Search Result Filters From Resource Content

What is this patent about?

’941 is related to the field of information retrieval and, more specifically, to techniques for refining search results by automatically generating and presenting relevant filters to the user. The background acknowledges that users often struggle to formulate precise queries and may need to refine their searches iteratively. Existing filtering mechanisms often rely on manually curated filters, which are difficult to maintain and may not capture emerging trends or specific user needs.

The underlying idea behind ’941 is to dynamically generate search filters based on the content of the resources that are initially returned in response to a user's query. The system analyzes the text and metadata of these resources to identify frequently occurring keywords and phrases. These keywords are then evaluated as potential filters, taking into account factors such as their frequency, relevance to the original query, and diversity of the resulting filtered result sets. The goal is to provide users with a set of automatically generated filters that can help them quickly narrow down their search and find the information they are looking for.

The claims of ’941 focus on a method, system, and computer-readable medium for generating and providing query filters. The core process involves receiving a query, identifying responsive resources, extracting keywords from those resources, determining a set of candidate filters from those keywords, and then selecting a subset of those candidate filters to present to the user as query filters . The selection of query filters is based on a term prominence threshold , which considers the position of the terms within the resources.

In practice, the system first identifies resources relevant to the initial query. It then extracts keywords and phrases from these resources. A crucial step is determining the 'term prominence' of each candidate filter. This involves analyzing where the terms appear within the resources – for example, terms in titles or headings are considered more prominent than those in the body text. Only candidate filters that meet a certain prominence threshold are selected as query filters and presented to the user. When a user selects a filter, the search results are updated to display only the content matching the selected filter.

This approach differs from traditional methods that rely on pre-defined or manually curated filters. By dynamically generating filters based on the content of the search results, the system can adapt to changing trends and provide more relevant and specific filtering options. The term prominence analysis ensures that the suggested filters are based on the most important and relevant terms within the search results, improving the user experience and helping them find the information they need more efficiently. This automated approach reduces the need for manual intervention and allows the system to scale more effectively.