Definition and Extraction of Causal Relations for QA on Fault Diagnosis of Devices

摘要

Causal relations in ontology should be defined based on the inference types necessary to solve the tasks specific to application, as well as domain. In this paper, we present a model to define and to extract causal relations for application ontology, which is targeted, as a case study, to serve a Question-Answering (QA) system on fault-diagnosis of electronic devices. In the first phase, causal categories are defined by identifying the generic inference patterns of QA on fault-diagnosis. In the second, the semantic relations between concepts in the corpus denoting the causal categories are defined as causal relations. In the third, instances of causal relations are extracted using the lexical patterns from the definitional statements of terms in domain, and extended with information from thesaurus. On the evaluation by domain experts, our model shows precision of 92.3% in classifying relations at the definition phase and precision of 80.7% in identifying causal relations at the extraction phase.