Heuristics for Recovery from Residual Ambiguity and Incongruity in the Semantic Interpretations of Texts

In this paper we present the heuristic algorithms at the core of a text meaning analyzer based on the ontological-semantic approach (e.g., Nirenburg and Raskin 2004). The full analyzer takes as input natural language texts and through many levels of analysis produces text meaning representations (TMRs) formulated in a specially developed ontology-based metalanguage that covers semantic as well as discourse and pragmatic information. We concentrate here on the process of deriving basic semantic " who-did-what-to-whom " dependency structure from a text. In presenting the material, we concentrate on the motivation for the introduction of specific algorithms (methods) and on their heuristic content.