International Journal of Language Education Volume 8.
Number 4, 2024, pp.
ISSN: 2548-8457 (Prin.
2548-8465 (Onlin.
Doi: https://doi.
org/10.
26858/ijole.
Assessing AI-Powered Translation Quality: Insights from the Translation of the Farewell Sermon Amal Abdelsattar Metwally King Khalid University.
Saudi Arabia Email: aamotwly@kku.
Wagdi Rashad Ali Bin-Hady Hadhramout University.
Yemen Email: wagdyrashad@hu.
Eisa Asiri King Khalid University.
Saudi Arabia Email: ealeisa@kku.
Received: 16 May 2024 Reviewed: 22 July 2024 Accepted: 24 October 2024 Published: 30 December 2024 Abstract Artificial Intelligence (AI) powered translation influxes in the translation industry.
Yet, questions have arisen about the quality of AI-powered translation systems (AIPTS.
Therefore, this study aims to evaluate the translation quality generated by ChatGPT for the Farewell Sermon.
Using House's .
model of translation quality assessment (TQA), the study analyzed the Farewell Sermon focusing on text, register analysis of field, tenor mode, and genre.
The language/text analysis revealed that AIPTSs produced lexical and syntactic inaccuracies that hindered capturing the sermon's intended religious voice.
In register.
Field, the analysis pointed out that the AIPTSs captured the broad themes of the Farewell Sermon, but often missed subtleties in religious terminology.
AIPTSs also adopted superficial literal translation of the sentence structures.
In Tenor, the analysis revealed that the Prophet's role as a moral and spiritual guide was somewhat diluted in the translation.
Likewise.
AIPTSs struggled to maintain the Prophet's authoritative and compassionate stance.
The AIPTSs maintained general coherence but faced challenges with cohesive devices such as conjunctions, pronouns, and references.
Furthermore, the microanalysis revealed an amount of 81 overt translation errors that the AIPTSs committed while translating the Farewell Sermon into English, the most frequent were "creative translation" with 24 errors, "not translated" with 15 errors, "distortion of meaning" with 9 errors, "slight change in meaning" with 8 errors and "breach of the SL system" with 7 errors.
"Significant change in meaning" and "cultural filtering" were less noticeable with 6 and 4 errors, respectively.
Additionally, the analysis revealed that House's .
model is to some extent suitable for assessing the quality of AIPTSs, one type of error was technical and beyond the model focus The researchers called it "software intervention".
Software interventions can either positively or negatively affect the translation depending on whether the error improves readability without distorting Vol.
No.
4, 2024
Metwally.
Bin-Hady, & Asiri meaning or inaccuracies.
The study recommends the importance of tuning the AIPTSs applications to prevent the inducing of technical systems.
Keywords: Early AIPTSs.
cultural and religious terms.
Farewell Sermon.
translation quality assessment Introduction AIPTSs have approached the translation industry by processing huge content (Valente.
As the AIPTSs are not error-free, researchers are motivated to focus on the challenges that resulted and related to imprecision, biases, and translation ethics (Valente, 2024.
Zaki & Ahmed.
Many models are used for assessing the translation quality, the most comprehensive is HouseAos .
which is based on Halliday's systematic functional theory (Alghamdi, 2016.
Hakami.
Youssef, 2.
Previous studies applied HouseAos .
7, 2.
model to evaluate the translation quality of political texts (Hakami, 2.
, literary texts (Al-Aizari, 2.
even religious texts between Arabic and English (Alghamdi, 2016.
Khoshafah, 2.
Using HouseAos model for evaluating AIPTSs has not yet been examined.
Translating Islamic teachings to the many Muslims worldwide is necessary (Ahmed, 2.
We in this study orient the translations between Arabic and English language pairs.
It is widely accepted that Arabic and English are rooted in different grammatical and stylistic features, yet this variation makes the process of conveying meaning between the two languages complex (Al-Kenai.
The complexity becomes more vital when a machine does the task of translation.
response to this complexity, scholars like Halahla .
, and Zabadi .
recommended translators employ strategies and footnotes to explain unfamiliar terms.
Previous studies focusing on translating the Prophet's (PBUH) teachings are extremely uncommon.
A study analyzed the translation of deixis in Shahih Al-Bukhari (Pratiwi, 2.
AlGhamdi .
examined the translation quality of AuThe Sealed NectarAy focusing on religious and cultural The researchers highlighted evaluating one of the prophetic teachings, i.
Khutbatul-WadA.
AuThe Farewell SermonAy from Arabic into English due to its fundamental principle for Muslims and for being not analyzed yet.
The Arabic concept.
Khutbatul-WadA, has been rendered into the AuFarewell SermonAy (Farooq, 2.
, and AuFarewell OrationAy (Ahmed, 2.
The Farewell Sermon has been the topic of many studies.
Farooq .
probed the significance of the Farewell sermon on Islamic Syed et al.
examined the Farewell Sermon from a critical discourse analysis perspective showing the ideology and the textuality of the Prophet PBHU.
Nevertheless, minimal studies assessed the translation quality of the Farewell Sermon.
Ahmed .
compared three translations of the Farewell Sermon to check the variation among them.
This shortage encourages the researchers to examine the quality of AIPTSs to translate cultural and religious terms of the Farewell Sermon from Arabic into English in the Farewell Sermon.
This study is designed to examine the quality and accuracy of AIPTSs, focusing on their ability to convey the core message of culturally and religiously significant texts.
By using the Farewell Sermon as a case study, this analysis seeks to uncover specific challenges in AI translations, particularly where misinterpretations could distort the meaning or impact of the original message.
This study seeks to address the following questions:
To what extent do AIPTSs handle cultural and religious elements within the Farewell Sermon? What overt translation errors are more prevalent while translating the Farewell Sermon into English using AIPTSs? Is HouseAos .
model suitable for assessing AIPTSs? Vol.
