Wizard of Shopping: Target-Oriented E-commerce Dialogue Generation with Decision Tree Branching
Xiangci Li * Zhiyu Chen Jason Ingyu Choi, Nikhita Vedula Besnik Fetahu Oleg Rokhlenko Shervin Malmasi · 2025
Read Wizard of Shopping: Target-Oriented E-commerce Dialogue Generation with Decision Tree Branching by Xiangci Li * Zhiyu Chen Jason Ingyu Choi, Nikhita Vedula Besnik Fetahu Oleg Rokhlenko Shervin Malmasi (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-00969.txt.
Description
Seven authors build TRACER, a system that writes realistic shopping dialogues so chat-based shopping assistants can be trained. Such assistants should grasp what a customer wants, ask clarifying questions, and land on the right product. Human-made training dialogues are tiny and rarely tie into real product search. TRACER instead has large language models generate the talks, but it grounds each talk in a plan first.
The plans come from a decision tree that maps the shortest search path to the target product, which keeps each dialogue aimed at a findable goal. The team also releases Wizard of Shopping, the first target-oriented dataset of its kind, with 3,600 conversations across three shopping areas. Human judges rated the talks natural and coherent, and models trained on them performed well on later tasks.
Catalog record
| CREATOR | Xiangci Li * Zhiyu Chen Jason Ingyu Choi, Nikhita Vedula Besnik Fetahu Oleg Rokhlenko Shervin Malmasi |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 82 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-00969