GUI-Net-Crawler / README.md
Bofeee5675's picture
Update README.md
67e346c verified
---
license: apache-2.0
language:
- en
tags:
- VLM
---
# How to use this data?
After download this repo, use `cat` to get zip file:
```bash
cat baidu_wiki_part_* > merge.zip
```
Then simply, unzip this zip file
```bash
unzip merge.zip
```
# What is in this data?
## Image(Screenshot)
Raw images are in `images` folder.
```bash
/wikihow$ ls data/images | head -5
1111-4.jpg
111-15.jpg
1-draw-7.png
20200613_130717.jpg
22-19.jpg
```
## Index page
Index page is a collection of web urls. This is how we start to crawl these websites.
```bash
wikihow$ cat data/index/index.txt | head -5
https://www.wikihow.com/Category:Amusement-and-Theme-Parks
https://www.wikihow.com/Category:Artwork
https://www.wikihow.com/Category:Books
https://www.wikihow.com/Category:Celebrities
https://www.wikihow.com/Category:Concerts
```
For examples, above shows we start to crawl tutorials from these 5 links.
## Summary file
Summary file is title and actual webpage
```bash
cat data/summary/Computer-Networking.json | head -10
[
{
"title": "How to\nOvercome Your Fear of Roller Coasters",
"url": "https://www.wikihow.com/Overcome-Your-Fear-of-Roller-Coasters"
},
{
"title": "Six Flags Flash Pass: Everything You Need to Know",
"url": "https://www.wikihow.com/Use-a-Flash-Pass-at-Six-Flags"
},
{
```
In this step, we get titles and links for each tutorials.
## Trajectory file
Trajectory file is trajectory we follow certain heuristic to process the HTML
```bash
cat data/steps/Add-a-Bookmark-in-Microsoft-Word/steps_processed.json | head -10
{
"introduction": "CategoriesComputers and ElectronicsSoftwareOfficeMicrosoft OfficeMicrosoft WordHow to Add a Bookmark in Microsoft Word",
"sections": [
{
"headline": "Adding a Bookmark",
"steps": [
{
"text": "Select the place in the text you wish to bookmark. You can either highlight a block of text or click to place your cursor at the start of a paragraph.",
"image_src": "v4-460px-Add-a-Bookmark-in-Microsoft-Word-Step-1-Version-4.jpg"
},
```
For examples, above shows a part of trajectory. They are organized as stepwise text-image. The image are downloaded as dicussed in previous documentation.
## Other notes
Baidu Experience is slightly different compared to wikihow. But they do follow the similar structure.
Source code to generate these data is in this [repo](https://github.com/TongUI-agent/TongUI-Crawler)