Hi, Last week I started a nepomuk based project called *resourcebrowser*. Here is an update.
Resource browser as the name suggests is a browser using which you can browse nepomuk resources. The user is provided with a search bar, through which the resource browsing is carried out. The user can search for resources using either the name or content or type of the resource.(Just like dolphin's search bar, but searching is extended to all resources,not just files) . Along with it, the user is provided with recommendations related to the resources. The recommendations are currently being provided if the resource name is present in other file's content. I plan to extend it, such that the most frequently occurred keywords present in one file are matched with other files and recommendations are shown in the near future.I need to do some NLP during this keyword extraction so that I can stem the content and ignore all English articles(a,the,is etc). I'm also thinking about usage based recommendation (Ivan's contour does the same. May be I can use his code). There are filters provided on the left hand side of the browser, which can be used to fine tune the search results. (Just like dolphin's filter panel.But I have added Nepomuk::Utils::Facet::createPriorityFacet along with other facets).Using this priority facet user can search for a movie in his computer that he has never watched.Or, he could use it to find out the unused files in his computer and delete them if needed. The user click tracking is done, so that the resource that is accessed the most pops out on top of the search results.(Nepomuk::Resource::usageCount() will be incremented every time). Since contour's recommendation engine tracks many more data , I should take a look at it. Since the usage analysis of the user is carried out, I can use it for filtering 'most important' from priority facet (which is doing nothing afaik). There are few extra features, one of which is where the user is provided with a GUI module to manually create a link between 2 resources. I have provided a list which shows all linked resources, which will be populated whenever the user selects any resource from main body.Unlinking can be done just by selecting unlink from context menu. Another feature is for the removal of duplicates of resources, again to save space. The ultimate aim of this project is to cut down the navigation time of reaching to any resource, similar to how a web search engine works. That is, the user gets the resource he wants immediately, preferably at the top of the search results. Future Enhancements: 1. Improved recommendation with respect to both content based and usage based. 2. The user can search using natural language queries eg. Play the movie i saw last nigh Or more complicated queries like, “ show me the C++ files I worked on last night “ 3. If combined with a voice service, it would turn out to more practically useful with respect to the common man as he would want to just talk to his computer to do his tasks for him. Yes, it is something similar to siri of iphone.(Just thinking big , I know its complexity). Finally, this project is yet to be named and I welcome your suggestions. I'm also open to any suggestions regarding this project,future enhancements etc. Thanks a lot :) resourcebrowser can be found in my scratch repo: git:// anongit.kde.org/scratch/hegde/resourcebrowser.git -- Regards - Phaneendra Hegde IRC-Nick: pnh (#nepomuk-kde,#kde-in)
_______________________________________________ Nepomuk mailing list [email protected] https://mail.kde.org/mailman/listinfo/nepomuk
