Exclusion of HTML navigation elements in webcrawl

As users of WDS are currently not able to use SDU on crawled web documents, it is impossible to use the field definition and exclusion function WDS provides to exclude unwanted results from the index.

In our case, website navigation and footers create noise that leads to bad query results. Inclusion of positive examples through enrichments and relevancy training was not enough to offset the noise. In most cases, stopwords cannot be used, as the navigation elements are still terms we need to be able to search for. The bad query performance has been noticed by customers and we're having trouble explaining to them that exclusion of those elements is apparently not implemented.

Example:

The URL to be crawled is https://www.dgm.org/muskelerkrankungen/ which is a website about muscular diseases in German. The navigation bar containing all the possible muscular diseases will remain the same for all pages below this path level, navigation does not show up in the URL. For https://www.dgm.org/muskelerkrankungen/amyotrophe-lateralsklerose-als which is the top page about ALS, the index will still include all the navigation for the website. In the case of a search for another muscular disease, the ALS subpage will still be shown, only because it contains the navigation. Since navigation does not show up in the URL, it cannot be excluded by using url processing settings. Same goes for common footers about GDPR regulations and such.

Depending on how websites are crawled, it could be made possible in the following ways:

• identify certain HTML elements by tag at ingestion time and give users an option for exclusion of HTML elements (processing settings for collections)

• make SDU availabe for webcrawl documents to use the existing field filter function

• exclusion by default at ingestion time