By Timo Selvaraj
Enterprise customers struggle with the ability to tune relevance for their unique requirements and business use cases. While enterprise search products provide an out of the box relevance setting for their use, in most cases this does not provide the necessary tuning levers to adjust relevance for different use cases especially when using internal versus external facing search or searching across different repositories. While customers are tuning relevance similar to what Google Web Search suggests for SEO rankings, this may not help with larger sites with documents/urls with varying topics/products/information and for multi-site search where different use cases may require variations in relevance and/or adjusting in real time for providing accurate search ranking.
SearchBlox aims to address this challenge using a variety of features built into the platform :
Creating the baseline relevance without any tuning applied
SearchBlox comes with out of the box relevance tuning values, however you can get a baseline without any relevance tuning applied to see how the search rankings perform. With no relevance tuning applied, SearchBlox results are ranked purely by the search algorithm based on terms found, term frequency, field length and the position of the terms within the document/url/content/data.
Get a baseline by applying the following parameter to the search request to get results with and without tuning parameters :
Relevance Tuning Applied: http://localhost:8080/searchblox/servlet/SearchServlet?query=time&tune=true
Relevance Tuning Not Applied : http://localhost:8080/searchblox/servlet/SearchServlet?query=time&tune=false
Once the baseline is established for your specific use case, then the following settings can be changed to suit your specific use case and requirements. Please note that you can apply different sets of tuning parameters for different use cases on the same collections/index data.
Read more on relevance tuning your SearchBlox Enterprise Search on our developer site.
SearchBlox also supports Collection/URL biasing/boosting for favoring one collection/set of urls over others by changing the Boosting value under Collections Settings.
Contact us to learn more about solving your unique relevance tuning challenges.