CRIS

Permanent URI for this communityhttps://scripta.up.edu.mx/handle/20.500.12552/1

Browse

Settings

Author: search.filters.author.González Mora, José GuillermoHas files: Yes

Search Results

Now showing 1 - 2 of 2
  • Loading...
    Thumbnail Image
    Some of the metrics are blocked by your 
    consent settings
    Item type:Publication,
    Pre-trained Models for Grammatical Error Correction in Healthcare-Specific Text
    (Springer Nature Switzerland, 2025)
    González Mora, José Guillermo
    ;
    Ponce, Hiram  
    Grammatical Error Correction (GEC) is a common Natural Language Processing task and has been studied extensively in the field. There is an added complexity when trying to correct domain-specific text that is uncommon in general-purpose corpora. In this paper, we present a comparative study of three different approaches using pre-trained language models for Grammatical Error Correction in healthcare-specific text. We evaluated the performance of all proposals using the GLEU score metric, which allows a quantitative comparison of all methods. We utilized two different problem settings: first, a sequence-to-sequence pre-trained T5 model, followed by a fine-tuning process over a small set of examples. The second model is a pre-trained large language model, for which we test both zero-shot and few-shot in-context learning. The study shows how the fine-tuned T5 model is capable of exceeding the performance shown by the LLM tested. With this result, we show how smaller encoder-decoder models can solve domain-specific tasks with fewer parameters than a purely generative pre-trained LLM. ©The authors ©Springer.
  • Loading...
    Thumbnail Image
    Some of the metrics are blocked by your 
    consent settings
    Item type:Publication,
    Contrastive Steering Vectors for Autoencoder Explainability
    (MDPI, 2025)
    González Mora, José Guillermo
    ;
    Ponce, Hiram  
    ;
    Martinez-Villaseñor, Lourdes  
    Generative models, particularly autoencoders, often function as black boxes, making it challenging for non-expert users to effectively control the generation process and understand how inputs affect outputs. Existing methods for improving interpretability and control frequently require specific model training regimes or labeled data, limiting their applicability. This work introduces a novel approach to enhance the controllability and explainability of generative models, specifically tested on autoencoders with entangled latent spaces. We propose using a semi-supervised contrastive learning setup to learn steering vectors. These vectors, when added to an input’s latent representation, effectively manipulate specific attributes in the generated output without conditional training of the model or attribute classifiers, thus being applicable to pretrained models and avoiding compound classification errors. Furthermore, we leverage these learned steering vectors to interpret and explain the decoding process of a target attribute, allowing for efficient exploration of feature dimension interactions and the construction of an interpretable plot of the generative process, while lowering scalability limitations of perturbation-based Explainable AI (XAI) methods by reducing the search space. Our method provides an efficient pathway to controllable generation, offers an interpretable result of the model’s internal mechanisms, and relates the interpretations to human-understandable explanation questions. ©The authors ©MDPI AG ©Electronics.