Special Issue on: AI driven Information Discovery in eCommerce
• 大类 : 工程技术 - 3区
• 小类 : 计算机：信息系统 - 3区
Search, ranking, and recommendation have applications ranging from traditional web search, to document databases, to vertical search systems. In the age of big data, eCommerce websites have accumulated large amounts of user personal information and behavioral data. Moreover, human-generated and machine-generated business data has been experiencing an exponential growth. This calls for sophisticated technologies from a wide spectrum of areas including information retrieval, machine learning, artificial intelligence, statistics, econometrics, and psychology, to explore how to effectively take advantage of such high-volume data to drive sales and user experience.
In this special issue we will explore approaches for search, recommendations, business analytics, computational advertising, and other related aspects of Information Discovery in the eCommerce domain. The task is superficially the same as web-page search (fulfill a user's information need), but how this is achieved is very much different. On leading eCommerce websites (such as eBay, Flipkart, Amazon, and Alibaba), the traditional web-page ranking features are either not present or are present in a very different form. The entities that need to be discovered (the information that fulfills the need) might be unstructured, associated with structure, semi-structured, or have facets such as: price, ratings, title, description, seller location, and so on. Domains with such facets raise interesting research challenges such as a) relevance and ranking functions that take into account the tradeoffs across various facets with respect to the input query b) recommendations based on entity similarity, user location (e.g. shipping cost). These challenges require an inherent understanding of product attributes, user behavior, and the query context. Unlike document and web search, product sites are also characterized by the presence of a dynamic inventory with a high rate of change and turnover, and a long tail of query distribution.
Outside of search but still within Information Retrieval, the same feature in different domains can have radically different meaning. For example, in email filtering the presence of “Ray-Ban” along with a price is a strong indication of spam, but within an auction setting this likely indicates a valid product for sale. Another example is natural language translation; company names, product names, and even product descriptions do not translate well with existing tools. Similar problems exist with knowledge graphs that are not customized to match the product domain. In addition to the above topics, this special issue will also focus on AI and machine learning enhanced business analytics approaches for understanding online shopping and consumer behaviors. Another area of focus is computational modeling and analysis of advertising and other promotional forms in eCommerce. The main objective of this special issue is to publish an up-to-date high-quality set of papers that deal with AI driven information discovery in the eCommerce domain.
All journal submissions will be reviewed by at least three reviewers recruited by the editors of the special issue. We expect to accept about 10 papers.
The special issue relates to all aspects of eCommerce search and recommendations. Research topics and challenges that are usually encountered in this domain include:
Machine learning techniques such as online learning and deep learning for eCommerce applications
Semantic representation for users, products, and services & Semantic understanding of queries
Structured data and faceted search, for example, converting unstructured data to its structured form
The use of domain specific facets in search and other IR tasks, and how those facets are chosen
Query intent, suggestion, and auto-completion
Temporal dynamics for Search and Recommendation
Models for relevance and ranking for multi-faceted entities
Recall-oriented search for eCommerce including deterministic sorting of results lists (e.g. price low to high)
Click models for eCommerce domain
Session aware, and session-oriented search and recommendation
Construction and use of knowledge graph, and ontologies for search and recommendations
Personalization & contextualization, and the use of personal facets such as age, gender, location etc.
Indexing and search in a rapidly changing environment (for example, an auction site)
Efficiency and scalability
Diversity in product search and recommendations
Strategies for resolving extremely low (or no) recall queries
The use of external features such as reviews and ratings in ranking
User interfaces (mobile, desktop, voice, etc.) and personalization
Reviews and sentiment analysis
The use of social signals in ranking and beyond
The balance between business requirements and user requirements (revenue vs relevance)