dchaplinsky
commited on
Commit
·
c378c7a
1
Parent(s):
7983a3a
Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,29 @@
|
|
1 |
---
|
2 |
license: mit
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: mit
|
3 |
+
task_categories:
|
4 |
+
- question-answering
|
5 |
+
pretty_name: Every Prompt
|
6 |
+
size_categories:
|
7 |
+
- 1M<n<10M
|
8 |
+
multilinguality:
|
9 |
+
- multilingual
|
10 |
---
|
11 |
+
|
12 |
+
## Every Prompt
|
13 |
+
Every Prompt is a data driven approach to mine instructions from the web.
|
14 |
+
It contains more than million FAQs and HowTos from all around the world in the structured format.
|
15 |
+
It also has basic pre-processing to calculate the length of the useful text and identify the language of that text with the help of [GCLD3](https://github.com/google/cld3)
|
16 |
+
|
17 |
+
It relies on the [Web Data Commons](http://webdatacommons.org) dataset to find the seed list of sites with [**HowTo**](https://schema.org/HowTo) and [**FAQPage**](https://schema.org/FAQPage) items.
|
18 |
+
The general pipeline looks like this:
|
19 |
+
* Download 1.6TB of structured data from webdatacommons to identify the pages with the structured data we need (wget/parallel). That gives us 1,985,925 seed pages
|
20 |
+
* Crawls the seed pages and tries to extract structured data using [extruct](https://pypi.org/project/extruct/#description) package. That lefts around 1,358,638 pages which are alive and well-formed.
|
21 |
+
* Extracts only the relevant structured data of the HowTo/FAQPage type with the help of jmespath. That boils down to 1,266,926 json documents.
|
22 |
+
* Extracts the textual information out of the structure to identify the language of the text, length of the textual data and the text/data ratio.
|
23 |
+
|
24 |
+
The resulting dataset you can use by applying the filtering on the language and amount of the text. You need to convert the structured data into instructions yourself.
|
25 |
+
You'll need to apply extra cleansing/evaluation of the instructions you've got because you know, internet is still full of crap
|
26 |
+
|
27 |
+
## License
|
28 |
+
**Code** of the project has MIT license.
|
29 |
+
|