Upload README.md with huggingface_hub

README.md

@@ -47,9 +47,10 @@ dataset_info:
     - [Data Instances](#data-instances)
     - [Data Fields](#data-fields)
   - [Example Usage (Python)](#example-usage-python)
-  - [Intended Use](#intended-use)
+  - [Intended Use][def]
   - [Dataset Statistics](#dataset-statistics)
     - [Token Statistics](#token-statistics)
+    - [Quality Signal: CCNet Perplexity](#quality-signal-ccnet-perplexity)
   - [License](#license)
   - [Aknowledgements](#aknowledgements)
   - [Citation](#citation)
@@ -130,9 +131,7 @@ Estimated using [sentencepiece](https://github.com/google/sentencepiece)   token
 - Estimated total tokens: ~1.9B tokens
 - Average number of tokens per document: ~4 120 tokens
 
-## Quality Signal: CCNet Perplexity
-
-### Overview
+### Quality Signal: CCNet Perplexity
 
 `CCNet Perplexity` is a linguistic quality indicator used to measure how close a given text is to a high-quality reference corpus, typically Wikipedia.
 It is commonly employed in large-scale dataset filtering pipelines for language model training, in order to identify noisy, malformed, or out-of-domain content.
@@ -141,7 +140,7 @@ In this dataset, the score is computed using the open-source implementation prov
 
 The value is exposed in the `quality_signals` field under the key `ccnet_perplexity`.
 
-### Score Interpretation
+#### Score Interpretation
 
 - **Low perplexity (~100–300)**
   Text is linguistically close to Wikipedia:
@@ -163,11 +162,11 @@ Lower perplexity indicates that the text is more likely under the Wikipedia-trai
 
 ⚠️ A high perplexity score does not necessarily imply low semantic value, but rather a **linguistic distance** from the Wikipedia domain.
 
-### Observations for the Wikisource FR Dataset
+#### Observations for the Wikisource FR Dataset
 
 For this dataset:
 
-![CCNet Perplexity Histogram](wikisource_fr_ccnet_perplexity_histogram.png)
+![CCNet Perplexity Histogram](images/wikisource_fr_ccnet_perplexity_histogram.png)
 
 - **Median CCNet Perplexity: 298.84**
 
@@ -210,3 +209,6 @@ Many thanks to the [Wikimedia Foundation](https://wikimediafoundation.org/) for
     url    = "https://huggingface.co/datasets/LeMoussel/wikisource_fr"
 }
 ```
+
+
+[def]: #intended-use