No.
4, 2024
International Journal of Language Education Literature review The translation of Classical Arabic presents numerous challenges due to the language's unique structure, rich vocabulary, and deep cultural and historical significance (Akan et al.
, 2.
This review of literature explores the complexities involved in translating Classical Arabic texts into English, highlighting the linguistic and cultural obstacles faced by translators.
Additionally, the review refers to Islamic religious texts and discourse, examining the specific challenges inherent in translating these religion and culture-rich materials.
Furthermore, it discusses the development of AI technology and its application in this domain, focusing on both its evolution and the difficulties it encounters when translating sacred Islamic Arabic texts, which require a deep understanding of context and cultural nuances to maintain their integrity and meaning.
Moreover.
House's .
model is examined in the conceptual framework, highlighting its significance in assessing translation quality in sensitive texts, particularly Islamic religious discourse.
Translating classical Arabic into English Classical Arabic, with its intricate grammar, stylistic features, and use of metaphor, often resists direct equivalence in English, making it difficult to preserve the original meaning and aesthetic value (Awad, 2.
Furthermore, the cultural and historical context embedded within Classical Arabic texts requires translators to have not only linguistic proficiency but also a deep understanding of the historical and cultural nuances (Al-Amer et al.
, 2.
to provide accurate and contextually appropriate translations.
Due to the inherent linguistic and cultural differences between Arabic .
source languag.
and target languages like English, translating Arabic discourse, especially Classical Arabic, presents several difficulties (Abdul-Raof, 2001.
Akan et , 2.
These challenges highlight the need for translators to employ strategies and adopt a culturally sensitive approach to maintain the authenticity and depth of the original text (AbdulRaof, 2.
Moreover.
Arabic stylistics is characterized by syndetic coordination, inflectional endings, ellipsis, flexible word order, repetition, preposing, and postposing, which other languages cannot replicate (Al-Jurjani, 2.
Additionally, beyond rhetoric, translators face obstacles with cultural references, including ecological elements, material culture .
, food and clothin.
, social practices, and religious Abdul-Raof .
and other scholars (Akmadhan, 2002/2014.
Gazalah, 2002/2.
highlighted how early translators, lacking familiarity with Arabic culture, misinterpreted the names of religious groups mentioned in the QurAoan, such as translating AuSabeansAy /AacUA/ inaccurately as those who worship angels or change their religion, and AuMagiansAy /AA/ as pagans or gentiles.
Translators inevitably face the complex challenge of navigating between two distinct discourse worlds when translating Islamic religious texts into Western languages.
The task is complicated by the need to bridge ideological, social, political, economic, and cultural differences, all deeply embedded within Islamic teachings (Al-Mohannadi, 2.
Translators must therefore adopt culturally sensitive and contextually informed strategies to convey the multifaceted nature of Islamic discourse accurately.
Translating Islamic religious discourse Translating Islamic religious discourse involves unique challenges due to the complex interplay of cultural, linguistic, and religious elements embedded within the texts.
Moreover, specific features of Islamic discourse, such as emotively charged language, cultural references, and religious terminology .
, terms like "A "Afor prayer or "A "aAfor jiha.
, add further complexity (AlGhamdi, 2.
These terms often carry connotations and implications that do not Vol.
No.
4, 2024
Metwally.
Bin-Hady, & Asiri directly translate into English, making it difficult to convey their meaning without additional explanation or context .
Additionally.
Islamic discourse heavily employs aesthetic elements, including metaphors, figures of speech, and phonetic and semantic devices, which are often rooted in the Quran and Hadith .
Prophetic Traditio.
These elements are difficult to translate while preserving their original impact and meaning (AlGhamdi, 2.
Highlighting the challenges faced by translators when translating Islamic discourse.
Holt .
studied Islamic religious discourse and its He compared the translated book AoMilestonesAo to its Arabic source text.
A"Ua eA "AEnA, and found that the translation is overwhelmed by Arabicness .
using Arabic words and expression.
indicating that this foreignising strategy is the dominant approach for this text.
Likewise.
Zabadi .
examined the translation of 17 Friday Sermons, revealing that translating religious texts is a challenging task.
The study showed that Islamic discourse is filled with emotive expressions, culture-specific religious features, structural and semantic repetition, aesthetic aspects, and linguistic and semantic voids due to differences between the two languages in phonetics, morphology, and syntax.
Similarly.
Halahla .
explored the challenges of translating religious discourse from Arabic into English and identified several factors that often lead to translation loss, including Arabic stylistics, the 'emphatic negation' produced by restricted expression, and the difficulty in interpreting rhetorical features like proverbs, metaphors, and These factors can lead to a loss of stylistic effect, as seen in the translation of a sentence like "A " u u Ainto "There is no God except Allah.
" Furthermore.
Khoshafah .
assessed the quality of the translation by Pickthall of Surat Yunus into English.
The study applied HouseAos .
model comparing the two texts' register levels .
ield, tenor, and mod.
The study revealed that despite the great efforts of the translation to safeguard the poetic quality of the original text, lexical and syntactic differences were noticeable between the two texts.
The study concluded by stating that the richness and depth of the Holy Quran could not be captured by any translation.
AI-powered tools in Islamic texts translation AI offers promising tools for translating Islamic texts by addressing the challenges of classical Arabic complex structure.
Atwell et al.
discussed the potential of AI and Natural Language Processing (NLP) for the translation and analysis of religious texts.
In this research, the Quranic Arabic Corpus, a comprehensive annotated resource of the Quran that includes morphological tagging, syntactic parsing, and efforts to build a deep semantic understanding of the text was introduced.
This project aimed to enhance the understanding of the Quran's linguistic and grammatical structures to serve as a foundational step toward AI-assisted translation efforts.
These resources could facilitate accurate and nuanced translation of the Quran and other religious texts by capturing complex linguistic features and subtle meanings encoded in classical Arabic.
