Abstract Syntax Networks for Code Generation and Semantic Parsing

Maxim Rabinovich, Mitchell Stern and Daniel Klein

EECS Department
University of California, Berkeley
Technical Report No. UCB/EECS-2019-172
December 17, 2019

http://www2.eecs.berkeley.edu/Pubs/TechRpts/2019/EECS-2019-172.pdf

Tasks like code generation and semantic parsing require mapping unstructured (or partially structured) inputs to well-formed, executable outputs. We introduce abstract syntax networks, a modeling framework for these problems. The outputs are represented as abstract syntax trees (ASTs) and constructed by a decoder with a dynamically-determined modular structure paralleling the structure of the output tree. On the benchmark HEARTHSTONE dataset for code generation, our model obtains 79.2 BLEU and 22.7% exact match accuracy, compared to previous state-of-the-art values of 67.1 and 6.1%. Furthermore, we perform competitively on the ATIS, JOBS, and GEO semantic parsing datasets with no task-specific engineering.

Advisor: Michael Jordan


BibTeX citation:

@mastersthesis{Rabinovich:EECS-2019-172,
    Author = {Rabinovich, Maxim and Stern, Mitchell and Klein, Daniel},
    Title = {Abstract Syntax Networks for Code Generation and Semantic Parsing},
    School = {EECS Department, University of California, Berkeley},
    Year = {2019},
    Month = {Dec},
    URL = {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2019/EECS-2019-172.html},
    Number = {UCB/EECS-2019-172},
    Abstract = {Tasks like code generation and semantic parsing require mapping unstructured (or partially structured) inputs to well-formed, executable outputs. We introduce abstract syntax networks, a modeling framework for these problems. The outputs are represented as abstract syntax trees (ASTs) and constructed by a decoder with a dynamically-determined modular structure paralleling the structure of the output tree. On the benchmark HEARTHSTONE dataset for code generation, our model obtains 79.2 BLEU and 22.7% exact match accuracy, compared to previous state-of-the-art values of 67.1 and 6.1%. Furthermore, we perform competitively on the ATIS, JOBS, and GEO semantic parsing datasets with no task-specific engineering.}
}

EndNote citation:

%0 Thesis
%A Rabinovich, Maxim
%A Stern, Mitchell
%A Klein, Daniel
%T Abstract Syntax Networks for Code Generation and Semantic Parsing
%I EECS Department, University of California, Berkeley
%D 2019
%8 December 17
%@ UCB/EECS-2019-172
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2019/EECS-2019-172.html
%F Rabinovich:EECS-2019-172