Poetry Translation and AI: "Suicide in the Trenches" By Siegfried Sassoon

INTERNATIONAL JOURNAL OF SOCIAL SCIENCE HUMANITY & MANAGEMENT RESEARCH ISSN (print) 2833-2172, ISSN (online) 2833-2180 Volume 04 Issue 11 November 2025 Impact Factor: 6.79 ,10.58806/ijsshmr.2025v4i11n01DOI: Page No. 2059-2063 IJSSHMR, Volume 04 Issue 11 November 2025 www.ijsshmr.com Page 2059 Poetry Translation and AI: “Suicide in the Trenches” By Siegfried Sassoon Daniele Duscovich Alma Mater Europaea ABSTRACT: Translating poetry is not about linguistic accuracy (Malmkjær, 2020), and if Torop finds that translation has not to be a mere word-by-word replacement (Torop, 2024), but has to deal with the transmission of whole experience of the original text, this is more the ever the case for poetry. The translation process is about getting the real voice and the real intention of the author, in prose, in drama and in poetry. It is about understating real cultural aspects, historical references and much more. Moreover, in poetry there is the challenge of the rhythm, metric and the rhetorical devices. AI has made big steps forward in the field of language and translation, but it is a fact that limitations persist in analysing and processing symbols, hidden messages and context, as well as – just to make an example – alliterations, assonances and metaphors. The aim of this study is to identify some of the challenges and main issues of Artificial Intelligence while translating poetry. Qualitative research will be conducted to analyse the role of AI in dealing with semantical, rhetorical and poetical aspects, so that translation choices do not prejudice any dimension. For this purpose, a contrastive analysis will be conducted between a Chat GPT-generated translation and a human-translated version of the same original text. This study focus on the poem Suicide in the Trenches by Siegfried Sassoon. Afterwards, the investigation will be about the analysis and interpretation of the poem, referring to the translation process into Italian itself. Considerations will be made in order to understand if AI was able to produce a translation which does need human intervention. Nevertheless, more empirical research is needed so that this study and its findings are generalised and positive and negative aspects in the translation process with AI are ultimately identified. KEYWORDS: Translation, Poetry, Chat GPT, AI, Sassoon I. “THE ART OF FAILURE” Poetry is as ancient as humanity and its birth coincides with the birth of language itself, which is creative and poetic per se. Considering that language is the tool for communication par excellence, it is doubtless that the origins of poetry are strictly connected with the necessity of analysing feelings, of interacting socially and of expression images. Analysing the language, Bühler identified three primary functions (Bühler, 1934): expressive, conative and representational function. Moreover, Jakobson was influenced by Bühler’s model and added more functions (Jakobson, 1960: 470), among which the poetic function, which according to Duranti (Donati, 1997) focuses on “the message for its own sake”. This is particularly the case for poetry. In fact, poetry translation is a challenge which has many requirements, including emotional imagery, rhetorical devices and intricate structure (Venuti, 2017) Starting from the assumption that translation is not about mere mechanics or linguistics, translation is an interand intra-cultural activity where also cultural and extralinguistic factors are central (Bassnett, 1993: 136). This awareness has led to the so-called Cultural Studies propose a multidisciplinary approach to translation and a methodology which is not prescriptive. The consequence of this new kind of orientation shows that translations belong first and foremost to the target culture and, for this reason, the focus shifts from the original text to the target text. Italian author and translator Umberto Eco called translation “the art of failure” (Eco, 2003: 54) and probably for good reason. This is mostly the case for poetry, where cultural, artistic and linguistic elements have to be translated accordingly. Poetry translating is a subtle activity making use compromises and unloyalties. Most of the semantic traits (even if often impoverished) are maintained, whereas others (phonology, rhythm) get necessarily lost. This is the reason why translating poetry is very hard, sometimes impossible. After all, the aim of every translator of poems is to demonstrate that Robert Frost, when saying “Poetry is what gets lost in translation” (Brooks, C., & Warren, R. P., 1961), was wrong. In the past, people thought that only poets could translate poems. Nowadays, it is commonly thought that translators specialised in this particular field are as good at translating and, in fact, as Italian translator Giovanni Nadiani (2009) stated in an