Emojilingo: Harnessing AI to Translate Words into Emojis Francesca Chiusaroli1,∗,† , Federico Sangati2,† , Johanna Monti3,† , Maria Laura Pierucci3 and Tiberio Uricchio5 1 University of Macerata, Italy 2 Okinawa Institute of Science and Technology, Okinawa, Japan 3 Università Orientale di Napoli 4 University of Macerata, Italy 5 University of Pisa, Italy Abstract This paper presents an AI experiment of translation into emoji conducted on a glossary from Dante Alighieri’s Divine Comedy. The experiment is part of a project aiming to build up an automated emoji-based pivot language providing an interlingua as a tool for linguistic simplification, accessibility, and international communication: Emojilingo (emojilingo.org). The present test involves human (Emojitaliano) and machine (Chat-GPT) translations in a comparative analysis in order to devise an automated integrated model highlighting emojis’ expressive ability in transferring senses, clarifying semantic obscurities and ambiguities, and simplifying language. A first evaluation highlights Chat-GPT’s ability to deal with a classic archaic literary vocabulary, also raising issues on managing criteria for better grasping the meanings and forms and about the multicultural extent of content transfer. Keywords Emoji, Intersemiotic Translation, Emojitaliano, Emojilingo, Large Language Models 1. Introduction The use of emoji has considerably increased over time, and besides complementing written texts in online mes- Consisting today in 3,782 icons, regularly updated by sages and posts as a nice means to express feelings and Unicode Consortium,1 the emoji international catalogue emotional statuses, emojis are also used to completely contains signs for facial expressions, human ges- replace verbal language statements [1, 2]. tures, people activities, jobs, plants, animals, Experiments have been carried out to explore the fea- food, objects, symbols of travel, places, sibility of using emojis as language to convey meanings flags, numbers, and geometrical forms. through emoji-only translations. Notable examples in- While the visual content seems to be able to provide clude the popular Emoji Dick project, the translation into an encyclopedic list with universal significance, ideally emoji of Moby Dick [3], or Wonderland [4], an emoji capable of conveying language-independent meanings, poster created in 2014 to reproduce the full story of Lewis the interpretation of emojis is, on the contrary, highly Carroll’s Alice in Wonderland. These earliest experiments, arbitrary. They are strongly subject to ambiguities and however, lack codification and, as such, cannot be con- variations due to linguistic, cultural, and personal speci- sidered as a language, that is, a shared system in the ficities. Saussurean sense [5]. The first translation, in fact, was crowdsourced in a free and creative way, while the second CLiC-it 2024: Tenth Italian Conference on Computational Linguistics, one was an individual and literal translation experiment Dec 04 — 06, 2024, Pisa, Italy from English. ∗ Corresponding author. A concrete attempt to create a truly codified emoji † These authors contributed equally. language can be represented by Emojitaliano [6].2 Emoji- Envelope-Open f.chiusaroli@unimc.it (F. Chiusaroli); taliano is an emoji code originated from a crowdsourcing federico.sangati@gmail.com (F. Sangati); jmonti@unior.it (J. Monti); m.pierucci6@unimc.it (M. L. Pierucci); experiment initiated by a social community, specifically tiberio.uricchio@unipi.it (T. Uricchio) created to share a common emoji language able to coun- GLOBE https://docenti.unimc.it/f.chiusaroli (F. Chiusaroli); teract the natural polysemy of emojis.3 https://fede.sangati.me (F. Sangati); Born with the translation of Collodi’s Pinocchio, The https://unifind.unior.it/resource/person/2569 (J. Monti); Story of a Puppet [7] (figure A.1), the structure and glos- https://docenti.unimc.it/m.pierucci6 (M. L. Pierucci); https://www.dii.unipi.it/user/1906 (T. Uricchio) sary of Emojitaliano have been afterwards usefully reap- Orcid 0000-0003-1923-3974 (F. Chiusaroli); 0000-0001-6088-415X plied for the translation of texts of different genres such (F. Sangati); 0000-0002-4563-5988 (J. Monti); 0000-0003-3637-2757 (M. L. Pierucci); 0000-0003-1025-4541 (T. Uricchio) 2 https://www.treccani.it/vocabolario/emojitaliano_ © 2024 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). res-2f30d44e-89c2-11e8-a7cb-00271042e8d9_%28Neologismi%29 1 3 https://unicode.org/emoji/charts/full-emoji-list.html https://www.scritturebrevi.it 1 CEUR ceur-ws.org Workshop ISSN 1613-0073 Proceedings