stanford-oval/genie-toolkit

Genie

This repository hosts Genie, a toolkit which allows you to quickly create new semantic parsers that translate from natural language to a formal language of your choice.

Genie was described in the paper:

Genie: A Generator of Natural Language Semantic Parsers for Virtual Assistant Commands
Giovanni Campagna (*), Silei Xu (*), Mehrad Moradshahi, Richard Socher, and Monica S. Lam
In Proceedings of the 40th ACM SIGPLAN Conference on Programming Language Design and Implementation (PLDI 2019), Phoenix, AZ, June 2019.

If you use Genie in any academic work, please cite the above paper.

Installation

Genie depends on additional libraries, including the ThingTalk library and the GenieNLP machine learning library. See doc/install.md for details and installation instructions.

License

This package is covered by the Apache 2.0 license. See LICENSE for details. Note that this package depends on several nodejs modules by third-parties, each with their own license. In particular, some modules might have licensing requirements that are more restrictive than Genie's. It is your responsability to comply with Genie's copyright license, as well as all licenses of included dependencies.

Reproducing The Results In Our Papers

To reproduce the machine learning results in Stanford papers that use Genie (including the PLDI 2019 paper and the ACL 2020 paper), please use the associated artifacts, available for download from our website. The artifact includes all the necessary datasets (including ablation and case studies), pretrained models and evaluation scripts. Please follow the instructions in the README file to reproduce individual experiments.

Using Genie

Genie Concepts

Genie is a synthesis-based tool to build dialogue agents. Genie is based on the Genie template language, which succintly defines a space of synthesized sentences. Genie can use the template language to generate a dataset, then sample a subset of sentences to paraphrase using crowdsourcing. Commonly, the template language is paired with a skill definition, entered in a repository like Thingpedia, which defines the APIs available to the dialogue agent.

A Turnkey Solution For Genie+Almond

A all-in-one solution to use Genie to extend ThingTalk with new skills and new templates is provided by almond-cloud. Please refer to its documentation for installation instructions.

After installation, administrators can create new natural language models, trigger automated training and deploy the trained models to any Almond system.

Manual Genie Usage

If one wants to avoid the complexity of setting up a database and web server, it is possible to invoke Genie manually from the command-line, and have it manipulate datasets stored as TSV/CSV files.

A number of tutorials are included in the doc/ folder, describing common Genie usage.

NOTE: Genie assumes all files are UTF-8, and ignores the current POSIX locale (LC_CTYPE and LANG enviornment varialbes). Legacy encodings such as ISO-8859-1 or Big5 are not supported and could cause problems.

Modifying ThingTalk

If you want to also extend ThingTalk (with new syntax or new features) you will need to fork and modify the library, which lives at https://github.com/stanford-oval/thingtalk. After modifying the library, you can use npm link to point the almond-cloud installation to your library. You must make sure that only one copy of the ThingTalk library is loaded (use npm ls thingtalk to check).