United States

Literary translation requires balancing target-language fluency with faithfulness to the source. Recent large language models (LLMs) often produce fluent translations, but it remains unclear whether fluency corresponds to semantic preservation in literary text. We examine this relationship using 130,486 translated paragraphs from 106 novels in 16 source languages, including human, Google Translate, and TranslateGemma translations. Fluency is measured as original-likeness with a translationese classifier trained on paragraph part-of-speech n-grams, and faithfulness with the automatic translation evaluation metric COMET-KIWI. We control for paragraph length and find a consistent negative correlation between fluency and faithfulness. The pattern appears for both human and Google Translate, but is weaker and often non-significant for TranslateGemma. These results show that segment length matters for automatic evaluation and suggest a tradeoff between fluency and faithfulness in literary translation.

ACL 2026

Fluency and Faithfulness in Human and Machine Literary Translation

literary translation

large language models

quality estimation

natural language processing

machine translation

### Welcome to 2025 Annual Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics

Welcome to the 2025 meeting of the Nations of the Americas Chapter of the Association for Computational Linguistics! I am proud to help organize the first NAACL conference to carry the new name of our organization, one that emphasizes inclusion for all of the Americas. I am also pleased to welcome you to Albuquerque, New Mexico, a state whose unique blend of cultural influences will make for an excellent backdrop for NAACL 2025, especially with this year’s special theme on NLP in a Multicultural World. 
**[Continue reading...](https://drive.google.com/file/d/1jX-qGhqVSZZCIrAnJaz798pflu5Irrdn/view?usp=sharing)**

*- Colin Cherry, Google, NAACL 2025 General Chair* 

[![](https://assets.underline.io/markdown_image/1/image/b087f8a4dc5816d6e1a6514e59c59ac3.png)](https://drive.google.com/file/d/1T96GzPqObXrMln2BMByCSXTSizjTg69P/view?usp=sharing)

You need to log in with the email address you registered with. Access credentials have been sent to your email. 

Please be sure to check your spam and other email folders if you do not see an email confirmation right away.

Please log in to explore this event.

To access NAACL 2025 event page you are required to register. Please follow [**this link**](https://2025.naacl.org/registration/registration/) to register. Access will depend on your registration type.

Please register!

NAACL 2025

Welcome to 2025 Annual Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics

workshop paper

It is my great pleasure and honor to welcome you to the 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026), held in San Diego, California, USA, July 2–7, 2026. ACL 2026 continues our fields tradition of excellence in scholarship, innovation, and inclusivity, and I am deeply grateful to the many volunteers who have worked tirelessly to bring this event to life.

It is exciting to see the many developments in the field reflected at our conference, such as the advancement of large language models (born out of our community) into many application directions. This is not without challenges, partly due to unsolved problems, and partly due to misuse of the technology. All this is reflected well at ACL 2026. The conference also reflects that the field is not only about advancing technology, but also the science of fundamentally understanding language and how it is use by many different speaker communities. 
[Full Message from General Chair Philipp Koehn](https://drive.google.com/file/d/1QiJFOmMjvDgsDo3i1XzlQnImxe3RU_Cy/view?usp=sharing) 
<html><button style="display: inline-flex; align-items: center; justify-content: center; white-space: nowrap; border-radius: 9999px; font-weight: bold; background: #7c3aed; color: white; font-family: 'Space Grotesk', sans-serif; height: 40px; font-size: 16px; padding: 0 20px; border: none; cursor: pointer" onclick="window.open('https://drive.google.com/file/d/1QiJFOmMjvDgsDo3i1XzlQnImxe3RU_Cy/view?usp=sharing')">Conference Handbook</button></html>

To access this event page, you need to log in with the **email address you registered with**. Access credentials will be sent to your email from Underline - subject line "Welcome to ACL 2026". Please be sure to check your spam email folder if you do not see an email confirmation right away.

To access this event page, you are required to register.
Please complete your registration to continue.

We recommend reading [**the registration information**](https://2026.aclweb.org/registration/) first.

Registration Required

The 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026) will take place in San Diego, California, from July 2 - 7, 2026.

We investigate narrative agency in hu-
man–LLM creative co-writing, asking who
drives story development in turn-based collabo-
ration. Using a new corpus of human–LLM co-
written stories, we apply sentiment and seman-
tic modeling to quantify affective alignment
and semantic novelty in turn-taking, and direc-
tional measures to assess which agent shapes
narrative progression. Our results show asym-
metric influence: human turns introduce greater
semantic novelty and are more likely to shape
subsequent developments, whereas LLM con-
tributions predominantly elaborate on human-
introduced elements. At the sentiment level,
alignment is also asymmetric, but more bidirec-
tional: LLMs exhibit stronger turn-level emo-
tional adaptation than humans, but both agents
track each other’s emotional valence and LLMs
show an independent tendency to more pos-
itive emotional baselines. These findings in-
dicate a complementary division of labor in
human–LLM co-writing, where humans drive
narrative innovation and direction, while LLMs
act as adaptive amplifiers that sustain coherence
and elaborate emerging narratives.