Francesca Chiusaroli et al. CEUR Workshop Proceedings 1–8 as the technical declaratory prose of the Italian Constitu- The paper is organised as follows: section 2 introduces tion (figure A.2), the Manifesto of non-hostile communi- the Emojilingo project Parole di Dante, the subject being cation (figure A.6), the narrative prose of classic moral translations in emoji of 365 words (Parole di Dante) from tales (i.e., The Wolf and the Lamb, in figure A.5), Gia- Dante’s poem Divine Comedy. Section 3 presents the como Leopardi’s lyrical poem L’infinito 4 (The infinite, in AI translation experiment carried out with two versions figure A.4). of Chat-GPT (3.5 and 4) [15] on the 365 Dante’s words, Emojitaliano is based on the assessment of conven- with a focus on the method and descriptions of some tional meanings and syntax, capable of guaranteeing the examples. Section 4 provides an evaluation of the results, sharing of sense by means of intersemiotic translation, also obtained through AI models and through a similarity beyond subjective interpretations.5 Emojitaliano pro- matrix, and the closing section includes conclusions and vides a grammatical structure and a shared vocabulary ideas on future work. which can be expanded and re-shared with each new translation[8].6 Recent experiments have opened new research hori- 2. Emojilingo: Parole di Dante zons in evaluating the capability of large language mod- The Emojilingo project is presented here as a follow els (LLMs) to translate words or text into emojis. This up of Emojitaliano. The general idea is that, through is predicated on the assumption that, given LLMs are the Emojitaliano community as control group, LLMs trained on extensive corpora sourced from the internet, technologies can develop and speed up the processes they have been exposed to emojis and are able to grasp of translation, enable wider and easier dissemination of the semantics of emojis [9]. Recently, Text2Emoji [10] the code, overcome the barriers of natural languages. The was proposed as an automatic translator, based on a large Emojilingo.org website republishes some Emojitaliano text-emoji parallel corpus, created by prompting the LLM, translations with English versions (see also Appendix A). Chat-GPT (OpenAI, 2023) and EmojiLM, a sequence-to- Our translation method pursues a program of concep- sequence model specialised in the text-emoji bidirectional tual linguistic simplification which can clarify linguistic translation. Another translation experiment involving meanings for the needs of international communication emojis, conducted by [11] is Emojinize. This experiment as well as for plain language policies [16]. The outcome leverages the power of LLMs to translate text by consid- never aims to replace the original sources, but rather in- ering both prior and subsequent contexts, which differs tends to provide a vehicular code useful to approach and from next-token prediction. Emojinize disambiguates directly understand words in any language [17, 18]. synonyms based on context, unlike a static lookup ta- The current work focuses on a new translation, based ble, and harnesses the expressive power of combining on Dante’s poem Divine Comedy, titled Parole di Dante multiple emojis. (Dante’s words). It is well known that Dante’s vocabulary Among the experiments, a first attempt using Chat- may be difficult to understand for foreign speakers, and GPT to learn the Emojitaliano grammar was also carried that a similar difficulty may occur for Italian speakers to- out in 2023 by the Emojitaliano research group. Assum- day, too, because of the many archaic or disused poetical ing the fundamental role of a conventional syntax as a words in the poem. Consequently, we believe that trans- basis for each shared code [12, 13], the aim was to verify lating this vocabulary into emojis can help mitigate these the ability of LLMs to learn and reapply the Emojitaliano comprehension difficulties and facilitate understanding. grammar rules to produce translations of Pinocchio on Parole di Dante consists of 365 emojis which are the its own [14]. Emojitaliano translations of 365 Dante’s words. The In this paper we present a follow-up experiment of source, together with the original context and explana- automatic translation into emoji, focused on special vo- tory comments, was published during 2021 as a daily so- cabulary. Chat-GPT’s translations of an authorial lexicon cial media dissemination event by the Italian Accademia have been tested and then compared to the corresponding della Crusca.7 On that occasion, through the partici- human solutions.The purpose is to test LLMs capabili- pation by the Emojitaliano Twitter/X community,8 the ties in autonomously rendering complex vocabulary, in #emojitaliano and #scritturebrevi social community pro- the horizon of building a translation tool into emoji as a duced emoji matches for all the 365 words, one per day. means of language simplification: the general project and Parole di Dante (Dante’s words) is therefore a glossary of the conlang itself are named Emojilingo and available Dante Alighieri’s Divine Comedy translated into emojis, online on emojilingo.org. with the corresponding Italian words. 