tamnd commited on
Commit
393e4b3
·
verified ·
1 Parent(s): e30b098

Publish shard CC-MAIN-2026-08/00020

Browse files
README.md CHANGED
@@ -32,7 +32,7 @@ configs:
32
 
33
  Open Index is a large-scale web text dataset built from [Common Crawl](https://commoncrawl.org). Every page goes through a pipeline that extracts the main content from raw HTML, converts it to clean Markdown using [trafilatura](https://github.com/adbar/trafilatura), and packages the result into Parquet files with full WARC metadata preserved.
34
 
35
- The dataset currently includes crawl **CC-MAIN-2026-08** with **407,490 documents across 20 shards**. We plan to add more snapshots over time.
36
 
37
  Open Index is released under the **Open Data Commons Attribution License (ODC-By) v1.0**, the same license used by Common Crawl.
38
 
@@ -166,27 +166,27 @@ Empty conversions (pages where trafilatura could not extract meaningful content)
166
 
167
  ### Compression Ratios
168
 
169
- Numbers below are actual measurements summed across all 20 files of CC-MAIN-2026-08 (407,490 pages total), projected to the full crawl of 100,000 WARC files.
170
 
171
- | Stage | 20 files (measured) | 100,000 files (projected) | Reduction |
172
  |---|---|---|---|
173
- | Raw WARC (.warc.gz, downloaded) | ~16.2 GB | ~83 TB | — |
174
- | HTML extracted (uncompressed) | 51.9 GB | ~295 TB | — |
175
- | Packed markdown WARC (.md.warc.gz) | ~771.3 MB | ~3.7 TB | **-98.5%** vs HTML |
176
- | Final Parquet (Zstd level 19) | 537.0 MB | ~2.9 TB | **-30.4%** vs packed WARC |
177
 
178
- The big win is the HTML → Markdown step: trafilatura strips all tags, scripts, styles, navigation, and ads, keeping only the main content. This cuts 51.9 GB of uncompressed HTML down to 1.6 GB of markdown — a **98.5% reduction** — before any file-level compression is applied. Parquet with Zstd level 19 then compresses the markdown a further 67.3%.
179
 
180
- End to end: ~16.2 GB of raw gzipped WARCs becomes **537.0 MB of Parquet** — a **96.8% total reduction** — containing 407,490 clean markdown documents.
181
 
182
  ### Processing Times
183
 
184
- Pipeline timings across 20 shards of CC-MAIN-2026-08:
185
 
186
  ```
187
  Pack (download + HTML→MD) ░░░░░░░░░░░░░░░░░░░░░░░░ total — avg —
188
  Export (Parquet) ░░░░░░░░░░░░░░░░░░░░░░░░ total — avg —
189
- Publish (HuggingFace) ░░░░░░░░░░░░░░░░░░░░░░░░ total 1m 21s avg 4s
190
  ```
191
 
192
  ### Personal and Sensitive Information
 
32
 
33
  Open Index is a large-scale web text dataset built from [Common Crawl](https://commoncrawl.org). Every page goes through a pipeline that extracts the main content from raw HTML, converts it to clean Markdown using [trafilatura](https://github.com/adbar/trafilatura), and packages the result into Parquet files with full WARC metadata preserved.
34
 
35
+ The dataset currently includes crawl **CC-MAIN-2026-08** with **427,107 documents across 21 shards**. We plan to add more snapshots over time.
36
 
37
  Open Index is released under the **Open Data Commons Attribution License (ODC-By) v1.0**, the same license used by Common Crawl.
38
 
 
166
 
167
  ### Compression Ratios
168
 
169
+ Numbers below are actual measurements summed across all 21 files of CC-MAIN-2026-08 (427,107 pages total), projected to the full crawl of 100,000 WARC files.
170
 
171
+ | Stage | 21 files (measured) | 100,000 files (projected) | Reduction |
172
  |---|---|---|---|
173
+ | Raw WARC (.warc.gz, downloaded) | ~17.0 GB | ~83 TB | — |
174
+ | HTML extracted (uncompressed) | 54.3 GB | ~295 TB | — |
175
+ | Packed markdown WARC (.md.warc.gz) | ~814.2 MB | ~3.7 TB | **-98.5%** vs HTML |
176
+ | Final Parquet (Zstd level 19) | 565.5 MB | ~2.9 TB | **-30.5%** vs packed WARC |
177
 
178
+ The big win is the HTML → Markdown step: trafilatura strips all tags, scripts, styles, navigation, and ads, keeping only the main content. This cuts 54.3 GB of uncompressed HTML down to 1.7 GB of markdown — a **98.5% reduction** — before any file-level compression is applied. Parquet with Zstd level 19 then compresses the markdown a further 67.4%.
179
 
180
+ End to end: ~17.0 GB of raw gzipped WARCs becomes **565.5 MB of Parquet** — a **96.8% total reduction** — containing 427,107 clean markdown documents.
181
 
182
  ### Processing Times
183
 
184
+ Pipeline timings across 21 shards of CC-MAIN-2026-08:
185
 
186
  ```
187
  Pack (download + HTML→MD) ░░░░░░░░░░░░░░░░░░░░░░░░ total — avg —
188
  Export (Parquet) ░░░░░░░░░░░░░░░░░░░░░░░░ total — avg —
189
+ Publish (HuggingFace) ░░░░░░░░░░░░░░░░░░░░░░░░ total 1m 26s avg 4s
190
  ```
191
 
192
  ### Personal and Sensitive Information
charts/compression_pie.png CHANGED
charts/rows_chart.png CHANGED
charts/size_chart.png CHANGED
charts/timing_chart.png CHANGED
data/CC-MAIN-2026-08/00020.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a5d094a2a707d5103122cc0211151e43a8a28d0cec7d9510ef7cf5a8cbb684fc
3
+ size 29947204
stats.csv CHANGED
@@ -18,4 +18,5 @@ CC-MAIN-2026-08,15,19509,2602820538,92725066,29817208,2026-03-13T05:46:20Z,0,0,2
18
  CC-MAIN-2026-08,16,19637,2631647654,95718985,31017617,2026-03-13T05:46:30Z,0,0,6
19
  CC-MAIN-2026-08,17,19286,2551077727,91540128,29028673,2026-03-13T05:46:40Z,0,0,7
20
  CC-MAIN-2026-08,18,19552,2597892662,93434179,29872907,2026-03-13T05:46:51Z,0,0,7
21
- CC-MAIN-2026-08,19,19176,2531689800,93034579,29963080,2026-03-13T05:46:51Z,0,0,0
 
 
18
  CC-MAIN-2026-08,16,19637,2631647654,95718985,31017617,2026-03-13T05:46:30Z,0,0,6
19
  CC-MAIN-2026-08,17,19286,2551077727,91540128,29028673,2026-03-13T05:46:40Z,0,0,7
20
  CC-MAIN-2026-08,18,19552,2597892662,93434179,29872907,2026-03-13T05:46:51Z,0,0,7
21
+ CC-MAIN-2026-08,19,19176,2531689800,93034579,29963080,2026-03-13T05:46:59Z,0,0,5
22
+ CC-MAIN-2026-08,20,19617,2599556257,95774763,29947204,2026-03-13T05:46:59Z,0,0,0