Directional Alignment and Narrative Agency in Human–LLM Co-Writing

AI-driven language technologies are increasingly used in hiring, but they may encode and reproduce harmful social stereotypes. Prior work often studies bias mitigation methods in isolation and outside realistic application settings. We examine the combined effects of data-level and model-level debiasing in a hiring-related context, using Norwegian-language academic bios and a proxy STEM/non-STEM classification task. Specifically, we study masking sensitive information, GenWriter-based rewrites \citep{soundararajan2025genwriter}, and adversarial debiasing \citep{han2021diverse,han2022towards}. We evaluate these interventions using downstream task performance, group fairness metrics, intrinsic bias tests based on WEAT \citep{caliskan2017semantics}, and measures of gender leakage from hidden representations. We find that combining masking, GenWriter rewrites, and adversarial debiasing substantially reduces gender leakage while maintaining or improving downstream performance. However, effects on fairness gaps and intrinsic bias are mixed, underscoring the need for downstream, context-sensitive evaluation of bias mitigation methods in hiring-related NLP.

Bias Mitigation in Hiring-Related NLP: Interactions Between Masking, Rewriting, and Adversarial Debiasing

While digitized corpora have transformed the study of intellectual transmission, current methods rely heavily on lexical text reuse detection, capturing verbatim quotations but fundamentally missing paraphrases and complex implicit engagement. This paper evaluates semantic search in 18th-century intellectual history through the reception of John Locke’s foundational work. Using expert annotation grounded in a semantic taxonomy, we examine whether an off-the-shelf semantic search pipeline can surface meaning-level correspondences overlooked by lexical methods. Our results demonstrate that semantic search retrieves substantially more implicit receptions than lexical baselines. However, linguistic diagnostics also reveal a "lexical gatekeeping" effect, where retrieval remains partially constrained by surface vocabulary overlap. These findings highlight both the potential and the limitations of semantic retrieval for analyzing the circulation of ideas in large historical corpora.

Matching Meaning at Scale: Evaluating Semantic Search for 18th-Century Intellectual History through the Case of Locke

How did the thematic repertoire of early English-language science fiction change as the genre consolidated between 1818 and 1930? Using a corpus of 238 public-domain texts, we apply temporally binned latent Dirichlet allocation (LDA), comparing models with and without Authorless preprocessing (which probabilistically downweights author-specific vocabulary). Cross-period topic alignments exceed a permutation null baseline, indicating continuity in topic structure over time. Full-corpus LDA can produce comparable topic quality, but only temporal binning enables diachronic alignment; within the binned setting, Authorless reduces author concentration and modestly increases the share of thematic topics without materially reducing coherence. Four high-continuity topic chains -- centered on mobility, affect, planetary scale, and scientific knowledge -- suggest a shift from earlier romantic and speculative concerns toward more consolidated technoscientific forms. These chains generate interpretable hypotheses about the literary history of early science fiction, and the workflow supports diachronic analysis in small, author-skewed corpora.

Tracing Thematic Change in Early English-Language Science Fiction, 1818-1930

This study investigates whether a high-quality, 19-label named entity recogniser for medieval Latin charters can be constructed using only a few hundred annotated sentences. The authors introduce "semantic scaffolding," an innovation that utilizes richly descriptive English label phrases as prompts to activate latent multilingual knowledge within the model. This is paired with a custom span-based architecture utilizing XLM-ROBERTa-large, 4-head attention pooling to handle long property descriptions, and a hybrid loss system including Asymmetric Focal-Dice and InfoNCE contrastive terms. Results demonstrate that semantic scaffolding enables fine-tuned GLiNER to reach 80.8% overlap F1, while the custom architecture achieves 83.4% overlap F1 using only 298 training sentences. Significantly, the paper provides an empirical demonstration that domain-specific pre-training on medieval Latin offers no performance advantage once task-specific fine-tuning is applied. While the model excels at frequent categories like PER (95.7% F1) and LOC (93.5% F1), challenges persist for rare, position-dependent legal categories such as LEG (53.1% F1) and TRANS (52.6% F1).

Twenty’s Plenty: Semantic Scaffolding and Span Architecture for 19-Label NER in Medieval Latin Charters

Motherboard’s Mother Tongue is a computational linguistics and artistic research project that explores a Large Language Model’s (LLM) vocal production of glossolalia-like language. Glossolalia, colloquially known as ‘speaking in tongues,’ consists of the human production of seemingly unintelligible utterances. It is, by its nature, difficult to annotate accurately with linguistic features relevant for natural language. The glossolalia-producing system demonstrated here consists of the interaction of 1) a ‘nonsense’ linguistic corpus 2) a micro-controller based environmental data stream and 3) a fine-tuned LLM. While discussing some philosophical and artistic considerations of machinic glossolalia, we also address some methodological considerations for NLP. Using the artistic project as a case study, we argue that machinic glossolalia presents a ‘stress test’ that could inform some creative redirections of NLP methods and reconsiderations of definitions held by the subfield.

