ClueWeb22 API

Getting Started

This API can be used with the ClueWeb22 dataset. Depending on which data formats are available, the structure of ClueWeb22 will look like this:

• ClueWeb22/html
  • The html directory contains the html of the webpage at the time that it was crawled. The html is in WARC format.
• ClueWeb22/txt
  • The text directory contains only the text (no html markup) from html nodes that were marked as primary content. The text is in json format.
• ClueWeb22/vdom
  • The vdom directory contains semantic annotations such as table, list, etc and node annotations such as the x and y position of the node. The vdom is in protobuf format, and requires this API to access. The protobuf files are stored in zip archives.
• ClueWeb22/inlink
  • The inlink directory contains the links that point to each ClueWeb document. Some documents may not have inlinks. The inlinks are in json format.
• ClueWeb22/outlink
  • The outlink directory contains the links each ClueWeb document point to. Some documents may not have outlinks. The outlinks are in json format.

Using the ClueWeb22Api

Each method of the API requires certain types of data, which are noted. The API must be initialized with a ClueWeb22 ID and the root directory of the ClueWeb22 dataset. The API relies on the directory structure originally provided with the dataset.

To get started initialize the API:

from ClueWeb22Api import ClueWeb22Api

cw22id = clueweb22-en0000-00-00004
root_path = [BASE_DIRECTORY]/ClueWeb22
clueweb_api = ClueWeb22Api(cw22id, root_path)

How to get the HTML

This method returns the HTML for a ClueWeb22 document ID.
Directories accessed: ClueWeb22/html

html = clueweb_api.get_html_from_warc()

Note: The ClueWebAPI is not required to get the html from the WARC files.

How to get the Primary Content of the page with HTML annotation

This function returns the primary content of the page and HTML annotations such as Title, Table, and List. The annotations are returned as a list with the node ids and the offsets in the text where they begin and end.
Directories accessed: ClueWeb22/html and ClueWeb22/vdom

clueweb_api.get_primary_content_with_annotations()

How to get Node Features

These functions return the vdom node features such as position x, position y, and list item.
Directories accessed: ClueWeb22/html and ClueWeb22/vdom

How to get Node Features by Node ID

This function returns just the node IDs and their vdom features.

clueweb_api.get_node_features()

How to get Node Features with Node IDs and Node text

This function returns the node ID, node text, and node vdom features. This will take more time than getting the node vdom features alone. There is one optional parameter: is_primary which is set by default to True. Setting the is_primary parameter returns only nodes that are considered primary to the page.

clueweb_api.get_node_features_with_text(is_primary=True)

How to get Primary Text

These functions return the clean text of what is considered the primary content of the page in json format.
Directories accessed: ClueWeb22/txt

clueweb_api.get_clean_text()

How to get Inlinks

This function returns the inlinks for a page in json format.
Directories accessed: ClueWeb22/inlink

clueweb_api.get_inlinks()

How to get Outlinks

This function returns the outlinks for a page in json format.
Directories accessed: ClueWeb22/outlink

clueweb_api.get_outlinks()

How to get topics

This function returns the topics assigned to a page as a list.
Directories accessed: ClueWeb22/vdom

clueweb_api.topics()