interview by Treccani, that the translator, as such, is the first, truly critical reader of the work to be translated and, therefore, knows what to look for, quantitatively verifying the hypotheses and working strategies critically identified and the concrete translation options Poetry Translation and AI: “Suicide in the Trenches” By Siegfried Sassoon IJSSHMR, Volume 04 Issue 11 November 2025 www.ijsshmr.com Page 2060 implemented. Only by doing so can they substantiate their ‘translation poetics’, aimed at meeting the ‘authorial poetics’. But the question now is: Is AI a poet after all? Since the advent and advancements of Artificial Intelligence (AI), the translation field has faced many and profound transformations in the translation world. AI started to be developed in Translation Studies back in the 1950s, where rule-based systems were created and relied on linguistic rules and dictionaries to create translations. Later on, in the 1990s Statistical Machine Translation (SMT) appeared, whereas it used statistical models to translate texts. The latest and most recent approach is the so-called Neural Machine Translation (NMT), which is based on neural networks for the purposes of translating a text (Hutchins, 2000). Literary texts are considered to be “open”, meaning that they can be given different interpretations and whose translations leads necessarily to “losses” caused by the impossibility of comprehending and conveying the full message of the original author. On the other side, technical texts are “closed”, meaning that their meaning and interpretation cannot be misinterpreted and the message is tied to the function of the texts and its target. So, it is possible to assume that the goal of translating a technical text is to reproduce integrally the information of the source text, without any loss or modification of the content (Scarpa, 2008: 63-70). According to this idea, considering the fact that it is domain-specific (Moslem, Way, Haque, Kelleher, 2022), NMT can be applied only and exclusively to the necessity of specialised translation, particularly because of its nature and characteristics. Linguistic accuracy is just inefficient for the translation of a poem (Malmkjær, 2020) and, in fact, if AI is able to translate many thousands of words in short time, it is also true that AI seems to be not completely capable of facing rhetorical devices, requiring a human “hand” to step in (Tymoczko, 2007). As stated above, translating poetry requires compromise and balance between linguistic and semantic accuracy, cultural aspects and poetic language. And since AI process is not adequately able to combine these three elements, translation poetry by using AI can be considered the very “art of failure”. Nonetheless, Literary Machine Translation (LMT) is a fact and, as the development of AI and machine learning continues, scholars are endeavouring to integrate corpora of poems, so that human translators’ work and reediting could be reduced. Many researchers in the field of Translation Studies, among whom Ana Guerberof Arenas and Toral (2022), believe it is completely impossible that a machine could translate a literary text, but LMT is nowadays more and more relevant with several research results, which should be analysed. II. “SUICIDE IN THE TRENCHES” BY SIEGFRIED SASSOON The Great War was a watershed moment between the 18th-century ideals of progress and optimism and the fragmentation of culture (Weidner, 1992) theorised by T. S. Eliot and the break with the past. War has been perceived as a trauma and dissociation remained as a model of an incomprehensible event, focused both on the trenches’ bloodbath and moments of long and eternal wait. As Eliot noted, the result is a sense fragmentation: in the daily experience in the trenches, back at home for the survivors, but also – in a wider sense – in culture itself. A group of poets who were historically close to the experimentation of Modernism, but stylistically far from it were the so-called “War Poets”. These poets shared their youth and the tragic experience in the trenches of the World War One during 1915 and 1917. They analysed the conflict, discussing topics such as hope, war destruction, in the view of poems with intimate tone, where the enormous waste of human lives is described. Sassoon is one of the biggest exponents of the group of the War Poets, and since – as Sassoon noted – “My real biography is my poetry. All the sequence of my development is there.” (Roberts, 2014, 98), the cultural and social elements in his texts are fundamental. He was wounded at war and he became a strong pacifist. In 1917, he immediately wrote a famous letter “Finished with the War: A Soldier’s Declaration” against war, which – with the support of the pacificist and philosopher Bertrand Russel – was read in front of the British Parliament, where he was “not protesting against the military