Social and cultural restrictions for movie titles translation: why AI will never do it

Full Text

Introduction Since the beginning of the 21st century, globalization processes (also known as homogenization and standardization (Pooch, 2016: 24)) have become particularly relevant, defining trends in global development in many fields, from economics itself to cultural studies and linguistics (Featherstone, 1992; Lang, 2006; Pooch, 2016). These processes of globalization have generated interest in the scientific study of such issues as the dominance of certain cultures and languages in the world. These issues fit into the problematic of symbolic and cultural capital, which P. Bourdieu addressed back in the late 1990s, introducing the concept of the “language market” (Bourdieu 1982, 1986). Today, English is undoubtedly the leader in this market. Despite the fact that in the global information space “there is always translation, adaptation, or ‘indigenization’ of the receiving culture” (Pooch, 2016: 25), the dominance of English leads to a unification of the global semiotic field. One manifestation of this unification is the widespread use of machine translation technologies, which are being implemented in many fields. Machine translation technologies mostly operate in language pairs with English, so in rare language pairs, even machine translators automatically use English as the intermediary language. Removing human involvement from the translation process in some areas and relying on English as the intermediary leads to the loss of significant cultural meanings that only a human can perceive, since in different languages, “cultures may well travel and move around the world, but ethnicity is still about the maintenance of social boundaries, something which remains a powerful force in the current phase of globalization” (Featherstone, 1992: 12). In 2022 a Google engineer Blake Lemoine claimed that LaMDA (Language Model for Dialogue Applications) - Google’s system for building chatbots based on its most advanced large language models - was sentient, i. e. it was capable of feeling empathy [58]. That was the conclusion that Blake Lemoine arrived at after having indepth conversations with LaMDA. Consciousness and sentience, when ascribed to a language model, mean that it can understand the feelings of a person with whom it is having a conversation, as well as interpret human emotions and therefore identify them. Large-scale implementation of such a technology into machine translation has the potential to revolutionise translation industry by transforming the role of a human translator. Post-editing machine translation may become his primary function not only in the areas where texts are predominantly clichéd and have well-established equivalents in the source language and the target language (such as legal documents, technical texts, manuals, etc.), but also in the spheres where texts are required to express multiple meanings and trigger an emotional response (Toral, Wieling, Way, 2018). The latter encompasses literary fiction (Toral, Wieling, Way, 2018; Matusov, 2019; Constantine, 2019), audiovisual products as well as marketing texts, including film titles. A film title is not only the name of the film that gives an idea about its plot. They “are created to draw attention to a film” (Haidegger, 2015: 425) and to urge the viewer to watch the film, i. e. to pay for it (Bae, Kim, 2019: 100). It could be argued that a film title contains a manipulative element that encourages the target audience to make the choice that the producers want them to make. In other words, “title should function as a means to market a film successfully” (Haidegger, 2015: 426). Thus, linguistically speaking, everything about the film title is of high importance - the choice of lexical units with all the meanings and connotations, their grammatical form (Haidegger, 2015: 425), the connection between the film title and the plot. If one looks at a film title in the original as a black and white text, then one might consider its lexical and grammatical aspects in conjunction with the content of the film and the parameters of the target audience (age, gender, social or professional status, creed, preferences, etc.) to be the key means of attracting attention of the viewer. However, when it comes to the translated version of the film title, these factors may be insufficient to adapt it successfully to a foreign film market. Since a film title’s main function is to sell the film in a foreign market, it would not suffice to resort to interlingual translation without adapting it to the foreign culture. One should consider cultural features of the host culture as well as social and political context that will accommodate the film along with its title. Moreover, the title should be engaging and marketable (Krasina, Moctar, 2020: 293-294). Indisputably, these factors may vary depending on the language pairs involved in translation or, what is more important, on the two cultures and the differences that have to be reconciled in