Commonsense Knowledge in Wikidata

Abstract

Wikidata and Wikipedia have been proven useful for reasoning in natural language applications, like question answering or entity linking. Yet, no existing work has studied the potential of Wikidata for commonsense reasoning. This paper investigates whether Wikidata contains commonsense knowledge which is complementary to existing commonsense sources. Starting from a definition of common sense, we devise three guiding principles, and apply them to generate a commonsense subgraph of Wikidata (Wikidata-CS ). Within our approach, we map the relations of Wikidata to ConceptNet, which we also leverage to integrate Wikidata-CS into an existing consolidated commonsense graph. Our experiments reveal that: 1) albeit Wikidata-CS represents a small portion of Wikidata, it is an indicator that Wikidata contains relevant commonsense knowledge, which can be mapped to 15 ConceptNet relations; 2) the overlap between Wikidata-CS and other commonsense sources is low, motivating the value of knowledge integration; 3) Wikidata-CS has been evolving over time at a slightly slower rate compared to the overall Wikidata, indicating a possible lack of focus on commonsense knowledge. Based on these findings, we propose three recommended actions to improve the coverage and quality of Wikidata-CS further.

PaCo: Preconditions Attributed to Commonsense Knowledge

“…glass) rather than entities (e.g. a specific brand of glass) (Ilievski et al, 2020b). There has been various efforts in capturing and conceptualizing such commonsense knowledge (e.g.…”

Dimensions of Commonsense Knowledge

“…Hence, in this paper, we make an assumption that the knowledge types in these sources suffice, or at least have previously sufficed, and can express the desired knowledge. The diversity of knowledge expressed by the relational-other dimension, as pointed out also in [25], might be an indicator for additional, latent dimensions hidden behind the vagueness of this dimension.…”

“…The content of Wikidata is collaboratively created by humans, as well as other existing sources. Given the vast number of statements in Wikidata and its sizable set of over 7 thousand relations, we consider its Wikidata-CS commonsense subset, as extracted in [25].…”

“…In addition, a more semantic resource would allow us to define constraints over the knowledge and detect anomalies and contradictory knowledge, which can be argued to define the boundaries of the knowledge that can be obtained. It is reasonable that such boundaries exist, as commonsense knowledge is characterized by commonness of its concepts and restricted set of relations [25]. Further, by organizing by dimensions also allows us to describe strengths (and/or weakenesses) of resources.…”

“…Efforts to consolidate commonsense knowledge across sources [26,16,45] have managed to bring these sources closer, which has shown impact on commonsense QA tasks [38]. In [25], we provide heuristics for defining the boundaries of commonsense knowledge, in order to extract such subset from one of the largest available graphs today, Wikidata [62]. Yet, these efforts have limited success, and many consolidation questions are left open.…”