is more of an
enabler; and accordingly should be enabled – not hobbled.
REFERENCES
1.
Ethnologue,
‘How many languages are there in the world?’ Ethnologue (2025). Accessed at https://www.ethnologue.com/insights/how-many-languages/,
on 20 November 2025, at 14:18 hrs GMT.
2.
Wiktionary, foreign language (2025). Accessed at https://en.wiktionary.org/wiki/foreign_language#:~:text=Noun,any%20experience%20of%20foreign%20languages,
on 20 November 2025, at 14:35 hrs GMT.
3.
Ibid.
4.
Oni
Samuel Boluwatife, ‘The Impact of AI on the Translation Industry’ ResearchGate (April 2025), at p. 3. Accessed
at https://www.researchgate.net/profile/Adams-Williams/publication/391050035_The_Impact_of_AI_on_the_Translation_Industry/links/6808f808bd3f1930dd633b43/The-Impact-of-AI-on-the-Translation-Industry.pdf?origin=publication_detail&_tp=eyJjb250ZXh0Ijp7ImZpcnN0UGFnZSI6InB1YmxpY2F0aW9uIiwicGFnZSI6InB1YmxpY2F0aW9uRG93bmxvYWQiLCJwcmV2aW91c1BhZ2UiOiJwdWJsaWNhdGlvbiJ9fQ&__cf_chl_tk=2YXBPEKtBeKtbcTEXqL3FR5StS3lhWn9a0isZpuyur0-1763491193-1.0.1.1-caL4jDp_IYbSayur9JlV7hHN34X4ZL5EtrjU5KFwjD4,
on 18 November 2025, at 21:40 hrs GMT.
5.
Ibid.
6.
Google
Cloud, ‘Artificial intelligence (AI): a
simple-to-understand guide’. Accessed at https://cloud.google.com/learn/what-is-artificial-intelligence,
on 16 November 2025, at 13:00 hrs GMT.
7.
According
to Mia Comic, ‘The simplest guide to neural machine translation’ lokalise
(2025). Accessed at https://lokalise.com/blog/neural-machine-translation/,
on 17 November 2025, at 08:30 hrs GMT, neural
machine translation is a type of AI that translates text from one language to
another. It learns how to translate by studying huge amounts of real-world
text, in multiple languages. NMT
processes entire sentences at once, considering the whole context to produce
more fluent and accurate translations, unlike older methods that break
sentences into smaller parts. It is the leading technology behind modern
translation services like Google
Translate, instant movie subtitles, translated news articles or public
documents or translated product descriptions from another language.
8.
Notable examples of large language models are like ChatGPT, Gemini and Claude. See Wikipedia,
‘Large language model’ (2025). Accessed at https://en.wikipedia.org/wiki/Large_language_model, on 17 November
2025, at 07:57 hrs GMT. They are advanced AI systems that train on massive amounts of
text data to understand, generate and manipulate human language using deep
learning techniques, particularly transformer architectures, to learn language
patterns and context, enabling them to perform tasks like answering questions,
summarizing text and translating languages.
9.
Oni
Samuel Boluwatife, supra note 4, at 9.
10.
Ibid.
11.
Nicolo Machiavelli, The Prince (1513), Planet PDF, at 17-19.
12.
Mia
Comic, supra note 7.
13.
Ibid.
14.
Oni
Samuel Boluwatife, supra note 4, at 6.
15.
Ibid., at 8-9.
16.
Ibid., at 6.
17.
Ibid., at 7.
18.
Ibid., at 4.
19.
Ibid., at 9.
20.
Ibid.
21.
Ibid., at 10.
22.
Ibid., at 6.
23.
Ibid., at 10.
24.
Mia
Comic, supra note 7.
25.
Oni
Samuel Boluwatife, supra note 4, at
10.
26.
Nick
Schäferhoff, ‘7 Typical Problems in Machine Translation (+ How to Solve Them)’ TranslatePress (2025). Accessed at https://translatepress.com/problems-in-machine-translation/,
on 20 November 2025, at 15:30 hrs GMT.
27.
Oni
Samuel Boluwatife, supra note 4, at 11.
28.
Ibid., at 7.
29.
Ibid., at 8.
30.
Ibid., at 11.
31. Ibid., at 5.
No comments:
Post a Comment