the process of film title adaptation. Thus, studies show that, for instance, when translating film titles from English into Thai, the dominating strategy is to create a new title in the Thai language that conveys the main idea of the film instead of translating the original one (Kettongma, 2024: 210), whereas, when translating from French into Russian, the priority is given to the adequacy of the title and its correspondence to the genre, plot and target audience of the film (Anissimov, Borissova, Konson, 2019: 453). The findings of academic research lead us to the conclusion that it is difficult to compile an exhaustive list of all translation strategies used when adapting a film title in a particular language pair. If such a list could be compiled, it would be possible to identify patterns of translating film titles and use them as a foundation for training artificial intelligence to translate the titles of audiovisual works. However, scientific research into translation of marketing texts, including film title adaptation, and the practice of translation do not align. The translation process presupposes careful analysis of the factors that may influence translation decisions but escape the attention of a researcher who views a foreign- language film title as a result of translation and marketing efforts. Thus, this research focuses on the study of translation strategies from a practical viewpoint, taking into account cultural, social and political contexts. Methodology and Material This article is of analytical nature, and it is primarily based on the observations derived from the practice of translating film titles, which can be found on the Internet in the public domain in large quantities. This is accounted for by the fact that scientific understanding of film title localisation strategies remains an open question for research in linguistics, and while seeking an answer one should start with marketing. According to audiovisual translation practitioners, there are no available technologies that can translate a film title without referring to the content of the audiovisual work in question. The current machine translation systems render film titles as stand- alone sentences that are not contextualised. The context for a film title comprises not only the film itself, but also the target audience as a set of cultural and social parameters. The cultural, historical or political context in which the film is released is also of great importance. This article is based on reviews of Russian streaming platforms and translation companies 2, which share their observations on strategies for translating English-l anguage titles in the context of localising British and American films to the Russian market. But before discussing the process of translation and localisation that inherently requires human participation, it is necessary to review the strengths and weaknesses of machine translation diachronically. AI Translation Algorithm and Where it Comes from Artificial Intelligence technologies are far from being the first generation of machine and automated translation. The advent of machine translation dates back to the 50s of the 20th century, and the idea for this method of translation is rooted in the practical necessity to automate human labour. By that time the amount of scientific and technical information in the world that needed to be translated had been growing rapidly, and human translation resources were insufficient. Thus, people came up with the idea of automating translation process (Sin-wai, 2014; DeCamp, Zetzsche, 2014). Historically speaking, there have been five main approaches to machine translation. Rule-based Machine Translation This can be represented as follows: dictionary + grammar = translation. This type of machine translation functions according to the same principle as human translation. It relies heavily on detailed dictionaries and grammars of both languages involved in translation (Jones et al., 2012: 1364). The first machine translation systems were rule-based and were generally not highly successful. This was due to, firstly, low flexibility (rules and grammars were fixed, making it difficult to adapt to new language constructions and features); secondly, high labour intensity (creating and maintaining extensive rule sets required significant resources and time); thirdly, problems with lexical polysemy (polysemy of lexical units was a difficult task, as the choice of the correct meaning depended on the context, which was sometimes difficult to take into account in predefined rules) (Shiwen, Xiaojing, 2014). 2 Zayacz, A. (2016, May 6). Kak lokalizuyut nazvaniya fil’mov v Rossii [How film titles are localised in Russia]. Film.ru. Retrieved 28 July 2025, from https://www.film.ru/articles/kinoslovar-trudnosti-perevoda Trudnosti perevoda: pochemu russkii variant nazvaniya fil’ma chasto ne sootvetstvuet originalu? [How film titles are localised in Russia. Difficulties of translation: why does the Russian version of a film title often not correspond to the original?]