4 https://www.scritturebrevi.it/?submit=Search&s=emojitaliano 5 The Emojitaliano grammar and glossary are registered on @emoji- 7 talianobot on Telegram; 3998 members are signed up to @emojital- https://accademiadellacrusca.it/it/dante 8 ianobot as of October 16th, 2024. The main accounts involved are @fchiusaroli and @stellissa/Maria 6 https://www.scritturebrevi.it/emojitalianobot Stella Bottai. 2 Francesca Chiusaroli et al. CEUR Workshop Proceedings 1–8 Table 1 Translation experiment with Chat-GPT 3.5 (IT) and Chat-GPT 4 (IT) of Dante’s “Divina Commedia”. For details see section 3.1. Term Verse AC Explanation Emojitaliano GPT-3.5 (IT) GPT-4 (IT) GPT-4 (IT) Explanation The pig emoji represents glut- Come talvolta stanno a ‘Mangione, ingordo, beone’, tony and greed, while the plate riva i burchi, / che parte latinismo derivato da lurcare emoji recalls the act of overeat- lurco sono in acqua e parte in ‘mangiare avidamente’. Nel ing. These two emoji together terra, / e come là tra li verso Dante attribuisce symbolise a “lurk”-a spoiled and Tedeschi lurchi “lurchi” a “li tedeschi”. gluttonous individual. A Latinism used by Dante to The ‘baiulo’ is a professional fig- Di quel che fé col baiulo refer to the ‘bearer’ of the sign ure who boasts of transporting seguente, / Bruto con of the empire, i.e., the emperor: something, so I chose the emoji Cassio ne l’inferno la- Dante was convinced that the baiulo of a worker (rural but represent- tra, / e Modena e Pe- universal institution of the an- ing the ethics of work) and a suit- rugia fu dolente. (Par- cient empire of Rome contin- case to represent the transporta- adise Vi, 73) ued even in his time, with Ar- tion. rigo VII. Poi l’addentar con più Battuta sarcastica dei diavoli a An emoji depicting a person di cento raffi, / disser: un barattiere; accaffare è voce climbing or scaling, followed by “Coverto convien che accaffi plebea dal significato certo di ’ a downward arrow to represent qui balli, / sì che, se ‘arraffare, acciuffare’, ma d’in- the action of falling or plummet- puoi, nascosamente ac- certa provenienza. ing. caffi” During a university course exercise,9 the glossary was machine would be author’s name and title of the work, later provided with the English matches.10 A comparison instead. This approach allowed us to test Chat-GPT’s between emoji solutions and English correspondences “autonomous” ability to handle this special lexicon di- was discussed in the classroom, bringing out the feasibil- rectly. ity of a chain translation method “text-emoji-text”. As a subsequent step, the research group tried a LLMs experi- 3.1. Translation Examples ment of translation from Italian and English into emojis. The resulting combination of human and AI translations, In table 1 we present some examples of the translation appropriately selected as will be shown below, is Parole experiment with Chat-GPT 3.5 and Chat-GPT 4 for the di Dante in Emojilingo.11 rendering of some terms, either rare or unusual, or now dismissed. The columns in the table are the following: 3. The AI translation experiment • the original Dante’s term; In this section we present the translation experiment of • the original verse containing the word; the 365 Italian terms from Dante’s Comedy with Chat- • the explanation by the Accademia della Crusca GPT. On the Emojilingo website, the translations chosen (AC); by Chat-GPT 4 (from the final evaluation explained in section 4.3) are available. 