A Generic Dialog Agent for Information Retrieval Based on Automated Planning Within a Reinforcement Learning Platform

Document Type



With easy availability of large data sets online, like product catalogs and open data, a common business problem is to allow users to search them for information using natural interfaces. Dialog systems provide such an interface where a user can type or speak to the system and ask for information, and the system navigates the ambiguity of request, the complexity of content (size, hierarchy, schema) and usage considerations (response time, dialog length) to create a series of conversation leading to the system providing user the appropriate information. However, current learning-based methods to build a dialog agent require large training data, are data specific, and hard to scale while a user’s interaction spans querying of multiple data sources. In this paper, we present a novel and generic approach for dialog for information retrieval based on automated planning within a reinforcement learning (RL)-based platform, ParlAI. The approach allows us to seamlessly scale to new data sources and to explore various planning and RL integration strategies. For instance, the planner performs search for response strategies that are controlled, goal-oriented, and across multiple turns without prior training data, while the RL is used to automate selection of data sources during an interaction. One can also just use the RL for end-to-end training or select a data source and use only the planner. We demonstrate the viability of our approach using the large data sets of UNSPSC and ICD-10, and a simple phone directory.

APA Citation

Pallagani, V. & Srivastava, B. (2021, August). A generic dialog agent for information retrieval based on automated planning within a reinforcement learning platform. Bridging the Gap Between AI Planning and Reinforcement Learning (PRL).