Hugging Face's logo

Ihor
/
gliner-biomed-base-v1.0

Token Classification
GLiNER
PyTorch
English
NER
GLiNER
information extraction
encoder
entity recognition
biomed
Model card Files
xet
Community
2
nielsr HF Staff commited on
Commit
9e1cfb4
·
verified ·
1 Parent(s): 6144b26

Add link to Github repository

Browse files

This PR adds a direct link to the Github repository in the model card for easier access to the code and training pipelines.

Files changed (1) hide show
  1. README.md +37 -34
README.md CHANGED
@@ -1,11 +1,15 @@
1
  ---
2
- license: apache-2.0
3
- language:
4
- - en
5
- library_name: gliner
6
  datasets:
7
  - knowledgator/GLINER-multi-task-synthetic-data
8
  - knowledgator/biomed_NER
 
 
 
 
 
 
9
  pipeline_tag: token-classification
10
  tags:
11
  - NER
@@ -14,11 +18,8 @@ tags:
14
  - encoder
15
  - entity recognition
16
  - biomed
17
- base_model:
18
- - microsoft/deberta-v3-base
19
- metrics:
20
- - f1
21
  ---
 
22
  # GLiNER-BioMed
23
 
24
  **GLiNER** is a Named Entity Recognition (NER) model capable of identifying any entity type using a bidirectional transformer encoders (BERT-like). It provides a practical alternative to traditional NER models, which are limited to predefined entities, and Large Language Models (LLMs) that, despite their flexibility, are costly and large for resource-constrained scenarios.
@@ -26,6 +27,8 @@ metrics:
26
 
27
  **GLiNER-biomed**, developed in collaboration with [DS4DH](https://www.unige.ch/medecine/radio/en/research-groups/1035teodoro) from the University of Geneva, introduces a specialized suite of efficient open biomedical NER models based on the GLiNER framework. GLiNER-biomed leverages synthetic annotations distilled from large generative biomedical language models to achieve state-of-the-art zero-shot and few-shot performance in biomedical entity recognition tasks.
28
 
 
 
