A Usage-centric Take on Intent Understanding in E-Commerce

Published 10/8/2024 by Wendi Zhou, Tianyi Li, Pavlos Vougiouklis, Mark Steedman, Jeff Z. Pan

Overview

The paper presents a usage-centric approach to intent understanding in e-commerce, focusing on how customers use product information to inform their purchasing decisions.
It introduces the FolkScope framework, which analyzes user interactions with product pages to infer their purchase intent.
The paper evaluates FolkScope on a real-world e-commerce dataset and discusses its potential applications in areas like product search and recommendation.

Illustration of intent understanding's usage-centric paradigm.

1/3

Original caption: Figure 1: A graphic illustration of the usage-centric paradigm of intent understanding.

Maximum MRR for bought-product recovery.

1/2

Model	Clothing Accuracy	Electronics Accuracy
FolkScope	0.192	0.263
FolkScope - Properties	0.116	0.166
FolkScope + GPT	0.187	0.257

Original caption: Table 1: MRRmaxsubscriptMRR\text{MRR}_{\max}MRR start_POSTSUBSCRIPT roman_max end_POSTSUBSCRIPT for bought-product-recovery task.

Query	Clothing	Electronics
Exact Match	16%	22%
Partial Match (>= 10 Hits)	73%	63%

Original caption: Table 2: The ratio of hit being the first in the prediction list and not in the top-10 of the prediction list.

Plain English Explanation

When shopping online, customers often have a specific intent in mind - whether they're looking to buy a product, research it, or just browse. Understanding this intent is crucial for e-commerce companies to provide a better user experience and drive more sales.

The researchers propose a "usage-centric" approach, which looks at how customers interact with product pages to infer their purchase intent. Instead of just analyzing search queries or click-through rates, this method examines things like how long a customer spends on a page, which sections they focus on, and whether they add items to their cart.

The key idea is that these usage patterns can reveal a lot about a customer's underlying goals and mindset. For example, someone who carefully reads product specifications and reviews is likely researching the item, while someone who quickly bounces between pages may just be browsing.

The researchers developed a framework called FolkScope to analyze these usage patterns and infer purchase intent. They evaluated it on real-world e-commerce data and found that it could outperform traditional intent detection methods.

This usage-centric approach could have important applications in areas like product search and recommendation. By understanding a customer's true intent, e-commerce sites can surface the most relevant products, provide personalized recommendations, and ultimately drive more sales.

Technical Explanation

The paper introduces a "usage-centric" framework for intent understanding in e-commerce, called FolkScope. Instead of relying solely on textual signals like search queries, FolkScope analyzes how customers interact with and navigate product pages to infer their underlying purchase intent.

The key idea is that usage patterns - such as time spent on a page, which sections a user focuses on, and whether items are added to the cart - can reveal a lot about a customer's goals and mindset. For example, a customer who carefully reads product reviews and specifications is likely researching the item, while someone who quickly bounces between pages may just be browsing.

FolkScope captures these usage patterns through a combination of interaction logs, page content analysis, and neural network models. It first extracts a set of usage features from the user's interactions, such as time on page, scroll depth, and interactions with specific page elements. These features are then fed into a neural classifier that predicts the user's purchase intent (e.g. research, browse, buy).

The researchers evaluated FolkScope on a large-scale e-commerce dataset, comparing its performance to traditional intent detection methods that rely on textual signals alone. They found that the usage-centric approach could significantly outperform these baselines, demonstrating the value of incorporating user behavior data into intent understanding.

Critical Analysis

The paper presents a novel and promising approach to intent understanding in e-commerce, but there are a few potential limitations and areas for further research:

Generalizability: The evaluation was conducted on a single e-commerce dataset, so more research is needed to assess how well the FolkScope framework would generalize to other domains or shopping platforms.
Explainability: While the neural network model can accurately predict purchase intent, it may be a "black box" that does not provide much insight into the underlying decision-making process. Developing more interpretable models could be valuable for e-commerce practitioners.
Dynamic Intent: The paper treats intent as a static classification, but in reality customer intent can shift dynamically during the shopping journey. Incorporating these temporal dynamics could lead to more nuanced and accurate intent modeling.
Multimodal Integration: The current approach focuses on usage patterns, but combining these with other signals like textual queries, product metadata, and user profiles could potentially yield even stronger intent understanding.

Overall, the usage-centric perspective introduced in this paper represents an important step forward in intent understanding for e-commerce. By focusing on how customers interact with and navigate product information, it opens up new avenues for delivering more personalized and efficient shopping experiences.

Conclusion

This paper presents a novel "usage-centric" approach to intent understanding in e-commerce, centered around analyzing how customers interact with and navigate product pages. By capturing usage patterns like time on page, scroll depth, and interaction with specific page elements, the FolkScope framework can infer a customer's underlying purchase intent - whether they are researching a product, browsing, or ready to buy.