conduct of the War, but against the political errors and insincerities for which the fighting men are being sacrificed”. In 1917, while he was in the Scottish sanatory of Craiglockhart, his collections of poems were published: The Old Huntsman (1917), Counter-Attack and Other Poems (1918), The War Poems of Siegfried Sassoon (1919), where war and its hypocrisies are described extremely frankly. The style is based on regular rhymes, colloquial language and satirical tone. The poem which is analysed in this study is “Suicide in the Trenches”, which was published in the collection Counter-Attack and Other Poems, is focused on the Sassoon’s experience as a soldier in the trenches. As it is understandable, war is the main topic of the poem, even if others aspects of the war are not negligible: dirt, death and social e civil responsibility. I knew a simple soldier boy Who grinned at life in empty joy, Slept soundly through the lonesome dark, And whistled early with the lark. Poetry Translation and AI: “Suicide in the Trenches” By Siegfried Sassoon IJSSHMR, Volume 04 Issue 11 November 2025 www.ijsshmr.com Page 2061 In winter trenches, cowed and glum, 5 With crumps and lice and lack of rum, He put a bullet through his brain. No one spoke of him again. You smug-faced crowds with kindling eye Who cheer when soldier lads march by, 10 Sneak home and pray you'll never know The hell where youth and laughter go. As a trained reader can see, in the poem, the author uses numerous rhetorical devices and there are many examples of figurative language, among which includes apostrophes (l. 9) or metaphors (l. 11). For the purposes of this study, a translation by Francesca Coin (2010) was used. She sociologist Francesca Coin, who obtained a PhD at the Georgia State University, USA and till September 2022 she worked as Associate Professor at University of Lancaster, UK. The other translation was generated using Chat GPT. For the purpose of this study, the two translation will be systematically compared. Translation by Francesca Coin Translation by Chat GPT Conoscevo un giovane soldato che ghignava senza gioia al fato la notte dormiva nel timore fischiava con le allodole all’aurora. Nelle trincee invernali tra tristezza e timore con i crampi i pidocchi senza liquore si è infilato un proiettile nella testa del suo nome da allora poco resta. Voi folle compiaciute ed eccitate che guardate i soldati marciare e tifate andate a casa pregate Dio di non sapere in quale inferno finiscono gioia e speme. Conoscevo un semplice soldatino, che sorrideva alla vita con gioia ingenua, dormiva profondamente nel buio solitario e fischiava presto insieme all’allodola. Nelle trincee d’inverno, abbattuto e cupo, tra granate, pidocchi e mancanza di rum, si sparò un colpo in testa. Nessuno parlò più di lui. Voi, folle dal volto sereno e dallo sguardo acceso, che applaudite quando i giovani soldati marciano, strisciate a casa e pregate di non conoscere mai l’inferno dove muoiono la gioventù e il riso. At first sight, the two translations seem not to have many differences, but a deeper and more attentive analysis shows interest elements. First of all, the Source Text (ST) is composed by three quatrains and this is also applied in the two translations. What is different is the metric. The ST is written using iambic pentameter, miming the sound of marching. On the other hand, the text by Coin uses irregular a blank vers (irregular nineand eleven-syllable lines) as well as Chat GPT’s version. In the ST there is an only exception: line 8. The metrical variation of “No one spoke of him again.” helps to emphasise the ordinariness of the suicide committed by the “young soldier boy”. In all texts punctuation remained the same. The AABB rhyme is used in the ST and this prosodic element reminds of a nursery rhyme, which creates irony in dealing with a tragic and shocking event, using a child-like verse. Coin tries to apply the same logic and, using perfect rhymes and assonances, obtains the same result. On the contrary, the AI translation seems to have disregarded this aspect and, unfortunately, this important element representing the author’s intention gets lost. Starting with lines 1 and 2, it is important to highlight the sense of melancholy and foreboding which Sassoon tries to obtain using the Past-Simple-form “I knew”. In this case, both translations were able to convey the same message also in Italian, using the tense Imperfetto “Conoscevo”. What is different is the way in which the “simple soldier boy” was translated. Sassoon intended to emphasise how the identity of the poor soldier was meaningless for the purposes of the war, and in fact “No one spoke of him again” after the tragic event. The lack of identity makes also this very soldier representational of the hundreds of thousands of soldiers with the same fate during WW1. Coin decides to translate it just with “un giovane