Likewise.
Zaid and Bennoudi .
explored the growing impact of AI tools, such as ChatGPT, in handling complex religious texts, and compared their performance against human translators.
Through qualitative and comparative analysis of seven religious texts, they assessed translation quality in terms of structure, word choice, readability, and accuracy.
The study found that while AI tools provided efficient translations, they failed to match the depth and cultural sensitivity of human translation.
Translating farewell sermon into English The Farewell Sermon has been the topic of dozens of studies.
These studies investigated this topic from different perspectives.
KhAn et al.
studied the Farewell Sermon from a Vol.
No.
4, 2024
International Journal of Language Education critical discourse analysis perspective.
The study revealed that Farewell Sermon composes the fundamental principle of Islamic teaching which criminalized racism, racial discrimination and inequality among people.
Furthermore.
Ullah .
investigated the historical impact of the Farewell Sermon on societies.
The study affirmed that Farewell Sermon is incomparable in its ruling for human welfare and liberating human beings.
The study also confirmed the immortal implications of the Sermon for contemporary societies.
When it comes to translation, the Farewell Sermon has been studied by very few studies.
Ahmed .
in a comparative study analysed the translation of the Farewell Sermon by three translations focusing on the complexities stemming from old and new challenges which affected the translators to perform the translation task.
The findings indicated that two translators succeeded in conveying the translation by using scholarly techniques while the third was not due to the influence of her ideology, which prioritized adapting the texts for a global audience over preserving their original meaning.
No single study to the researchers' knowledge evaluates the AIPTS of the Farewell Sermon.
This makes the study at hand worth investigating.
Conceptual framework Translating religious texts requires some theories and models that set the necessity of including both linguistic and cultural contexts.
Zabadi .
and Halahla .
emphasized the need for translators to be thoroughly acquainted with both Islamic and Western cultural contexts to bridge these gaps effectively.
Alghamdi .
adopted HouseAos model for assessing the quality of cultural-specific terms in the AoSealed NuctaraAo.
Alqefari .
reported that Aotext-type theoryAo is more suitable for analyzing the rendering of religious texts.
On the contrary, studies also focused on assessing the quality of AIPTSs.
Early research focused on the accuracy of AIPTSs by identifying the semantic mismatching between the two language pairs (Koponen, 2.
HouseAos .
model of TQA is widely used in translation studies.
The model compared the original text and the translated text under three categories, i.
, .
ield, tenor, and mod.
analyzed the translation at three levels: words, syntactic relations, and textual meanings of the text (House, 2.
House proposed two terminologies for evaluating the translation quality.
The first is Auovert translationAy which refers to the strangeness of the translation and the domination of SL.
The second term is Aucovert translationAy where the translation focuses on the communication message of the SL to convey to the target language (Hakami, 2.
This model has been applied by various researchers who assessed the Arabic-English translation (Al-Aizari, 2023.
Alghamdi, 2016.
Hakami.
Youssef, 2.
Al-Aizari .
and Algamadi .
applied the model to assess the translation quality of literary texts.
Algamadi .
is among the few studies we came across which applied HourseAos .
model in assessing the translation quality of prophetic religious texts.
Hakami .
used it for assessing political Despite the comprehensiveness of House's .
model for assessing translation (Hakami.
Youssef, 2.
, it is not free from faults (Al-Aizari, 2023.
Algamadi, 2.
For instance.
Al-Aizari .
found that some overt errors could not be classified according to the seven criteria suggested by HouseAos model.
Research method Research design A corpus design was used to study the AIPTSs register accuracy of the Farewell Sermon due to its profound religious, and cultural significance in Islamic history.
The sermon, as such, its universal message serves as a criterion for Islamic moral values and legal tenets, making it a critical Vol.
No.
4, 2024
Metwally.
Bin-Hady, & Asiri text for assessing the precision and accuracy of translation towards detecting errors in translation and providing possible recommendations.
Sample and data collection The Farewell Sermon has been chosen as a significant text to test the capabilities of AIPTSs.
This compilation of sermons was delivered by the Prophet Muhammad PBUH during his pilgrimage in 10 Ah across various locations including Mina.
Muzdalifa, and Arafat, addressing over 100,000 believers observing the Hajj.
Furthermore, the Farewell Sermon is characterized by its Classical Arabic source language and dense religious vocabulary (Farooq, 2017, p.
, which presents intrinsic translation challenges.
The sermonAos entangled language structure and culturally specific expressions, such as references to concepts like taqwa .
God-consciousnes.
, riba .
, and Jahiliyyah .
the pre-Islamic er.
, require precise and culturally informed translation for proper interpretation for non-Arabic speakers.
Thus, this text offers a rich and intricate corpus to evaluate whether AIPTSs can handle such nuanced, sacred religious content while preserving its original meaning (Farooq, 2017, p.
The corpus is composed of 1335 The source text is Arabic comprising 498 words and extracted from http://hrlibrary.
edu/arab/IS-4.
html (Bassiouni, 2.
while the English corpus includes 837 words built by translating the ST using ChatGPT 4 to study the translation of cultural and religious Data analysis In evaluating the quality of AIPTSs, particularly in the context of the specified religious text, namely The Farewell Sermon.
House's .
TQA paradigm is employed.
This paradigm incorporates its emphasis on functional equivalence and cultural sensitivity between the source and target texts, ensuring that the meaning, tone, and purpose of the original text are preserved in translation Figure 1:
Creating ST/TT Profiles Defining ST Function Comparing ST and TT Error Identification Classification Figure 1.
The systematic four-step evaluation process developed and adopted in this study The analysis involves creating detailed profiles of both ST and TT to determine their respective textual functions.
This includes defining the intended function of the original Arabic sermon and then comparing it to the profile of the AI-powered English translation.
