tamnd commited on
Commit
18e7ca4
·
verified ·
1 Parent(s): 8855f7b

Publish shard CC-MAIN-2026-08/00380

Browse files
Files changed (3) hide show
  1. README.md +9 -9
  2. data/CC-MAIN-2026-08/00380.parquet +3 -0
  3. stats.csv +2 -1
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 **24,217,431 documents across 1253 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
 
@@ -164,28 +164,28 @@ Empty conversions (pages where trafilatura could not extract meaningful content)
164
 
165
  ### Compression Ratios
166
 
167
- Numbers below are actual measurements summed across all 1253 files of CC-MAIN-2026-08 (24,217,431 pages total), projected to the full crawl of 100,000 WARC files.
168
 
169
- | Stage | 1253 files (measured) | 100,000 files (projected) | Reduction |
170
  |---|---|---|---|
171
- | Raw WARC (.warc.gz, downloaded) | ~1015.6 GB | ~83 TB | — |
172
  | HTML extracted (uncompressed) | 2.9 TB | ~295 TB | — |
173
  | Packed markdown WARC (.md.warc.gz) | ~51.4 GB | ~3.7 TB | **-98.3%** vs HTML |
174
  | Final Parquet (Zstd level 19) | 34.7 GB | ~2.9 TB | **-32.5%** vs packed WARC |
175
 
176
  The big win is the HTML → Markdown step: trafilatura strips all tags, scripts, styles, navigation, and ads, keeping only the main content. This cuts 2.9 TB of uncompressed HTML down to 109.3 GB of markdown — a **98.3% reduction** — before any file-level compression is applied. Parquet with Zstd level 19 then compresses the markdown a further 68.3%.
177
 
178
- End to end: ~1015.6 GB of raw gzipped WARCs becomes **34.7 GB of Parquet** — a **96.6% total reduction** — containing 24,217,431 clean markdown documents.
179
 
180
  ### Processing Times
181
 
182
- Pipeline timings across 1253 shards of CC-MAIN-2026-08:
183
 
184
  ```
185
- Download (raw WARC) ████████████████████████ total 96h 6m 35s avg 4m 36s
186
  Convert (HTML → MD) █░░░░░░░░░░░░░░░░░░░░░░░ total 6h 2m 48s avg 17s
187
- Export (Parquet) █████░░░░░░░░░░░░░░░░░░░ total 23h 18m 0s avg 1m 6s
188
- Publish (HuggingFace) █░░░░░░░░░░░░░░░░░░░░░░░ total 7h 51m 53s avg 22s
189
  ```
190
 
191
  ### Dataset Charts
 
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 **24,236,778 documents across 1254 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
 
 
164
 
165
  ### Compression Ratios
166
 
167
+ Numbers below are actual measurements summed across all 1254 files of CC-MAIN-2026-08 (24,236,778 pages total), projected to the full crawl of 100,000 WARC files.
168
 
169
+ | Stage | 1254 files (measured) | 100,000 files (projected) | Reduction |
170
  |---|---|---|---|
171
+ | Raw WARC (.warc.gz, downloaded) | ~1016.4 GB | ~83 TB | — |
172
  | HTML extracted (uncompressed) | 2.9 TB | ~295 TB | — |
173
  | Packed markdown WARC (.md.warc.gz) | ~51.4 GB | ~3.7 TB | **-98.3%** vs HTML |
174
  | Final Parquet (Zstd level 19) | 34.7 GB | ~2.9 TB | **-32.5%** vs packed WARC |
175
 
176
  The big win is the HTML → Markdown step: trafilatura strips all tags, scripts, styles, navigation, and ads, keeping only the main content. This cuts 2.9 TB of uncompressed HTML down to 109.3 GB of markdown — a **98.3% reduction** — before any file-level compression is applied. Parquet with Zstd level 19 then compresses the markdown a further 68.3%.
177
 
178
+ End to end: ~1016.4 GB of raw gzipped WARCs becomes **34.7 GB of Parquet** — a **96.6% total reduction** — containing 24,236,778 clean markdown documents.
179
 
180
  ### Processing Times
181
 
182
+ Pipeline timings across 1254 shards of CC-MAIN-2026-08:
183
 
184
  ```
185
+ Download (raw WARC) ████████████████████████ total 96h 13m 56s avg 4m 36s
186
  Convert (HTML → MD) █░░░░░░░░░░░░░░░░░░░░░░░ total 6h 2m 48s avg 17s
187
+ Export (Parquet) █████░░░░░░░░░░░░░░░░░░░ total 23h 21m 18s avg 1m 7s
188
+ Publish (HuggingFace) █░░░░░░░░░░░░░░░░░░░░░░░ total 7h 52m 19s avg 22s
189
  ```
190
 