12 • the crowdsourced Emojitaliano translation; The very large database on which the LLM architecture is based makes us assume that the machine knows the • the Chat-GPT 3.5 (IT) translation; original text, belonging to the world literary canon, and • the Chat-GPT 4 (IT) translation, and we may also assume that it knows the English version of the work, as well as it will presumably have available • its explanation by Chat-GPT 4 (IT) itself (trans- multilingual Dante’s glossaries and commentaries. Un- lated into English for dissemination). like human translators, who translated on the basis of the explanations provided by the Accademia della Cr- 3.2. Methods usca, we decided that the only input to be given to the For our experiment we chose two models of Chat-GPT: 9 https://docenti.unimc.it/f.chiusaroli/courses/2023/28680 Chat-GPT 3.5 (turbo-0125), and Chat-GPT 4 (0613) as our 10 Sourced from https://dante.princeton.edu. reference models, to examine the differences in machine 11 https://emojilingo.org/parole_di_dante_about translations between the two models. In the second phase 12 In addition, the full data is publicly available both on of the project, we compared and evaluated these two an online spreadsheet (https://docs.google.com/spreadsheets/d/ versions against the human translations (Emojitaliano) . 13vkH3a-C0OpVTm9r5daFg_y0MN8lPASwGICaa72zaGg and on GitHub (https://github.com/EmojiLingo/emojilingo.github.io/tree/ To automatically translate the words into emojis with main/_chatgpt)). Chat-GPT we adopted a zero-shot prompting approach 3 Francesca Chiusaroli et al. CEUR Workshop Proceedings 1–8 using OpenAI APIs.13 Despite the archaic and often ob- 4.2. Preliminary study scure vocabulary, as already mentioned, no preliminary According to a preliminary evaluation, it is immediately training was provided and the only context given in the apparent that both versions of Chat-GPT can provide prompt was the reference to the work’s title. The prompt interesting translation solutions of Dante’s words and was provided both in Italian and English (the latter using are able to motivate their choices in a meaningful way. English terms), although the final evaluation was done One initial observation is that the translation solutions using the Italian version. The English prompt is here provided by the two Chat-GPT models often include mul- provided: tiple emojis, with Chat-GPT 3.5 doing so 88% of the time I will give you a word from Dante’s Divine and Chat-GPT 4 at 81%. In contrast the Emojitaliano Comedy and ask you to invent a trans- shows a higher tendency to use single emojis, doing so lation in emoji. Respond with a single in 49% of the cases. translation in 2 lines of plain text (with- The two Chat-GPT versions rarely provide the same out formatting): translations, except for some terms related to animals, such as ‘colubro’ or ‘lonza’. In some instances, particu- - translation into emoji larly those involving realia (e.g., ‘eagle’, ‘angel’, ‘book’, - brief explanation of the choice. ‘galaxy’), the translations provided by Chat-GPT align The word is ‘{term}’. with those given by human translators. In most cases, however, the solutions generated by In the next section we will present some comments as Chat-GPT 4 differ, as do the accompanying explanations. well as some evaluation remarks. The differences between the translations provided by the two versions of Chat-GPT are most often largely disparate. For example, the phrase ‘dolenti note’ is trans- 4. Evaluation lated by Chat-GPT 3.5 as and by Chat-GPT 4 as . Additionally, there are differences in the order- 4.1. Initial Comments ing of emojis, as observed in the translation of ‘occhi di bragia’, where Chat-GPT 3.5 uses and Chat-GPT The following are some initial comments on the examples 4 uses . In other instances, while both versions reported in Table 1. include a common emoji, they are paired with different lurco All the translations use an animal to represent the additional emojis; for instance, ‘inanellare’ is translated negative qualities expressed in the text, likely due to a by Chat-GPT 3.5 as and by Chat-GPT 4 as ; plausible interference from the English word ‘lurk’. Chat- ‘colubro’ is rendered with the snake in all cases, but GPT 3.5 focuses on the environmental nocturnal context Emojitaliano adds the skull to convey the accurate instead of the vice of gluttony. The choice of a specific an- meaning of the poisonous animal, as derived from the Ac- imal, as the pig or the wolf, to convey negative semantic cademia della Crusca comment. In some cases, Chat-GPT values, reflects an Eurocentric view, which raises issues translations correctly grasp the core idea of the word but for the multilingual and multicultural reception. dismiss the figurative strength of the original: ‘intuarsi’, meaning ‘intimate and deep understanding’ and ‘inter- baiulo The human translator reproduces the complex penetration between minds’, is in fact one of Dante’s meaning of the archaic word for “Emperor” as “bearer of original coinages (see also ‘infuturarsi’, etc.). Chat-GPT the sign of Empire”, while Chat-GPT translates it more 3.5 versions of ‘intuarsi’ as and appear not simply as the action