Browse Prior Art Database

Automated Category Lookup Algorithm for E-commerce Platform to Improve Search Result Relevancy Disclosure Number: IPCOM000240633D
Publication Date: 2015-Feb-13
Document File: 4 page(s) / 72K

Publishing Venue

The Prior Art Database


Disclosed is a probabilistic category lookup algorithm to improve e-commerce search relevancy when shoppers enter multiple search keywords. This algorithm adds intelligence/cognitive power to the search to identify the intent of the shopper.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 40% of the total text.

Page 01 of 4

Automated Category Lookup Algorithm for E-commerce Platform to Improve Search Result Relevancy

E-commerce platforms utilize search engines that return results based on the total number of matched terms between the shopper's query and the products' indexed fields. E-commerce websites face two types of challenges when the search contains multiple keywords. One is that a results set is too large, because a focus on matching 'any' one of the shopper's keywords and the indexed documents returns too many results; therefore, the shopper sees products that do are not related to the desired items .

Another problem is that the results set is too small. A focus to return products with indexed fields containing all or part of the search query might be too limiting, depending on what product information is indexed and what search phrases the shopper enters . These poor results sets inhibit the store's ability to generate revenue.

To compensate for the aforementioned drawbacks, systems use search rules are used such as specifying the distance between the words in the index, pre-defining a set of synonyms/replacement terms/business rules for common keyword searches, or boosting the weight of indexed fields. However, this involves a lot of manual work, and can still produce misleading search results because the search is still based on the matching scheme, and not an understand of the shopper's intent or needs. In addition, common search approaches focus on searching at the product level. This is a time-consuming process of looking at all products being indexed within the store at run-time.

To improve the store's conversion rate, algorithms are needed that improve search relevancy for multiple-keyword searches on an e-commerce platform.

The core novel idea is to use a search algorithm that, instead of counting matched terms, determines which category within the store contains the product mentioned in shopper's search query. In response to a shopper's multiple-keyword search query on an e-commerce user interface (UI), the system uses relevancy to the search criteria to determine which products to return as well as the relative order within the results list. Thus, results are in order of relevancy based on the user's search query. This greatly affects the store's conversion rate.

Once the category is determined, the search system can use it to influence the returned results by applying the following three options:

Scoping (filtering) or boosting results using the known category Suggesting related products when a single Stock Keeping Unit (SKU)/product identifier (ID) is entered
Assigning certain categories higher importance than others, and then organizing the categories into different boost groups


Page 02 of 4

that are associated with different boost factors (e.g., when 'notebook' is searched, boost the Electronic category higher than the Stationary category )

With this search algorithm, the system returns relevant results for shoppe...