A cognitive interactionist sentence parser with simple recurrent networks

Sentence parsing has a long history in the research fields of machine learning and natural language processing. The state-of-the-art technologies used to tackle this task include those based on statistical language learning. In the meantime, human sentence parsing has attracted massive research efforts for decades in the field of cognitive psychology. A range of behaviouristic experiments verify that the interactionist approach is a sensible and effective way to simulate the human parsing mechanism. This paper proposes a novel and effective sentence parser, the Cognitive Interactionist Parser (CIParser), which incorporates the cognitive interactionist approach with semantic information and simple recurrent networks to extend and enrich the technologies for sentence parsing. Considering the parsing efficiency, CIParser processes the semantic information of nouns and verbs in current stage. The performance of the Cognitive Interactionist Parser is evaluated using elaborately designed experiments using the noted SUSANNE Corpus. The experimental results demonstrate that the Cognitive Interactionist Parser surpasses two state-of-the-art statistical parsers in two classical measures, Precision and Recall, of Information Retrieval (IR).