Additionally, the researchers systematically identify, categorize, and evaluate errors within the translation to assess its quality and accuracy.
This model involves the analysis and comparison of an original text and its translation on three distinct levels, i.
Language/text, register including .
ield, tenor, and mod.
, and genre.
HouseAos .
model utilizes the concepts of field, tenor, and mode to analyze the interplay between text and context, thereby providing a comprehensive framework for assessing translation quality.
Results and discussion This section examines the AIPTSs of the Farewell Sermon through HouseAos .
TQA This analysis follows a systematic four-step evaluation process.
Vol.
No.
4, 2024
International Journal of Language Education RQ1: To what extent do AIPTSs handle cultural and religious elements within the Farewell Sermon? Table 1 offers a comparison of the TT produced by AI and the ST, examining the key components of the register .
ield, tenor, and mod.
and genre as outlined in House's model.
highlights how certain aspects in the translation may diverge from the original text.
Table 1.
Texts' register analysis of the Farewell Sermon in terms of cultural and contextual depth.
Register Field Criteria Source text Target text Subject Matter Sermon .
radition and Sermon spirituality versus modernis.
Modernism Social Action General and Popular General and Popular Tenor Author's Farewell Sermon by Prophet AIPTSs (ChatGPT) Provenance Mohammed (PBUH) and Stance Social Role Asymmetrical Asymmetrical Relationship Social Attitude Formal and intimate Intimate Mode Medium Complex Complex Participation Complex Monologic Genre Sermon Table 1 provides a structured comparison of the ST and TT profiles based on House's TQA model, examining key elements including text/language, register and genre.
Within the register dimension, the field shows both texts as "General and Popular," with the subject matter being a however, the source text integrates elements of "tradition and spirituality" versus "modernism," which the AIPTSs may not fully reflect due to its reliance on contemporary language For tenor, the AIPTSs maintain an "asymmetrical" social role relationship, aligning with the authoritative tone of a sermon.
Nevertheless, there is a shift in social attitude, as the target text's tone becomes more "intimate" rather than the "formal and intimate" stance of the source text.
Mode is described as "Complex" for the source text and "Monologic" for the target text, highlighting the significant linguistic and contextual complexity, especially given the religious text's cultural and spiritual importance.
Lastly, in genre, while both maintain the sermonic form, the TTAos modern interpretation may impact its spiritual resonance, and the cultural fidelity required to convey the sermon's original depth and purpose.
As indicated in Table 1, the analysis follows a systematic evaluation across three dimensions: text/language, register, and genre.
Each dimension contributes a specific lens to assessing the AIPTSs' accuracy and cultural sensitivity.
Text/language The text/language dimension focuses on analyzing the original and translated texts at the linguistic level, including vocabulary, syntax and grammatical structures.
This analysis focuses on how the AIPTSs capture subtle meanings, particularly when translating culturally bound expressions and religious terminologies.
For instance, a term like "A "UiAwas inaccurately translated as "annulled," which lacks the intended force of AuabolishedAy in English, impacting the textAos overall meaning.
Another issue in the syntactic structure is represented in the failure to convey conditional forms.
This is clear in the translation of the ST " AeOI eU eOI I u I UAA A "OO e AaOAinto AuIf they do, then Allah has given you permission to isolate them in their Vol.
No.
4, 2024
Metwally.
Bin-Hady, & Asiri The TT translates "A( "IAhas permitte.
as "has given you permission," making the structure verbose and indirect.
Furthermore, the TT lacks the specific conditional tone that defines the original phrase.
A more concise and accurate rendering would be "If they do so.
Allah permits you to confine them to their rooms," which preserves the conditional structure and succinctness of the original.
The language/text dimension analysis shows that literal and syntactic inaccuracies hinder the AIPTSsAo ability to capture the sermonAos intended religious voice, clarity, and cultural weight of the text.
To sum up, the AIPTSs often failed to capture subtle meanings, particularly when translating culturally bound expressions and religious terminologies.
These findings align with Zaid and Bennoudi .
who pointed out that despite the AIPTSs' efficiency in translations, they lack the capability of humans to capture the depth of cultural sensitivity.
Register The register dimension, incorporating field, tenor, and mode, examines the relationship between the text and its social and communicative context.
More precisely, this dimension investigates the relationship between the Sermon text and context, specifically how the AI translation preserves the textAos social functions, relational tone, and intended communication form.
Field The field dimension explores the textAos subject matter, content, and the main social actions conveyed in the sermon.
The main themes of the Farewell sermon are presented.
The main themes of the Farewell Sermon include advising people to obey Allah, the sanctity of blood, property, and honour, the annulment of pre-Islamic usury and revenge, the rights and duties of women, brotherhood among believers, holding fast to the Quran and Sunnah, the unity of humanity and equality, and matters of inheritance.
The analysis reveals that the AIPTSs captured the broad themes of the Farewell Sermon, including obedience to Allah and adherence to Islam values, social justice and equality, and moral conduct, but often missed subtleties in religious terminology.
For example, the culturally specific term "taqwa" (God-consciousnes.
was translated as Aupiety,Ay which lacks the full depth intended in the sermon.
The AIPTSs also adopted superficial literal translation of the sentence structures which produce ill-formed sentences for the target audience.
Lexical differences The AIPTSs have committed some violations while translating some lexical items without paying attention to context.
The Arabic word AuA AyUiAin the original text indicates a little different connotation.
In the example below.
AuA AyOuI ca ai Ahas been translated into AuAnd verily, the usury of the pre-Islamic era is annulledAy where the word in question AuA AyAis rendered into AuannulledAy.
This rendering does not convey the intended meaning.
The word AuannulledAy connotates Auto reduceAy, while the intended meaning is to Auuproot or stopAy, therefore, the most suitable is AuabolishedAy as mentioned by some translations of the Farewell Sermon by Ismail K.
Poonawala .
