Abstract
Commonsense knowledge is essential for many AI applications, including thosein natural language processing, visual processing, and planning. Consequently,many sources that include commonsense knowledge have been designed andconstructed over the past decades. Recently, the focus has been on largetext-based sources, which facilitate easier integration with neural (language)models and application on textual tasks, typically at the expense of thesemantics of the sources. Such practice prevents the harmonization of thesesources, understanding their coverage and gaps, and may hinder the semanticalignment of their knowledge with downstream tasks. Efforts to consolidatecommonsense knowledge have yielded partial success, but provide no clear pathtowards a comprehensive consolidation of existing commonsense knowledge. The ambition of this paper is to organize these sources around a common setof dimensions of commonsense knowledge. For this purpose, we survey a widerange of popular commonsense sources with a special focus on their relations.We consolidate these relations into 13 knowledge dimensions, each abstractingover more specific relations found in sources. This consolidation allows us tounify the separate sources and to compute indications of their coverage,overlap, and gaps with respect to the knowledge dimensions. Moreover, weanalyze the impact of each dimension on downstream reasoning tasks that requirecommonsense knowledge, observing that the temporal and desire/goal dimensionsare very beneficial for reasoning on current downstream tasks, whiledistinctness and lexical knowledge have little impact. These results revealfocus towards some dimensions in current evaluation, and potential neglect ofothers.