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. The model was developed for the MentalRisk 2023 Depression (MR23-D) shared task and builds upon the Longformer-es-mental-large foundation model.

The ICE methodology restructures the training data at the context level, enabling the model to learn from progressively expanding user message histories rather than complete user timelines. This training strategy better reflects real-world early detection scenarios, where predictions must be generated with limited and evolving evidence.

The model is based on the Longformer architecture and supports input sequences of up to 4096 tokens, allowing it to integrate evidence distributed across long and heterogeneous user histories. It has been fine-tuned to detect depression-related risk signals from Spanish user-generated content under both early detection and full-context evaluation settings.

Developed by: ELiRF group, VRAIN (Valencian Research Institute for Artificial Intelligence), Universitat Politècnica de València
Shared by: ELiRF
Model type: Transformer-based sequence classification model (Longformer)
Language: Spanish
Base model: Longformer-es-mental-large
License: Same as base model

Uses

This model is intended for research purposes in early mental health risk detection, specifically for depression.

Direct Use

The model can be used directly for early detection of depression risk from Spanish user-generated content, generating predictions incrementally as new user messages become available.

Downstream Use

Early risk detection for depression
User-level mental health screening
Comparative studies on early detection methodologies
Research on temporally-aware and incremental NLP models

Out-of-Scope Use

Automated intervention systems without human supervision
Use on languages other than Spanish
High-stakes decision-making affecting individuals’ health or safety

ICE Methodology

Incremental Context Expansion (ICE) is a training methodology designed for early and incremental mental health detection tasks. Instead of training models on full user histories, ICE generates multiple incremental contexts per user, each corresponding to a partial message history.

This approach allows the model to:

Learn from incomplete and early evidence
Reduce detection latency
Improve robustness under early detection evaluation metrics

ICE modifies the dataset construction process while keeping the standard fine-tuning pipeline unchanged.

Bias, Risks, and Limitations

Training data originates from social media platforms and may contain demographic, cultural, and linguistic biases.
Automatically translated texts may introduce translation artifacts or subtle semantic shifts.
Depression detection is inherently subjective and context-dependent.
The model does not provide explanations or clinical interpretations of its predictions.

How to Get Started with the Model

python

from transformers import AutoTokenizer, AutoModelForSequenceClassification

tokenizer = AutoTokenizer.from_pretrained("ELiRF/Longformer-es-m-large-ICE-MR23-D")
model = AutoModelForSequenceClassification.from_pretrained(
    "ELiRF/Longformer-es-m-large-ICE-MR23-D"
)

inputs = tokenizer(
    "Ejemplo de historial de mensajes relacionado con depresión.",
    return_tensors="pt",
    truncation=True,
    max_length=4096
)

outputs = model(**inputs)

Training Details

Training Data

The model was fine-tuned on the MentalRisk 2023 Depression (MR23-D) dataset. Training data was restructured using the ICE methodology, generating incremental user contexts from original user timelines.

Training Procedure

Base model: Longformer-es-mental-large
Fine-tuning strategy: ICE-based context-level training
Objective: Sequence classification
Training regime: fp16 mixed precision

Evaluation

Results

When evaluated on the MentalRisk 2023 Depression task, Longformer-es-m-large-ICE-MR23-D shows competitive performance and improves upon the state of the art under both early detection and full-context (user-level) evaluation settings.

Environmental Impact

Hardware type: NVIDIA A40 GPUs
Training time: several hours (fine-tuning)

Technical Specifications

Model Architecture and Objective

Architecture: Longformer (large)
Objective: Sequence classification
Maximum sequence length: 4096 tokens
Model size: approximately 435M parameters

Citation

This model is part of an ongoing research project. The associated paper is currently under review and will be added to this model card once the publication process is completed.

Model Card Authors

ELiRF research group (VRAIN, Universitat Politècnica de València)