Embedding Open-domain Common-sense Knowledge from Text

Embedding Open-domain Common-sense Knowledge from Text Travis Goodwin and Sanda Harabagiu Human Language Technology Research Institute University of Texas at Dallas Richardson, TX 75083-0688, USA {travis,sanda}@hlt.utdallas.edu Abstract Our ability to understand language often relies on common-sense knowledge – background information the speaker can assume is known by the reader. Similarly, our comprehension of the language used in complex domains relies on access to domain-specific knowledge. Capturing common-sense and domain-specific knowledge can be achieved by taking advantage of recent advances in open information extraction (IE) techniques and, more importantly, of knowledge embeddings, which are multi-dimensional representations of concepts and relations. Building a knowledge graph for representing common-sense knowledge in which concepts discerned from noun phrases are cast as vertices and lexicalized relations are cast as edges leads to learning the embeddings of common-sense knowledge accounting for semantic compositionality as well as implied knowledge. Common-sense knowledge is acquired from a vast collection of blogs and books as well as from WordNet. Similarly, medical knowledge is learned from two large sets of electronic health records. The evaluation results of these two forms of knowledge are promising: the same knowledge acquisition methodology based on learning knowledge embeddings works well both for common-sense knowledge and for medical knowledge Interestingly, the common-sense knowledge that we have acquired was evaluated as being less neutral than than the medical knowledge, as it often reflected the opinion of the knowledge utterer. In addition, the acquired medical knowledge was evaluated as more plausible than the common-sense knowledge, reflecting the complexity of acquiring common-sense knowledge due to the pragmatics and economicity of language. Keywords: Common-sense knowledge, medical domain knowledge, knowledge embedding 1. Introduction Our ability to understand language often relies on common- sense knowledge – background information the speaker can assume is known by the reader. When language is used to communicate, common-sense knowledge is not articulated in most of the cases. Hence, the availability of resources capturing or approximating common-sense knowledge is crucial. Multiple attempts have been made to capture gen- eral common-sense knowledge, such as the CYC project (Lenat and Guha, 1989) and ConceptNet (Speer and Havasi, 2013). Moreover, WordNet (Miller, 1995) captures lexico- semantic knowledge in English, while DBPedia (Auer et al., 2007) provides knowledge that was used for common- sense reasoning. These resources encode a large number of concepts without representing all their possible attributes. Moreover, another important limitation of these existing resources for common-sense knowledge stems from the lim- ited number of relation types spanning concepts. Some of the existing relations are taxonomic, e.g. IS-A, PART-OF, while others capture causality, e.g. ENABLE, CAUSE-OF, but they hardly represent all relation types we use when accessing common-sense knowledge. To address these two limitations, we took advantage of recent advances in open information extraction (IE) techniques and, more impor- tantly, of the advent of knowledge embeddings, which are multi-dimensional representations of concepts and relations. The use of knowledge embedding enabled us to consider lexicalized relations between concepts, allowing relational similarity to be easily identified, even if not grouped in …

Content truncated to 500 words