Recursos

Longformer-es-mental-large Longformer-es-mental-large

Longformer-es-mental-large is a Spanish domain-adapted language model designed for the analysis of mental health–related content in long user-generated texts.

RoBERTa-es-mental-large RoBERTa-es-mental-large

RoBERTa-es-mental-large is a Spanish domain-adapted language model specialized in mental health text analysis.

Longformer-es-mental-base Longformer-es-mental-base

Longformer-es-mental-base is the base-sized version of the Longformer-es-mental family, a Spanish domain-adapted language model designed for mental health text analysis on long user-generated content.

Longformer-es-m-large-ICE-MR23-ED Longformer-es-m-large-ICE-MR23-ED

Longformer-es-m-large-ICE-MR23-ED is a Spanish long-context language model for early detection of eating disorder risk, trained using the Incremental Context Expansion (ICE) methodology.

RoBERTa-es-m-large-ICE-MR23-ED RoBERTa-es-m-large-ICE-MR23-ED

RoBERTa-es-m-large-ICE-MR23-ED is a Spanish language model for early detection of eating disorder risk, trained using the Incremental Context Expansion (ICE) methodology.

Longformer-es-m-base-ICE-MR23-ED Longformer-es-m-base-ICE-MR23-ED

Longformer-es-m-base-ICE-MR23-ED is a Spanish long-context language model for early detection of eating disorder risk, trained using the Incremental Context Expansion (ICE) methodology.

Longformer-es-m-large-ICE-MR24-DD Longformer-es-m-large-ICE-MR24-DD

Longformer-es-m-large-ICE-MR24-DD is a Spanish long-context language model for mental disorder detection, trained using the Incremental Context Expansion (ICE) methodology.

RoBERTa-es-m-large-ICE-MR24-DD RoBERTa-es-m-large-ICE-MR24-DD

RoBERTa-es-m-large-ICE-MR24-DD is a Spanish language model for mental disorder detection, trained using the Incremental Context Expansion (ICE) methodology.

Longformer-es-m-base-ICE-MR24-DD Longformer-es-m-base-ICE-MR24-DD

Longformer-es-m-base-ICE-MR24-DD is a Spanish long-context language model for mental disorder detection, trained using the Incremental Context Expansion (ICE) methodology.

Longformer-es-m-large-ICE-MR23-D Longformer-es-m-large-ICE-MR23-D

Longformer-es-m-large-ICE-MR23-D is a Spanish long-context language model for early detection of depression risk, trained using the Incremental Context Expansion (ICE) methodology.

RoBERTa-es-m-large-ICE-MR23-D RoBERTa-es-m-large-ICE-MR23-D

RoBERTa-es-m-large-ICE-MR23-D is a Spanish language model for early detection of depression risk, trained using the Incremental Context Expansion (ICE) methodology.

Longformer-es-m-base-ICE-MR23-D Longformer-es-m-base-ICE-MR23-D

Longformer-es-m-base-ICE-MR23-D is a Spanish long-context language model for early detection of depression risk, trained using the Incremental Context Expansion (ICE) methodology.

dacsa Dataset for Automatic summarization of Catalan and Spanish newspaper Articles

The Dataset for Automatic summarization of Catalan and Spanish newspaper Articles (DACSA) corpus. It is a high-quality large-scale corpus that can be used to train summarization models for Catalan and Spanish. The data provides pairs of news article and its summary from different newspapers for both, the Catalan and the Spanish languages. Regarding the Catalan set, there are 725,184 sample pairs from 9 newspapers, regarding the Spanish set, the corpus provides 2,120,649 sample pairs from 21 newspapers.