29
  ### How to Use
30
  Install the official GLiNER library with pip:
31
  ```bash
@@ -77,32 +80,32 @@ We examined our models on 8 complex real-world datasets and compared them with o
77
  |------------------------|----------|----------------|------------------|-------------|
78
  | **Large models** | | | | |
79
  | [NuNER Zero](https://huggingface.co/numind/NuNER_Zero) | 40.87 | 21.79 | 13.94 | 33.67 |
80
- | [NuNER Zero span](https://huggingface.co/numind/NuNER_Zero-span) | 40.26 | 22.51 | 14.27 | 32.52 |
81
- | [GLiNER bio v0.1](https://huggingface.co/urchade/gliner_large_bio-v0.1) | 42.34 | 27.10 | 24.44 | 38.32 |
82
- | [GLiNER bio v0.2](https://huggingface.co/urchade/gliner_large_bio-v0.2) | 38.66 | 25.36 | 17.02 | 32.42 |
83
- | [GLiNER v1.0](https://huggingface.co/urchade/gliner_large-v1) | 47.77 | 29.60 | 21.13 | 40.78 |
84
- | [GLiNER v2.0](https://huggingface.co/urchade/gliner_large-v2) | 37.38 | 21.42 | 15.44 | 33.11 |
85
- | [GLiNER v2.1](https://huggingface.co/urchade/gliner_large-v2.1) | 48.04 | 29.75 | 28.20 | 43.43 |
86
- | [GLiNER news v2.1](https://huggingface.co/EmergentMethods/gliner_large_news-v2.1) | 48.99 | 31.79 | 33.77 | 45.13 |
87
- | [GLiNER v2.5](https://huggingface.co/gliner-community/gliner_large-v2.5) | 53.81 | 35.22 | 35.65 | 51.57 |
88
- | **[GLiNER-biomed](https://huggingface.co/Ihor/gliner-biomed-large-v1.0)** | **59.77**| **40.67** | **42.65** | **58.40** |
89
- | **[GLiNER-biomed-bi](https://huggingface.co/Ihor/gliner-biomed-bi-large-v1.0)** | 54.90 | 35.78 | 31.66 | 50.46 |
90
- | **Base models** | | | | |
91
- | [GLiNER v1.0](https://huggingface.co/urchade/gliner_medium-v1) | 41.61 | 24.98 | 10.27 | 31.59 |
92
- | [GLiNER v2.0](https://huggingface.co/urchade/gliner_medium-v2) | 34.33 | 24.48 | 22.01 | 30.58 |
93
- | [GLiNER v2.1](https://huggingface.co/urchade/gliner_medium-v2.1) | 40.25 | 25.26 | 14.41 | 32.64 |
94
- | [GLiNER news v2.1](https://huggingface.co/EmergentMethods/gliner_medium_news-v2.1) | 41.59 | 27.16 | 17.74 | 34.44 |
95
- | [GLiNER v2.5](https://huggingface.co/gliner-community/gliner_base-v2.5) | 46.49 | 30.93 | 25.26 | 44.68 |
96
- | **[GLiNER-biomed](https://huggingface.co/Ihor/gliner-biomed-base-v1.0)** | 54.37| **36.20** | **41.61** | 53.05 |
97
- | **[GLiNER-biomed-bi](https://huggingface.co/Ihor/gliner-biomed-bi-base-v1.0)** | **58.31** | 35.22 | 32.39 | **54.91** |
98
- | **Small models** | | | | |
99
- | [GLiNER v1.0](https://huggingface.co/urchade/gliner_small-v1) | 40.99 | 22.81 | 7.86 | 31.15 |
100
- | [GLiNER v2.0](https://huggingface.co/urchade/gliner_small-v2) | 33.55 | 21.12 | 15.76 | 28.78 |
101
- | [GLiNER v2.1](https://huggingface.co/urchade/gliner_small-v2.1) | 38.45 | 23.25 | 10.92 | 30.67 |
102
- | [GLiNER news v2.1](https://huggingface.co/EmergentMethods/gliner_small_news-v2.1) | 39.15 | 24.96 | 14.48 | 33.10 |
103
- | [GLiNER v2.5](https://huggingface.co/gliner-community/gliner_small-v2.5) | 38.21 | 28.53 | 18.01 | 36.88 |
104
- | **[GLiNER-biomed](https://huggingface.co/Ihor/gliner-biomed-small-v1.0)** | 52.53| **34.49** | **38.17** | 50.87 |
105
- | **[GLiNER-biomed-bi](https://huggingface.co/Ihor/gliner-biomed-bi-small-v1.0)** | **56.93** | 33.88 | 33.61 | **53.12** |
106
 
107
 
108
  ### Join Our Discord
 
1
  ---
2
+ base_model:
3
+ - microsoft/deberta-v3-base
 
 
4
  datasets:
5
  - knowledgator/GLINER-multi-task-synthetic-data
6
  - knowledgator/biomed_NER
7
+ language:
8
+ - en
9
+ library_name: gliner
10
+ license: apache-2.0
11
+ metrics:
12
+ - f1
13
  pipeline_tag: token-classification
14
  tags:
15
  - NER
 
18
  - encoder
19
  - entity recognition
20
  - biomed
 
 
 
 
21
  ---
22
+
23
  # GLiNER-BioMed
24
 
25
  **GLiNER** is a Named Entity Recognition (NER) model capable of identifying any entity type using a bidirectional transformer encoders (BERT-like). It provides a practical alternative to traditional NER models, which are limited to predefined entities, and Large Language Models (LLMs) that, despite their flexibility, are costly and large for resource-constrained scenarios.
 
27
 
28
  **GLiNER-biomed**, developed in collaboration with [DS4DH](https://www.unige.ch/medecine/radio/en/research-groups/1035teodoro) from the University of Geneva, introduces a specialized suite of efficient open biomedical NER models based on the GLiNER framework. GLiNER-biomed leverages synthetic annotations distilled from large generative biomedical language models to achieve state-of-the-art zero-shot and few-shot performance in biomedical entity recognition tasks.
29
 
30
+ Code and training pipelines: https://github.com/ds4dh/GLiNER-biomed
31
+
32
  ### How to Use
33
  Install the official GLiNER library with pip:
34
  ```bash
 