. Lexxis. Retrieved 28 July 2025, from https://lexxis.ru/articles/trudnosti-perevoda-pochemu-russkiy-variantnazvaniya-filma-chasto-ne-sootvetstvuet-originalu/#close Darbinyan, A. (2020, December 2). Kak pravil’no lokalizovyvat’ fil’my v Rossii? Kratkaya instruktsiya [How to localise films in Russia? Brief instructions]. Kinopoisk. https://www.kinopoisk.ru/media/article/4003560/ Blicz perevod. Pochemu perevody nazvanii fil’mov byvayut daleki ot originalov? [Why do translations of film titles tend to be far from the originals?]. Retrieved 28 July 2025, from https://blitz-perevod.ru/articles/pochemu-perevodynazvanii-filmov-byvaiut-daleki-ot-originalov 20 fil’mov s nepravil’no perevedennymi nazvaniyami smotret’ onlain [20 films with mistranslated titles]. @kino. Retrieved 28 July 2025, from https://kino.mail.ru/cinema/selection/2733_trudnosti_perevoda_filmi_s_nepravilnimi_nazvaniyami/ Example- based machine translation This machine translation method presupposes finding similar examples of sentence pairs in the source language and the target language. Example- based machine translation relies on data from parallel language corpora, and examples are extracted from large sets of bilingual texts. The mechanism of this type of translation is as follows: the computer algorithm analyses the sentence to be translated → compares this sentence with the parallel corpus → identifies segments in the parallel corpus that are similar to the sentence to be translated → extracts the necessary components from the corpus and constructs a sentence in the target language (Tak-ming, Webster, 2014; Maučec, Donaj, 2020). Statistical machine translation It boils down to the following scheme: parallel texts + statistics = translation. Thus, this type of translation is based on statistics and machine learning. The computer is given a database of parallel texts that consists of sentences paired with their translation into a foreign language. Trained on parallel texts, the computer records how a certain lexical unit was translated in this or that language environment, and how frequently that occurred. This way it analyses the statistical data of translation. The larger the database of parallel texts is, the higher is the translation quality of the algorithm that was trained on the basis of these texts (Yang, Min, 2014: 202). Hybrid machine translation This combines the first and the third approaches mentioned above, so it could be represented as follows: rules + statistics = translation. The combination of the two approaches can be implemented in various ways, but the two most common variants are: (1) integration of a statistical module into translation, and (2) integration of rules into a statistical model (Qun, Xiaojun, 2014). Neural machine translation This type of machine translation uses artificial neural networks that imitate human neural networks. The main advantage of neural machine translation is its ability to account for contextual relationships and grammar. In contrast to statistical machine translation, neural machine translation ensures a higher degree of adequacy and equivalence, and the translation outcome appears more authentic. This result is attributed to the fact that neural networks analyse individual language units and whole sentences within a given context, rather than selecting a translation equivalent in isolation, based on the frequency of its occurrence in a similar linguistic environment (Koehn, 2020; Maučec, Donaj, 2020). Back in the mid-2000s, many people made fun of how a computer was handling translation of texts and could scarcely believe that we would soon be questioning the relevance of translation as a profession. However, the situation has changed dramatically since neural machine translation was introduced in 2016. Neural models are capable of taking into account the context and semantics of sentences, allowing them to produce more accurate and natural sounding translations as compared to earlier versions of machine translation. Presently, neural machine translation is considered to be the most successful software solution in translation industry in a variety of different subject areas. The quality of present-day machine translation raises the issue of preserving translation as a profession and drives various experiments that are, to some extent, competitions between a machine translation system and a human translator. Let us consider one of them. In 2018, two years after the declared launch of neural machine translation, the technology started being actively adapted to translation tasks in different subject areas. Intento [59], a company which specialises in implementing AI technologies into machine translation to run various international platforms, has evaluated the effectiveness of neural machine translation. The Intento experiment involved translations from a neural network, a human translator and several machine translation systems. All participants of the experiment translated the same biomedical texts in the English- German language pair. Text fragments that were