Conversational Document Prediction to Assist Customer Care Agents

TL;DR

This paper investigates predicting relevant documents in customer care conversations, introduces a new dataset, and finds that a hybrid IR and deep learning approach offers optimal performance and practicality.

Contribution

It introduces a new public dataset for document prediction in customer care and evaluates hybrid IR and deep learning models for improved accuracy.

Findings

Hybrid IR+DL models outperform individual models.

The new dataset supports research in conversational document prediction.

Hybrid models balance accuracy and inference efficiency.

Abstract

A frequent pattern in customer care conversations is the agents responding with appropriate webpage URLs that address users' needs. We study the task of predicting the documents that customer care agents can use to facilitate users' needs. We also introduce a new public dataset which supports the aforementioned problem. Using this dataset and two others, we investigate state-of-the art deep learning (DL) and information retrieval (IR) models for the task. Additionally, we analyze the practicality of such systems in terms of inference time complexity. Our show that an hybrid IR+DL approach provides the best of both worlds.