of “carrying”, which simplifies but so poignant as the human literal solution seems more clarifies the direct meaning. expressing . Sometimes the Chat-GPT version succeeds in reproducing the sense more physically than accaffi Both the human and Chat-GPT 4 translations the human one, as for ‘trasumanar’ ‘to rise above the hu- convey the semantic value of rapid movement and ag- man’, by Chat-GPT 4 compared with the human gression, while the Chat-GPT 3.5 version emphasizes the version and Chat-GPT 3.5 translation . sarcastic tone used to depict despicable characters. The In a few cases, the translation solutions provided by issue of the symbolic representation of the animal icons both versions of Chat-GPT misinterpret the intended also emerges here. meaning of Dante’s word. For example, Chat-GPT 3.5 translates ‘zeba’ (meaning ‘goat’) as , erroneously con- flating it with a similar-looking word. Similarly, Chat- GPT 4 translates the term as , misinterpreting it as ‘cattle dung’. 13 https://platform.openai.com/docs/api-reference 4 Francesca Chiusaroli et al. CEUR Workshop Proceedings 1–8 Table 2 Results of the evaluation task by Chat-GPT 4. 1 1.0 Emojilingo GPT 3.5 GPT 4 Preferences 116 110 139 Percentage 31,8% 30,1% 38,1% 41 4.3. Chat-GPT 4 as evaluation agent To evaluate Chat-GPT 4’s ability to suggest the best trans- 0.8 lation solutions we organised an evaluation task run by 81 the model itself using the human crowdsourced transla- tion, the Chat-GPT 3 and the Chat-GPT 4 ones. Also, in this case we adopted a zero-shot prompting approach. The original Italian prompt is translated into English as follows: 122 I would like to ask you to evaluate 3 trans- lations from archaic Italian words extracted 0.6 from Dantes’s Divina Commedia into emoji. I will provide you with: Emojitaliano 162 - The Italian word - Emoji translation A - Emoji translation B - Emoji translation C I ask you to tell me which translation into 203 emoji do you prefer and why. Respond 0.4 with 2 lines of plain text (without format- ting) with the following info: - Choice: 243 - Explanation: Here you are: - Italian word: term - Translation A: emoji1 284 - Translation B: emoji2 0.2 - Translation C: emoji3 To ensure more reliable results we instructed the model 324 to perform 10 retries and select the most frequent answer. As the model was evaluating several translations ex ae- quo, we decided to reiterate the process until a difference was reached between the first and the second preferred 365 0.0 translation. The results of this evaluation task are shown GP o Ch .5 IT GP IT N at- 4 EN N in table 2. n Ch .5 E WI 4 Ch italia T3 Ch GPT T4 T3 T The data from the Chat-GPT 4 evaluation shows that GP oj GP at- Em at- at- Chat-GPT 4 has the highest preference score at 38.1%, at- Ch followed closely by Emojiitaliano at 31.8%, and Chat- GPT 3.5 at 30.1%. This suggests that Chat-GPT 4 was Figure 1: Similarity matrix between Emojitaliano and various generally rated more favourably compared to Chat-GPT versions of Chat-GPT engines. 3 and Emojitaliano. The distribution of proportions is essentially symmet- rical and balanced. The currently preferred translations have been compiled into a corpus validated as Emojilingo. 5 Francesca Chiusaroli et al. CEUR Workshop Proceedings 1–8 language: Emojilingo. Using a zero-shot prompting ap- 1.0 proach, both Chat-GPT versions (3.5 and 4) provided 0.9 tal ian o 1 0.1 0.13 0.12 0.13 0.36 emoji translations for 365 words extracted from Dante’s oji Comedy, along with explanations for the their own trans- Em 0.8 IT lation solutions. We also had Chat-GPT evaluate the 3.5 0.1 1 0.19 0.16 0.12 0.42 PT G 0.7 three different translations produced within the Emoji- at- Ch IT taliano project, alongside those produced by Chat-GPT. 4 0.13 0.19 1 0.14 0.18 0.53 PT 0.6 G at- The present experiment substantially succeeds in con- Ch EN firming AI easiness and ability to use emojis to convey .5 0.12 0.16 0.14 1 0.22 0.15 0.5 3 PT G at- linguistic meanings, also managing special and archaic Ch 0.4 EN vocabulary. We in fact tested the machine’s ability to 4 0.13 0.12 0.18 0.22 1 0.17 PT G at- handle denotative and connotative issues in the differ- Ch 0.3 N WI 0.36 0.42 0.53 0.15 0.17 1 ent translation choices, i.e. the translation solutions can 4 PT G at- 0.2 be multi-faceted, each one catching some of the many Ch o IT IT EN EN N ian WI .5 4 semantic features underlying words. In this sense most .5 4 tal 3 PT 4 3 PT PT PT G oji PT G at- G G Em at- G at- at- Ch at- Ch solutions may be acceptable, such as to demonstrate the Ch Ch Ch versatility of the emoji code to convey the senses. Figure 