The same word AuA AyUiAas mentioned in AuA AyOuI a ai UiAhas been translated by ChatGPT as AuannulledAy.
The context in the sermon referred to stopping avenge on bloodshed in the pre-Islamic days.
Therefore, the best accurate translation is AuunavengedAy as suggested by Poonawala .
This inconsistency in lexical representation affects the translationAos quality, as the English version does not fully communicate the sermon's religious and moral depth.
Vol.
No.
4, 2024
International Journal of Language Education Table 2.
Lexical differences Source text Translation Gross The usury of the pre-Islamic period is A OuI ca ai AAnd verily, the usury of the pre-Islamic era is annulled.
The blood feuds of the preAll bloodshed in the pre-Islamic days is to be left unavenged.
A OuI a ai UiAIslamic era are annulled.
Syntactic differences Similarly, the syntax of the source text and the AIPTSs is compared to identify whether there is mismatching in the rendering.
An example is as follows:
Table 3.
Syntactic differences Source text Translation A OuIAUA OuI a ai UiAThe first blood we start with is the A OE I e c I Ua c cUiAblood of Amir ibn Rabia ibn Harith A c a c Ue EaAibn Abdul-Muttalib.
O people, indeed Satan has A a cU a a uI EaI Adespaired that he will be UA U I Ue e u NAworshipped in this land of yours.
Gross All bloodshed in the pre-Islamic days is to be left unavenged.
The first such claim I revoke is that of Ibn Rabiah.
O people, indeed Satan despairs of ever being worshipped in this land of yours.
In the above examples (Table .
, the AIPTSs could not convey the message.
it blindly depended on the superficial literal translation of the sentence AuAAyOuI OE I e c I Ua c cUiA, the translation is as follow Authe first blood we start with is the blood of Amir ibn RabiaAy.
It is not clear what is meant by Austart withAy, this results in misunderstanding to the readers.
The suitable translation is the one which conveys the message to the readers.
A suggestion may be that Authe first such claim I revokeAy could be understood by the readers as it referred to what is meant by the first blood in the sermon.
Another instance of mismatching between the source and the AIPTSs is the rendering of AuA AyuI EaI U I Ue e u NAwhere the AI added that close to the sentence as in AuSatan has despaired that he will be worshipped in this land of yoursAy.
This style brings redundancy to the readers.
A suggestion is to use Auof ever being worshippedAy which is an acceptable style and widely used in the English language.
Tenor In examining the tenor, the researchers assessed whether the AIPTSs effectively communicated the tone of the sermon and the Prophet's authoritative and compassionate stance.
As such, this category considers the speakerAos social role and emotional tone, examining the relationship between Prophet Muhammad PBUH as the speaker and his followers as the audience.
Tenor focuses on two dimensions: the author's provenance and stance and the social relationship.
In the ST, the authorAos provenance and stance is a Prophet.
In contrast, the provenance stance of the TT is ChatGPT, a generative artificial intelligence (AI) chatbot, released in 2022 by OpenAI The ProphetAos tone in the original sermon is formal yet compassionate, aimed at instilling reverence and moral discipline.
However, the AI translation occasionally produced a neutral tone Vol.
No.
4, 2024
Metwally.
Bin-Hady, & Asiri that diluted the emotional resonance.
For example, "A( "a aAO peopl.
, used to establish an intimate connection, was translated in ways that lacked the full force of a spiritual leader addressing his followers.
Additionally, the emotional weight of phrases like " AOu Uea coA A( "AI advise you.
O servants of Allah, to fear Alla.
was softened, affecting the intended emotional Social relationship In assessing the AIPTSs of the sermon, it seems that the Prophet's role as a moral and spiritual guide was somewhat diluted in the translation, where certain terms lost their emphatic and connotative impact.
For instance, phrases denoting commands or moral obligations in the original text were rendered more neutrally, potentially reducing the emotional engagement intended for the TT audience.
The tenor explores the relationship between the participants.
In the Farewell Sermon, the tenor displays the relationship between the prophet Mohammed PBUH addressing over 100,000 believers observing the Hajj.
However, the Prophet PBUH is a leader, in the Farewell Sermon, the social relationship seems to be "asymmetrical" showing the believers what is for their benefit in this life and the hereafter.
These were shown through the use of illocutionary forces as in, "AAu AoOu Uea co AI advise you.
O servants of Allah, to fear AllahAy.
AuAAya a U c uA.
AuO people, listen to me attentivelyAy.
This translation emphasizes the importance of listening but loses the emotional depth embedded in "A "c uAwhich is not translated.
This phrase which was ignored in the AI translation conveys the ProphetAos PBUH earnest concern and his profound commitment to guiding humanity toward the Straight Path.
However, a violation of the role and rights between husbands and wives violates the organization.
In the ST, the Prophet PBUH started pointing out the rights of wives upon their husbands, while the TT initiated the rights of husbands over their wives as in AuO people, indeed you have rights over your women, and they have rights over you,Ay Ay AAya a uI aUu Uu oa U Ou U nA.
Level Speaker -listeners Asymmetric Table 4.
Social relationship
Function
GLOSS
Illocutionary force A a a uI aUuAO people, indeed Asserting A Uu oa U Ou UAyour wives have A nArights on you and you have rights over them.
O people, indeed you have rights over your women, and they have rights over you.
Social attitudes The last component of Tenor, a register component, is the social attitude.
The language of the SL is formal, classical, and idiomatic.
The Prophet PBUH delivered the farewell Sermon to more than one hundred thousand believers to be a social guide for them and whosoever come after them in this life.
The language used is characterized by intimacy which arouses the emotion of the Similarly, the language adopted by the AI-powered system is formal.
As such, the analysis of the AIPTSs of the Farewell Sermon revealed several discrepancies in conveying the tenor, tone, and emotional impact of the original text.
