September 2012
·
50 Reads
The motivation for this paper is to improve search on mobile devices. As search usually is performed in the same way as on the desktop, search providers focus on depth, i.e. to find documents about very specific topics. However, search on mobile devices is much more diverse and people often want to browse the web for interesting things around them, for news articles or for content that other people tagged as being valuable. The paper presents the implementation of a search engine optimized for mobile devices called Local Browsing. The application consists of a backend which is responsible for crawling the web and recommending datasets to users. The crawler fetches keywords, which are selected from web sites by frequency and windowing, further processed, filtered and weighted. The recommender on the other hand provides users with the best possible results based on several user inputs. These inputs include keywords, location, user history, general history and more. The recommendation is done by aggregation and weighting of matching results. The frontend is implemented independently and connects to the backend via the Local Browsing REST API. To find an optimized user interface, typical search use cases on mobile devices are analyzed and compared. As result we present a running prototype which is compared to different search providers. In comparison it outperforms other search engines for typical mobile searches. For general depth searches a comparison is difficult as our current dataset only covers a tiny bit of all the information available on the web.