Journal article 99 views
EMIL: Extracting Meaning from Inconsistent Language / Adam, Wyner
International Journal of Approximate Reasoning, Volume: 112, Pages: 55 - 84
Swansea University Author: Adam, Wyner
Accepted Manuscript under embargo until: 13th May 2020
Developments in formal and computational theories of argumentation reason with inconsistency. Developments in Computational Linguistics extract arguments from large textual corpora. Both developments head in the direction of automated processing and reasoning with inconsistent, linguistic knowledge...
|Published in:||International Journal of Approximate Reasoning|
Check full text
No Tags, Be the first to tag this record!
Developments in formal and computational theories of argumentation reason with inconsistency. Developments in Computational Linguistics extract arguments from large textual corpora. Both developments head in the direction of automated processing and reasoning with inconsistent, linguistic knowledge so as to explain and justify arguments in a humanly accessible form. Yet, there is a gap between the coarse-grained, semi-structured knowledge-bases of computational theories of argumentation and fine-grained, highly-structured inferences from knowledge-bases derived from natural language. We identify several subproblems which must be addressed in order to bridge the gap. We provide a direct semantics for argumentation. It has attractive properties in terms of expressivity and complexity, enables reasoning by cases, and can be more highly structured. For language processing, we work with an existing controlled natural language (CNL), which interfaces with our computational theory of argumentation; the tool processes natural language input, translates them into a form for automated inference engines, outputs argument extensions, then generates natural language statements. The key novel adaptation incorporates the defeasible expression ‘it is usual that’. This is an important, albeit incremental, step to incorporate linguistic expressions of defeasibility. Overall, the novel contribution of the paper is an integrated, end-to-end argumentation system which bridges between automated defeasible reasoning and a natural language interface. Specific novel contributions are the theory of ‘direct semantics’, motivations for our theory, results with respect to the direct semantics, an implementation, experimental results, the tie between the formalisation and the CNL, the introduction into a CNL of a natural language expression of defeasibility, and an ‘engineering’ approach to fine-grained argument analysis.