Research and Publications
Academic research is useful only when it becomes operational infrastructure. Pangeanic’s work in machine translation, multilingual data, speech corpora, evaluation, MTQE and model alignment feeds directly into the systems enterprises use today.
Explore Pangeanic’s academic publications, European R&D projects and scientific presentations to product development: AI Data Services, Datasets, Machine Translation, Deep Adaptive AI Translation, MTQE, Evaluation and AI QA and Model Alignment.
Our journey is simple: research produces methods, methods become product features, and product features are tested in operational use cases.
Applied research
Academic papers solve technical problems. Production AI requires solving workflow failures. Pangeanic’s research agenda has been shaped by practical constraints: multilingual data scarcity, translation quality estimation, terminology control, speech data generation, model adaptation, and the evaluation of outputs before they reach real users.
The value of research is not the publication itself. The value appears when research becomes a dataset, an engine, a control layer, a human review workflow, or a deployable system. We document that path publicly here.
Research profiles
These profiles connect research work, publications and technical presentations to Pangeanic’s current products in multilingual AI data, machine translation, MTQE, model alignment and sovereign AI systems.
Founder and research direction
CEO and founder of Pangeanic. Manuel’s research record connects Pangeanic’s early machine translation work with today’s product lines in adaptive MT, MTQE, multilingual AI data, enterprise document translation and sovereign deployment.
Google Scholar →Machine learning and data spaces
Head of Machine Learning. His work includes speech corpus generation, information retrieval and multimodal data spaces, connecting Pangeanic’s speech datasets, QA retrieval, data processing and AI platform work.
ELE2 speech corpus report →Speech and multilingual ASR
His work on speech corpora, code-switching transcription and multilingual text classification connects directly with Pangeanic’s speech datasets, ASR evaluation, language identification and multilingual NLP pipelines.
DAAIT and MTQE
Machine translation researcher presenting Pangeanic work on multi-agent MT, post-editing, MTQE and Deep Adaptive AI Translation at AMTA, LREC and AI4Culture contexts.
AMTA 2025 →LLMs and NLP research
Head of LLMs and AI Research. His publications cover sentiment analysis, persuasion detection, distant supervision, relation extraction, evaluation methodology and machine translation adaptation.
Google Scholar →European MT infrastructure
Research and engineering contributions connected to European machine translation infrastructure, NTEU data collection work and the creation of more than 500 neural machine translation engines for EU languages.
Google Scholar →From research to production
Pangeanic’s research record is most relevant when it explains what the company can build. The same work that produced speech corpora, neural MT engines, evaluation methods and adaptive translation research now supports commercial AI data workflows, MTQE, DAAIT and sovereign deployment.
01 · Speech and low-resource data
Research: Large speech corpus generation for the languages of Spain using data augmentation.
Product capability: Speech datasets, data augmentation, ASR evaluation and low-resource language collection.
Commercial path: Bespoke speech and audio datasets for public administrations, AI labs and media organizations.
02 · Neural MT infrastructure
Research: Large-scale multilingual corpus design and neural machine translation engine creation.
Product capability: Machine Translation, parallel corpora, domain adaptation and multilingual engine deployment.
Commercial path: Secure MT and multilingual AI data infrastructure for enterprise and public-sector workflows.
03 · Adaptive translation
Research: Incremental NMT adaptation, post-editing workflows and domain-specific machine translation.
Product capability: Deep Adaptive AI Translation using client terminology, translation memories and domain resources.
Commercial path: Adaptive translation workflows for organizations that need tone, terminology and domain control.
04 · MTQE and review routing
Research: Multi-agent translation, post-editing automation and quality estimation presented in AMTA and LREC contexts.
Product capability: MTQE, automatic corrective loops and review routing by quality threshold.
Commercial path: Lower review waste and stronger quality control in high-volume translation operations.
05 · Evaluation and NLP
Research: Distant supervision, noise reduction, sentiment analysis, classification and evaluation methodology.
