Timezone: »

 
Poster
A Retrieve-and-Edit Framework for Predicting Structured Outputs
Tatsunori Hashimoto · Kelvin Guu · Yonatan Oren · Percy Liang

Tue Dec 04 07:45 AM -- 09:45 AM (PST) @ Room 517 AB #154
in Tue Poster Session A »

For the task of generating complex outputs such as source code, editing existing outputs can be easier than generating complex outputs from scratch. With this motivation, we propose an approach that first retrieves a training example based on the input (e.g., natural language description) and then edits it to the desired output (e.g., code). Our contribution is a computationally efficient method for learning a retrieval model that embeds the input in a task-dependent way without relying on a hand-crafted metric or incurring the expense of jointly training the retriever with the editor. Our retrieve-and-edit framework can be applied on top of any base model. We show that on a new autocomplete task for GitHub Python code and the Hearthstone cards benchmark, retrieve-and-edit significantly boosts the performance of a vanilla sequence-to-sequence model on both tasks.

Keywords: Generative Models · Natural Language Processing
[ Paper] 

Author Information

Tatsunori Hashimoto (Stanford)
Kelvin Guu (Google)
Yonatan Oren (Stanford)
Percy Liang (Stanford University)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors