Skip to main content

METODOS DE LA LINGUISTICA COMPUTACIONAL PARA LA LEGIBILIDAD Y SIMPLIFICACION AUTOMATICA DEL DISCURSO MEDICO

PID2020-116001RA-C33

•
Funding agency name Agencia Estatal de Investigación
Funding agency acronym AEI
Program Programa Estatal de I+D+i Orientado a los Retos de la Sociedad
Call Proyectos de I+D+i (Generación de Conocimiento y Retos Investigación)
Call year 2020
State Plan 2017-2020 Plan Estatal de Investigación Científica y Técnica y de Innovación
Beneficiary centre AGENCIA ESTATAL CONSEJO SUPERIOR DE INVESTIGACIONES CIENTIFICAS (CSIC)

Resultados relacionados

Resultados totales (Incluyendo duplicados): 9
Encontrada(s) 1 página(s)

DIGITAL.CSIC. Repositorio Institucional del CSIC
Publication Artículo científico (JournalArticle)

Building a comparable corpus and a benchmark for Spanish medical text simplification

  • Campillos-Llanos, Leonardo
  • Terroba Reinares, Ana Rosa
  • Zakhir Puig, Sofía
  • Valverde Mateos, Ana
  • Capllonch Carrión, Adrián
Este artículo está sujeto a una licencia CC BY-NC-ND 4.0, [EN] We report the collection of the CLARA-MeD comparable corpus, which is made up of 24 298 pairs of professional and simplified texts in the medical domain for the Spanish language (>96M tokens). Texts types range from drug leaflets and summaries of product characteristics (10 211 pairs of texts, >82M words), abstracts of systematic reviews (8138 pairs of texts, >9M words), cancer-related information summaries (201 pairs of texts, >3M tokens) and clinical trials announcements (5748 pairs of texts, 451 690 words). We also report the alignment of professional and simplified sentences, conducted manually by pairs of annotators. A subset of 3800 sentence pairs (149 862 tokens) has been aligned each by 2 experts, with an average inter-annotator agreement kappa score of 0.839 (+- 0.076). The data are available in the community and contributes with a new benchmark to develop and evaluate automatic medical text simplification systems., [ES] Se describe la recogida del corpus comparable CLARA-MeD, formado por 24 298 pares de textos profesionales y simplificados de dominio médico en lengua española (>96M palabras). Los tipos de textos varían desde prospectos médicos y fichas técnicas de medicamentos (10 211 pares de textos, >82M palabras), resúmenes de revisiones sistemáticas (8138 pares de textos, >9M palabras), resúmenes de información sobre el cáncer (201 pares de textos, >3M palabras) y anuncios de ensayos clínicos (5748 pares de textos, 451 690 palabras). También presentamos el alineamiento de frases técnicas y simplificadas, realizado a mano por pares de anotadores. Un subconjunto de 3800 pares de frases (149 862 tokens) se han emparejado, con un acuerdo medio entre anotadores con valor kappa = 0.839 (_0.076). Los datos están disponibles en la comunidad y este nuevo recurso permite desarrollar y evaluar sistemas de simplificación automática de textos médicos., Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publication Artículo científico (JournalArticle)

MedLexSp – a medical lexicon for Spanish medical natural language processing

  • Campillos-Llanos, Leonardo
Este artículo está sujeto a una licencia CC BY 4.0, [Background] Medical lexicons enable the natural language processing (NLP) of health texts. Lexicons gather terms and concepts from thesauri and ontologies, and linguistic data for part-of-speech (PoS) tagging, lemmatization or natural language generation. To date, there is no such type of resource for Spanish., [Construction and content] This article describes an unified medical lexicon for Medical Natural Language Processing in Spanish. MedLexSp includes terms and inflected word forms with PoS information and Unified Medical Language System® (UMLS) semantic types, groups and Concept Unique Identifiers (CUIs). To create it, we used NLP techniques and domain corpora (e.g. MedlinePlus). We also collected terms from the Dictionary of Medical Terms from the Spanish Royal Academy of Medicine, the Medical Subject Headings (MeSH), the Systematized Nomenclature of Medicine - Clinical Terms (SNOMED-CT), the Medical Dictionary for Regulatory Activities Terminology (MedDRA), the International Classification of Diseases vs. 10, the Anatomical Therapeutic Chemical Classification, the National Cancer Institute (NCI) Dictionary, the Online Mendelian Inheritance in Man (OMIM) and OrphaData. Terms related to COVID-19 were assembled by applying a similarity-based approach with word embeddings trained on a large corpus. MedLexSp includes 100 887 lemmas, 302 543 inflected forms (conjugated verbs, and number/gender variants), and 42 958 UMLS CUIs. We report two use cases of MedLexSp. First, applying the lexicon to pre-annotate a corpus of 1200 texts related to clinical trials. Second, PoS tagging and lemmatizing texts about clinical cases. MedLexSp improved the scores for PoS tagging and lemmatization compared to the default Spacy and Stanza python libraries., [Conclusions] The lexicon is distributed in a delimiter-separated value file; an XML file with the Lexical Markup Framework; a lemmatizer module for the Spacy and Stanza libraries; and complementary Lexical Record (LR) files. The embeddings and code to extract COVID-19 terms, and the Spacy and Stanza lemmatizers enriched with medical terms are provided in a public repository., Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature. This work has been done under the NLPMedTerm project, funded by the European Union’s Horizon 2020 research program under the Marie Skodowska-Curie grant agreement no. 713366 (InterTalentum UAM), and the CLARA-MeD project (PID2020-116001RA-C33), funded by MCIN/AEI/10.13039/501100011033/, in project call: “Proyectos I+D+i Retos Investigación”., Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publication Artículo científico (JournalArticle)

