wzkariampuzha
commited on
Commit
·
8e354bc
1
Parent(s):
c609ced
Update README.md
Browse files
README.md
CHANGED
@@ -48,7 +48,7 @@ task_ids:
|
|
48 |
|
49 |
### Dataset Summary
|
50 |
|
51 |
-
EpiSet4NER is a bronze-standard dataset for epidemiological entity recognition of location, epidemiologic types (e.g. "prevalence", "annual incidence", "estimated occurrence"), and epidemiological rates (e.g. "1.7 per 1,000,000 live births", "2.1:1.000.000", "one in five million", "0.03%") created by the [Genetic and Rare Diseases Information Center (GARD)](https://rarediseases.info.nih.gov/), a program in [the National Center for Advancing Translational Sciences](https://ncats.nih.gov/), one of the 27 [National Institutes of Health](https://www.nih.gov/). It was labeled programmatically using spaCy NER and rule-based methods. This weakly-supervised teaching method allowed us to construct this imprecise dataset with minimal manual effort and achieve satisfactory performance on a multi-type token classification problem. The test set was manually corrected by 3 NCATS researchers and a GARD curator (genetic and rare disease expert).
|
52 |
|
53 |
An example of 'train' looks as follows.
|
54 |
```
|
@@ -69,16 +69,19 @@ The data fields are the same among all splits.
|
|
69 |
|
70 |
|name |train |validation|test|
|
71 |
|---------|-----:|----:|----:|
|
72 |
-
|EpiSet abstracts|456|114|50|
|
73 |
-
|EpiSet tokens |117888|31262|13910|
|
74 |
|
75 |
|
76 |
## Dataset Creation
|
77 |
-
![EpiSet Creation Flowchart](https://raw.githubusercontent.com/ncats/epi4GARD/master/EpiExtract4GARD/datasets/EpiCustomV3/EpiSet%20Flowchart%20FINAL.png
|
|
|
|
|
78 |
*Figure 1:* Creation of EpiSet4NER by NIH/NCATS
|
79 |
Comparing the programmatically labeled test set to the manually corrected test set allowed us to measure the precision, recall, and F1 of the programmatic labeling.
|
80 |
|
81 |
-
*Table 1:* Programmatic labeling of EpiSet4NER
|
|
|
82 |
| Evaluation Level | Entity | Precision | Recall | F1 |
|
83 |
|:----------------:|:------------------------:|:---------:|:------:|:-----:|
|
84 |
| Entity-Level | Overall | 0.559 | 0.662 | 0.606 |
|
@@ -92,7 +95,7 @@ Comparing the programmatically labeled test set to the manually corrected test s
|
|
92 |
|
93 |
An example of the text labeling:
|
94 |
![Text Labeling](https://raw.githubusercontent.com/ncats/epi4GARD/master/EpiExtract4GARD/datasets/EpiCustomV3/Text%20Labeling4.png)
|
95 |
-
*Figure 2:* Text Labeling using spaCy and rule-based labeling. Ideal labeling is bolded on the left. Actual programmatic output is on the right. [
|
96 |
|
97 |
### Curation Rationale
|
98 |
|
@@ -139,7 +142,7 @@ Assisting 25-30 millions Americans with rare diseases. Additionally can be usefu
|
|
139 |
|
140 |
### Dataset Curators
|
141 |
|
142 |
-
NIH GARD
|
143 |
|
144 |
### Licensing Information
|
145 |
|
|
|
48 |
|
49 |
### Dataset Summary
|
50 |
|
51 |
+
EpiSet4NER is a bronze-standard dataset for epidemiological entity recognition of location, epidemiologic types (e.g. "prevalence", "annual incidence", "estimated occurrence"), and epidemiological rates (e.g. "1.7 per 1,000,000 live births", "2.1:1.000.000", "one in five million", "0.03%") created by the [Genetic and Rare Diseases Information Center (GARD)](https://rarediseases.info.nih.gov/), a program in [the National Center for Advancing Translational Sciences](https://ncats.nih.gov/), one of the 27 [National Institutes of Health](https://www.nih.gov/). It was labeled programmatically using spaCy NER and rule-based methods. This weakly-supervised teaching method allowed us to construct this imprecise dataset with minimal manual effort and achieve satisfactory performance on a multi-type token classification problem. The test set was manually corrected by 3 NCATS researchers and a GARD curator (genetic and rare disease expert). It was used to train [EpiExtract4GARD](https://huggingface.co/ncats/EpiExtract4GARD), a BioBERT-based model fine-tuned for NER.
|
52 |
|
53 |
An example of 'train' looks as follows.
|
54 |
```
|
|
|
69 |
|
70 |
|name |train |validation|test|
|
71 |
|---------|-----:|----:|----:|
|
72 |
+
|EpiSet \# of abstracts|456|114|50|
|
73 |
+
|EpiSet \# tokens |117888|31262|13910|
|
74 |
|
75 |
|
76 |
## Dataset Creation
|
77 |
+
![EpiSet Creation Flowchart](https://raw.githubusercontent.com/ncats/epi4GARD/master/EpiExtract4GARD/datasets/EpiCustomV3/EpiSet%20Flowchart%20FINAL.png)
|
78 |
+
Other type
|
79 |
+
<img src="https://raw.githubusercontent.com/ncats/epi4GARD/master/EpiExtract4GARD/datasets/EpiCustomV3/EpiSet%20Flowchart%20FINAL.png" height="200">
|
80 |
*Figure 1:* Creation of EpiSet4NER by NIH/NCATS
|
81 |
Comparing the programmatically labeled test set to the manually corrected test set allowed us to measure the precision, recall, and F1 of the programmatic labeling.
|
82 |
|
83 |
+
*Table 1:* Programmatic labeling of EpiSet4NER
|
84 |
+
|
85 |
| Evaluation Level | Entity | Precision | Recall | F1 |
|
86 |
|:----------------:|:------------------------:|:---------:|:------:|:-----:|
|
87 |
| Entity-Level | Overall | 0.559 | 0.662 | 0.606 |
|
|
|
95 |
|
96 |
An example of the text labeling:
|
97 |
![Text Labeling](https://raw.githubusercontent.com/ncats/epi4GARD/master/EpiExtract4GARD/datasets/EpiCustomV3/Text%20Labeling4.png)
|
98 |
+
*Figure 2:* Text Labeling using spaCy and rule-based labeling. Ideal labeling is bolded on the left. Actual programmatic output is on the right. [\[Figure citation\]](https://pubmed.ncbi.nlm.nih.gov/33649778/)
|
99 |
|
100 |
### Curation Rationale
|
101 |
|
|
|
142 |
|
143 |
### Dataset Curators
|
144 |
|
145 |
+
[NIH GARD](https://rarediseases.info.nih.gov/about-gard/pages/23/about-gard)
|
146 |
|
147 |
### Licensing Information
|
148 |
|