This research explores Cultural Transcreation (CT) for East Asian languages, focusing primarily on Mandarin Chinese (ZH) and the customer service (CS) market. We combined Large Language Models (LLMs) with prompt engineering to develop a CT product that, aligned with the Augmented Translation concept, enhances multilingual CS communication, enables professionals to engage with their target audience effortlessly, and improves overall service quality. Through a series of preparatory steps, including guideline establishment, benchmark validation, iterative prompt refinement, and LLM testing, we integrated the CT product into the CS platform, assessed its performance, and refined prompts based on a pilot feedback. The results highlight its success in empowering agents, regardless of linguistic or cultural expertise, to bridge effective communication gaps through AI-assisted cultural rephrasing, thus achieving its market launch. Beyond CS, the study extends the concept of transcreation and prompt-based LLM applications to other fields, discussing its performance in the language conversion of website content and advertising.

We present how at Unbabel we have been using Large Language Models to apply a Cultural Transcreation (CT) product on customer support (CS) emails and how we have been testing the quality and potential of this product. We discuss our preliminary evaluation of the performance of different MT models in the task of translating rephrased content and the quality of the translation outputs. Furthermore, we introduce the live pilot programme and the corresponding relevant findings, showing that transcreated content is not only culturally adequate but it is also of high rephrasing and translation quality.

This paper presents the findings from the third edition of the Chat Translation Shared Task. As with previous editions, the task involved translating bilingual customer support conversations, specifically focusing on the impact of conversation context in translation quality and evaluation. We also include two new language pairs: English-Korean and English-Dutch, in addition to the set of language pairs from previous editions: English-German, English-French, and English-Brazilian Portuguese.We received 22 primary submissions and 32 contrastive submissions from eight teams, with each language pair having participation from at least three teams. We evaluated the systems comprehensively using both automatic metrics and human judgments via a direct assessment framework.The official rankings for each language pair were determined based on human evaluation scores, considering performance in both translation directions—agent and customer. Our analysis shows that while the systems excelled at translating individual turns, there is room for improvement in overall conversation-level translation quality.

Task-oriented conversational datasets often lack topic variability and linguistic diversity. However, with the advent of Large Language Models (LLMs) pretrained on extensive, multilingual and diverse text data, these limitations seem overcome. Nevertheless, their generalisability to different languages and domains in dialogue applications remains uncertain without benchmarking datasets. This paper presents a holistic annotation approach for emotion and conversational quality in the context of bilingual customer support conversations. By performing annotations that take into consideration the complete instances that compose a conversation, one can form a broader perspective of the dialogue as a whole. Furthermore, it provides a unique and valuable resource for the development of text classification models. To this end, we present benchmarks for Emotion Recognition and Dialogue Quality Estimation and show that further research is needed to leverage these models in a production setting.

This work proposes an approach to use Part-Of-Speech (POS) information to automatically detect context-dependent Translation Units (TUs) from a Translation Memory database pertaining to the customer support domain. In line with our goal to minimize context-dependency in TUs, we show how this mechanism can be deployed to create new gender-neutral and context-independent TUs. Our experiments, conducted across Portuguese (PT), Brazilian Portuguese (PT-BR), Spanish (ES), and Spanish-Latam (ES-LATAM), show that the occurrence of certain POS with specific words is accurate in identifying context dependency. In a cross-client analysis, we found that ~10% of the most frequent 13,200 TUs were context-dependent, with gender determining context-dependency in 98% of all confirmed cases. We used these findings to suggest gender-neutral equivalents for the most frequent TUs with gender constraints. Our approach is in use in the Unbabel translation pipeline, and can be integrated into any other Neural Machine Translation (NMT) pipeline.

This paper describes the research developed at Unbabel, a Portuguese Machine-translation start-up, that combines MT with human post-edition and focuses strictly on customer service content. We aim to contribute to furthering MT quality and good-practices by exposing the importance of having a continuously-in-development robust Named Entity Recognition system compliant with General Data Protection Regulation (GDPR). Moreover, we have tested semiautomatic strategies that support and enhance the creation of Named Entities gold standards to allow a more seamless implementation of Multilingual Named Entities Recognition Systems. The project described in this paper is the result of a shared work between Unbabel ́s linguists and Unbabel ́s AI engineering team, matured over a year. The project should, also, be taken as a statement of multidisciplinary, proving and validating the much-needed articulation between the different scientific fields that compose and characterize the area of Natural Language Processing (NLP).

In this work, we propose a Named Entity handling approach to improve translation quality within an existing Natural Language Processing (NLP) pipeline without modifying the Neural Machine Translation (NMT) component. Our approach seeks to enable fast delivery of such improvements and alleviate user experience problems related to NE distortion. We implement separate NE recognition and translation steps. Then, a combination of standard entity masking technique and a novel semantic equivalent placeholder guarantees that both NE translation is respected and the best overall quality is obtained from NMT. The experiments show that translation quality improves in 38.6% of the test cases when compared to a version of the NLP pipeline with less-developed NE handling capability.

This paper presents an analysis of discourse markers in two spontaneous speech corpora for European Portuguese - university lectures and map-task dialogues - and also in a collection of tweets, aiming at contributing to their categorization, scarcely existent for European Portuguese. Our results show that the selection of discourse markers is domain and speaker dependent. We also found that the most frequent discourse markers are similar in all three corpora, despite tweets containing discourse markers not found in the other two corpora. In this multidisciplinary study, comprising both a linguistic perspective and a computational approach, discourse markers are also automatically discriminated from other structural metadata events, namely sentence-like units and disfluencies. Our results show that discourse markers and disfluencies tend to co-occur in the dialogue corpus, but have a complementary distribution in the university lectures. We used three acoustic-prosodic feature sets and machine learning to automatically distinguish between discourse markers, disfluencies and sentence-like units. Our in-domain experiments achieved an accuracy of about 87% in university lectures and 84% in dialogues, in line with our previous results. The eGeMAPS features, commonly used for other paralinguistic tasks, achieved a considerable performance on our data, especially considering the small size of the feature set. Our results suggest that turn-initial discourse markers are usually easier to classify than disfluencies, a result also previously reported in the literature. We conducted a cross-domain evaluation in order to evaluate the robustness of the models across domains. The results achieved are about 11%-12% lower, but we conclude that data from one domain can still be used to classify the same events in the other. Overall, despite the complexity of this task, these are very encouraging state-of-the-art results. Ultimately, using exclusively acoustic-prosodic cues, discourse markers can be fairly discriminated from disfluencies and SUs. In order to better understand the contribution of each feature, we have also reported the impact of the features in both the dialogues and the university lectures. Pitch features are the most relevant ones for the distinction between discourse markers and disfluencies, namely pitch slopes. These features are in line with the wide pitch range of discourse markers, in a continuum from a very compressed pitch range to a very wide one, expressed by total deaccented material or H+L* L* contours, with upstep H tones.

This paper presents a linguistic revision process of a speech corpus of Portuguese broadcast news focusing on metadata annotation for rich transcription, and reports on the impact of the new data on the performance for several modules. The main focus of the revision process consisted on annotating and revising structural metadata events, such as disfluencies and punctuation marks. The resultant revised data is now being extensively used, and was of extreme importance for improving the performance of several modules, especially the punctuation and capitalization modules, but also the speech recognition system, and all the subsequent modules. The resultant data has also been recently used in disfluency studies across domains.