Model Overview

This model, developed by Pablo Cano Ortiz, is a fine-tuned Gemma 3.4B variant specifically created for research into linguistic differences between healthy controls and patients with psychosis. It represents the "psychosis model" within a larger project aiming to compare language patterns between these two groups using metrics like perplexity.

Key Capabilities

Psychosis Simulation: Generates text simulating responses from patients with psychosis, particularly in the context of clinical interviews.
Linguistic Research: Designed to facilitate the study of language differences associated with psychosis.
Population-Specific Text Generation: Enables experiments in generating text characteristic of a specific population group.

Training Details

The model was trained on a specialized dataset consisting of question-answer pairs extracted from the Discourse-UWO clinical interview dataset.

Intended Use Cases

Research in clinical Natural Language Processing (NLP).
Language modeling studies focused on psychosis-related speech.
Analysis of speech patterns in psychosis.
Experiments involving the generation of population-specific text.

Limitations

Research-Only: Strictly for research purposes; not intended for diagnostic or clinical decision-making.
Dataset Bias: The training dataset is limited and imbalanced, which may introduce specific biases into the model's outputs.

Citation

Users of this model are requested to cite the associated undergraduate thesis by Pablo Ramon Cano Ortiz, titled "Towards Explainable AI for Psychosis Detection Through Clinical Language Modeling" (2026).

Overview