soldato” (Italian for “A young soldier”), which seems to be the most appropriate version. Chat GPT, on the other hand translates this phrase as “un semplice soldatino”, using a hypocorism expressing endearment. In this case, also considering the topics with which the poem deals, seems inappropriate and out of context. Linguistically speaking, it is a good choice, but – as noted – not appropriate for the context. The sibilance of “slept soundly” in line 3 gets lost in both translations. The original intention of the author was to represent the nights before the war, full of effortless dreams and comfort. The symbol of the “lark” of line 4 is maintained in the Target Texts (TT), probably unconsciously. The lark is a symbol for the English countryside, associated with the life not at war. The adverb “soundly” is interestingly skipped by Coin, who translates line 3, changing slightly the meaning of the original text. In her case Poetry Translation and AI: “Suicide in the Trenches” By Siegfried Sassoon IJSSHMR, Volume 04 Issue 11 November 2025 www.ijsshmr.com Page 2062 “through the lonesome dark” becomes “nel timore” (Italian for “in dread/fear”). This probably has probably to do with metric reasons. Chat GPT, instead, translates it literally with “profondamente” (Italian for “deeply”, “soundly”), which is (again) correct, even if impoverish the Italian text, considering also the facts that also adverbs in modern poetry becomes choices about style, having to do with rhythm and poetic tone (Segre, 2008). Apart from not-noteworthy lexical choices and differences, in lines 5 and 6, it is important to note the use of polysyndeton by Sassoon to highlight the endless and motionless situation of the trenches. Unfortunately, neither Coin nor Chat GPT expressed the same idea, focusing more on the “content”, as if their TT were a paraphrase. In lines 8 and 9, Sassoon deals with the topic of suicide and death. To create a violent tone echoing the sound of a gun explosion, the author uses the alliteration plosive letters “p” and “b” in the line “Put a bullet in his brain”. The effect of this line is extreme and takes the reader in the trenches, in front of the “simple soldier boy”, watching the scene in first person. In the ST and in the TT by Chat GPT the scene is further made real by the end-stopped line. In this case, Coin decided to disregard punctuation, but probably with a result which is – for poetic reasons – less appropriate and attentive than Chat GPT’s translation. Focusing on lines 9-12, the first aspect to analyse is the apostrophe of “You smug-faced with kindling eye”. The second person plural gives responsibility to those “smug-faced crowds” who were so enthusiast of sending “young soldier boy[s]” to war, becoming this way somehow complicit of the massacre of WW1. The accusatory tone of Sassoon conveyed by the sibilance of lines 9 and 10 gets unfortunately lost in both translations. The metaphor of the “kindling eye” is also treated interestingly. The author accuses those at home of having lit the flames of destructions, leading to the enormous number of casualties. “Sguardo acceso” in Chat GPT’s version refers to the same semantic field (probably unconsciously, if unconscious can be applied to AI), whereas Coin translates it with “eccitate” (Italian for “excited”). Again, this choice has for sure to do with rhythmic and metric choices. In line 11, the author says “and pray”. This verb has to be intended more in rhetorical, than religious sense. After the horrors WW1, as Nietzsche had anticipated in “The Gay Science”, people thought that “Gott is tot”, that God was dead (Nietzsche, 1882). In this sense, the choice of Coin of writing “e pregate Dio” (Italian for “and pray God”) is anachronistic and out of context. The conclusion of the poem of all versions, ST and TTs, is characterised by the use of the Imperative mode and the personification of “Youth” and “Laughter”. Coin’s version changes slightly the semantic field, using “goia e speme” (Italian for “joy and hope”), but the message conveyed is the same. To note the use of the poetic and archaic term “speme” instead of “speranza”, almost as if the translator need to give a solemn and archaic tone to her translation. III. CONCLUSION It could sound very simplistic, but the ultimate purpose of Artificial intelligence, as such, is to be “intelligent” and to simulate humans in the way in which it works. In essence, it is supposed to replicate human reasoning in machines, delivering it with typically human abilities, such as problem-solving, learning, processing and reasoning. Focusing more on translation and the translation process, AI can analyse, translate and process big amounts of words and text, improving accuracy and efficiency. The point is that AI works based on algorithms and rules and, for this reason, suits perfectly to the so-called “close” texts, those of technical