translated with significant differences were submitted for manual review to Logrus IT [60] - a large international high-tech company that provides services of translation, localisation, testing and language consulting to the business sector. Logrus IT linguists received the source texts in English, their anonymised translations into German from 13 machine translation systems, as well as a human translation that was considered exemplary under the terms of the experiment. None of the experts knew which translation they were checking - a machine translation or a human translation [61]. Logrus IT linguists used their own quality assessment methodology to analyse the translations. It included criteria such as the degree of translation adequacy, readability, terminological consistency with the subject matter of the text, style, as well as a scale of translation errors, which range in severity from critical to minor. A comparison of translations performed by a neural network, a human translator and several machine translation systems revealed that the human translation was not rated as exemplary by the experts. In a sense, this experiment demonstrated that neural machine translation, i. e. a human- designed translation system, was able to outperform a human translator. One could say that the student has surpassed the master. Although neural machine translation has made significant advances in recent years, it is still far from ideal. AI continues to stumble over the complexity of a natural language. The most challenging type of translation for the system is so far the translation of literary fiction. Only a human translator is capable of perceiving the meaning of prose and poetry that is concealed in a wide range of stylistic elements of a text. Machine algorithms often translate word for word, following the rules of language, i. e. translating words - the external linguistic form of a literary work, rather than sense - the inner semantic content of a literary text. Automatically translating marketing texts, including film titles, presents another serious challenge to translation systems and it is closely linked to the machine learning technology. Contemporary machine translation algorithms are trained on the database of written texts. This means that they require a training database that the algorithm uses to record certain morphological, lexical, grammatical, syntactic patterns in multiple languages or in a certain language pair. Machine translation systems rely on these patterns to generate the most adequate version of translation, taking into account the formal aspects of the context that is analysed by AI. When it comes to the translation of a film title, such an analysis proves to be quite difficult, since its context encompasses not only the film or TV series, but also the culture of the country of localisation (Anissimov, Borissova, Konson, 2019; Kiran, 2023; Jiang, Yin, 2024), as well as the social and political context (Crane, 2014; Zhang, 2023). It entails ideological and censorship requirements that accompany the release of an audiovisual work in a foreign market. In addition, an audiovisual work may be a film adaptation of a literary text of the same title (the play by Anton Chekhov The Seagull 1895 → TV drama The Seagull 2018) or of a different title (Anton Chekhov’s short story The Duel 1891 → TV drama The Duel 2010), or it may be an adaptation of a work of fiction (the 2013 tragicomedy The Double based on Dostoyevsky’s novella of the same title The Double (1846)). In both cases, the title of the film should serve as a clear reference to the title of the literary text. To understand the film’s semantic content, the marketing expert, who is translating the film title, should watch the film or at least read the script or the screenplay. Furthermore, the marketing expert/translator has to consider the distinctive features of the target audience for whom the film is made and, therefore, who the film’s title should be addressed to. Thus, it is only through an analysis of these factors - the social and political context, the culture of the country of localisation, the intertextual links of the audiovisual work, the target audience - that the most adequate decision can be made about the choice of one or another variant of the translation of the film title. Naturally, one can assume that if modern AI technologies are capable of translating and dubbing audiovisual works without human participation (e. g. browser- based Yandex video translation [62], machine translation platforms for dubbing and neural network systems for voice-over such as Rask AI [63], Veed [64], HeyGen [65], Kapwing [66], AI Video Translator [67], Clideo [68] etc.), then neural machine translation will be able to easily ‘watch’ a film, analyse it, and offer several variants of translation of the film title taking into account the content of the film. Nevertheless, neural machine translation does not go further than the audiovisual work itself when analysing the film. The neural machine translation process includes