Enhancing the understanding of clinical trials with a sentence-level simplification dataset

  • Campillos-Llanos, Leonardo
  • Bartolomé, Rocío
  • Terroba Reinares, Ana R.
Este artículo está sujeto a una licencia CC BY-NC-ND 4.0, [EN] We introduce a dataset with 1200 manually simplified sentences (144 019 tokens) from clinical trials in Spanish. A total of 1040 announcements from the European Clinical Trials Register (EudraCT) were analyzed to select sentences with ambiguities or exceeding 25 words. Simplification criteria were devised in an annotation guideline, which is released publicly along with the dataset. We obtained two versions: syntactically simplified sentences, and sentences with syntactic and lexical simplification. We report a quantitative, a qualitative and a human evaluation, in which three independent evaluators assessed the grammaticality/fluency, semantic adequacy and overall simplification. Results show that the resource is suitable for advancing research on automatic simplification of medical texts., [ES] We introduce a dataset with 1200 manually simplified sentences (144 019 tokens) from clinical trials in Spanish. A total of 1040 announcements from the European Clinical Trials Register (EudraCT) were analyzed to select sentences with ambiguities or exceeding 25 words. Simplification criteria were devised in an annotation guideline, which is released publicly along with the dataset. We obtained two versions: syntactically simplified sentences, and sentences with syntactic and lexical simplification. We report a quantitative, a qualitative and a human evaluation, in which three independent evaluators assessed the grammaticality/fluency, semantic adequacy and overall simplification. Results show that the resource is suitable for advancing research on automatic simplification of medical texts., Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publication Artículo científico (JournalArticle)

Cross-Linguistic Disease and Drug Detection in Cardiology Clinical Texts: Methods and Outcomes

  • Styll, Patrick
  • Campillos-Llanos, Leonardo
  • Kusa, Wojciech
  • Hanbury, Allan
This paper presents our approach to the MultiCardioNER lab at CLEF2024, focusing on disease detection in Spanish texts and drug detection in Italian, Spanish, and English texts. We enhance model performance through several strategies: (1) fine-tuning on automatically translated TREC Clinical Trials admission notes using Masked Language Modeling (MLM); (2) data augmentation with translated MTSamples processed through a Spanish medical lexicon (MedLexSp) for accurate vocabulary matching; and (3) employing sliding windows with overlap to improve data capture. Additionally, we use transfer learning with a clinical trials corpus (CT-EMB-SP) to refine the outcomes. We further fine-tune several already established disease and drug extraction models to leverage their extensive vocabulary and compare their performance to models trained from scratch. Our methods and experiments demonstrate notable improvements in multilingual clinical NER, as evidenced by our track results., Leonardo Campillos-Llanos’ work is conducted in the CLARA-MeD project (PID2020-116001RA-C33), funded by MICIU/AEI/10.13039/501100011033/, in call Proyectos I+D+i Retos Investigación., Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publication Artículo científico (JournalArticle)

Entity normalization in a Spanish medical corpus using a UMLS‑based lexicon: findings and limitations

