Update README.md
Browse files
README.md
CHANGED
@@ -4,4 +4,78 @@ language:
|
|
4 |
- en
|
5 |
size_categories:
|
6 |
- n<1K
|
7 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
4 |
- en
|
5 |
size_categories:
|
6 |
- n<1K
|
7 |
+
---
|
8 |
+
|
9 |
+
## sfia-9-chunks Dataset
|
10 |
+
|
11 |
+
### Overview
|
12 |
+
|
13 |
+
The `sfia-9-chunks` dataset is a derived dataset from [`sfia-9-scraped`](https://huggingface.co/datasets/Programmer-RD-AI/sfia-9-scraped). It uses sentence embeddings and hierarchical clustering to split each SFIA-9 document into coherent semantic chunks. This chunking facilitates more efficient downstream tasks like semantic search, question answering, and topic modeling.
|
14 |
+
|
15 |
+
### Chunking Methodology
|
16 |
+
|
17 |
+
We employ the following procedure to generate chunks:
|
18 |
+
|
19 |
+
```python
|
20 |
+
from sentence_transformers import SentenceTransformer
|
21 |
+
from sklearn.cluster import AgglomerativeClustering
|
22 |
+
|
23 |
+
MODEL_NAME = "all-MiniLM-L12-v2"
|
24 |
+
|
25 |
+
def get_chunks(document: str) -> list[str]:
|
26 |
+
model = SentenceTransformer(MODEL_NAME)
|
27 |
+
sentences = document.split(". ")
|
28 |
+
embs = model.encode(sentences)
|
29 |
+
clustering = AgglomerativeClustering(n_clusters=None, distance_threshold=1.0).fit(
|
30 |
+
embs
|
31 |
+
)
|
32 |
+
chunks = []
|
33 |
+
for label in set(clustering.labels_):
|
34 |
+
group = [
|
35 |
+
sentences[i]
|
36 |
+
for i in range(len(sentences))
|
37 |
+
if clustering.labels_[i] == label
|
38 |
+
]
|
39 |
+
chunks.append(". ".join(group))
|
40 |
+
return chunks
|
41 |
+
```
|
42 |
+
|
43 |
+
* **Model:** `all-MiniLM-L12-v2` for efficient sentence embeddings.
|
44 |
+
* **Clustering:** Agglomerative clustering with a distance threshold of `1.0` to dynamically determine the number of semantic groups.
|
45 |
+
|
46 |
+
### Dataset Structure
|
47 |
+
|
48 |
+
This dataset consists of a flat list of semantic text chunks derived from the SFIA-9 documents. When loaded, each example is an object with a single field:
|
49 |
+
|
50 |
+
* `chunks` (`string`): One coherent semantic chunk extracted via hierarchical clustering.
|
51 |
+
|
52 |
+
### Usage Example
|
53 |
+
|
54 |
+
```python
|
55 |
+
from datasets import load_dataset
|
56 |
+
|
57 |
+
dataset = load_dataset("Programmer-RD-AI/sfia-9-chunks")
|
58 |
+
|
59 |
+
for example in dataset:
|
60 |
+
print(example["chunks"])
|
61 |
+
```
|
62 |
+
|
63 |
+
### License
|
64 |
+
|
65 |
+
This dataset is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.
|
66 |
+
You can view the full license at: [https://creativecommons.org/licenses/by/4.0/](https://creativecommons.org/licenses/by/4.0/)
|
67 |
+
|
68 |
+
### Citation
|
69 |
+
|
70 |
+
If you use this dataset in your research, please cite:
|
71 |
+
|
72 |
+
```bibtex
|
73 |
+
@misc{ranuga_disansa_gamage_2025,
|
74 |
+
author = {Ranuga Disansa Gamage},
|
75 |
+
title = {sfia-9-chunks (Revision 035dc41)},
|
76 |
+
year = 2025,
|
77 |
+
url = {https://huggingface.co/datasets/Programmer-RD-AI/sfia-9-chunks},
|
78 |
+
doi = {10.57967/hf/5747},
|
79 |
+
publisher = {Hugging Face}
|
80 |
+
}
|
81 |
+
```
|