Product capability: Evaluation datasets, model diagnostics, data quality workflows and AI QA.
Commercial path: Higher-quality data and better evaluation for task-specific multilingual models.
06 · Human feedback and alignment
Research: Annotation, ranking, evaluation, preference data and human review across multilingual projects.
Product capability: PECAT, RLHF, model alignment and human-governed AI operations.
Commercial path: Human-supervised data and alignment workflows for enterprise AI systems.
Selected publications and presentations
The entries below show how research outputs connect to Pangeanic’s current commercial and technical capabilities.
Manuel Herranz, Marina Albert · LREC 2026 Industry Day
Presentation of Pangeanic’s Deep Adaptive AI Translation work, connecting LLM-based translation, terminology adaptation, domain control and multi-agent language workflows.
LREC 2026 programme →Marina Albert Girona · AMTA 2025
Presentation connected to Pangeanic’s development of DAAIT, MTQE, automatic post-editing and quality routing through specialized translation agents.
AMTA 2025 programme →Marina Albert Girona · AI4Culture 2025, Vilnius
Workshop on machine translation and CAT tooling developed in the context of the Europeana and AI4Culture ecosystem, connecting cultural-sector AI with practical translation workflows.
AI4Culture resource →José Miguel Herrera, Moisés Barrios · European Language Equality 2
Technical report connecting Pangeanic’s data generation work to speech datasets, data augmentation and low-resource language support.
Read report →Pangeanic, KantanMT, Tilde · Connecting Europe Facility
Large-scale European machine translation project involving multilingual data collection and the creation of more than 500 neural translation engines for EU languages.
NTEU project →Mercedes García Martínez, Manuel Herranz, Amando Estela, Laurent Bié, A. Franco · IWLT
Early evidence of Pangeanic’s platform approach to industrial neural machine translation, engine management and client-facing MT workflows.
Read paper →Miguel Domingo, Mercedes García Martínez, Álvaro Peris, Alexandre Helle, Amando Estela, Laurent Bié, Francisco Casacuberta, Manuel Herranz
Research into adaptive NMT and professional post-editing, directly connected to Pangeanic’s long-standing work on human feedback, model adaptation and translation productivity.
Read paper →José Miguel Herrera, Barbara Poblete, Denis Parra · ECIR 2018
Research on ranking and information retrieval, relevant to semantic search, QA retrieval and retrieval-based AI workflows.
Springer →J. M. de la Torre Vilariño, Juan Luis García Mendoza, A. Petrucci · EAMT 2023
Case study on domain adaptation in luxury e-commerce translation, relevant to Pangeanic’s Deep Adaptive AI Translation and client-specific MT work.
L. Bustio-Martínez, V. Herrera-Semenets, Juan Luis García Mendoza et al. · IWAIPR 2023
Research on automatic detection of textual patterns using machine learning, connected to classification, evaluation and content intelligence workflows.
DOI →Juan Luis García Mendoza, L. Villaseñor Pineda, D. Buscaldi, L. Bustio-Martínez, F. Orihuela Espina · MICAI 2022
Work on noise reduction for relation extraction, relevant to data quality, annotation consistency and NLP evaluation.
DOI →Juan Luis García Mendoza, L. Villaseñor Pineda, F. Orihuela Espina · Procesamiento del Lenguaje Natural
Evaluation methodology work relevant to Pangeanic’s AI QA, benchmarking, data quality and model evaluation services.
DOI →From paper to deployment
Research becomes commercially useful when it lowers operational uncertainty. Pangeanic’s research record connects directly to production systems that handle multilingual data, machine translation, evaluation, human feedback and controlled deployment.
This is the reason our proof structure links academic work, European projects, scientific events, products and use cases in one sequence.
Research-driven AI systems
Pangeanic helps organizations move from multilingual data and research prototypes to production AI workflows with governance, human review and secure deployment.