recognising diegetic speech → transcribing it → generating machine translation of the transcript → voicing of the translation → superimposing the voiced text on a video sequence (Martín- Mor, Sánchez- Gijón, 2016; Sulubacak, et al. 2020). When given a transcript of the speech in the original or in translation, a machine algorithm can analyse it and provide several variants of the title in the source language or the target language. Why does this scenario of creating an original and/or translated film title seem questionable? Firstly, because when translating a film, the machine translates a written text with no reference to the film’s video sequence and its sound or music tracks, which also carry a semantic and emotional message. Hence the visual component (video sequence), audio component (music) and the plot line (conceptual content) of the film will also be disregarded in the process of creating the film title. Secondly, at the present stage of development of machine translation technology, neural machine translation does not provide an integrated option of analysing the social and political agenda and current affairs as well as the cultural features of the localisation market. AI technologies are trained on the databases that have already been created. Therefore, they do not reflect the reality in its dynamic aspect. Only a human translator can take into consideration the nuances of the ever-changing reality, selecting the most adequate strategy for translating a film title, while also considering all the necessary factors. Movie Titles Translation Strategies Having analysed strategies for translating film titles that streaming companies and translation agencies resort to in their practice (including social, political, and cultural localisation factors), we identified the following groups of marketing strategies for localising film titles. Expanding a well-known franchise Marketing experts claim that sometimes it is quite difficult to sell a film to the audience without a brand affiliation. Therefore, localising a film title may involve a well-established franchise or film. Thus, for example, the 2019 horror film Polaroid was titled Пункт назначения: Смайл [Final Destination: Smile] in Russian. Another example of the same strategy is the 2019 feature film Dark Phoenix that was localised as Люди Икс: Тёмный Феникс [X-men: Dark Phoenix]. A literal translation of the title - Dark Phoenix - might seem recognisable only to the fans of the franchise, who are well-versed in the X-Men universe, but not to those who might have once watched one of the previous films. It turned out that the Russian word братва [gang] was a popular choice for rendering titles of cartoons and films for children from English into Russian. After the 2004 cartoon Shark Tale hit the Russian big screen under the title Подводная братва [Underwater gang], other localised film titles started to feature the word братва [gang]: Over the Hedge (2006) → Лесная братва [Forest gang]; Delhi Safari (2011) → Братва из джунглей [Gang from the Jungle]; Alpha and Omega (2010) → Альфа и Омега: клыкастая братва [Alpha and Omega: fang gang]; Joey and Ella (2021) → Прыгучая братва [Bouncy gang]. As we can see, the original film titles were in no way related to each other and represented independent stories, not united by the same franchise, however, in the Russian language they seem to be artificially linked together during localisation process. This marketing move brought together films for children about animal adventures under the umbrella of the word братва [gang]. Thus, this strategy involves analysing already existing translated film titles that have similar genre and plot features and have been successfully localised in the host market in terms of reaching their target audience. Minimising the risk of politicising the title This strategy presupposes avoiding situations where viewers may see political subtext in the film title. This is the case when the word America ordinarily disappears from the localised title when the film is released in Russia. For instance, the film, well-known to the Russian audience as Первый мститель [The first avenger], is originally called Captain America. The First Avenger (2011); while American Sniper, released in 2014, was distributed in Russian cinemas as Снайпер [Sniper], losing its American affiliation. Another case to support this strategy is the film title Siberia (2018), which originally was accompanied by a slogan A dangerous land breeds a dangerous man (Fig. 1). The film tells the story of an American gem dealer who travels to Siberia to make a deal. It was released in Russia under the title Профессионал [Professional] (Fig. 2), while the slogan was adapted to fit the context Один против русской мафии [Alone against Russian mafia] (Fig. 2). It should be mentioned that the localised slogan is not available on popular Russian streaming services (Okko, Kinopoisk) and could be found only by searching images. Fig. 1. Original movie poster Fig. 2. Localised movie poster for Siberia in English for