Trinity: An Extensible Synthesis Framework for Data Science

Abstract

In this demo paper, we introduce Trinity, a general-purpose framework that can be used to quickly build domain-specific program synthesizers for automating many tedious tasks that arise in data science. We illustrate how Trinity can be used by three different users: First, we show how endusers can use Trinity’s built-in synthesizers to automate data wrangling tasks. Second, we show how advanced users can easily extend existing synthesizers to support additional functionalities. Third, we show how synthesis experts can change the underlying search engine in Trinity. Overall, this paper is intended to demonstrate how users can quickly use, modify, and extend the Trinity framework with the goal of automating many tasks that are considered to be the “janitor” work of data science. PVLDB Reference Format: Ruben Martins, Jia Chen, Yanju Chen, Yu Feng, and Isil Dillig. Trinity: An Extensible Synthesis Framework for Data Science. PVLDB, 12(12): 1914-1917, 2019. DOI: https://doi.org/10.14778/3352063.3352098