Generating Rich Product Descriptions for Conversational E-commerce Systems

TL;DR

This paper introduces eBERT, a sequence-to-sequence model that generates natural, spoken language product descriptions from webpage titles, improving voice-based e-commerce interactions.

Contribution

The paper presents eBERT, a novel pre-trained BERT-based model specifically designed for generating conversational product descriptions in e-commerce.

Findings

eBERT outperforms baseline models in experiments

Human evaluations favor eBERT-generated descriptions

Deployed successfully in real-world applications

Abstract

Through recent advancements in speech technologies and introduction of smart assistants, such as Amazon Alexa, Apple Siri and Google Home, increasing number of users are interacting with various applications through voice commands. E-commerce companies typically display short product titles on their webpages, either human-curated or algorithmically generated, when brevity is required. However, these titles are dissimilar from natural spoken language. For example, "Lucky Charms Gluten Free Break-fast Cereal, 20.5 oz a box Lucky Charms Gluten Free" is acceptable to display on a webpage, while a similar title cannot be used in a voice based text-to-speech application. In such conversational systems, an easy to comprehend sentence, such as "a 20.5 ounce box of lucky charms gluten free cereal" is preferred. Compared to display devices, where images and detailed product information can be presented to users, short titles for products which convey the most important information, are necessary when interfacing with voice assistants. We propose eBERT, a sequence-to-sequence approach by further pre-training the BERT embeddings on an e-commerce product description corpus, and then fine-tuning the resulting model to generate short, natural, spoken language titles from input web titles. Our extensive experiments on a real-world industry dataset, as well as human evaluation of model output, demonstrate that eBERT summarization outperforms comparable baseline models. Owing to the efficacy of the model, a version of this model has been deployed in real-world setting.