Siberia in Russian Source: Kinopoisk. Source: Kinopoisk. Retrieved 21 July 2025, Retrieved 21 July 2025, from https://www.kinopoisk.ru/film/1038843/ from https://www.kinopoisk.ru/film/1038843/ posters/page/1/ posters/page/1/ Adding sexual subtext From a marketing point of view, in order to attract the attention of a certain target audience, erotic or sexual innuendo should be imparted into the title of the film. This localisation strategy could be seen in the case of the 2011 film Friends with Benefits. The word combination in the original film title is a set expression that could be translated into Russian as Друзья с привилегиями [Friends with privileges] or Друзья с бонусами [Friends with bonuses], referring to friends who have an intimate relationship but are not boyfriend and girlfriend or husband and wife. Nevertheless, the film was released in Russia under the title Секс по дружбе [Sex for friendship’s sake]. The original title implies the same, albeit in a less direct fashion, so the film title was translated adequately. At the same time another possible variant of translation - Друзья с привилегиями [Friends with privileges] - would not give any hint to the Russian- speaking viewers as to the implicit meaning of this phrase. Thus, the strategy that was used in the process of localisation is to make the sexual subtext of the original title explicit in translation. Another case in point is the film No Strings Attached (2010) that could be translated as Без всяких обязательств [without any commitment]. But the Russian audience was offered a more explicit film title - Больше чем секс [More than sex]. Although this localisation strategy is used quite often, it requires some caution. In November 2020 The Boss Baby: Family Business was promoted in Russia as Босс-молокосос 2 [Boss-milk sucker 2]. The poster in English read Playtime is over, which could be rendered into Russian as Игры закончились [Games came to an end]. Instead, the slogan was localised as Новая соска - просто бомба [New baby pacifier is a bomb/sensation/smash]. The Russian viewers have spotted in this phrase not just a sexual or erotic subtext, but propaganda of paedophilia, and criticised the film distributor [69]. Lowering speech register This strategy is characterised by the use of colloquial vocabulary that could be considered offhand or even rude, express a negative attitude and include evaluative characteristics. For instance, the 2013 film Identity Thief that may be translated as Кража личности [Stealing identity] (Fig. 3), has nothing in its title that may suggest emotional, evaluative or rude connotations. Meanwhile the Russian localised film title read Поймай толстуху, если сможешь [Catch the fatso if you can] (Fig. 4), which included an offensive colloquial word толстуха, referring to a fat woman. Adding slang This localisation strategy involves adding slang vocabulary to the titles of films and TV series. It can be illustrated by the Russian localised film titles that ended up with the word перец, which literally means “pepper”, but is also used figuratively by younger generation to denote someone cool [70]: Superbad (2007) → SuperПерцы [Super dudes]; Your Highness (2010) → Храбрые перцем [Brave dudes]; Last Vegas (2013) → Starперцы [Star dudes]. In the case of Superbad, the inclusion of the slang word перец (the plural form - перцы) in the Russian film title directly indicates that the main characters of the comedy are young people: SuperПерцы [Super dudes]. The word перец in the localised title functions as a substitute for bad, which could refer to sexually provocative behaviour [71] portrayed in the film. Fig. 3. Original movie poster for Identity Thief in Fig. 4. Localised movie poster for Identity Thief English in Russian Source: Kinopoisk. Source: Kinopoisk. Retrieved 21 July 2025, Retrieved 21 July 2025, from https://www.kinopoisk.ru/film/606698/ from https://www.kinopoisk.ru/film/606698/ posters/page/1/ posters/page/1/ As for the film Your Highness, its Russian counterpart, Храбрые перцем, also includes the word перец, however, its use seems to be intentionally ambiguous. It may mean “a soldier who is serving while violating the regulations” [72], which fits the plot line as one of the main characters is a knight who tries to avoid problems by any means necessary. It may invoke associations with the phrase храбрые сердцем [brave- hearted] as it sounds very similar in Russian. Other interpretations are also possible. And in the final example illustrating the strategy of adding slang to the translated film title, Last Vegas, a comical effect is achieved through a play on words as the Russian title Starперцы sounds very similar to the slang word старпёр, designating an old man with a derogatory connotation: старик, пенсионер, старикашка 17. Thus, the original title, Last Vegas, is referring to the city in the USA Las Vegas and at the same time hinting at the age of the protagonists, while the localised title of the film, Starперцы, mirrors the main storyline - an attempt by several retired people to