It was found that the AIPTSs struggled to maintain the ProphetAos authoritative and compassionate stance.
Furthermore, specific connotations were softened, as seen in the translation of "A( "eaiAa grave moral breac.
as a "lewd act," which diminished the term's severity in the Islamic context.
Additionally, key phrases such as "A( "NAwe praise Hi.
and "A( "uAbu.
were left untranslated which indicates the AIPTSs inability to Vol.
No.
4, 2024
International Journal of Language Education accurately represent or stay fully true to the original text.
To sum up, the analysis shows that the Prophet's role as a moral and spiritual guide was somewhat diluted in the translation.
Likewise.
AIPTSs struggled to maintain the ProphetAos authoritative and compassionate stance.
Mode .
edium & connectivit.
Mode evaluates the medium and textual connectivity, focusing on cohesion and coherence.
The Farewell Sermon is a speech delivered in spoken mode by the Prophet Mohammed PBHU as farewell guidance and provisions for Muslims.
The ST has been well-structured and shows high The transition from one sentence to another goes smoothly.
Moreover, the punctuation system in the written version of this sermon is crucial for accurately reflecting the original spoken form, especially given that it was originally delivered orally by the Prophet Muhammad PBUH.
Since oral speech relies on pauses, intonation, and emphasis to convey meaning and emotional weight, written punctuation in the translation serves as a vital tool to approximate these spoken features.
The AIPTSs maintained general coherence but faced challenges with cohesive devices such as conjunctions, pronouns, and references.
In the phrase "A eaA.
A" canA, punctuation, such as the ellipsis (".
"), conveys a pause that reflects the Prophet's appeal for reflection and the weight of his words.
This phrase, being translated as "Have I conveyed the message? O Allah, bear witness," would lose much of its impact without the pause indicated by the ellipsis.
In the original Arabic, this pause invites listeners to consider the significance of the message, creating a moment of formal spiritual observation before invoking Allah as a witness.
Genre dimension: religious sermon characteristics While the register categories .
ield, tenor, and mod.
capture the relationship between text and micro-context, genre helps in understanding the textAos macro-context within its linguistic and cultural community.
The genre dimension considers the Farewell Sermon as a religious text, demanding faithfulness to both its spiritual tone and culturally embedded messages.
As a genre, religious sermons require precise handling of metaphorical language, cultural references, and an understanding of connotations unique to the text's community.
The AIPTSs, however, struggled to maintain these genre-specific attributes.
(ST) .
AOuI Cu ai Ui a i Ooai OU Oe U a o caUa O Oe aUi cUA The deeds of ignorance .
are fabricated, except for the protection of the Kaaba, providing water, and acts of intentional murder punishable by retaliation equivalent to killing one hundred (AIPT) For instance, the translation of "A "i OoaiAas "protection of the Kaaba, providing water" fails to capture the culturally and religiously significant connotations of these terms.
In Arabic, "A "iAsignifies a sacred custodianship of the Kaaba, rooted in honour, piety, and ancestral duty, while "A "oaiArefers to the ceremonial responsibility of ensuring pilgrims are provided with sustenance during their rites.
These terms embody themes of collective duty and reverence for sacred traditions.
The simplified AI translation reduces them to functional tasks, losing their symbolic depth and communal significance.
This example highlights AIAos difficulty in maintaining genre fidelity in religious texts, where cultural context and deeper meanings are essential.
Vol.
No.
4, 2024
Metwally.
Bin-Hady, & Asiri Furthermore, the poetic quality of the Arabic text, marked by rhythmic repetition, emphatic expressions, and a natural flow that reinforces key messages, is diminished in the AI translation.
In Arabic, repetition is more than a stylistic device.
it emphasizes core teachings, reinforces the messageAos urgency, and facilitates memorization and reflection.
For instance, the repetitive structure in phrases like "A eaA.
A " canAtranslated into AoHave I conveyed the message? O Allah, bear witnessAo creates a rhythmic resonance that intensifies the Prophet's appeal.
The language/text analysis revealed that AIPTSs produced lexical and syntactic inaccuracies which hindered to capture of the sermon's intended religious voice.
These findings revealed that AIPTSs are still in need of development and it is not wise to depend on them entirely for performing translations.
These findings echoed with Zaid and Bennoudi .
who confirmed that AI tools struggle in rendering cultural sensitivities.
This result ensures translators to discard the worry that AI will replace them.
This idea is confirmed by Zaid and Bennoudi .
who showed that human translation still dominates translating complex ideas and nuances.
In register, the analysis pointed out that the AIPTSs captured the broad themes of the Farewell Sermon, but often missed subtleties in religious terminology.
This finding echoes Albalawi and Ghafour .
who pointed out that AI-based translation sufficiently interprets the basis of the Qur'anic texts, yet it could not capture the depth of meaning with the text layers.
Furthermore, in Tenor, the analysis revealed that the Prophet's role as a moral and spiritual guide does not maintain the Prophet's authoritative and compassionate stance.
This contradicts House .
who reported the necessity of safeguarding the source texts' purpose while rendering.
sum up, the AI translation did not consistently maintain these genre-specific attributes, reducing the nature of the text's religious and cultural fidelity.
RQ2: What overt translation errors are more prevalent while translating the Farewell Sermon into English using AIPTSs? The error frequency as shown in Table 5 highlights the varied types of errors encountered in the AIPTSs, illustrating the complexity and challenges inherent in translating a culturally significant text.
Table 5.
Frequency of overt errors produced by the AIPTSs Error Type Frequency Not translated Creative translation Distortion of meaning Slight change in meaning Significant change in meaning Cultural filtering Breach of the SL system Software intervention (Positiv.
Software intervention (Negativ.
Total The analysis highlights that 81 overt errors were detected, categorized as shown in Table The most frequent error type was AuCreative translation,Ay with 24 occurrences, indicating a tendency for the AIPTSs system to take liberties with the original text, often resulting in translations that deviate from a strictly literal interpretation.
