Predicting Latent Structured Intents from Shopping Queries

Abstract

In online shopping, users usually express their intent through search queries. However, these queries are often ambiguous. For example, it is more likely (and easier) for users to write a query like "high-end bike" than "21 speed carbon frames jamis or giant road bike". It is challenging to interpret these ambiguous queries and thus search result accuracy suffers. A user oftentimes needs to go through the frustrating process of refining search queries or self-teaching from possibly unstructured information. However, shopping is indeed a structured domain, that is composed of category hierarchy, brands, product lines, features, etc. It would be much better if a shopping site could understand users' intent through this structure, present organized information, and then find the items with the right categories, brands or features.In this paper we study the problem of inferring the latent intent from unstructured queries and mapping them to structured attributes. We present a novel framework that jointly learns this knowledge from user consumption behaviors and product metadata. We present a hybrid Long Shortterm Memory (LSTM) [10] joint model that is accurate and robust, even though user queries are noisy and product catalog is rapidly growing. Our study is conducted on a largescale dataset from Google Shopping, that is composed of millions of items and user queries along with their click responses. Extensive qualitative and quantitative evaluation shows that the proposed model is more accurate, concise, and robust than multiple possible alternatives. In terms of information retrieval (IR) performance, our model is able to improve the quality of current Google Shopping production system, which is a very strong baseline.

An End-to-End Solution for Named Entity Recognition in eCommerce Search

“…Wu et al (Wu et al 2017) identify product attributes intended by user queries by treating this as a multilabel text categorization problem. They train character-and word-level bidirectional-LSTMs (BiLSTM) jointly with a product-attribute auto-encoder, using implicit user feedback and no hand-labeled training data.…”

“…Therefore, it is impossible to determine how much labeling or engineering cost the current system relies on as their description is not truly end-to-end. Among these papers, three papers (Wu et al 2017;Majumder et al 2018;Wen et al 2019) explore deep learning for NER in eCommerce search. Two papers (Wu et al 2017;Wen et al 2019) perform an evaluation on queries in production.…”

A Regularised Intent Model for Discovering Multiple Intents in E-Commerce Tail Queries

“…The supervised baseline models we compare against are Multinomial Logistic Regression (LR), the Linear Chain CRF from the query intent understanding work in [11,19], and the Bi-LSTM-CRF from [10]. The recent work [26] on understanding intent in Google shopping queries is not applicable in our setting since it focuses on a different problem of understanding overall query intent and not token level attribute labelling as ours. These supervised baseline models were trained on the click-log labelled data set D L with elastic-net regularisation whose hyper-parameters were selected by 3-fold cross-validation with F 1 score as the performance metric.…”

“…However, the CRF auto-encoder has difficulty scaling to the label space for query intent understanding that is much larger than that for POS tagging. The most recent work on understanding intent of e-commerce search queries is described in [26] for Google shopping. However, it is not applicable in our setting since it focuses on a different problem of understanding overall query intent and not token level attribute labelling.…”

Attention improves concentration when learning node embeddings

“…In the specific context of e-commerce there have been works conducting an empirical study using LSTM networks to map queries to structured attributes [29], as well as works that consider the more specific problem of ranking query reformulations [24,22]. As opposed to the former work, our latent space model AttEST allows for arbitrary downstream tasks on queries while having a theoretical grounding.…”