191
  ### Dataset Charts
data/CC-MAIN-2026-08/00380.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3418a8f379f00f3be02ee3eeeced54d76c585bf1b3f3754daf82eed451696aca
3
+ size 28624484
stats.csv CHANGED
@@ -349,6 +349,7 @@ CC-MAIN-2026-08,376,19603,2573100392,94744832,31392835,2026-03-14T02:02:32Z,474,
349
  CC-MAIN-2026-08,377,19653,2608016900,94790831,30143007,2026-03-14T02:12:35Z,526,0,101,14
350
  CC-MAIN-2026-08,378,19520,2557252949,91815354,29665998,2026-03-14T02:24:41Z,543,0,167,14
351
  CC-MAIN-2026-08,379,19426,2552876640,91839182,29849311,2026-03-14T02:34:20Z,420,0,159,13
 
352
  CC-MAIN-2026-08,401,19441,2586676962,94334130,29618260,2026-03-13T11:41:00Z,0,0,33,0
353
  CC-MAIN-2026-08,402,19446,2552785517,96654791,30560442,2026-03-13T11:49:21Z,0,443,57,0
354
  CC-MAIN-2026-08,403,19320,2579048144,91406292,29341198,2026-03-13T11:56:46Z,0,414,29,0
@@ -590,7 +591,7 @@ CC-MAIN-2026-08,797,19361,2587106167,92002417,29439496,2026-03-14T02:10:35Z,454,
590
  CC-MAIN-2026-08,798,19332,2555419003,99434220,31645658,2026-03-14T02:21:57Z,477,0,178,24
591
  CC-MAIN-2026-08,799,19275,2559716233,94060891,29424540,2026-03-14T02:30:35Z,422,0,120,16
592
  CC-MAIN-2026-08,800,19353,2574122589,98621234,30172589,2026-03-14T02:41:05Z,458,0,165,18
593
- CC-MAIN-2026-08,801,19437,2573915480,93218254,30052830,2026-03-14T02:44:35Z,0,0,212,0
594
  CC-MAIN-2026-08,851,19575,2612873485,97000364,30911294,2026-03-13T06:41:39Z,163,0,37,204
595
  CC-MAIN-2026-08,852,19334,2546015208,94959847,30205044,2026-03-13T06:45:52Z,183,0,57,11
596
  CC-MAIN-2026-08,853,19694,2567973975,95901932,30303494,2026-03-13T06:50:12Z,150,0,28,80
 
349
  CC-MAIN-2026-08,377,19653,2608016900,94790831,30143007,2026-03-14T02:12:35Z,526,0,101,14
350
  CC-MAIN-2026-08,378,19520,2557252949,91815354,29665998,2026-03-14T02:24:41Z,543,0,167,14
351
  CC-MAIN-2026-08,379,19426,2552876640,91839182,29849311,2026-03-14T02:34:20Z,420,0,159,13
352
+ CC-MAIN-2026-08,380,19347,2574961073,89057601,28624484,2026-03-14T02:46:25Z,441,0,198,0
353
  CC-MAIN-2026-08,401,19441,2586676962,94334130,29618260,2026-03-13T11:41:00Z,0,0,33,0
354
  CC-MAIN-2026-08,402,19446,2552785517,96654791,30560442,2026-03-13T11:49:21Z,0,443,57,0
355
  CC-MAIN-2026-08,403,19320,2579048144,91406292,29341198,2026-03-13T11:56:46Z,0,414,29,0
 
591
  CC-MAIN-2026-08,798,19332,2555419003,99434220,31645658,2026-03-14T02:21:57Z,477,0,178,24
592
  CC-MAIN-2026-08,799,19275,2559716233,94060891,29424540,2026-03-14T02:30:35Z,422,0,120,16
593
  CC-MAIN-2026-08,800,19353,2574122589,98621234,30172589,2026-03-14T02:41:05Z,458,0,165,18
594
+ CC-MAIN-2026-08,801,19437,2573915480,93218254,30052830,2026-03-14T02:44:35Z,0,0,212,26
595
  CC-MAIN-2026-08,851,19575,2612873485,97000364,30911294,2026-03-13T06:41:39Z,163,0,37,204
596
  CC-MAIN-2026-08,852,19334,2546015208,94959847,30205044,2026-03-13T06:45:52Z,183,0,57,11
597
  CC-MAIN-2026-08,853,19694,2567973975,95901932,30303494,2026-03-13T06:50:12Z,150,0,28,80