This error type suggests that the AI Vol.
No.
4, 2024
International Journal of Language Education occasionally prioritizes fluency or readability over precise equivalence, which may lead to altered meanings that impact the text's cultural or functional integrity.
The following most frequent error type is AuNot translated,Ay with 15 occurrences which indicates the AI system's limitations in fully translating complex phrases and religious concepts.
Other prominent issues included distortion of meaning .
Moreover, 7 errors were found due to not adhering to the syntactic, grammatical, or linguistic rules unique to the source language .
breach of the SL syste.
Furthermore, instances of change in meaning ranged from slight change of meaning .
to significant change of meaning .
detected case.
Lacking the ability to recognize and preserve culturally specific meanings, expressions, or references, the AIPTSs have committed 4 cultural filtering errors.
The overt errors identified in the AI following HouseAos TQA model .
along with the technical errors detected by the researchers and called Aosoftware interventionAo either positive or negative are identified below:
AoNot translatedAo error occurs when portions of the original text are completely omitted in the translation, resulting in an incomplete rendition of the source message.
For instance, the sentence Aa cU a a U c u eO O U oau cU Ua e i A.
was translated into "Now then.
O people, listen to me attentively, for I do not know if I will meet you again after this year in this place of mine," lacks an equivalent in the translated Arabic text.
With 15 instances of omitted segments, this recurring error significantly reduces the completeness and flow of the translation, especially problematic in speeches or messages where each line may convey important instructions or sentiments.
Creative translations occur when the translator translates the ST somehow freely by adding some extra words/ information which did not exist in the original ST.
Creative translations, occurring 24 times, employ stylistic choices to enhance readability or resonate better with the target audience.
For example, "The Arabic phrase UAOuI aI o uUo I n aO OA to imply that "time has returned to its original form," is translated into " Indeed, time has completed its cycle like the day Allah created the heavens and the earth".
This AI translation creatively conveys the cyclical notion in a way that resonates with English-speaking audiences, capturing the undertone of the original.
However, this creative approach may slightly shift the tone by adding a sense of finality that isnAot explicitly present in Arabic.
Distortion of Meaning refers to those mistakes which result in complete distortion of the meaning of the ST.
It further involves a significant misrepresentation of the original intent, often causing misunderstanding.
For instance, translating "A "Oe aUi cUAin AOuI Cu ai Ui aA Ai Ooai OU Oe U a o caUa O Oe aUi cUAinto Aoequivalent to killing one hundred camelsAo.
This translation distorts the meaning by inaccurately interpreting the cultural and legal context of "A( "aUi cUAliterally "one hundred camels").
In the original context, "one hundred camels" refers to the traditional compensation or "blood money" for certain offenses, not an equivalence in severity to "killing one hundred camels.
" Occurring 9 times, such distortions can alter the meaning, especially problematic in religious or legal contexts where precise language is necessary to avoid As for the Aoslight change in meaningAo error, the translation keeps most of the original message intact but alters nuances that subtly shift the intended meaning.
This means that there is a little distortion of meaning, partial transference of meaning or not complete faithfulness to ST.
but this change in meaning is not so severe as to impair communication.
An example is the phrase "A Ao Ou Uea Atranslated into AoI advise you.
O servants of AllahAo.
While the translation retains the general sense of guidance, the Arabic phrase "A "OuAcarries a tone closer to "I counsel you" or "I enjoin upon you," which implies a stronger, more formal encouragement rather than mere advice.
Vol.
No.
4, 2024
Metwally.
Bin-Hady, & Asiri Occurring 8 times, these subtle shifts may influence the readerAos interpretation of moral implications, affecting their understanding of the original concepts.
Significant change in meaning materializes when there is a big difference between the ST and the TT.
Significant change in meaning involves a substantial shift from the original intent, often altering the core message.
For example, " AOO e AaOAin AeOI eU eOI I u IA A UA OO e AaO OAc ca U a eAis translated into isolate them in their rooms.
This translation deviated from the intended meaning adds an element of physical separation beyond the original context and implies confinement, which isn't present in the Arabic text.
With 6 occurrences, these shifts affect the impact of the text, particularly in legal or moral statements where precise interpretation is essential.
Cultural Filtering is committed when there are some cultural phrases, words or local names and titles, which are untranslatable.
In such cases, usually, the translator tries to find some alternative equivalents according to target culture and intended readers.
It also involves omitting or modifying concepts that could be misinterpreted by the target audience.
In the example " AeA AU cUO uae U A cUAu a cUA," which is translated literally as " So do not turn back as disbelievers, striking the necks of one another," the phrase "striking the necks of one another" is an idiomatic expression in Arabic, symbolizing severe conflict or even violence.
While this adjustment appears 4 times, which can aid readability, it risks obscuring cultural context or altering the intended moral weight of the message.
Breach of the SL system: this category is recognized when the TT has deviated from the norms, syntax or grammatical rules of the ST.
It involves failing to maintain the structural integrity of the source language, often leading to changes in syntax or parts of speech and sometimes leading to distortion of meaning.
For example.
A a cU a a uI aUu Uu oa U Ou U nAtranslated into 'O people, indeed you have rights over your women, and they have rights over you' resulted in a different structure.
This translation alters the original sentence structure, where the Arabic construction emphasizes mutual rights with a specific sequence.
This error occurred in 7 instances throughout the translation, causing sometimes shift from action-oriented verbs to static nouns .
translating the verb "A( "NAwe prais.
, into a nou.
which disrupts the dynamism of the original, affecting the tone and possibly altering the reader's perception of the original intent or sentiment.
In the analysis of the translation, another type of fault, termed a "software intervention," was identified, with two subcategories: positive and negative which will be discussed under RQ3.
Figure 2.