2: Similarity matrix between all models. Within this broad faculty of choice, however, some options seem quite critical, due to the dissimilarity of cultural values expressed by the languages, and by the 4.4. Similarity Matrix emojis themselves. That is, a main consequence of us- ing AI for translation, also in emojis, is the reaffirma- In figure 1 we report the similarity matrix between all tion of the crucial challenge in international translation: Emojitaliano 365 values and the corresponding values the need for careful attention to specific cultural dimen- provided by all Chat-GPT engines (both for Italian and sions during localization [19]. Cultural values underlie English).14 In figure 2 we report the similarity matrix texts, words and languages, as, for example, a ‘pig’ is an between all model pairs. In both figures, we include the ‘occidental’ symbol for negative concepts as ‘dirt’ and selection of the Chat-GPT 4 evaluation agent presented ‘gluttony’ (as in ‘lurco’), while the animal has a totemic in section 4.3, using the label Chat-GPT 4 WIN. or sacred value elsewhere; likewise, colors, or gestures, The similarity score between two strings, is computed take on cultural values according to societies and can- 15 using the Levenshtein distance: 𝛿. It is defined as the not be accorded univocal international meanings. The minimum number of single-character edits (insertions, choice of an icon as and international multilingual sign deletions or substitutions) required to change one word cannot override cultural peculiarities. Finally, cultural into the other. It is then normalized (i.e., divided by the vocabularies may vary on the basis of literary contexts maximum length of either strings). Finally the similarity and textual genres, often conveying suggestions related is obtained as 1 − 𝛿. The similarity is 1 (black) when the to signifiers that are now lost. Given the conservative two strings are identical, and 0 (white) when they have structure of poetical language, emoji translations may no emojis in common. In the heatmap of figure 2 the therefore need to move beyond the broadness of interlin- similarity between two models is computed as the mean gua to fully convey meanings by reproducing linguistic between all the 365 term-pairs similarities. signs ‘verbatim’ (es. ‘intuarsi’ ): that is, the lit- eral solution, usually ruled out from the perspective of an international semantic code, becomes substantial to 5. Conclusion recover the cultural dimension of a literary text [20]. Spe- In this paper we presented a translation experiment into cial care is therefore required in selecting corresponding emojis using two versions of Chat-GPT, to compare them matches in emoji so that they do not conflict with recep- with a human version, already available, realized within tion in different countries and societies and so that they the framework of the Emojitaliano experience. The present do succeed in reaching the core content of the original, project focuses on an integrated translation program, that which is the main purpose of ‘the emojilingua’. combines both human (Emojitaliano) and automated ap- Future research will always need a human evaluation proaches, as a basis for a constructed emoji-based pivot of automated outcomes, carried on by a team with exten- sive expertise in cross-cultural perspectives, and with a 14 deep understanding of cultural values of emojis. This will For clarity we include also Emojitaliano in the the first column, help to limit unrestricted creativity and ensure a wide which is all black as it is identical to the original values used as reference. common comprehension of Emojilingo, and its highest 15 https://en.wikipedia.org/wiki/Levenshtein_distance exportability. 6 Francesca Chiusaroli et al. CEUR Workshop Proceedings 1–8 Acknowledgement Conference on Computational Linguistics, CLiC-it 2015, 2015. URL: https://api.semanticscholar.org/ This work has been funded by the European Union - CorpusID:191807868. NextGenerationEU under the Italian Ministry of Univer- [13] F. Chiusaroli, Da emojipedia a pinocchio in emoji- sity and Research (MUR) National Innovation Ecosystem taliano: l’“emojilingua” tra scritture e riscritture, in: grant ECS00000041 - VITALITY - CUP D83C22000710005. Homo Scribens 2.0. Scritture ibride della modernità, The views and opinions expressed in the paper are the S. Lubello (Ed.), 2019. authors’ alone and remain entirely their responsibility. [14] F. Chiusaroli, T. Uricchio, J. Monti, M. L. Pierucci, F. Sangati, GPT-based Language Models meet Emo- jitaliano: A Preliminary Assessment Test between