CauseNet aims at creating a causal knowledge base that comprises all human causal knowledge and to separate it from mere causal beliefs, with the goal of enabling large-scale research into causal inference.
CauseNet: Towards a Causality Graph Extracted from the Web
Causal knowledge is seen as one of the key ingredients to advance artificial intelligence. Yet, few knowledge bases comprise causal knowledge to date, possibly due to significant efforts required for validation. Notwithstanding this challenge, we compile CauseNet, a large-scale knowledge base of claimed causal relations between causal concepts. By extraction from different semi- and unstructured web sources, we collect more than 11 million causal relations with an estimated extraction precision of 83\% and construct the first large-scale and open-domain causality graph. We analyze the graph to gain insights about causal beliefs expressed on the web and we demonstrate its benefits in basic causal question answering. Future work may use the graph for causal reasoning, computational argumentation, multi-hop question answering, and more.
Download
We provide three versions of our causality graph CauseNet:
CauseNet-Full: The complete dataset
CauseNet-Precision: A subset of CauseNet-Full with higher precision
CauseNet-Sample: A small sample dataset for first explorations and experiments without provenance data
Statistics
Relations Concepts File Size CauseNet-Full 11,609,890 12,186,195 1.8GB CauseNet-Precision 199,806 80,223 135MB CauseNet-Sample 264 524 54KB
Data Model
The core of CauseNet consists of causal concepts which are connected by causal relations. Each causal relation has comprehensive provenance data on where and how it was extracted.
Examples of Causal Relations
Causal relations are represented as shown in the following example. Provenance data is omitted.
{ "causal_relation" : { "cause" : { "concept" : "disease" }, "effect" : { "concept" : "death" } } }
For CauseNet-Full and CauseNet-Precision, we include comprehensive provenance data. In the following, we give one example per source.
For relations extracted from natural language sentences we provide:
surface : the surface form of the sentence, i.e., the original string
: the surface form of the sentence, i.e., the original string path_pattern : the linguistic path pattern used for extraction
ClueWeb12 Sentences
clueweb12_page_id : page id as provided in the ClueWeb12 corpus
: page id as provided in the ClueWeb12 corpus clueweb12_page_reference : page reference as provided in the ClueWeb12 corpus
: page reference as provided in the ClueWeb12 corpus clueweb12_page_timestamp : page access data as stated in the ClueWeb12 corpus
{ "causal_relation" :{ "cause" :{ "concept" : "smoking" }, "effect" :{ "concept" : "disability" } }, "sources" :[ { "type" : "clueweb12_sentence" , "payload" :{ "clueweb12_page_id" : "urn:uuid:4cbae00e-8c7f-44b1-9f02-d797f53d448a" , "clueweb12_page_reference" : "http://atlas.nrcan.gc.ca/site/english/maps/health/healthbehaviors/smoking" , "clueweb12_page_timestamp" : "2012-02-23T21:10:45Z" , "sentence" : "In Canada, smoking is the most important cause of preventable illness, disability and premature death." , "path_pattern" : "[[cause]]/N \t -nsubj \t cause/NN \t +nmod:of \t [[effect]]/N" } } ] }
Wikipedia Sentences
wikipedia_page_id : the Wikipedia page id
: the Wikipedia page id wikipedia_page_title : the Wikipedia page title
: the Wikipedia page title wikipedia_revision_id : the Wikipedia revision id of the last edit
: the Wikipedia revision id of the last edit wikipedia_revision_timestamp : the timestamp of the Wikipedia revision id of the last edit
: the timestamp of the Wikipedia revision id of the last edit sentence_section_heading : the section heading where the sentence comes from
: the section heading where the sentence comes from sentence_section_level : the level where the section heading comes from