UX-comments UX-comments

Access to dataset ELiRF/UX-comments is restricted. You must have access to it and be authenticated to access it. Please log in.

dacsa-multilingual-summaries dacsa-multilingual-summaries

Access to dataset ELiRF/dacsa-multilingual-summaries is restricted. You must have access to it and be authenticated to access it. Please log in.

clasum CLASum Dataset

The CLASum is built for abstractivity characterization and measurement in the summarization task. It contains 200 document-summary pairs extracted from CNN/DailyMail and XSUM datasets. Each summary has been annotated for 11 summary-related questions by 3 different annotators.

montero2k Hate Speech Against Women Politicians

The "Montero2K" dataset is composed by 2,239 tweets, selected on the specific criteria that each message had garnered more than 50 likes or retweets during the specified period. Human experts manually labeled these tweets with “hate” and “improper language” categories.

NASES NASca and NASes: Two Monolingual Pre-Trained Models for Abstractive Summarization in Catalan and Spanish

Most of the models proposed in the literature for abstractive summarization are generally suitable for the English language but not for other languages. Multilingual models were introduced to address that language constraint, but despite their applicability being broader than that of the monolingual models, their performance is typically lower, especially for minority languages like Catalan. In this paper, we present a monolingual model for abstractive summarization of textual content in the Catalan language. The model is a Transformer encoder-decoder which is pretrained and fine-tuned specifically for the Catalan language using a corpus of newspaper articles. In the pretraining phase, we introduced several self-supervised tasks to specialize the model on the summarization task and to increase the abstractivity of the generated summaries. To study the performance of our proposal in languages with higher resources than Catalan, we replicate the model and the experimentation for the Spanish language. The usual evaluation metrics, not only the most used ROUGE measure but also other more semantic ones such as BertScore, do not allow to correctly evaluate the abstractivity of the generated summaries. In this work, we also present a new metric, called content reordering, to evaluate one of the most common characteristics of abstractive summaries, the rearrangement of the original content. We carried out an exhaustive experimentation to compare the performance of the monolingual models proposed in this work with two of the most widely used multilingual models in text summarization, mBART and mT5. The experimentation results support the quality of our monolingual models, especially considering that the multilingual models were pretrained with many more resources than those used in our models. Likewise, it is shown that the pretraining tasks helped to increase the degree of abstractivity of the generated summaries. To our knowledge, this is the first work that explores a monolingual approach for abstractive summarization both in Catalan and Spanish.

mt5-base-dacsa-es mT5 (base model), fine-tuned on the Dataset for Automatic summarization of Catalan and Spanish newspaper Articles (DACSA) dataset for Spanish

The mT5 model was presented in mT5: A massively multilingual pre-trained text-to-text transformer by Linting Xue, Noah Constant, Adam Roberts, Mihir Kale, Rami Al-Rfou, Aditya Siddhant, Aditya Barua, Colin Raffel. The base version of the mT5 model is pre-trained in 101 languages, including English, Spanish, Italian, Catalan and other ones.

mbart-large-cc25-dacsa-es mBART (large-cc25 model), fine-tuned on the Dataset for Automatic summarization of Catalan and Spanish newspaper Articles (DACSA) dataset for Spanish

The mBART model was presented in Multilingual Denoising Pre-training for Neural Machine Translation by Yinhan Liu, Jiatao Gu, Naman Goyal, Xian Li, Sergey Edunov Marjan Ghazvininejad, Mike Lewis, Luke Zettlemoyer. The large-cc25 version of the mBART model is pre-trained in 25 languages, including English, Spanish, Italian, and other ones.

NASCA NASca and NASes: Two Monolingual Pre-Trained Models for Abstractive Summarization in Catalan and Spanish

Most of the models proposed in the literature for abstractive summarization are generally suitable for the English language but not for other languages. Multilingual models were introduced to address that language constraint, but despite their applicability being broader than that of the monolingual models, their performance is typically lower, especially for minority languages like Catalan. In this paper, we present a monolingual model for abstractive summarization of textual content in the Catalan language. The model is a Transformer encoder-decoder which is pretrained and fine-tuned specifically for the Catalan language using a corpus of newspaper articles. In the pretraining phase, we introduced several self-supervised tasks to specialize the model on the summarization task and to increase the abstractivity of the generated summaries. To study the performance of our proposal in languages with higher resources than Catalan, we replicate the model and the experimentation for the Spanish language. The usual evaluation metrics, not only the most used ROUGE measure but also other more semantic ones such as BertScore, do not allow to correctly evaluate the abstractivity of the generated summaries. In this work, we also present a new metric, called content reordering, to evaluate one of the most common characteristics of abstractive summaries, the rearrangement of the original content. We carried out an exhaustive experimentation to compare the performance of the monolingual models proposed in this work with two of the most widely used multilingual models in text summarization, mBART and mT5. The experimentation results support the quality of our monolingual models, especially considering that the multilingual models were pretrained with many more resources than those used in our models. Likewise, it is shown that the pretraining tasks helped to increase the degree of abstractivity of the generated summaries. To our knowledge, this is the first work that explores a monolingual approach for abstractive summarization both in Catalan and Spanish.