References Automation and Creativity, in: Proceedings of the [1] M. Danesi, The Semiotics of Emoji: The Rise of Ninth Italian Conference on Computational Lin- Visual Language in the Age of the Internet, Blooms- guistics, Venice, 2023. bury Publishing, London, New York, 2017. [15] OpenAI, J. Achiam, S. Adler, S. Agarwal, L. Ahmad, [2] V. Evans, The emoji code : how smiley faces, love …, B. Zoph, Gpt-4 technical report, 2024. URL: https: hearts and thumbs up are changing the way we //arxiv.org/abs/2303.08774. arXiv:2303.08774 . communicate, Michael O’Mara Books Limited, Lon- [16] R. Willerton, Plain Language and Ethical Action: A don, 2017. Dialogic Approach to Technical Content in the 21st [3] H. Melville, Emoji Dick, Or, The Whale, Lulu.com, Century, ISSN, Taylor & Francis, 2015. URL: https: 2010. URL: https://www.emojidick.com. //books.google.co.jp/books?id=9lWsCQAAQBAJ. [4] J. Hale, Wonderland, https://www.joehale.info/ [17] F. Chiusaroli, Emoji e semplificazione linguistica, in: visual-poetry/wonderland.html, 2014. Accessed: Comunicare il patrimonio culturale. Accessibilità 2024-10-17. comunicativa, tecnologie e sostenibilità, FrancoAn- [5] F. de Saussure, Course in General Linguistics, Duck- geli, 2021, pp. 164–193. worth, London, [1916] 1983. (trans. Roy Harris). [18] C. Bliss, Semantography (Blissymbolics): A Simple [6] J. Monti, F. Chiusaroli, F. Sangati, Emojitaliano: A System of 100 Logical Pictorial Symbols, which Can Social and Crowdsourcing Experiment of the Cre- be Operated and Read Like 1+2, A logical writing for ation of a Visual International Language, in: M. M. an illogical world, Semantography (Blissymbolics) Soares, E. Rosenzweig, A. Marcus (Eds.), Design, Publications, 1965. User Experience, and Usability: UX Research and [19] J. Monti, F. Chiusaroli, Il codice emoji da oriente a Design, Springer International Publishing, Cham, occidente : standard unicode e dinamiche di in- 2021, pp. 426–441. ternazionalizzazione, in: RILA : Rassegna Ital- [7] F. Chiusaroli, J. Monti, F. Sangati, Pinocchio in Emo- iana di Linguistica Applicata, Bulzoni, Roma, 2017, jitaliano, Apice Libri, 2017. pp. 83–101. URL: http://digital.casalini.it/10.1400/ [8] J. Monti, F. Sangati, F. Chiusaroli, M. Benjamin, 256182. S. Mansour, Emojitalianobot and emojiworldbot - [20] F. Chiusaroli, Alla ricerca della traduzione univer- new online tools and digital environments for trans- sale. L’infinito leopardiano tra lingue artificiali e lation into emoji, in: Proceedings of Third Italian codici non verbali, Macerata, eum, 2022, pp. 73–100. Conference on Computational Linguistics (CLiC-it 2016), 2016. [9] A. Srivastava, A. Rastogi, A. Rao, A. A. M. Shoeb, A. Abid, A. Fisch, …, Z. Wu, Beyond the imita- tion game: Quantifying and extrapolating the ca- pabilities of language models, 2023. URL: https: //arxiv.org/abs/2206.04615. arXiv:2206.04615 . [10] L. Peng, Z. Wang, H. Liu, Z. Wang, J. Shang, Emojilm: Modeling the new emoji language, 2023. URL: https://arxiv.org/abs/2311.01751. arXiv:2311.01751 . [11] L. H. Klein, R. Aydin, R. West, Emojinize: Enriching any text with emoji translations, 2024. URL: https: //arxiv.org/abs/2403.03857. arXiv:2403.03857 . [12] F. Chiusaroli, La scrittura in emoji tra dizionario e traduzione, in: Proceedings of the Second Italian 7 Francesca Chiusaroli et al. CEUR Workshop Proceedings 1–8 A. Emojitaliano Works ♾ ⌊⏳♾⌋ 🧍 '⌊🥰🥰⌋ 👇 ⌊⛰👶⌋ 🏜 , ➕👇🌳,➰ '🗝 🌅 🔭 . ☝ 🧍 '🧎 ➕ '🔮 , ➕ '📽 🌌 ♾🪂🌳,➕🔕 🪐,➕🐚🔆; 👉〰 ❤ '😱 . 🔜 🧍 '🎧 🌬 '⌊🍃🎶⌋ 🙌 ⌊🌿🌿⌋ , 🧍 '⚖ 🔕 ♾ 👉 ⌊🔔🌬⌋ : 🧍 '⌊💭🌟⌋ ⌊⏳♾⌋ ➕ ⌊🌗🌗⌋ ⌊🍂🍂⌋ , ➕ 🌗 📆 ➕🍃,➕🎶.🔜🙌🌌 💭 '⌊🌊🌪⌋ : 🧍 '🥰 '⌊🙆🌪⌋ 👇 🌊. Figure A.1: “Pinocchio” in Emojitaliano Emojitaliano 📜" ⌊☝ 📃 ☝ 📃 ⌋ ⌊🔝 🔝 ⌋ 💬 ' " '🈶 ' ⭐ ⚖ ⌊⚓ ◀ ⌋ ⤵ ⌊✋ 🛠 🖊 ⌋. 👑 '🈶 ⏩ 💬 w ⭐ '🈶 ' ➕ '🈶 ⌊🙅 ✂ ⌋, ➕ '⌊🗣 👍 ⌋ ➕ '⌊🏹 🆙 ⌋ 💬 • ⭐ '⌊🏹 🆙 ⌋ 📈 ⏩ 📚 ➕ ⌊🔍 🔎 ⌋ ⌊🔬 ⚗ ⌋ ➕ Figure A.4: L’“Infinito” in Emojitaliano ⌊👬 👭 ⌋, ↪ '🔧 👉 📎 ⌊〽 〽 ⌋ ➕ 📎 ⌊📐 📐 ⌋ ⏩ ⌊🔋 🔋 ⌋ ⏩ ⌊🏘 🏘 ⌋; ⭐ '⌊➡ 🔩 ⌋ ⤵ ⌊🔧 🔧 ⌋ ↪ '🎣 ⌊💻 🚀 ⌋. ⭐ '🛡 🏞 ➕ 💰 ⌊🏛 📖 ⌋ ➕ ⌊🎨 🎼 ⌋ ⏩ 📜. ◀ ⌊🏘 👑 ⌋ ⌊⭕ ⤴ ⌋ 📑 ⌊💯 🔝 ⌋; ⭐ '⌊➡ ↔ ⌋ ⌊☝ ☝ ⌋ ⌊👬 👭 👑 ⌋. ➕ ⌊🔧 🔧 ⌋ ⏩ ⌊R 📝 ⌋ ⌊⏩ 👈 ⌋ ▶ ⌊💪 💪 ⌋ ⏩ 🔋 💬 < ➕ ⏩ ⌊⭕ ⤴ ⌋. 💬 ⌊' › ⌋ ⭐ '⌊🗣 👍 ⌋ ➕ '⌊🗣 ✋ ⌋ ⌊💪 💪 ⌋ ⌊🙅 ⚔ 🙅 ⚔ ⌋ ⏩ ⌊R 📝 ⌋ " '⌊🈶 ↔ ⌋ ▶ ⌊R 📝 R 📝 ⌋ ⏩ ⌊☝ 🎓 📖 ⌋ 👫 , ➕ ↔ ⌊' 🕴 ⌋ ➕ ⤵ ⌊🔗 🔗 ⌋ ⌊F F ⌋ ⤵ 📩 ⏩ 💬 ~ 🌍 ↪ 💯 ⌊👬 👭 👬 👭 ⌋ '⌊🗣 👍 ⌋. 📊 ⌊R 📝 ⌋ ⏩ 👫 '➰ , ➕ ⭐ '⌊🗣 ☝ ⌋ 🙇 ⏩ ⌊☝ ☝ ⌋ ⌊🙅 📴 🙅 📴 ⌋ ⭐ '🛡 📎 ⌊R 📝 R 📝 ⌋ ⌊🆗 🆗 ⌋ ⌊🚻 🚻 ⌋ ⌊👤 🌍 ⌋ '⌊🈶 ↔ ⌋ ⌊R 📝 R 📝 ⌋ ➕ ⌊✍ ✍ ⌋ ⌊🌍 🌍 ⌋ ⏩ 💞 🗳,💶 ➕ F. ⌊🗣 🔤 🗣 🔤 ⌋. '☝ . ⌊👤 🌎 ⌋, ↪ ⤵ 🏘 ⌊⏩ 👈 ⌋ 🙅 '⌊🈶 📎 ⌋ 💪 🔩 ⏩ ⌊🗽 🗽 ⌋ ⌊⚖ ⚖ ⌋ ↪ 📜 " '⌊🗣 ✋ ⌋, ⌊🈶 📎 ⌋ 💪 💬 N 💬 ‚ ▶ ⌊⛺ 🤗 ⌋ ⤵ 🌐 ⏩ ⭐ " ↔ ⌊🎲 🎲 ⌋ ↪ ⌊R 📝 ⌋ 💯 ⌊👬 👭 ⌋ '⌊🈶 📎 ⌋ 💎 F ↔ ➕ '🈶 ⌊↔ ↔ ⌋ 🆚 ⌊🏘 👑 ⌋ " ➕ ⛪ '🈶 , 👤 ⤵ ⌊⭕ ⌋ ⌊⏩ 👈 ⌋ '⌊🗣 ☝ ⌋. ⭐ " '👎 ⌊🚪 👉 ⌋ ⏩ ⌊👤 🌍 ⌋ ↙ ⌊🚫 🚫 ⌋ ⌊R 📝 ⌋, 📴 ⌊🙅 ↔ ⌋ ↙ ⌊👗 👔 ⌋, 🎭 , ⌊🗣 🔤 ⌋, 🛐 , ⌊📴 ⛓ 📴 ⛓ ⌋ ➕ ⌊👑 👑 ⌋. ⌊🤝 ✝ 🤝 ✝ ⌋ '☝ ⌊🖇 🖇 ⌋ ⌊🗳 🗳 ⌋. ⌊💭 💭 ⌋ ⌊🗳 🗳 ⌋, ⌊📊 📊 ⌋ ⌊⏩ 👤 ⏩ 👤 ⌋ ➕ ⌊📊 📊 ⌋ 🎯 ⛪ ➕ ⭐ " . ⌊➿ 🎲 ➿ 🎲 ⌋ ⏩ ⌊🤝 ✝ 🤝 ✝ ⌋ 🎯 ⛪ ➕ ⭐ " ⌊👍 ◀ 👍 ◀ ⌋ ◀ < ⌊🍰 🍰 ⌋ 🙅 💬 ⌊' ' ⌋ ⌊F F ⌋. 