Evaluated on a real-world e-commerce dataset, FolkScope demonstrated superior performance compared to traditional intent detection methods that rely solely on textual signals. This usage-centric perspective has significant potential applications in areas like product search, recommendation, and personalization, as e-commerce companies strive to better understand and cater to customer needs and goals.

While further research is needed to address limitations like generalizability and explainability, this paper represents an important step forward in leveraging customer behavior data to enhance the online shopping experience. By focusing on how users interact with product information, it offers a fresh and insightful angle on the critical challenge of intent understanding.

Full paper

Loading PDF viewer...

Read original: arXiv:2402.14901

This summary was produced with help from an AI and may contain inaccuracies - check out the links to read the original source documents!

Follow @aimodelsfyi on 𝕏 →

Related Papers

IntentionQA: A Benchmark for Evaluating Purchase Intention Comprehension Abilities of Language Models in E-commerce

Wenxuan Ding, Weiqi Wang, Sze Heng Douglas Kwok, Minghao Liu, Tianqing Fang, Jiaxin Bai, Junxian He, Yangqiu Song

Enhancing Language Models' (LMs) ability to understand purchase intentions in E-commerce scenarios is crucial for their effective assistance in various downstream tasks. However, previous approaches that distill intentions from LMs often fail to generate meaningful and human-centric intentions applicable in real-world E-commerce contexts. This raises concerns about the true comprehension and utilization of purchase intentions by LMs. In this paper, we present IntentionQA, a double-task multiple-choice question answering benchmark to evaluate LMs' comprehension of purchase intentions in E-commerce. Specifically, LMs are tasked to infer intentions based on purchased products and utilize them to predict additional purchases. IntentionQA consists of 4,360 carefully curated problems across three difficulty levels, constructed using an automated pipeline to ensure scalability on large E-commerce platforms. Human evaluations demonstrate the high quality and low false-negative rate of our benchmark. Extensive experiments across 19 language models show that they still struggle with certain scenarios, such as understanding products and intentions accurately, jointly reasoning with products and intentions, and more, in which they fall far behind human performances. Our code and data are publicly available at https://github.com/HKUST-KnowComp/IntentionQA.

6/17/2024

🤔

Understanding the User: An Intent-Based Ranking Dataset

Abhijit Anand, Jurek Leonhardt, V Venktesh, Avishek Anand

As information retrieval systems continue to evolve, accurate evaluation and benchmarking of these systems become pivotal. Web search datasets, such as MS MARCO, primarily provide short keyword queries without accompanying intent or descriptions, posing a challenge in comprehending the underlying information need. This paper proposes an approach to augmenting such datasets to annotate informative query descriptions, with a focus on two prominent benchmark datasets: TREC-DL-21 and TREC-DL-22. Our methodology involves utilizing state-of-the-art LLMs to analyze and comprehend the implicit intent within individual queries from benchmark datasets. By extracting key semantic elements, we construct detailed and contextually rich descriptions for these queries. To validate the generated query descriptions, we employ crowdsourcing as a reliable means of obtaining diverse human perspectives on the accuracy and informativeness of the descriptions. This information can be used as an evaluation set for tasks such as ranking, query rewriting, or others.

9/2/2024

Centrality-aware Product Retrieval and Ranking

Hadeel Saadany, Swapnil Bhosale, Samarth Agrawal, Diptesh Kanojia, Constantin Orasan, Zhe Wu

This paper addresses the challenge of improving user experience on e-commerce platforms by enhancing product ranking relevant to users' search queries. Ambiguity and complexity of user queries often lead to a mismatch between the user's intent and retrieved product titles or documents. Recent approaches have proposed the use of Transformer-based models, which need millions of annotated query-title pairs during the pre-training stage, and this data often does not take user intent into account. To tackle this, we curate samples from existing datasets at eBay, manually annotated with buyer-centric relevance scores and centrality scores, which reflect how well the product title matches the users' intent. We introduce a User-intent Centrality Optimization (UCO) approach for existing models, which optimises for the user intent in semantic product search. To that end, we propose a dual-loss based optimisation to handle hard negatives, i.e., product titles that are semantically relevant but do not reflect the user's intent. Our contributions include curating challenging evaluation sets and implementing UCO, resulting in significant product ranking efficiency improvements observed for different evaluation metrics. Our work aims to ensure that the most buyer-centric titles for a query are ranked higher, thereby, enhancing the user experience on e-commerce platforms.

10/22/2024

Semantic In-Domain Product Identification for Search Queries

Sanat Sharma, Jayant Kumar, Twisha Naik, Zhaoyu Lu, Arvind Srikantan, Tracy Holloway King

Accurate explicit and implicit product identification in search queries is critical for enhancing user experiences, especially at a company like Adobe which has over 50 products and covers queries across hundreds of tools. In this work, we present a novel approach to training a product classifier from user behavioral data. Our semantic model led to >25% relative improvement in CTR (click through rate) across the deployed surfaces; a >50% decrease in null rate; a 2x increase in the app cards surfaced, which helps drive product visibility.

5/30/2024