As a technology company, Wayfair encourages all team members to roll up their sleeves and innovate. We take immense pride in how our work impacts customers and contributes to the business’s success. We also derive great satisfaction when a team member is recognized by the larger technology community.
Someone like our own Senior Data Scientist Yan Chen, who is attending the upcoming 44th annual European Conference on Information Retrieval (ECIR). ECIR is widely regarded as the premier European event when it comes to presenting new research in the area of information retrieval.
At this year’s gathering, Yan is presenting her research paper titled “ WANDS: Dataset for Product Search Relevance Assessment.” The impetus for her research began with an examination of search relevance, the measurement of the relationship between a user’s queries and products returned in search results.
Search relevance touches each of us every day. Think about when you shop on an eCommerce site. You make a query, and then it’s the search engines’ job to find and present relevant products among millions of options. The more relevant the results, the happier we are.
But as Yan discovered, determining the relevancy of results becomes more challenging when the scale of the data being searched grows. Specifically, this size makes it difficult to create relevance-focused evaluation datasets manually. After looking into other potential options, such as click logs, she decided to create an open-source e-commerce product dataset that could be used to fairly and accurately evaluate the relevancy of e-commerce product search engines. That’s the Wayfair Annotation DataSet, also known as WANDS.
Yan’s paper delves into the full scope of her work in creating the biggest publicly available search relevance dataset in the e-commerce domain. The paper also includes experimental results showing how effective WANDS is at improving the scalability of human annotation efforts and demonstrates how WANDS is effective in evaluating and discriminating between different search models.
Working at a company offering more than 33 million products from 23,000 suppliers, we can all appreciate the importance of providing our customers with the most relevant results for all of their searches. Our ability to deliver this relevance consistently is part of the larger commitment to offering the best experiences possible and this could not be achieved without the efforts of our teams and people like Yan.
Let’s all wish her good luck in Norway at ECIR 2022!