80
  |------------------------|----------|----------------|------------------|-------------|
81
  | **Large models** | | | | |
82
  | [NuNER Zero](https://huggingface.co/numind/NuNER_Zero) | 40.87 | 21.79 | 13.94 | 33.67 |
83
+ | [NuNER Zero span](https://huggingface.co/numind/NuNER_Zero-span) | 40.26 | 22.51 | 14.27 | 32.52 \t|
84
+ | [GLiNER bio v0.1](https://huggingface.co/urchade/gliner_large_bio-v0.1) \t| 42.34 | 27.10 | 24.44 \t| 38.32 \t|
85
+ | [GLiNER bio v0.2](https://huggingface.co/urchade/gliner_large_bio-v0.2) \t| 38.66 | 25.36 \t| 17.02 \t| 32.42 \t|
86
+ | [GLiNER v1.0](https://huggingface.co/urchade/gliner_large-v1) \t| 47.77 | 29.60 \t| 21.13 \t| 40.78 \t|
87
+ | [GLiNER v2.0](https://huggingface.co/urchade/gliner_large-v2) \t| 37.38 | 21.42 \t| 15.44 \t| 33.11 \t|
88
+ | [GLiNER v2.1](https://huggingface.co/urchade/gliner_large-v2.1) \t| 48.04 | 29.75 \t| 28.20 \t| 43.43 \t|
89
+ | [GLiNER news v2.1](https://huggingface.co/EmergentMethods/gliner_large_news-v2.1) \t| 48.99 | 31.79 \t| 33.77 \t| 45.13 \t|
90
+ | [GLiNER v2.5](https://huggingface.co/gliner-community/gliner_large-v2.5) \t| 53.81 | 35.22 \t| 35.65 \t| 51.57 \t|
91
+ | **[GLiNER-biomed](https://huggingface.co/Ihor/gliner-biomed-large-v1.0)** \t| **59.77**| **40.67** \t| **42.65** \t| **58.40** |
92
+ | **[GLiNER-biomed-bi](https://huggingface.co/Ihor/gliner-biomed-bi-large-v1.0)** \t| 54.90 | 35.78 \t| 31.66 \t| 50.46 \t|
93
+ | **Base models** \t| \t| \t| \t| \t|
94
+ | [GLiNER v1.0](https://huggingface.co/urchade/gliner_medium-v1) \t| 41.61 | 24.98 \t| 10.27 \t| 31.59 \t|
95
+ | [GLiNER v2.0](https://huggingface.co/urchade/gliner_medium-v2) \t| 34.33 | 24.48 \t| 22.01 \t| 30.58 \t|
96
+ | [GLiNER v2.1](https://huggingface.co/urchade/gliner_medium-v2.1) \t| 40.25 | 25.26 \t| 14.41 \t| 32.64 \t|
97
+ | [GLiNER news v2.1](https://huggingface.co/EmergentMethods/gliner_medium_news-v2.1) \t| 41.59 | 27.16 \t| 17.74 \t| 34.44 \t|
98
+ | [GLiNER v2.5](https://huggingface.co/gliner-community/gliner_base-v2.5) \t| 46.49\t| 30.93 \t| 25.26 \t| 44.68 \t|
99
+ | **[GLiNER-biomed](https://huggingface.co/Ihor/gliner-biomed-base-v1.0)** \t| 54.37| **36.20** \t| **41.61** \t| 53.05 |
100
+ | **[GLiNER-biomed-bi](https://huggingface.co/Ihor/gliner-biomed-bi-base-v1.0)** \t| **58.31**\t| 35.22 \t| 32.39 \t| **54.91** \t|
101
+ | **Small models** \t| \t| \t| \t| \t|
102
+ | [GLiNER v1.0](https://huggingface.co/urchade/gliner_small-v1) \t| 40.99\t| 22.81 \t| 7.86 \t| 31.15 \t|
103
+ | [GLiNER v2.0](https://huggingface.co/urchade/gliner_small-v2) \t| 33.55\t| 21.12 \t| 15.76 \t| 28.78 \t|
104
+ | [GLiNER v2.1](https://huggingface.co/urchade/gliner_small-v2.1) \t| 38.45\t| 23.25 \t| 10.92 \t| 30.67 \t|
105
+ | [GLiNER news v2.1](https://huggingface.co/EmergentMethods/gliner_small_news-v2.1) \t| 39.15\t| 24.96 \t| 14.48 \t| 33.10 \t|
106
+ | [GLiNER v2.5](https://huggingface.co/gliner-community/gliner_small-v2.5) \t| 38.21\t| 28.53 \t| 18.01 \t| 36.88 \t|
107
+ | **[GLiNER-biomed](https://huggingface.co/Ihor/gliner-biomed-small-v1.0)** \t| 52.53| **34.49** \t| **38.17** \t| 50.87 |
108
+ | **[GLiNER-biomed-bi](https://huggingface.co/Ihor/gliner-biomed-bi-small-v1.0)** \t| **56.93**\t| 33.88 \t| 33.61 \t| **53.12** \t|
109
 
110
 
111
  ### Join Our Discord