  • Báez, Pablo
  • Campillos-Llanos, Leonardo
  • Núñez, Fredy
  • Dunstan, Jocelyn
Entity normalization is a common strategy to resolve ambiguities by mapping all the synonym mentions to a single concept identifer in standard terminology. Normalizing medical entities is challenging, especially for languages other than English, where lexical variation is considerably under-represented. Here, we report a new linguistic resource for medical entity normalization in Spanish. We applied a UMLS-based medical lexicon (MedLexSp) to automatically normalize mentions from 2000 medical referrals of the Chilean Waiting List Corpus. Three medical students manually revised the automatic normalization. The inter-coder agreement was computed, and the distribution of concepts, errors, and linguistic sources of variation was analyzed. The automatic method normalized 52% of the mentions, compared to 91% after manual revision. The lowest agreement between automatic and automatic-manual normalization was observed for Finding, Disease, and Procedure entities. Errors in normalization were associated with ortho-typographic, semantic, and grammatical linguistic inadequacies, mainly of the hyponymy/hyperonymy, polysemy/metonymy, and acronym-abbreviation types. This new resource can enrich dictionaries and lexicons with new mentions to improve the functioning of modern entity normalization methods. The linguistic analysis ofers insight into the sources of lexical variety in the Spanish clinical environment related to error generation using lexicon-based normalization methods. This article also introduces a workfow that can serve as a benchmark for comparison in studies replicating our analysis in Romance languages, This study was supported by ANID Postdoctoral FONDECYT 3210395 and FONDECYT 11201250, Basal Funds for Center of Excellence FB210005 (CMM), Millennium Science Initiative Program ICN2021_004 (iHealth) and ICN17_002 (IMFD), and partly supported by the CLARA-MeD project (PID2020-116001RA-C33, funded by MICIU/AEI/10.13039/501100011033/ in call “Retos de investigación”), Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publication Otros (Other)

CLARA-MeD Tool – A System to Help Patients Understand Clinical Trial Announcements and Consent Forms in Spanish

  • Campillos-Llanos, Leonardo
  • Ortega Riba, Federico
  • Terroba Reinares, Ana Rosa
  • Valverde Mateos, Ana
  • Capllonch Carrión, Adrián
We present an NLP web-based tool to help users understand consent forms (CFs) and clinical trial announcements (CTAs) in Spanish. For complex word identification, we collected: 1) a lexicon of technical terms and simplified synonyms (14 465 entries); and 2) a glossary (70 547 terms) with explanations from sources such as UMLS, the NCI dictionary, Orphadata or the FDA. For development, we extracted entities from 60 CTAs, 60 CFs and 60 patient information documents (PIDs). To prepare definitions for new terms, we used ChatGPT and experts validated them (28.99% needed to be fixed). We tested the system on 15 new CTAs, 15 CFs, and 15 PIDs, and we achieved an average F1 score of 82.91% (strict match) and of 94.65% (relaxed). The tool is available at: http://claramed.csic.es/demo., CLARA-MeD project (PID2020-116001RA-C33) funded by AEI/MICIU/10.13039/50 1100011033/, Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publication Artículo científico (JournalArticle)

Hybrid natural language processing tool for semantic annotation of medical texts in Spanish

  • Campillos-Llanos, Leonardo
  • Valverde Mateos, Ana
  • Capllonch Carrión, Adrián
Background Natural language processing (NLP) enables the extraction of information embedded within unstructured texts, such as clinical case reports and trial eligibility criteria. By identifying relevant medical concepts, NLP facilitates the generation of structured and actionable data, supporting complex tasks like cohort identification and the analysis of clinical records. To accomplish those tasks, we introduce a deep learning-based and lexicon-based named entity recognition (NER) tool for texts in Spanish. It performs medical NER and normalization, medication information extraction and detection of temporal entities, negation and speculation, and temporality or experiencer attributes (Age, Contraindicated, Negated, Speculated, Hypothetical, Future, Family_member, Patient and Other). We built the tool with a dedicated lexicon and rules adapted from NegEx and HeidelTime. Using these resources, we annotated a corpus of 1200 texts, with high inter-annotator agreement (average F1 = 0.841% ± 0.045 for entities, and average F1 = 0.881% ± 0.032 for attributes). We used this corpus to train Transformer-based models (RoBERTa-based models, mBERT and mDeBERTa). We integrated them with the dictionary-based system in a hybrid tool, and distribute the models via the Hugging Face hub. For an internal validation, we used a held-out test set and conducted an error analysis. For an external validation, eight medical professionals evaluated the system by revising the annotation of 200 new texts not used in development., Results In the internal validation, the models yielded F1 values up to 0.915. In the external validation with 100 clinical trials, the tool achieved an average F1 score of 0.858 (± 0.032); and in 100 anonymized clinical cases, it achieved an average F1 score of 0.910 (± 0.019)., Conclusions The tool is available at https://claramed.csic.es/medspaner . We also release the code (https://github.com/lcampillos/medspaner) and the annotated corpus to train the models., The first version of MEDSPANER was developed under the Marie Curie COFUND grant agreement no. 713366 (InterTalentum UAM). The training of the Transformer-based models and their integration into the new version was conducted under the CLARA-MeD project (PID2020-116001RA-C33), funded by MICIU/AEI/10.13039/501100011033/, in project call: “Proyectos I+D+i Retos de investigación"., Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publication Artículo científico (JournalArticle)

