Iterative Utterance Segmentation for Neural Semantic Parsing

Neural semantic parsers usually fail to parse long and complex utterances into correct meaning representations, due to the lack of capturing compositional structures in utterances. To address this issue, we present a novel framework for boosting neural semantic parsers via iterative utterance segmentation. Given an input utterance, our framework iterates between segmenting a span from the utterance and parsing the span through a base neural semantic parser. Then, we compose these intermediate parsing results into the final meaning representation. Considering the usual absence of labeled data for utterance segmentation, we propose to search for latent optimal segmentation points via cooperative training of segmentation and parsing models. Experiments on GEO, COMPLEXWEBQUESTIONS and FORMULAS show that our framework can consistently improve performances of neural semantic parsers in different domains.