mt5-base-dacsa-ca mT5 (base model), fine-tuned on the Dataset for Automatic summarization of Catalan and Spanish newspaper Articles (DACSA) dataset for Catalan

The mT5 model was presented in mT5: A massively multilingual pre-trained text-to-text transformer by Linting Xue, Noah Constant, Adam Roberts, Mihir Kale, Rami Al-Rfou, Aditya Siddhant, Aditya Barua, Colin Raffel. The base version of the mT5 model is pre-trained in 101 languages, including English, Spanish, Italian, Catalan and other ones.

mbart-large-cc25-dacsa-ca mBART (large-cc25 model), fine-tuned on the Dataset for Automatic summarization of Catalan and Spanish newspaper Articles (DACSA) dataset for Catalan

The mBART model was presented in Multilingual Denoising Pre-training for Neural Machine Translation by Yinhan Liu, Jiatao Gu, Naman Goyal, Xian Li, Sergey Edunov Marjan Ghazvininejad, Mike Lewis, Luke Zettlemoyer. The large-cc25 version of the mBART model is pre-trained in 25 languages, including English, Spanish, Italian, and other ones.

NASca-es NASca-es

Chest-COVID Model description

modelname: roberta-base-biomedical-clinical

RoBERTa-large-bne-MR23-D RoBERTa-large-bne-MR23-D

SPA-Sentences: SPA-Sentences: Entrenamiento y evaluación de sistemas de reconocimiento de escritura manuscrita en español

La tendencia mundial a la automatización de procesos es mayor, si cabe, cuando se trata de migrar documentos en papel a soporte digital. Además dela digitalización propiamente dicha, el procesado y transcripción eficaz de documentos, en muchos casos manuscritos y/o con baja resolución, permiten explotar una información que de otro modo no sería viable. Todas las herramientas para la transcripción de texto manuscrito se basan en sistemas deaprendizaje automático que, en última instancia, requieren el uso de corpus para su entrenamiento. Aunque existen corpus de escritura manuscrita moderna para distintas lenguas como el inglés o francés, no sucede lo mismo para el español que difiere lo suficiente para requerir un corpus específico. SPA-Sentences es un corpus de frases manuscritas en español para el entrenamiento y evaluación de sistemas de reconocimiento deescritura en lengua española. El corpus consta de frases manuscritas extraídas de 1.617 formularios producidos por el mismo número de escritores. Hay un total de 13,691 frases que contienen alrededor de 100,000 instancias de palabras, con un vocabulario de 3,288 palabras. Estos datos permiten realizar un entrenamiento eficaz de los sistemas de reconocimiento. Los ficheros del corpus comprenden las imágenes escaneadas de los formularios así como información de su segmentación en líneas y su transcripción manualmente supervisada. También se proporciona un conjunto de programas en Python para extraer las informaciones a partir de los ficheros de la imagen y sus correspondientes ficheros xml. Gracias a SPA-Sentences es posible entrenar y poner a punto sistemas de reconocimiento de escritura manuscrita para alfabetos latinos y, en particular, en español, así como evaluar el sistema de reconocimiento en condiciones estándar para la comunidad científica.

NoisyOffice NoisyOffice

Corpus intended to do cleaning (or binarization) and enhancement of noisy grayscale printed text images using supervised learning methods. Noisy images and their corresponding ground truth provided.