Frequency of errors Accordingly, it is apparent that the most prevalent overt translation errors in the Farewell Sermon when translated via AIPTSs include creative translations, not translated .
, and distortion of meaning.
Omissions occur when culturally dense phrases are left out entirely while distortions happen when religious or cultural meanings are misinterpreted, such as in the example Vol.
No.
4, 2024
International Journal of Language Education where "one hundred camels" is mistranslated as a severe consequence rather than a legal The findings of this study reported that AIPTSs can handle some cultural and religious elements within the Farewell Sermon with reasonable accuracy but often struggle with nuances, especially when dealing with culturally specific terms and religious expressions.
These findings are supported by some studies, .
Al-Kaabi et al.
, 2024.
Taghian, 2024.
Zaid & Bennoudi, 2.
Taghian .
for example found that AI-based tools fail to capture the pragmatic, contextual, and cultural dimensions in translating two words succession in the Prophet's Zaid and Bennoudi .
pointed out that AI provided unprecedented accuracy in translation, but some challenges hinder the system in translating cultural and ideological language To them, such challenges should be characterized to enhance the AIPTSs' performance.
Similarly.
Al-Kaabi et al.
reported that AI translation suffered in translating cultural, idiomatic, and rhetorical language elements while translating literary work.
Their findings recommend the necessity of training translators on techniques for developing AI-powered RQ3: Is HouseAos .
model suitable for assessing AIPTSs? House's .
TQA model is found to be to some extent suitable for assessing AIPTSs of culturally and religiously significant texts.
Many previous studies reported the capability of House's .
model in assessing the translation.
This is evident by the many studies which applied this model.
they described it as a comprehensive model (Al-Aizari, 2023.
Alghamdi, 2016.
Hakami.
Youssef, 2.
However, previous studies focused on the suitability of House's model in assessing human translation, this study applies the model for assessing AIPTSs.
The model's emphasis on situational and cultural adequacy aligns well with the shortcomings identified in AI translations, such as the breach of the SL system, cultural filtering and the varying degrees of misinterpretation .
slight change in meaning, significant change in meaning, distortion of meanin.
, creative translation and instances where no translation is provided.
By assessing both the communicative function and the linguistic form.
House's model provides a comprehensive framework to evaluate how effectively AI systems meet the complex demands of translating religious and cultural texts.
Nevertheless, a modification to the model could improve its ability to handle all errors by adding a "software intervention" category to assess technical issues that may positively or negatively impact translation quality.
The positive software intervention, occurring 4 times in the corpus, refers to instances where additions are made.
This is frequent in adding AoverilyAo to enhance emphasis.
On the contrary, negative software interventions, also occurring 4 times, involve alterations that inadvertently distort the accuracy of the translated text.
Examples include inserting "That is the correct religion" in the translation of " AOuI U U ua U U e uoaA A a cUi IAUA I n aO OA.
" which is not present in the original, and misusing punctuation, as in placing a question mark after "O Allah, bear witness" ("A)" can eaA.
These findings were found to agree with previous studies (Khoshafah, 2024.
AlGhamdi, 2.
which support the view that HouseAos model can be applied to the translation of religious texts.
AIPTSs failed to match the depth and cultural impact of human translation (Zaid & Bennoudi, 2.
Altogether, these results emphasize the need for enhanced AIPTSs to better address cultural and contextual nuances, particularly for texts rich in cultural or religious Statement of quality The total word count for the text is 1,335 words, with the original Arabic version comprising 498 words and the English translation expanding to 837 words.
This increase in word Vol.
No.
4, 2024
Metwally.
Bin-Hady, & Asiri count in the English version reflects the challenges of translating from Arabic, where additional words are often required to capture the full meaning and nuances of the original.
The AIPSs have some violations in translating religious-based texts.
In this study, some violations were noticeable in rendering lexical terms where the AIPSs did not pay any attention to the TT contexts.
This finding indicated the AIPSs are still not fully prepared to select the most suitable equivalent of lexical items at various contexts.
Moreover, the findings agree with Al-AizariAos .
findings regarding the limitations of House's .
model when applied to AIPTSs.
While the model provides a comprehensive framework for evaluating human translations, it lacks the flexibility to fully capture certain types of errors specific to AI translations, particularly in handling overt errors and nuances that AI systems struggle with.
Similar to Al-AizariAos observations, we found that some AI-specific errors, identified by the researchers as "software interventions," which could not be effectively classified within the model's established criteria, highlighting the need for an adapted framework that better accounts for the unique characteristics of AI translation outputs.
Conclusion AIPTSs fall short in capturing the sermon's full cultural and emotional depth, diminishing its aesthetic richness and spiritual influence in the TT.
While AI provides a foundational translation, human post-translation and refinement are necessary for achieving a nuanced, cultural awareness, and contextually faithful rendering of the significant religious text, the Farewell Sermon.
The analysis indicated that House's .
model is effective for assessing the quality of AI-driven translation systems, providing a framework to identify various translation errors, however, the model was found to have a deficiency in handling 'software interventions'.
These software interventions refer to technical errors introduced by the AI system itself, often due to limitations in algorithmic processing, the impact of stored data, or specific technical constraints within the software.
Recognizing such faults as separate from linguistic or contextual errors underscores the need for TQA model refinement.
As such, the study strongly recommends the continual tuning and upgrading of AI translation applications to mitigate these technical issues.
Addressing these technical faults not only improves translation accuracy but also enhances user trust and the reliability of AIPTSs across diverse applications.
Furthermore, these findings demonstrate the need for humanized intervention against AIPTSs to ensure faithful and culturally sensitive translation of complex religious texts.
Declaration of conflicting interest The authors declare that there is no conflict of interest in this work.
Funding acknowledgements The authors extend their appreciation to the Deanship of Research and Graduate Studies at King Khalid University for funding this work through Large Group Project under grant number .
/1.
References