Integrating Cross-Modal Context for PP Attachment Disambiguation

Correct prepositional phrase (PP) attachment is a notoriously hard problem in natural language parsing. Despite the inherent structural ambiguity it introduces, human communication still succeeds in most cases with remarkable robustness. The reason for this is that in processing natural language humans also integrate information from sources other than the linguistic material alone. Humans access additional knowledge to enrich the semantic specification that guides syntactic and semantic disambiguation. One important source of additional knowledge for humans is cross-modal context as construed from sensory perception. Motivated by effects during human sentence processing we propose to integrate contextual knowledge into syntactic parsing to support PP-attachment disambiguation. We hypothesise that integrating cross-modal context into syntactic constraint dependency parsing will significantly and substantially improve the accuracy of PP attachment disambiguation.