Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

User Willingness-aware Sales Talk Dataset

Asahi Hentona, Jun Baba, Shiki Sato, Reina Akama

TL;DR

This work tackles the lack of willingness-grounded data in sales dialogue research by constructing the first user willingness–aware dataset captured under ecologically valid conditions. It identifies three core willingness types (engagement, information provision, and goal acceptance) and uses a Wizard-of-Oz setup to collect near-realistic, utterance-level willingness labels, paired with a purchase-intent metric. The authors analyze the data to reveal how willingness evolves within dialogues and propose strategies for early, middle, and final dialogue stages, then demonstrate that willingness-aware fine-tuning and progression-based dialogue strategies can improve purchase uplift in GPT-3.5-based systems, with GPT-4o providing a strong benchmark. The dataset and findings offer practical guidance for building sale-oriented dialogue systems that respond to real-time willingness signals, enabling more effective, user-centered e-commerce interactions.

Abstract

User willingness is a crucial element in the sales talk process that affects the achievement of the salesperson's or sales system's objectives. Despite the importance of user willingness, to the best of our knowledge, no previous study has addressed the development of automated sales talk dialogue systems that explicitly consider user willingness. A major barrier is the lack of sales talk datasets with reliable user willingness data. Thus, in this study, we developed a user willingness-aware sales talk collection by leveraging the ecological validity concept, which is discussed in the field of human-computer interaction. Our approach focused on three types of user willingness essential in real sales interactions. We created a dialogue environment that closely resembles real-world scenarios to elicit natural user willingness, with participants evaluating their willingness at the utterance level from multiple perspectives. We analyzed the collected data to gain insights into practical user willingness-aware sales talk strategies. In addition, as a practical application of the constructed dataset, we developed and evaluated a sales dialogue system aimed at enhancing the user's intent to purchase.

User Willingness-aware Sales Talk Dataset

TL;DR

This work tackles the lack of willingness-grounded data in sales dialogue research by constructing the first user willingness–aware dataset captured under ecologically valid conditions. It identifies three core willingness types (engagement, information provision, and goal acceptance) and uses a Wizard-of-Oz setup to collect near-realistic, utterance-level willingness labels, paired with a purchase-intent metric. The authors analyze the data to reveal how willingness evolves within dialogues and propose strategies for early, middle, and final dialogue stages, then demonstrate that willingness-aware fine-tuning and progression-based dialogue strategies can improve purchase uplift in GPT-3.5-based systems, with GPT-4o providing a strong benchmark. The dataset and findings offer practical guidance for building sale-oriented dialogue systems that respond to real-time willingness signals, enabling more effective, user-centered e-commerce interactions.

Abstract

User willingness is a crucial element in the sales talk process that affects the achievement of the salesperson's or sales system's objectives. Despite the importance of user willingness, to the best of our knowledge, no previous study has addressed the development of automated sales talk dialogue systems that explicitly consider user willingness. A major barrier is the lack of sales talk datasets with reliable user willingness data. Thus, in this study, we developed a user willingness-aware sales talk collection by leveraging the ecological validity concept, which is discussed in the field of human-computer interaction. Our approach focused on three types of user willingness essential in real sales interactions. We created a dialogue environment that closely resembles real-world scenarios to elicit natural user willingness, with participants evaluating their willingness at the utterance level from multiple perspectives. We analyzed the collected data to gain insights into practical user willingness-aware sales talk strategies. In addition, as a practical application of the constructed dataset, we developed and evaluated a sales dialogue system aimed at enhancing the user's intent to purchase.
Paper Structure (51 sections, 9 figures, 5 tables)

This paper contains 51 sections, 9 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Sales Dialogue Dataset
  4. Ecological Validity
  5. Dataset Design
  6. Envisioned Situation
  7. Participants.
  8. Beginning and end.
  9. Target users.
  10. User willingness.
  11. Construction Policies
  12. Participants.
  13. Beginning and end.
  14. Target users.
  15. User willingness.
  16. ...and 36 more sections

Figures (9)

  • Figure 1: Overview of collecting sales talk dialogue data.
  • Figure 2: Distributions of proportions of user willingness labels (positive, neutral, and negative) across individual dialogues.
  • Figure 6: Correlation coefficients between the proportion of user willingness labels and the improvement in the users' purchase intention.
  • Figure 7: Average user willingness scores at dialogue progression level.
  • Figure 10: Success dialogues
  • ...and 4 more figures