2020/05/28 by Ali Ahmadvand, Surya Kallumadi, Ahmadvand, Ali +5 · 1 citation
Computer Science · #Computation and Language (cs.CL) #Computer science #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Information Retrieval (cs.IR) #Information Retrieval and Search Behavior #Information retrieval #Process (computing) #Product (mathematics) #Query expansion #Ranking (information retrieval) #Sargable #Search engine #Set (abstract data type) #Text and Document Classification Technologies #Web query classification #Web search query #cs.CL #cs.IR
paper · pdf · doi:10.48550/arxiv.2005.13783
published in arXiv (Cornell University) (Cornell University) · SIGIR 2020
openalex publication_date 2020/05/28 · arxiv created 2020/05/29 · arxiv updated 2020/06/02 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/08
An accurate understanding of a user's query intent can help improve the\nperformance of downstream tasks such as query scoping and ranking. In the≠-commerce domain, recent work in query understanding focuses on the query to\nproduct-category mapping. But, a small yet significant percentage of queries\n(in our website 1.5% or 33M queries in 2019) have non-commercial intent\nassociated with them. These intents are usually associated with non-commercial\ninformation seeking needs such as discounts, store hours, installation guides,\netc. In this paper, we introduce Joint Query Intent Understanding (JointMap), a\ndeep learning model to simultaneously learn two different high-level user\nintent tasks: 1) identifying a query's commercial vs. non-commercial intent,\nand 2) associating a set of relevant product categories in taxonomy to a\nproduct query. JointMap model works by leveraging the transfer bias that exists\nbetween these two related tasks through a joint-learning process. As curating a\nlabeled data set for these tasks can be expensive and time-consuming, we\npropose a distant supervision approach in conjunction with an active learning\nmodel to generate high-quality training data sets. To demonstrate the\neffectiveness of JointMap, we use search queries collected from a large\ncommercial website. Our results show that JointMap significantly improves both\n"commercial vs. non-commercial" intent prediction and product category mapping\nby 2.3% and 10% on average over state-of-the-art deep learning methods. Our\nfindings suggest a promising direction to model the intent hierarchies in an≠-commerce search engine.\n