Flowris: Managing Data Analysis Workflows for Conversational Agent

Jiajia Sun, Juan Wang, Yueguo Chen, Xiongpai Qin

Abstract: Conversational agent has become a new way of conducting data analysis tasks, enabling people with different levels of analytic experience to interact with the system by providing natural language (NL) commands, choices, and parameters etc. However, flexibility becomes a challenge considering that users may want to integrate model services on the Web in such systems. To address it, we present Flowris, a prototype system that collects and manages provenance data for conversational agent. We will show how Flowris collects and manages provenance data, and how it supports the comparison and reproducibility of experiments.