have fun in the city, which is famous for its vibrant nightlife. Using the English word star in the Russian film title may also serve the function of attracting attention to the star-studded cast. Film titles that are localised using such strategies as Adding slang and Lowering speech register tend to appeal to a younger audience and therefore act as a filter to discourage those viewers for whom the film is not intended. Developing the story The film distributor may use the localisation strategy of developing the story by hinting at the scale of the film. As a result, words like history, chronicles and saga are added to the title. For example, this strategy was implemented while translating the title of a vampire film: Let Me In (2010) was localised for the Russian market as Впусти меня. Сага [Let me in. Saga]. The localised title serves as a reference to another famous vampire saga - Twilight. Another case in point - the 2024 film Furiosa: A Mad Max Saga, released in Russia under the title Фуриоса: хроники безумного Макса [Furiosa: chronicles of mad Max], which was a sequel to the 2015 film Mad Max: Fury Road, translated literally into Russian as Безумный Макс: дорога ярости [Mad Max: fury road]. The same strategy can be identified in the localisation of the following films and TV series: Masters of Science Fiction (2007) → Хроники будущего [Chronicles of the future]; Dylan Dog: Dead of Night (2010) → Хроники вампиров [Vampire chronicles]; Mortal Engines (2018) → Хроники хищных городов [Chronicles of predator cities]; The Secret Kingdom (2022) → Хроники Панголинов [Chronicles of the Pangolin]; The Creep Tapes (2024) → Ублюдские хроники [Chronicles of a creep]. Unlike the Franchise expansion strategy, the Strategy of developing the story presupposes incorporating into the film titles only those words that increase the artistic value of an audiovisual work, hinting that it may have a sequel, prequel or is a TV series. Using memes This method of localisation is arguably less common as it targets even more precisely the audience of the source language and culture. It could be observed in the translation of the 2011 film Salmon Fishing in Yemen (Fig. 5), which in Russian literally means Ловля лосося в Йемене, however, it was released under the title Рыба моей мечты [The fish of my dreams] (Fig. 6). Another example is the 2012 comedy Lola Versus, i. e. Лола против in Russian, which was localised as a colloquial expression Давай, до свидания! [Good riddance!]. Fig. 5. Original movie poster for Salmon Fishing in the Yemen in English Source: Kinopoisk. Retrieved 21 July 2025, from https://www.kinopoisk.ru/film/484559/ posters/page/1/ Fig. 6. Localised movie poster for Salmon Fishing in the Yemen in Russian Source: Kinopoisk. Retrieved 21 July 2025, from https://www.kinopoisk.ru/film/484559/ posters/page/1/ Fig. 7. Original movie poster for Saint Laurent in English Source: Kinopoisk. Retrieved 21 July 2025, from https://www.kinopoisk.ru/film/735178/posters/ page/1/ Fig. 8. Localised movie poster for Saint Laurent in Russian Source: Kinopoisk. Retrieved 21 July 2025, from https://www.kinopoisk.ru/film/735178/posters/ page/1/ Expanding the title This strategy is employed when it makes sense to use the original title, but it is not always sufficient for the target audience to grasp the message of the film. In this case the title is expanded to include an explanation. For instance, the 2014 film Saint Laurent (Fig. 7), with the name of a famous designer in the original title, in translation gained an extension in the form of a sentence: Сен- Лоран. Стиль - это я [Saint Laurent. I am style] (Fig. 8). It seems that the Russian film distributors deemed it necessary to clarify for the target audience that the film is about fashion. One more example of this localisation strategy is the film Elysium (2013), which in Russian had an explanatory note: Элизиум. Рай на земле [Elysium. Heaven on Earth]. Discussion Evidently, while localising a film, it is not always necessary to resort to certain strategies for translating the title. In some cases, the title is self-evident and does not require changes in translation. For example, when a film is based on a book with the same title: Harry Potter and the Philosopher’s Stone (2001) → Гарри Поттер и философский камень; Memoirs of a Geisha (2005) → Мемуары Гейши; The Great Gatsby (2013) → Великий Гэтсби. Another instance is when the title could easily be translated into the target language word by word. To name a few, Titanic (1997) → Титаник; Avatar (2009) → Аватар; Once Upon a Time in Hollywood (2019) → Однажды в Голливуде. These are examples of literal translations. Another case where translation as such is not performed is when the film title is transliterated. This may be due to the use of proper names in the film title or the inability to find a euphonious equivalent in the target language. This case can be illustrated by the example of the series Virgin River (2019 - present), which is broadcast in Russia under the title Виргин Ривер [Virgin River]. Sometimes