Executable semantic parsers map natural language utterances to meaning representations that can be executed in a particular context such as databases, knowledge graphs, robotic environment, and software applications. The field has become increasingly important as it allows users to seek information and control computer systems naturally and flexibly via interactive exchanges in natural language. We envision that practical semantic parsing systems need to be equipped with three core capabilities:
To this end, the problem of mapping well-formed, individual natural language utterances to formal representations has been studied extensively.
In comparison, semantic parsing in an interactive setup has received less attention until very recently. Furthermore, most of existing semantic parsers assume valid input only hence cannot detect ambiguous/invalid utterances and clarify them effectively. There is also less focus on explainablility and trustworthiness, where the system can explain its interpreted actions to the user for verification and feedback.
This workshop aims to bring together researchers and promote exciting work towards powerful, robust and reliable interactive executable semantic parsing systems. We seek submissions in two tracks:
|Yoav Artzi (Cornell)||Jonathan Berant (Tel Aviv University/AI2)||Richard Socher (Salesforce Research)||Dilek Hakkani-Tür (Amazon Alexa AI)|
|Alex Polozov (Microsoft Research)||Mirella Lapata (The University of Edinburgh)|
|Jonathan Berant (Tel Aviv University/AI2)||Graham Neubig (CMU)||Yunyao Li (IBM Research)||Caiming Xiong (Salesforce Research)|
|Dragomir Radev (Yale University)||Luke Zettlemoyer (University of Washington)|
|Ben Bogin (Tel Aviv University)||Srinivasan Iyer (University of Washington)||Victoria Lin (Salesforce Research)||Alane Suhr (Cornell University)|
|Panupong (Ice) Pasupat (Google)||Pengcheng Yin (CMU)||Tao Yu (Yale University)||Rui Zhang (Yale University -> Penn State University)|
|Victor Zhong (University of Washington)|