👉 '🈶 ☝ ⏩ ⭐ '⌊✊ 📤 ⌋ ⌊🚧 🚧 ⌋ ⌊💶 💶 ⌋ ➕ ⌊🚧 🚧 ⌋ ⌊F F ⌋, ↪ , '⌊📐 ⬅ ⌋ ⌊🔩 ⬅ ⌋ 🗽 ➕ ⚖ '⌊🗣 ☝ ⌋ ⌊➿ 🎲 ⌋ ⌊⏩ 📜 " ⌋. " '⌊🙅 🙅 ⌋ ⌊💣 🔫 ⌋ ↔ 🔧 ⏩ 👊 🆚 🗽 ⏩ ⏩ ⌊👬 👭 ⌋, '⛔ 📈 💯 ⏩ 👫 ➕ ⌊👐 🖇 ⌋ 🔩 ⏩ 💯 ⌊👬 👭 👬 👭 ⌋ ⌊🆕 🆕 ⌋ ➕ ↔ 🔧 ↗ '📤 ⌊👊 👊 ⌋ 💬 ˆ ⌊🌍 🌍 ⌋; " '⌊🗣 👍 ⌋, ⤵ ⌊📊 📊 ⌋ ↔ 📎 ⌊👫 🛠 🖊 👫 🛠 🖊 ⌋ ▶ 🗃 🗳 , 💶 ➕ F ⏩ ⌊🏘 " ⌋. 💯 ⌊🛐 🛐 ⌋ '🈶 ⌊↔ ⬅ ⌋ ⌊🗽 🗽 ⌋ 🆚 ⌊R 📝 ⌋. ⌊🛐 🛐 ⌋ ⌊🏘 👑 🏘 👑 ⌋ ⌊🆕 🆕 ⌋ ▶ ⌊📐 📐 ⌋ ⏩ 👑 , ↪ '🆘 ↗ 💬 j ↪ '🈶 ⌊🆕 🆕 ⌋ ◀ ⛪ '⌊🈶 📎 ⌋ 💪 ⏩ '⌊🗃 🔗 ⌋ ↔ ' ⌊R 📝 ⌋ ↪ '⌊🗣 ✋ ⌋ ☮ ➕ ⚖ 🎯 💯 ⭐ '⌊🗣 👍 ⌋ ▶ 💯 ⌊👬 👭 ⌋ 💪 ▶ ⌊✋ 🛠 🖊 ⌋ ➕ ⌊📄 📄 ⌋ ⌊⏩ 👈 ⌋, 🎲 🙅 '⌊👞 🆚 ⌋ ⌊R 📝 ⌋ " . ⌊👬 👭 👬 👬 👭 ⌋; " '⌊🏹 🆙 ⌋ ➕ '🆘 ⌊🔗 🔗 ⌋ ⌊🖇 🖇 ⌋ ⏩ ⌊🛐 🛐 ⌋ 🙅 ⌊⛪ ⛪ ⌋ 📎 ⭐ " '⌊🈶 ↔ ⌋ ⌊🌍 🌍 ⌋ ↪ '👀 ▶ 👇 ↗ . 🐑 ➕ ⌊🐎🐎⌋ '⌊🏹 🆙 ⌋ ⌊📊 📊 ⌋ ↪ '⌊➡ 🔩 ⌋ 👇 💪 . 💯 ⌊👬 👭 ⌋ '⌊🈶 📎 ⌋ 💪 ⏩ '➿ , ↔ ⌊🎲 🎲 ⌋ ⌊⏩ 👈 ⌋ ➕ ↔ ⌊R 📝 ⌋ '☝ ⌊👈 👈 ⌋ '⌊⚓ ⬅ ⌋ ⤵ 🤝 🤝 📎 ⌊🎩 🎩 ⌋ La pecora e i cavalli ⌊⏩ 👈 ⌋. 💬 ⌊' < ⌋ ⌊😊 👍 ⌋ ⌊⏩ 👈 ⌋, ' ⌊✋ 🛠 🖊 ⌋ 🔁 ' 🏷 ↪ '⌊🤝 🖇 ⌋ ▶ 📈 💶 🔁 💌 ⏩ F . 🏳 ⏩ ⭐ '🈶 " : 💚 , ⚪ ➕ ❤ , 📎 N ⌊🎞 🎞 ⌋ ⌊📍 📍 ⌋ 📎 ⌊📏 📏 ⌋ ⌊↔ ↔ ⌋. Una pecora tosata vide dei cavalli, $ 🐑 ✂ '⌊👀◀⌋ ⌊🐎🐎⌋ , di Marina Pierani & Emojitaliano uno dei quali tirava un pesante carro, $ 🐎 '⌊🚂◀⌋ $ 🚜 🏋 , un altro portava un grande carico $ 🐎 🆕 '⌊🚛◀⌋ $ 🚚 🔝 Figure A.2: “Italian Constitution” in Emojitaliano e un altro trasportava un uomo. ➕ $ 🐎 🆕 '⌊🚛◀⌋ $ 👱 . La pecora disse ai cavalli: 🐑 '⌊🗣◀⌋ ▶ ⌊🐎🐎⌋ : "Mi piange il cuore vedendo " ℹ '⌊😭🔝⌋ '⌊👀⬅⌋ come l'uomo tratta i cavalli". ⌊🤔⁉⌋ 👱 '🤴 ⌊🐎🐎⌋ " . I cavalli le dissero: ⌊🐎🐎⌋ '⌊🗣◀⌋ ▶ 🐑 : 🎭🌟 "Ascolta, pecora: " 👆 '⌊❗👂⌋ , 🐑 : per noi è penoso vedere ⌊ℹℹ⌋ '⌊😭🔝⌋ 🔜 '⌊👀⌋ ⌊🎯🕰🥾⌋ ⏩ ⌊⏳⌛⌋ 🧍' ⌊🚊🗺◀⌋ ▶ ⌊🌲🐺🦉⌋ ⌊🌌🌑⌋ che l'uomo, nostro signore, 👱 , 🎩 ⌊⏩ℹℹ⌋ , ❓ 🧍 '⌊🤷🎈◀⌋ 🛣 ⚖ . si fa un vestito '⌊🔧👈⌋ $ ⌊👕👖⌋ con la lana delle pecore, 📎 ⌊🧶🐑⌋ ⏩ ⌊🐑🐑⌋ 😢 🧍 '⌊🗣💬⌋ 🏋 ❓ 👇 ⌊🌲🐺🦉⌋ '🈶 🐅 ➕ 🪨 ➕ 🦾 mentre le pecore restano senza lana". 🔁 ⌊🐑🐑⌋ '📴 ⌊🧶🐑⌋ . " 🔝 🧍 '⌊💭🔛⌋ ➕ '⌊😱🔛⌋ ! Dopo aver sentito ciò, 🐑 🔜 '⌊👂◀⌋ 👇 , ⌊🧙⚗🧪⌋ 🔝 ↔ ☠ ; la pecora se ne fuggì nei campi. 🐑 '⌊💨◀⌋ 🆚 ⌊🌾🌾🌾🌾⌋ . ✋ 🧍 '⌊🙏‼⌋ '⌊🗣📖⌋ 💟 ↪ 🧍 '⌊🕵◀⌋ , 🧍 '⌊🗣▶⌋ 📜 🆕 🧍 '⌊👀◀⌋ . Corso di Storia della traduzione UniMC 2018-19 & Emojitaliano ! Emojitaliano Corso di Storia della traduzione UniMC a.a. 2020-21 Figure A.5: “La pecora e i cavalli” in Emojitaliano Figure A.3: “Divina Commedia” in Emojitaliano 8 Francesca Chiusaroli et al. CEUR Workshop Proceedings 1–8 ( Il Manifesto della comunicazione non ostile ( ⌊🌬❄⌋ ➕ 🌞 La tramontana e il sole 1. ⌊🕰🔭⌋ ⌊🌬❄⌋ ➕ 🌞 '⌊💬 Virtuale è reale Si bisticciavano un giorno il vento Dico o scrivo in rete solo cose che ho il coraggio di dire di persona. 💥◀⌋, della tramontana e il sole, 2. ⌊🌬❄⌋ '⌊🗣‼◀⌋ '🈶 🦾🔝 l'uno pretendendo d'esser più forte Si è ciò che si comunica 🆚 🌞, ➕ 🔁, dell'altro, 🔜 ⌊🚹🚹⌋ '⌊👀◀⌋ ⌊🚶🧳⌋ Le parole che scelgo raccontano la persona che sono: mi rappresentano. quando videro un viaggiatore che 3. '👞 📎 🧥 🧷. veniva innanzi, avvolto nel mantello. Le parole danno forma al pensiero Mi prendo tutto il tempo necessario a esprimere al meglio quel che penso. 🔜 2⃣ ⌊🚹💬💥🚹💬💥⌋ '⌊☝ I due litiganti decisero allora che 🗣◀⌋ ⌊🚹🚹⌋ '⌊💭👌◀⌋ sarebbe ritenuto più forte chi fosse 4. Prima di parlare bisogna ascoltare 🦾🔝 👤 '⌊✌◀⌋ '📴 🧥 ⏩ riuscito a levare il mantello al viaggiatore. Nessuno ha sempre ragione, neanche io. Ascolto con onestà e apertura. ⌊🚶🧳⌋. 5. Il vento di tramontana cominciò a Le parole sono un ponte ⌊🌬❄⌋ '⌊🎬◀⌋ '🌬 ⌊👊⬅⌋; soffiare con violenza; Scelgo le parole per comprendere, farmi capire, avvicinarmi agli altri. 🤷, ⌊🌬❄⌋ '⌊🌬◀⌋🔝, ⌊🚶 ma, più soffiava, più il viaggiatore si 6. 🧳⌋ '⌊🧷◀⌋ 🧥🔝; stringeva nel mantello; Le parole hanno conseguenze ⌊🔝➡⌋ 🏁 🌬 😥 '⌊☝◀⌋ tanto che alla fine il povero vento So che ogni mia parola può avere conseguenze, piccole o grandi. '🏳 ▶ 🎯. dovette desistere dal suo proposito. 7. Condividere è una responsabilità 🔜 🌞 '⌊🌅◀⌋ ⤴ 🌌, ➕ ⌊🔜 Il sole allora si mostrò nel cielo, e Condivido testi e immagini solo dopo averli letti, valutati, compresi. 🤏⌋ ⌊🚶🧳⌋, ↪ '⌊🥵◀⌋, poco dopo il viaggiatore, che sentiva caldo, si tolse il mantello. 8. '⌊📴◀⌋ 🧥. Le idee si possono discutere. Le persone si devono rispettare E la tramontana fu costretta così a ✔ ⌊🌬❄⌋ '⌊☝◀⌋ '⌊🗣👍⌋ Non trasformo chi sostiene opinioni che non condivido in un nemico da annientare. riconoscere che il sole era più forte di 🌞 '🈶 🦾 🔝🆚 ⌊🌬❄⌋. lei. 9. Gli insulti non sono argomenti Non accetto insulti e aggressività, nemmeno a favore della mia tesi. 10. Emojitaliano & Storia della traduzione UniMC a.a. 2021-22 Anche il silenzio comunica Quando la scelta migliore è tacere, taccio. Emojitaliano - paroleostili.it Figure A.6: “Manifesto della comunicazione non ostile” in Figure A.7: “La tramontana e il sole” in Emojitaliano Emojitaliano 9