Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Instant Answering in E-Commerce Buyer-Seller Messaging using Message-to-Question Reformulation

Besnik Fetahu, Tejas Mehta, Qun Song, Nikhita Vedula, Oleg Rokhlenko, Shervin Malmasi

TL;DR

The paper tackles the problem of slow, manual responses to buyer inquiries in e-commerce messaging by introducing M2Q, a sequence-to-sequence pipeline that reformulates verbose buyer messages into concise questions suitable for a federated QA system. By combining generative and extractive reformulation strategies, M2Q preserves user intent while enabling instant, automated answers and preserving privacy. A dedicated dataset of approximately $\sim 6k$ message–reformulation pairs supports offline evaluation, which, along with live online deployment, demonstrates substantial improvements in reformulation quality, QA understandability, and response rates, translating into higher purchase rates and reduced seller workload. The approach shows that distilling long messages into targeted questions can leverage existing QA resources to deliver fast, reliable buyer support at scale, with the Hybrid variant offering robust performance across varying message lengths.

Abstract

E-commerce customers frequently seek detailed product information for purchase decisions, commonly contacting sellers directly with extended queries. This manual response requirement imposes additional costs and disrupts buyer's shopping experience with response time fluctuations ranging from hours to days. We seek to automate buyer inquiries to sellers in a leading e-commerce store using a domain-specific federated Question Answering (QA) system. The main challenge is adapting current QA systems, designed for single questions, to address detailed customer queries. We address this with a low-latency, sequence-to-sequence approach, MESSAGE-TO-QUESTION ( M2Q ). It reformulates buyer messages into succinct questions by identifying and extracting the most salient information from a message. Evaluation against baselines shows that M2Q yields relative increases of 757% in question understanding, and 1,746% in answering rate from the federated QA system. Live deployment shows that automatic answering saves sellers from manually responding to millions of messages per year, and also accelerates customer purchase decisions by eliminating the need for buyers to wait for a reply

Instant Answering in E-Commerce Buyer-Seller Messaging using Message-to-Question Reformulation

TL;DR

The paper tackles the problem of slow, manual responses to buyer inquiries in e-commerce messaging by introducing M2Q, a sequence-to-sequence pipeline that reformulates verbose buyer messages into concise questions suitable for a federated QA system. By combining generative and extractive reformulation strategies, M2Q preserves user intent while enabling instant, automated answers and preserving privacy. A dedicated dataset of approximately message–reformulation pairs supports offline evaluation, which, along with live online deployment, demonstrates substantial improvements in reformulation quality, QA understandability, and response rates, translating into higher purchase rates and reduced seller workload. The approach shows that distilling long messages into targeted questions can leverage existing QA resources to deliver fast, reliable buyer support at scale, with the Hybrid variant offering robust performance across varying message lengths.

Abstract

E-commerce customers frequently seek detailed product information for purchase decisions, commonly contacting sellers directly with extended queries. This manual response requirement imposes additional costs and disrupts buyer's shopping experience with response time fluctuations ranging from hours to days. We seek to automate buyer inquiries to sellers in a leading e-commerce store using a domain-specific federated Question Answering (QA) system. The main challenge is adapting current QA systems, designed for single questions, to address detailed customer queries. We address this with a low-latency, sequence-to-sequence approach, MESSAGE-TO-QUESTION ( M2Q ). It reformulates buyer messages into succinct questions by identifying and extracting the most salient information from a message. Evaluation against baselines shows that M2Q yields relative increases of 757% in question understanding, and 1,746% in answering rate from the federated QA system. Live deployment shows that automatic answering saves sellers from manually responding to millions of messages per year, and also accelerates customer purchase decisions by eliminating the need for buyers to wait for a reply
Paper Structure (21 sections, 1 figure, 3 tables)

This paper contains 21 sections, 1 figure, 3 tables.

Table of Contents

  1. Introduction
  2. Background
  3. M2Q: Message-to-Question
  4. Generative Reformulation (A):
  5. Extractive Reformulation (B):
  6. Question Answering System (C):
  7. Dataset Construction
  8. Offline Experiments and Results
  9. Baselines and Our Approach
  10. Extractive Reformulation:
  11. Flan-T5-XXL:
  12. Vicuna-13B:
  13. Approach:
  14. Evaluation Strategies
  15. Results
  16. ...and 6 more sections

Figures (1)

  • Figure 1: M2Q approach overview. For each buyer message that may contain multiple questions/intents, M2Q reformulates them (using (B) and (C)) into a succinct question with the most salient intent (or conjunction of intents). These are sent to a federated QA system (C) for instant answering. If customers are unsatisfied with the response or receive none, the original message is forwarded to the seller for manual reply.