Paris: a Parallel Inference System 1

This paper presents an inferential system based on abductive interpretation of text. Inference to the best explanation is performed by the recognition of the most economic semantic paths produced by the propagation of markers on a very large linguistic knowledge base. The propagation of markers is controlled by their intrinsic propagation rules, devised from plausible semantic relation chains. An interpretation is inferred whenever two markers collide. The trajectories of the two colliding markers represent the abductive tree, where each marker propagation stands for an implication from the knowledge base. Markers propagate semantic constrains throughout the knowledge base, evaluating the plausibility of each inference. Using a very large knowledge base, our inferential system aims at producing interpretations accountable for commonsense reasoning. The novelty is that the inference rules model a large variety of implications, as suggested by the knowledge base relations. Textual implicatures are recognized as pragmatic inferences. 1 Basic Idea In this paper, we describe a system that extracts pragmatic inferences from text. The process of interpreting sentences in a text can be viewed as the process of providing the best explanatory relations between the concepts of that text. This system, called PARIS (PARallel Inference System), represents linguistic knowledge as a large semantic network that can be abstracted as nodes connected by relations. The abductive interpretation of text is performed in PARIS by searching in the knowledge base for the most economic sequences of connecting relations between the textual concepts. The recognition of connections between concepts is based on three types of restrictions: (1) the inference rules which are implemented as chains of selected relations that semantically link concepts in the knowledge base; (2) textual linguistic constraints, deening explicit relations between concepts, e.g. semantic case relations and (3) the actual knowledge base representation.