TranX

A general-purpose Transition-based abstract syntaX parser that maps natural language queries into machine executable source code (e.g., Python) or logical forms (e.g., lambda calculus). Online Demo.

System Architecture

For technical details please refer to our ACL '18 paper and EMNLP '18 demo paper. To cope with different domain specific logical formalisms (e.g., SQL, Python, lambda-calculus, prolog, etc.), TranX uses abstract syntax trees (ASTs) defined in the Abstract Syntax Description Language (ASDL) as intermediate meaning representation.

Figure 1 gives a brief overview of the system.

1. TranX first employs a transition system to transform a natural language utterance into a sequence of tree-constructing actions, following the input grammar specification of the target formal language. The grammar specification is provided by users in textual format (e.g., asdl/lang/py_asdl.txt for Python grammar).

2. The tree-constructing actions produce an intermediate abstract syntax tree. TranX uses ASTs defined under the ASDL formalism as general-purpose, intermediate meaning representations.

3. The intermediate AST is finally transformed to a domain-specific representation (e.g., Python source code) using customly-defined conversion functions.

File Structure tranX is mainly composed of two components:

1. A general-purpose transition system that defines the generation process of an AST z using a sequence of tree-constructing actions a_0, a_1, ..., a_T.
2. A neural network that computes the probability distribution over action sequences, conditional on the natural language query x, p(a_0, a_1, ..., a_T | x).

These two components are implemented in the following two folders, respectively:

• asdl defines a general-purpose transition system based on the ASDL formalism, and its instantiations in different programming languages and datasets. The transition system defines how an AST is constructed using a sequence of actions. This package can be used as a standalone library independent of tranX. See Section 2.2 of the technical report for details.

• model contains the neural network implementation of the transition system defined in asdl, which computes action probabilities using neural networks.See Section 2.3 of the technical report for details.

Here is a detailed map of the file strcuture:

├── asdl (grammar-based transition system)
├── datasets (dataset specific code like data preprocessing/evaluation/etc.)
├── model (PyTorch implementation of neural nets)
├── server (interactive Web server)
├── components (helper functions and classes like vocabulary)

Supported Language and Datasets

TranX officially supports the following grammatical formalism and datasets. More languages (C#) are coming!

Language	Transition System	Grammar Specification	Example Datasets
Python 2	asdl.PythonTransitionSystem	asdl/lang/py/py_asdl.txt	Django (Oda et al., 2015)
Python 3	asdl.Python3TransitionSystem	asdl/lang/py3/py3_asdl.simplified.txt	CoNaLa (Yin et al., 2018)
Lambda Calculus	asdl.LambdaCalculusTransitionSystem	asdl/lang/lambda_asdl.txt	ATIS, GeoQuery (Zettlemoyer and Collins, 2005)
Prolog	asdl.PrologTransitionSystem	asdl/lang/prolog_asdl.txt	Jobs (Zettlemoyer and Collins, 2005)
SQL	asdl.SqlTransitionSystem	asdl/lang/sql/sql_asdl.txt	WikiSQL (Zhong et al., 2017)

Evaluation Results

Here is a list of performance results on six datasets using pretrained models in data/pretrained_models

Dataset	Results	Metric
GEO	88.6	Accuracy
ATIS	87.7	Accuracy
JOBS	90.0	Accuracy
Django	77.2	Accuracy
CoNaLa	24.5	Corpus BLEU
WikiSQL	79.1	Execution Accuracy

Usage

TL;DR

git clone https://github.com/pcyin/tranX
cd tranX

bash ./pull_data.sh  # get datasets and pre-trained models

conda env create -f config/env/tranx.yml  # create conda Python environment.

./scripts/atis/train.sh 0  # train on ATIS semantic parsing dataset with random seed 0
./scripts/geo/train.sh 0  # train on GEO dataset
./scripts/django/train.sh 0  # train on django code generation dataset
./scripts/conala/train.sh 0  # train on CoNaLa code generation dataset
./scripts/wikisql/train.sh 0  # train on WikiSQL SQL code generation dataset

Web Server/HTTP API

tranX also ships with a web server for demonstraction and interactive debugging perpuse. It also exposes an HTTP API for online semantic parsing/code generation.

To start the web server, simply run:

source activate tranx
PYTHONPATH=../ python app.py --config_file config/server/config_py3.json

This will start a web server at port 8081 with ATIS/GEO/CoNaLa datasets.

HTTP API To programmically query tranX to get semantic parsing results, send your HTTP GET request to

http://<IP Address>:8081/parse/<dataset_name>/<utterance>

# e.g., http://localhost:8081/parse/atis/show me flight from Pittsburgh to Seattle

Conda Environments

TranX supports both Python 2.7 and 3.5. Please note that some datasets only support Python 2.7 (e.g., Django) or Python 3+ (e.g., WikiSQL). The main example conda environment (config/env/tranx.yml) supports Python 3, but we also provide one for Python 2 (config/env/tranx-py2.yml). You can export the enviroments using the following command:

conda env create -f config/env/(tranx.yml,tranx-py2.yml)

FAQs

How to adapt to a new programming language or logical form?

You need to implement the TransitionSystem class with a bunch of custom functions which (1) convert between domain-specific logical forms and intermediate ASTs used by TranX, (2) predictors which check if a hypothesis parse if correct during beam search decoding. You may take a look at the examples in asdl/lang/*.

How to generate those pickled datasets (.bin files)?

Please refer to datasets/<lang>/dataset.py for code snippets that converts a dataset into pickled files.

How to run without CUDA?

Simply remove the --cuda`` flag from the command line arguments. It is included by default in all scripts in the scripts` directory.

Reference

TranX is described/used in the following two papers:

@inproceedings{yin18emnlpdemo,
    title = {{TRANX}: A Transition-based Neural Abstract Syntax Parser for Semantic Parsing and Code Generation},
    author = {Pengcheng Yin and Graham Neubig},
    booktitle = {Conference on Empirical Methods in Natural Language Processing (EMNLP) Demo Track},
    year = {2018}
}

@inproceedings{yin18acl,
    title = {Struct{VAE}: Tree-structured Latent Variable Models for Semi-supervised Semantic Parsing},
    author = {Pengcheng Yin and Chunting Zhou and Junxian He and Graham Neubig},
    booktitle = {The 56th Annual Meeting of the Association for Computational Linguistics (ACL)},
    url = {https://arxiv.org/abs/1806.07832v1},
    year = {2018}
}

Thanks

We are also grateful to the following papers that inspire this work :P

Abstract Syntax Networks for Code Generation and Semantic Parsing.
Maxim Rabinovich, Mitchell Stern, Dan Klein.
in Proceedings of the Annual Meeting of the Association for Computational Linguistics, 2017

The Zephyr Abstract Syntax Description Language.
Daniel C. Wang, Andrew W. Appel, Jeff L. Korn, and Christopher S. Serra.
in Proceedings of the Conference on Domain-Specific Languages, 1997

We also thank Li Dong for all the helpful discussions and sharing the data-preprocessing code for ATIS and GEO used in our Web Demo.