| --- |
| license: mit |
| datasets: |
| - anthonyyazdaniml/gliner-biomed-pre-training |
| language: |
| - en |
| - multilingual |
| base_model: |
| - urchade/gliner_multi-v2.1 |
| pipeline_tag: token-classification |
| tags: |
| - gliner |
| --- |
| |
| # About |
| This is an initial test for multilingual biomedical NER model using GLiNER. The model uses multilingual GLiNER model (`urchade/gliner_multi-v2.1`) as the base model and fine-tuned with biomedical NER dataset. |
|
|
| ## Installation |
| This model uses GLiNER python library. You can install the library: |
| ``` |
| pip install gliner |
| ``` |
|
|
| ## Usage |
| To use the model, you can load the model and tokenizer using `GLiNER.from_pretrained()` method. Then, you can use the `predict_entities()` method to extract entities from the input text. |
|
|
| ```python |
| from gliner import GLiNER |
| model = GLiNER.from_pretrained("fjoeda/gliner-biomed-multi-test", load_tokenizer=True) |
| |
| text = "<your input text here>" |
| labels = ["entity #1", "entity#2"] |
| |
| entities = model.predict_entities(text, labels, threshold=0.5) |
| |
| for entity in entities: |
| print(entity["text"], "=>", entity["label"]) |
| ``` |
|
|
| ### Sample on English Text |
| Here is an example of how to use the model to extract entities from English text: |
|
|
| ```python |
| text = """ |
| Male patient, 68 years old, admitted from July 1, 2026 to July 4, 2026 |
| due to recurrent urination symptoms, weak flow, and pelvic pain |
| caused by Benign Prostatic Hyperplasia (BPH). |
| Transurethral resection of the prostate (TURP) was performed without complications. |
| Patient was prescribed continued medical therapy with tamsulosin 0.4 mg qd, |
| """ |
| |
| labels = ["Disease", "Procedure", "Medication", "Symptom", "Anatomy", "Medical Device"] |
| |
| entities = model.predict_entities(text, labels, threshold=0.5) |
| |
| for entity in entities: |
| print(entity["text"], "=>", entity["label"]) |
| ``` |
|
|
| output: |
| ``` |
| recurrent urination symptoms => Symptom |
| weak flow => Symptom |
| pelvic pain => Symptom |
| Benign Prostatic Hyperplasia (BPH) => Disease |
| Transurethral resection of the prostate (TURP) => Procedure |
| tamsulosin 0.4 mg qd => Medication |
| ``` |
|
|
|
|
| ### Sample on Indonesian Text |
| Since GLiNER multilingual model can recognize entities from non-English text without having explicitly trained on the language, here is an example of how to use the model to extract entities from Indonesian text: |
|
|
| ```python |
| text = """ |
| Pasien pria usia 68 tahun, dirawat dari 1 Juli 2026 hingga 4 Juli 2026 |
| karena gejala urinasi berulang, aliran lemah, dan nyeri panggul |
| yang disebabkan oleh Benign Prostatic Hyperplasia (BPH). |
| evaluasi klinis, USG, uroflowmetry, dan cystoscopy, dilakukan TURP tanpa komplikasi |
| serta diberikan terapi medis lanjutan dengan tamsulosin 0.4 mg qd, |
| antibiotik profilaksis, analgesik, dan lanjutan pengobatan hipertensi |
| """ |
| |
| labels = ["Disease", "Procedure", "Medication", "Symptom", "Anatomy", "Medical Device"] |
| |
| entities = model.predict_entities(text, labels, threshold=0.5) |
| |
| for entity in entities: |
| print(entity["text"], "=>", entity["label"]) |
| ``` |
|
|
| output: |
| ``` |
| urinasi berulang => Symptom |
| aliran lemah => Symptom |
| nyeri panggul => Symptom |
| Benign Prostatic Hyperplasia (BPH) => Disease |
| evaluasi klinis => Procedure |
| USG => Procedure |
| uroflowmetry => Procedure |
| cystoscopy => Procedure |
| TURP => Procedure |
| tamsulosin 0.4 mg qd => Medication |
| antibiotik profilaksis => Medication |
| analgesik => Medication |
| ``` |
|
|