the film’s title and its meaning originate from some well-known phrase, which should be preserved in translation. One such example is the 2014 film Ex Machina which was rendered into Russian as Из машины [From machine]. The film tells a story about a scientist who makes robots and, by implanting them with artificial intelligence, tries to ensure that his creatures are as close to humans as possible in their behaviour, thinking, sensations and feelings, level of social adaptation and empathy. The title Ex Machina refers to the ancient Greek Deus ex machina [73], that was used in Greek and Roman theatre setting. To play a deity that is capable of flying, the actor was suspended by a rope and brought on stage with a device similar to a modern crane (from Greek mēchanē). That gave rise to the expression Deus ex machina, which in Russian was traditionally rendered as Бог из машины [74] [God from machine]. Thus, the Russian localised title Из машины makes a reference to the already existing phrase in the target language. This article does not aspire to provide an exhaustive list of film title localisation strategies, however, it offers an overview of general tendencies concerning film title translation taking into consideration social and cultural aspects. Moreover, it is not feasible to give a complete list of marketing strategies for the localisation of film titles, since translation strategy is always connected to a multitude of external factors that are constantly changing and reflecting the time period of the film release in a foreign market. Although some of the film titles seem quite easy to translate, they are always localised by a human marketing expert, who has seen the film itself, or at least its trailer. Machine translation systems are unable to take into consideration all the linguistic connotations, the variety of characteristics of the target audience as well as localisation trends. Even a human may err, as we have seen in case of the localised cartoon title Босс-молокосос 2 [Boss-milk sucker 2]. Conclusion This study has shown that, despite the rapid progress of machine translation technologies, it is impossible to use them to translate such marketing texts as film titles, at least at the current stage of development of neural machine translation systems. A neural network can be trained to script the entire film and analyse its content, provide a summary, and suggest variants of the film title translation on the basis of diegetic speech scripts, summary and the key ideas of the film. Nevertheless, the neural network as of now is unable to analyse the video sequence and match it with diegetic speech, as well as the emotions conveyed by the film’s cast and musical background. Therefore, a film title suggested by an AI on the basis of diegetic speech scripts cannot perform all the functions that a film title has to perform to ensure a box office hit. Even a marketing expert, when translating a film title, relies on the film or its trailer if the film itself is not available due to some reasons. In addition, one has to consider the time period in which a film is translated and released: certain social, cultural, economic, political factors that affect translation may become relevant. AI is currently unable to take into account the dynamic nature of the context because it is trained on already existing databases. AI may be unaware of the events and their implications if they are very recent. It is likely to miss connotations and associations that are in the process of being established in the target language and culture. So far, the AI is lagging behind the present timeline, so it can not translate a film title in view of all the marketing and localisation requirements that humans bear in mind. The conclusions of this study can be regarded as a prospect for the development of AI technologies in the translation industry.

About the authors

Yulia M. Alyunina

RUDN University; Peter the Great Museum of Anthropology and Ethnography (Kunstkamera) of the Russian Academy of Sciences (MAE RAS)

Author for correspondence.
Email: alyunina.y@mail.ru
ORCID iD: 0000-0002-4970-5584
SPIN-code: 1972-7723

candidate of Philological Sciences and Ph.D. in Linguistics, senior lecturer and research fellow, Department of Foreign Languages, Faculty of Philology, RUDN University; research assistant, Laboratory of Museum Technologies, Peter the Great Museum of Anthropology and Ethnography (Kunstkamera) of the Russian Academy of Sciences

6 Miklukho-Maklaya st., Moscow, 117198, Russian Federation; 7-9 Universitetskaya Embankment, Saint Petersburg, 199034, Russian Federation

Elvira R. Ganeeva

Financial University under the Government of the Russian Federation

Email: erganeeva@fa.ru
ORCID iD: 0000-0001-6473-6576
SPIN-code: 9831-6986

candidate of Philological Sciences, associate professor, Department of Foreign Languages and Intercultural Communication, Faculty of International Economic Relations

49 Leningradsky Avenue, bldg. 2, Moscow, 125167, Russian Federation

References

Supplementary files