{ "causal_relation" :{ "cause" :{ "concept" : "human_activity" }, "effect" :{ "concept" : "climate_change" } }, "sources" :[ { "type" : "wikipedia_sentence" , "payload" :{ "wikipedia_page_id" : "13109" , "wikipedia_page_title" : "Global warming controversy" , "wikipedia_revision_id" : "860220175" , "wikipedia_revision_timestamp" : "2018-09-19T04:52:18Z" , "sentence_section_heading" : "Global warming controversy" , "sentence_section_level" : "1" , "sentence" : "The controversy is, by now, political rather than scientific: there is a scientific consensus that climate change is happening and is caused by human activity." , "path_pattern" : "[[cause]]/N \t -nmod:agent \t caused/VBN \t +nsubjpass \t [[effect]]/N" } } ] }
Wikipedia Lists
list_toc_parent_title : The heading of the parent section the list appears in
: The heading of the parent section the list appears in list_toc_section_heading : The heading of the section the list appears in
: The heading of the section the list appears in list_toc_section_level : The nesting level of the section within the table of content (toc)
{ "causal_relation" :{ "cause" :{ "concept" : "separation_from_parents" }, "effect" :{ "concept" : "stress_in_early_childhood" } }, "sources" :[ { "type" : "wikipedia_list" , "payload" :{ "wikipedia_page_id" : "33096801" , "wikipedia_page_title" : "Stress in early childhood" , "wikipedia_revision_id" : "859225864" , "wikipedia_revision_timestamp" : "2018-09-12T16:22:05Z" , "list_toc_parent_title" : "Stress in early childhood" , "list_toc_section_heading" : "Causes" , "list_toc_section_level" : "2" } } ] }
Wikipedia Infoboxes
infobox_template : The Wikipedia template of the infobox
: The Wikipedia template of the infobox infobox_title : The title of the Wikipedia infobox
: The title of the Wikipedia infobox infobox_argument : The argument of the infobox (the key of the key-value pair)
{ "causal_relation" :{ "cause" :{ "concept" : "alcohol" }, "effect" :{ "concept" : "cirrhosis" } }, "sources" :[ { "type" : "wikipedia_infobox" , "payload" :{ "wikipedia_page_id" : "21365918" , "wikipedia_page_title" : "Cirrhosis" , "wikipedia_revision_id" : "861860835" , "wikipedia_revision_timestamp" : "2018-09-30T15:40:21Z" , "infobox_template" : "Infobox medical condition (new)" , "infobox_title" : "Cirrhosis" , "infobox_argument" : "causes" } } ] }
Loading CauseNet into Neo4j
We provide sample code to load CauseNet into the graph database Neo4j.
The following figure shows an excerpt of CauseNet within Neo4j (showing a coronavirus causing the disease SARS):
Concept Spotting Datasets
For the construction of CauseNet, we employ a causal concept spotter as a causal concept can be composed of multiple words (e.g., “global warming”, “human activity”, or “lack of exercise”). We determine the exact start and end of a causal concept in a sentence with a sequence tagger. Our training and evaluation data is available as part of our concept spotting datasets: one for Wikipedia infoboxes, Wikipedia lists, and ClueWeb sentences. We split each dataset into 80% training, 10% development and 10% test set
Paper
CauseNet forms the basis for our CIKM 2020 paper CauseNet: Towards a Causality Graph Extracted from the Web. Please make sure to refer to it as follows:
@inproceedings { heindorf2020causenet, author = { Stefan Heindorf and Yan Scholten and Henning Wachsmuth and Axel-Cyrille Ngonga Ngomo and Martin Potthast } , title = { CauseNet: Towards a Causality Graph Extracted from the Web } , booktitle = {{ CIKM }} , publisher = {{ ACM }} , year = { 2020 } }
For questions and feedback please contact:
Stefan Heindorf, Paderborn University
Yan Scholten, Technical University of Munich
Henning Wachsmuth, Paderborn University
Axel-Cyrille Ngonga Ngomo, Paderborn University
Martin Potthast, Leipzig University
Licenses
The code is licensed under a MIT license. The data is licensed under a Creative Commons Attribution 4.0 International license.