Probing Product Description Generation via Posterior Distillation

Abstract

In product description generation (PDG), the user-cared aspect is critical for the recommendation system, which can not only improve user's experiences but also obtain more clicks. High-quality customer reviews can be considered as an ideal source to mine user-cared aspects. However, in reality, a large number of new products (known as long-tailed commodities) cannot gather sufficient amount of customer reviews, which brings a big challenge in the product description generation task. Existing works tend to generate the product description solely based on item information, i.e., product attributes or title words, which leads to tedious contents and cannot attract customers effectively. To tackle this problem, we propose an adaptive posterior network based on Transformer architecture that can utilize user-cared information from customer reviews. Specifically, we first extend the self-attentive Transformer encoder to encode product titles and attributes. Then, we apply an adaptive posterior distillation module to utilize useful review information, which integrates user-cared aspects to the generation process. Finally, we apply a Transformer-based decoding phase with copy mechanism to automatically generate the product description. Besides, we also collect a large-scare Chinese product description dataset to support our work and further research in this field. Experimental results show that our model is superior to traditional generative models in both automatic indicators and human evaluation.

Automatic product copywriting for e‐commerce

“…Similarly, Li et al (2020) constructed a set of aspects (i.e., salient attributes) and the corresponding keywords for each aspect, which, however, requires expensive expert annotations. An adaptive posterior transformer-based network Zhan et al (2021) has been designed to utilize relevant information from customer reviews extracted by an adaptive posterior dis-tillation module, where the product title and attribute information are considered as the main model input.…”

“…An adaptive posterior transformer‐based network Zhan et al. (2021) has been designed to utilize relevant information from customer reviews extracted by an adaptive posterior distillation module, where the product title and attribute information are considered as the main model input.…”

Intelligent online selling point extraction and generation for e‐commerce recommendation

“…Due to the diversity of e‐commerce platform services Zhan et al. (2021); Zhang et al. (2019); Chen et al.…”

Geração De Descrições De Produtos a Partir De Avaliações De Usuários Usando Um LLM

“…Em outra abordagem também com o framework Transformer, Zhan et al (2021) propõem um novo modelo com foco no aspectos que interessam aos usuários. Nesse sentido, assim como Novgorodov et al (2019), também usam as avaliações, mas não como fonte de informação e sim com o objetivo de incorporar na geração de texto os aspectos úteis aos usuários.…”

“…Dentre as técnicas existentes, temos uma grande diversidade em relação a quais dados são utilizados para gerar as descrições de forma automática, combinando múltiplas informações, como atributos fornecidos pelo anunciantes (Wang et al, 2017), o título do produto (Zhan et al, 2021), e até slogans publicitários (Zhang et al, 2022) . No nosso trabalho, seguimos a linha proposta por Novgorodov et al (2019) e utilizamos as avaliações deixadas por usuários do produto, entendendo que essa é uma fonte fonte rica de informações.…”