and scientific nature. But what about “open” texts? Because of its programming, AI lacks that sensibility which is typically “human”. Linguistically, it is very precise and accurate, but it is unable to find what is “behind” words. If one considers literature, and poetry in particular, AI cannot interpret the real intention of the original author, what their real purposes were. The aim of this study is to highlight the limitation AI faces while translating texts out of rules or fixed schemes. In the analysis above, we noticed that Chat GPT obtains quite good results in the word-to-word translation, where it lacks sometimes sympathy and context (as in the case of “soldatino”), which creates unfortunate results. The word-to-word translation has led several times to several losses. The human translation was in some cases farther from the original text, but the attempts of rendering the very intentions or imageries of the author it was more successful. And so, notwithstanding what said above, the version by Coin is extremely close to the Source Text in creating the same effect as in the original by Sassoon. Overall, the aim of this study was to analyse some of the limitations of Artificial Intelligence in applying it in poetry translation. Artificial Intelligence is “artificial” and, as such, it is significant to note that it has limitations in all those aspects which are “human”, as imagination, originality and in the pragmatic and contextual interpretation of the world, which means that – at this stage – human intervention is still needed. REFERENCES 1) Baker, M. 2001. Routledge Encyclopedia of Translation Studies. 2nd ed. Routledge. 2) Bassnett, S. 1993. Translation studies (2nd ed). Shanghai, Shanghai Foreign Language education press. 3) Bassnett, S. and Lefevere A. (eds) 1990. Translation, History, and Culture. London: Pinter. 4) Brooks, C., & Warren, R. P. (Eds.). 1961. Conversations on the craft of poetry. Holt, Rinehart and Winston. 5) Burkhanov, I. 2003. Translation: Theoretical Prerequisites. Rzeszw: University Press. Poetry Translation and AI: “Suicide in the Trenches” By Siegfried Sassoon IJSSHMR, Volume 04 Issue 11 November 2025 www.ijsshmr.com Page 2063 6) Bühler, K. 1934. Sprachtheorie. Oxford, England: Fischer. 7) Campbell, P. 2007. Siegfried Sassoon: A study of the war poetry. McFarland. 8) Catford, John Cunnison 2000. A Linguistic Theory of Translation. London: Oxford University Press. 9) Coin, F. 2010. Combattendo (nel) l’esercito: la lunga lotta dei veterani americani. DEP. DEPORTATE, ESULI, PROFUGHE, 13, 121-148. 10) Duranti, A. 1997. Linguistic anthropology. New York: Cambridge University Press. 11) Eco, U. 2003 Mouse or Rat: Translation as Negotiation. London: Phoenix/Orion Books. 12) Guerberof-Arenas, A., & Toral, A. 2022. Creativity in translation: Machine translation as a constraint for literary texts. Translation Spaces, 11(2), 184-212. 13) Hutchins, W. 2000. The first decades of machine translation: Early years in Machine translation. John Benjamins Publishing Company. 14) Jakobson, R. 1960. “Closing statements: Linguistics and Poetics”. In Sebeok, Thomas A. (ed.). Style in language. NewYork: M.I.T. 15) Lefevere, A. 1975. Translating Poetry: Seven Strategies and a Blueprint. 16) Malmkjær, K. 2020. Translation and creativity: Perspectives on literary translation. Routledge. 17) Malmkjær, K. 2020. The challenges of translating poetry in the age of AI. Poetry Translation Review, 28, 34-45. 18) Malmkjær, K. 2020. Translation and Creativity. The Routledge Handbook of Translation and Creativity. 19) Moslem Y., Way A., Haque R., Kelleher J. D. 2022. Domain-Specific Text Generation for Machine Translation. arXiv preprint arXiv:2208.05909. 20) Poems, C., Ruminations, R., Illuminations, L., & Progress, S. S. 1947. Siegfried Sassoon. Faber & Faber. 21) Robinson P. 2010. Poetry and Translation The Art of the Impossible. Liverpool University Press. Pp. 23-47. 22) Roberts, J. S. 2014. Siegfried Sassoon-The First Complete Biography of One of Our Greatest War Poets. Kings Road Publishing. 23) Scarpa, F. 2008. La traduzione specializzata. Un approccio didattico professionale. Milano, Hoepli. 24) Segre C. 2008. I segni e la critica. Fra strutturalismo e semiologia. Einaudi. Piccola biblioteca Einaudi. Nuova serie. 25) Somers, Harrold 2003. Translation memory systems. Computers and Translation. A translator’s guide. John Benjamins Publishing Company. 26) Torop, P. 2024. La traduzione totale: Tipi di processo traduttivo nella cultura. (B. Osimo, Trad.). Osimo Bruno. 27) Tymoczko, M. 2007. Translation and ideology: Encounters and clashes. St. Jerome Publishing. 28) Venuti, L. 2017. The translator’s invisibility: A history of translation (2nd ed.). Routledge. 29) Verdonk, P. 2013. The stylistics of poetry. 30) Weidner, S. 1992. TS Eliot: Poet and critic as historical theorist. The Student Historical Journal, 1993, 1992-3.