Artistic Interventions for NLP Annotation Challenges: The Stress Test of Machinic Glossolalia

This paper addresses a practical problem in computational literary history: retrieving adventure novels from a large digitized collection of French fiction where genre metadata are sparse and unreliable. We begin with supervised genre modeling based on a historically situated seed list of 101 adventure novels drawn from literary scholarship. We compare several classifiers and representations, and validate them against 364 independently labeled adventure novels from the Chapitres corpus. The best-performing model, HistGradientBoosting on mean paragraph embeddings, achieves strong external recall (81\%) despite the small training set. We then apply this model to the 12,176-novel Fictions litt\éraires de Gallica archive and refine the resulting candidate corpus through a graph-based post-processing step over a $k$-nearest-neighbor similarity graph. On the Chapitres benchmark, this graph correction produces negligible changes in retrieval performance, indicating that it is not a generally superior classifier. On Gallica, however, it yields a more cohesive and homogeneous candidate corpus and surfaces interpretable correction cases, including missed canonical adventure novels and excluded borderline texts. We therefore argue that graph-based correction is best understood not as a replacement for supervised classification, but as a heuristic for refining large, noisy archival corpora where exhaustive manual annotation is impossible.

In Search of Lost Adventure Novels: Supervised Genre Retrieval and Corpus Refinement in Gallica

This study examines narratives in which students describe challenges they faced in higher education due to low socioeconomic (SES) backgrounds and the strategies they used to overcome them. Using computational text analysis, we operationalize three educational theories, Paulo Freire’s Critical Pedagogy, Urie Bronfenbrenner’s Ecological Systems Theory, and Pierre Bourdieu’s Theory of Capital and Habitus to analyze patterns in these narratives. To strengthen the theory-to-method connection, we incorporate temporal timeline extraction, identifying ordered event sequences and tracking how challenges and forms of capital evolve across a student’s posting history. This temporal lens links theoretical categories (barriers, supports, forms of capital) to when they occur, highlighting moments for timely interventions. By combining theory-driven features with temporal analysis, we evaluate the explanatory capacity of each framework and demonstrate how computational methods can quantitatively examine qualitative lived experience at scale, supporting interdisciplinary research on equity in education.

Computational Modeling of Educational Theory in Low-Socioeconomic Contexts

The diachronic evolution from Latin to the Romance languages involved a restructuring of the grammatical gender system from a tripartite configuration (masculine, feminine, neuter) to a bipartite one (masculine, feminine). In this work, we introduce an interpretable deep learning framework to investigate this phenomenon at both lexical and contextual levels. First, we show that conventional tokenization strategies are insufficiently robust for this low-resource historical setting, and that our proposed tokenizer improves performance over these baselines. At the lexical level, we evaluate the contribution of morphological features to grammatical gender prediction. Subsequently, at the contextual level, we analyze how different part-of-speech categories contribute to grammatical gender prediction.

Lost in Translation? Exploring the Shift in Grammatical Gender from Latin to Occitan

Part-of-speech (POS) tagging for Medieval Romance languages remains challenging due to orthographic variation, morphological complexity, and limited annotated resources. This paper presents a systematic empirical evaluation of large language models (LLMs) for POS tagging across three medieval varieties: Medieval Occitan, Medieval Catalan, and Medieval French. We compare traditional rule-based and statistical taggers with modern open-source LLMs under zero-shot prompting, few-shot prompting, monolingual fine-tuning, and cross-lingual transfer learning settings.

Experiments on historically grounded datasets show that LLM-based approaches consistently outperform traditional taggers, with fine-tuning and multilingual training yielding the largest improvements. In particular, cross-lingual transfer learning substantially benefits under-resourced varieties, while targeted bilingual training can outperform broader multilingual configurations for specific target languages. The results highlight the importance of linguistic proximity and dataset characteristics when designing transfer strategies for historical NLP.

These findings provide empirical insights into the applicability of modern neural methods to medieval text processing and provide practical guidance for deploying LLM-based POS tagging pipelines in digital humanities research. All code, models, and processed datasets are released for reproducibility.

Downloads

Next from ACL 2026

Directional Alignment and Narrative Agency in Human–LLM Co-Writing

Similar lecture

How Good Are LLMs for Literary Translation, Really? Literary Translation Evaluation with Humans and LLMs

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from ACL 2026

Directional Alignment and Narrative Agency in Human–LLM Co-Writing

Similar lecture

How Good Are LLMs for Literary Translation, Really? Literary Translation Evaluation with Humans and LLMs

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Downloads