Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Forecasting Live Chat Intent from Browsing History

Se-eun Yoon, Ahmad Bin Rabiah, Zaid Alibadi, Surya Kallumadi, Julian McAuley

TL;DR

This paper represents each browsing history as a text sequence of page attributes and use the ground-truth class labels to fine-tune pretrained Transformers to generate fine-grained intents.

Abstract

Customers reach out to online live chat agents with various intents, such as asking about product details or requesting a return. In this paper, we propose the problem of predicting user intent from browsing history and address it through a two-stage approach. The first stage classifies a user's browsing history into high-level intent categories. Here, we represent each browsing history as a text sequence of page attributes and use the ground-truth class labels to fine-tune pretrained Transformers. The second stage provides a large language model (LLM) with the browsing history and predicted intent class to generate fine-grained intents. For automatic evaluation, we use a separate LLM to judge the similarity between generated and ground-truth intents, which closely aligns with human judgments. Our two-stage approach yields significant performance gains compared to generating intents without the classification stage.

Forecasting Live Chat Intent from Browsing History

TL;DR

This paper represents each browsing history as a text sequence of page attributes and use the ground-truth class labels to fine-tune pretrained Transformers to generate fine-grained intents.

Abstract

Customers reach out to online live chat agents with various intents, such as asking about product details or requesting a return. In this paper, we propose the problem of predicting user intent from browsing history and address it through a two-stage approach. The first stage classifies a user's browsing history into high-level intent categories. Here, we represent each browsing history as a text sequence of page attributes and use the ground-truth class labels to fine-tune pretrained Transformers. The second stage provides a large language model (LLM) with the browsing history and predicted intent class to generate fine-grained intents. For automatic evaluation, we use a separate LLM to judge the similarity between generated and ground-truth intents, which closely aligns with human judgments. Our two-stage approach yields significant performance gains compared to generating intents without the classification stage.
Paper Structure (11 sections, 3 figures, 3 tables)

This paper contains 11 sections, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Dataset
  3. Methods
  4. Problem Formulation
  5. Two-Stage Intent Prediction
  6. Evaluation
  7. Results
  8. Classification Results
  9. Generation Results
  10. Related Work
  11. Further Discussion

Figures (3)

  • Figure 1: Example of a user browsing pages in an online store and then calling for a live chat agent. Out goal is to predict the user's intent, which is the reason for requesting assistance.
  • Figure 2: Method overview. The components with dashed borders represent our two-stage approach. Intent classification is trained and evaluated using ground-truth class labels; intent generation is evaluated based on raw intents.
  • Figure 3: Longformer+ model architecture for intent classification. Removing the optional embeddings (dotted lines) becomes the original Longformer.