multiclinner_enigma_cz_disease_robeczech-morph
Czech clinical Named Entity Recognition model for the
DISEASE entity
type, fine-tuned from
ufal/robeczech-base on the
Czech portion of the
MultiClinAI 2026 IberLEF shared task.
Developed by
Team Enigma at the Faculty of Mathematics and Informatics,
Sofia University.
Summary
| |
|---|
| Task | Token classification (BIO), single entity type |
| Entity type | DISEASE (diseases and disorders) |
| Language | Czech (cs) |
| Base model | ufal/robeczech-base |
| Architecture | Transformer (softmax) |
| Augmentation | Morphological synonyms (curated + Wikidata) |
| Training data | MultiClinAI Czech train + dev combined (1,258 documents) plus augmentation |
| Test F1 (strict) | 0.6552 (char F1 0.7737) |
Quick start
1from transformers import AutoTokenizer, AutoModelForTokenClassification, pipeline
2
3repo = "SU-FMI-AI/multiclinner_enigma_cz_disease_robeczech-morph"
4
5tokenizer = AutoTokenizer.from_pretrained(repo)
6model = AutoModelForTokenClassification.from_pretrained(repo)
7
8ner = pipeline(
9 "token-classification",
10 model=model,
11 tokenizer=tokenizer,
12 aggregation_strategy="simple",
13)
14
15text = "Pacient byl přijat s hypertenzí a podstoupil koronarografii."
16for span in ner(text):
17 print(span)
Each predicted span is a dictionary with keys entity_group,
score, word, start, end.
Label set
The model predicts a single entity type using the BIO tagging scheme:
| ID | Label | Meaning |
|---|
| 0 | O | outside any entity |
| 1 | B-DISEASE | beginning of a DISEASE mention |
| 2 | I-DISEASE | inside a DISEASE mention |
Intended use
- Extracting
DISEASE mentions from Czech clinical text (discharge
summaries, case reports, medical records).
- Building block for ensembles (this model was deployed as part of an
ensemble in the original submission).
- A starting point for further fine-tuning on related Czech biomedical
corpora.
Out-of-scope use
- Other languages. Although the XLM-RoBERTa based variants share a
multilingual encoder, the classification head was only trained on Czech.
- Other domains. The model was not exposed to non-clinical text, social
media or layperson descriptions.
- Clinical decision making. This is a research artifact. Do not use it
as the sole input to any clinical decision.
Training data
- Source. MultiClinAI Czech NER: 1,006 train documents
and 252 dev documents per entity type in BRAT standoff format, derived
from the DisTEMIST, SympTEMIST and MedProcNER corpora translated and
annotation-projected to Czech. For the final submission models the gold
train + dev sets are merged into a single training partition (no
held-out validation).
- Augmentation. Morphological synonym replacement. Rare entity surface forms (occurring at most 5 times in training) are substituted with morphological variants drawn from a curated Czech medical synonym dictionary. The dictionary combines manually curated morphological paradigms for Czech medical terms with Wikidata concept labels for Czech medical entities, totalling roughly 1,400 synonym entries across about 900 morphological families. Approximately 2,200 augmented documents per entity type are appended to the 1,258 document gold train+dev partition.
- Tokenisation. SentencePiece tokenizer inherited from the base model.
Training procedure
| Hyperparameter | Value |
|---|
| Base model | ufal/robeczech-base |
| Head | SOFTMAX |
| Optimiser | AdamW |
| Learning rate | 2e-5 |
| Batch size | 64 |
| Epochs | 10 |
| Max sequence length | 512 |
| Input granularity | Sentence-level |
| Warmup ratio | 0.1 |
| Weight decay | 0.01 |
| Seed | 42 |
| Mixed precision | fp16 (CUDA) |
Token classification head: backbone hidden states are projected through a linear layer to BIO logits and decoded greedily (argmax per token).
Evaluation
Held-out development set
Best dev-set entity-level F1 observed during development: 0.698.
MultiClinAI Czech, official blind test set
Run name in the official MultiClinAI ranking: robeczech-morph_cz_disease.
| Metric | Strict | Character-level |
|---|
| Precision | 0.6741 | 0.7967 |
| Recall | 0.6373 | 0.7520 |
| F1 | 0.6552 | 0.7737 |
Strict matching requires the predicted span to exactly match a gold span
(same start, end, and type). Character-level matching gives partial credit
for overlapping spans.
Related models
Other models for the same entity type:
SU-FMI-AI/multiclinner_enigma_cz_disease_robeczech-para: robeczech-base, Morphological synonyms + GPT-4.1-mini paraphrase, SOFTMAX head, test F1 = 0.6494.SU-FMI-AI/multiclinner_enigma_cz_disease_xlmr-crf: xlm-roberta-base, Morphological synonyms (curated + Wikidata), CRF head, test F1 = 0.6381.SU-FMI-AI/multiclinner_enigma_cz_disease_xlmr-para: xlm-roberta-base, Morphological synonyms + GPT-4.1-mini paraphrase, SOFTMAX head, test F1 = 0.6283.SU-FMI-AI/multiclinner_enigma_cz_disease_xlmr-os1: xlm-roberta-base, Morphological synonyms + 1x oversample of entity-bearing docs, SOFTMAX head, test F1 = 0.6440.
License
Released under the
apache-2.0
license. Base-model and dataset licenses apply to their respective
artifacts.
Code and resources
Training code, augmentation pipeline, ablation log and evaluation scripts
are available in the project's GitHub repository:
https://github.com/TeogopK/MultiClinAI-Czech.
