UNIVERSITY OF JYVÄSKYLÄ
Mobile Search – Social Network Search
Using Mobile Devices
1st IEEE International Peer-to-Peer for Handheld Devices Workshop
IEEE CCNC, Las Vegas, 12th of January 2007
Mikko Vapa, research student, [email protected]
With co-authors Pedro Tiago, Niko Kotilainen,
Heikki Kokkinen and Jukka K. Nurminen (Nokia P2P Team)
Department of Mathematical Information Technology
University of Jyväskylä, Finland
www.mit.jyu.fi/cheesefactory
UNIVERSITY OF JYVÄSKYLÄ
Background
• Mobile phones' computational power has been improving
approaching the capabilities of general purpose computers
• Nowadays it is possible to host a web site on a mobile device
• It is also expected that the number of mobile web sites will
outnumber the static web servers
• Recently, there has been a growing interest in how to explore
the mobile phone capabilities in the web search context and how
to merge them with existing phone functionalities
[Johan Wikman, Ferenc Dosa, and Mikko Tarkiainen. Personal website
on a mobile phone. Technical report, Nokia Research Center, 2006]
UNIVERSITY OF JYVÄSKYLÄ
Mobile Search
• Mobile Search is a system for social network search on a mobile
device
• Prototype was implemented on top of Drupal content
management system running on Mobile Apache/Raccoon mobile
web server
• Based on pure peer-to-peer architecture offers scalability,
efficiency, resilience to failures and privacy at a higher degree
than centralized solutions
[Choon-Hoong, Nutanong and Buyya, Peer-to-Peer Networks for
Content Sharing, Peer-to-Peer Computing: Evolution of a Disruptive
Technology, 2005]
UNIVERSITY OF JYVÄSKYLÄ
Features
• Allows executing searches to the contents of mobile devices
using a web interface
• Searches through social network defined by the addressbooks of
the mobile devices
• Manages access rights for different kind of contents (calendar
data, photos, blogs etc.) using motto:
“I only display what I want to who I want”
• Can also search normal Drupal websites
UNIVERSITY OF JYVÄSKYLÄ
New Search Concepts
•
•
Manual multi-hopping
– Users search one graph level of their social network at a time
usually starting from their neighbors
– Every time a user issues a search query the mobile device forwards
it to all the neighbors of the user
– The neighbors answer back by returning a result set and a list of
their neighbors
– If the user who issued the query is not satisfied by the results he
can always ask new results from the next level neighbors as long as
there are non-visited nodes in the network
Automatic multi-hopping
– A sorting algorithm decides which of the non-visited nodes are
queried further thus avoiding the need for user decision
– Automatically sorting the non-visited nodes leads to tradeoff
between search accuracy and easiness of searching suggesting
that both manual and automatic multi-hopping should be available
for the user
UNIVERSITY OF JYVÄSKYLÄ
Benefits
•
Compared to centralized web search engines:
– Mobile Search provides access to rare personal data relevant to
people close in the social network
– The contents indexed by Mobile Search might not be referenced
anywhere but still they are searchable
– Real-time - Does not provide outdated links
– Highly distributed, decentralized and no single point of failure
– Mobile Search can utilize websites’ internal search functionalities
– Search is executed within the limits of access control rights providing
means to search non-public data (internal search among friends etc.)
– However, social network search is not suited to find popular content
• But, it's a powerful mechanism in restricted topic set environment
[Mislove, Gummadi, and Druschel, Exploiting social networks for internet
search, Proceedings of the 5th Workshop on Hot Topics in Networks,
2006]
UNIVERSITY OF JYVÄSKYLÄ
Drupal Prototype of Mobile Search
• Drupal is an open-source content management system for
managing and publishing several types of content
• Prototype is logically divided to local web search engine and
metacrawler parts
– Local web search engine is a search service, which
manages the search index of the mobile device
– Metacrawler is a search service, which uses other local web
search engines for getting the results and combines different
result sets into one
• Metacrawler was built as a weakly coupled component on top of
Drupal local web search engine
– Features automatic multi-hopping and result interleaving
– Differs from blog aggregators because content is being
searched and a set of queried nodes is not fixed
UNIVERSITY OF JYVÄSKYLÄ
Drupal Prototype of Mobile Search
• Drupal tac_lite module and Drupal module were used as
fundamental elements in the prototype
– These modules allow setting content access rules and to
process user authentication in distributed fashion without any
central servers
• An extra component that allows to do queries to local mobile
phone content such as location, address book and meeting data
was implemented
– This feature was built as a simple proof of concept
– However, the prototype is also able to gather search results
from unmodified Drupal web sites
UNIVERSITY OF JYVÄSKYLÄ
User Interface
UNIVERSITY OF JYVÄSKYLÄ
Technical Limitations
• The current implementation is single threaded because Mobile
Apache/Raccoon web server doesn't support multiple threads
[Wikman, Mobile web server - eurooscon presentation, 2006]
[Wikman and Dosa, Providing http access to web servers running on
mobile phones, Technical report, Nokia Research Center, 2006]
• Single-threaded nature of the metacrawler is a drawback
• This has a negative impact on response time because site
crawling is done in a serial way
• A multi-threaded implementation would speed up the system
considerably
UNIVERSITY OF JYVÄSKYLÄ
Future Work
•
•
Query forwarding/node sorting algorithms should be considered though
in a different setting than previous studies
– Algorithms like K-Random walk, Expanding Ring and hybrids using
NeuroSearch neural network should be considered
• Requires collecting some search usage statistics
Also one interest is the usability of search results, and new paradigms
of displaying different types of information and user interaction
– Web 2.0 may not be fully suitable for mobile device paradigm of
interaction
– This could also be an excellent opportunity to use a query language
applied to this type of systems for example an adaptation of
webSQL
[Mendelzon et al., Querying the world wide web, Int. J. on Digital Libraries,
1997]
• Would likely create a bigger interoperability and homogenization
in this type of systems with easier deployment of new
functionalities
UNIVERSITY OF JYVÄSKYLÄ
Future Work
• Mobile Search can be extended by creating different ways of
accessing the content, one entry point could be tags
• Tags work as links between content categorized similarly
• At each hop the user gets the list of contents tagged in a similar
way by nodes in its neighborhood
• Searching
Portugal would
give six results,
but then the
user might
continue the
search via
Lisboa tag and
finds the
Trolley image
UNIVERSITY OF JYVÄSKYLÄ
Conclusions
• Mobile Search complements traditional web search engines
• It gives the user means to explore the neighbors’ contents by
traveling to the friends network topology
• It covers a multitude of environments not covered by the
centralized solutions
• One of the main advantages in relation to current centralized
social network sites is the possibility to manage the site without
interference from an external entity
– Currently in a normal social network site a user can only
display or use modules made available by a third entity
– With Mobile Search approach it is possible to merge different
social network sites that cover different topics and create a
social network "melting pot”