Complex Word Identification for Lexical Simplification in Spanish Texts for Patients

  • Ortega Riba, Federico
  • Campillos-Llanos, Leonardo
  • Samy, Doa
[EN] This work describes the task of complex word identification (CWI) in Spanish medical texts for patients. Identifying complex words is the first step in lexical simplification, which aims to overcome the language gap between patients and healthcare professionals, enable access to information, and ensure unambiguous terminology for effective and clear communication. As part of the task, we created a medical complex words annotation guideline and compiled a corpus consisting of 225 texts (162575 tokens). A total of 18203 complex words (single and multi-words) were manually labeled, each text being annotated by two linguists with high interannotator agreement (F1 = 84.42%). The corpus was utilized to train two machine learning classifiers (Support Vector Machines and Logistic Regression) as baselines, in addition to seven deep learning transformer models. The models were selected by considering two factors: language (Spanish and multilingual) and domain (general or medical). The final results on the test set achieve an overall average F1 score of 79.02 (±0.65) for the transformer model with the best performance., [ES ] Este artículo describe la tarea de identificación de palabras complejas en textos médicos en español para pacientes. Este es el primer paso para la simplificación léxica, cuyo objetivo es superar la barrera lingüística entre pacientes y profesionales sanitarios, permitir el acceso a la información y garantizar una terminología sin ambigüedades y una comunicación clara y eficaz. Se ha creado una guía de anotación y se ha compilado un corpus de 225 textos (162575 tokens). Se anotaron 18203 palabras complejas (entidades simples como multipalabra), siendo cada texto revisados por dos lingüistas, y alcanzando un alto valor de acuerdo entre anotadores (F1 = 84.42%). El corpus se ha empleado para entrenar modelos de aprendizaje automático (máquinas de soporte vectorial y regresión logística) como referencia, y siete modelos de aprendizaje profundo basados en transformers. Estos modelos fueron seleccionados considerando dos factores: idioma (español o multilingüe) y dominio (general o medico). Los experimentos finales muestran una puntuación F1 de 79.02 (±0.65) para el modelo transformer con mejores resultados., Peer reviewed
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publication Artículo científico (JournalArticle)

Transformer-based relation extraction and concept normalization using an annotated clinical trials corpus

  • Campillos-Llanos, Leonardo
  • Valverde Mateos, Ana
  • Capllonch Carrión, Adrián
  • González-Quevedo, David
  • López-Urbán, María Rosa
  • Hernando-Tundidor, Soledad
  • Zakhir Puig, Sofía
  • Heras, Jónathan
Healthcare professionals manually review electronic health records to select patients who meet eligibility criteria of clinical trials. Natural language processing offers a complement for this task, although few initiatives exist in Spanish. We present version 3 of the CT-EBM-SP corpus of 1200 clinical trials (292173 tokens), annotated with 23 entity types and 18 relation types, covering Unified Medical Language System (UMLS) semantic groups, drug-related information, temporal data, and negation/speculation. We encoded 11 attributes (e.g., event temporality and experiencer status) and normalized entities to UMLS Concept Unique Identifiers. The corpus contains 87037 entities, including nested and discontinuous entities, 16597 attributes and 68206 relationships. Inter-annotator agreement (IAA) achieved average F1 values of 0.861 (entities), 0.810 (attributes), and 0.791 (relations). 81.75% of entities were normalized (IAA: F1 = 0.966). We benchmarked this dataset by fine-tuning Transformer models for relation extraction (RE) and medical concept normalization (MCN). In RE, the average F1 ranged from 0.858 to 0.879, and for MCN, the accuracy at rank 1 was 0.896. The corpus and models are publicly available., This work was conducted in the CLARA-MED project (PID2020-116001RA-C33), the ExPlain4Health project (PID2024-158912NB-I00), and partly (JH) by project PID2020-115225RB-I00, all funded by MCIN/AEI/10.13039/501100011033, in project call: Proyectos I+D+i Generación de conocimiento (Retos Investigación). Sofía Zakhir was funded by a JAE Intro 2021 scholarship at CSIC., Peer reviewed