Researchers claim to have developed a new software that can fix 90
per cent of broken links in the web of data - provided the resources are
still on the site's server.
Everyone knows the frustration of following a link to an interesting
web site only to discover the target page is no longer there and to be
presented with an error page, Iranian researchers said.
However, more frustrating and with wider implications for science,
healthcare, industry and other areas is when machines communicate and
expect to find specific resources that turn out to be missing or
dislocated from their identifier.
This can cause problems when a computer is processing large amounts
of data in a financial or scientific analysis, for instance, researchers
added.
If the resource is still on the servers, then it should be
retrievable given a sufficiently effective algorithm that can recreate
the missing links.
Computing engineers Mohammad Pourzaferani and Mohammad Ali
Nematbakhsh of the University of Isfahan explained that previous efforts
to address the issue of broken links in the web of data have focused on
the destination point.
This approach has two inherent limitations. First, it homes in on a
single point of failure whereas there might be wider issues across a
database. Secondly, it relies on knowledge of the destination data
source.
The team introduced a method for fixing broken links based on the
source point of links and a way to discover the new address of the
digital entity that has become detached.
Their method creates a superior and an inferior dataset which lets
them create an exclusive data graph that can be monitored over time in
order to identify changes and trap missing links as resources become
detached.
"The proposed algorithm uses the fact that entities preserve their
structure event after movement to another location. Therefore, the
algorithm creates an exclusive graph structure for each entity," said
Pourzaferani.
"When the broken link is detected the algorithm starts its task to
find the new location for detached entity or the best similar candidate
for it.
"To this end, the crawler controller module searches for the
superiors of each entity in the inferior dataset, and vice versa. After
some steps the search space is narrowed and the best candidate is
chosen," said Pourzaferani.
Researchers tested the algorithm on two snapshots of DBpedia within which are contained almost 300,000 person entities.
Their algorithm identified almost 5,000 entities that changed between the first and second snapshot recorded some time later.
The algorithm relocated 9 out of 10 of the broken links. The details
are reported in the International Journal Web Engineering and
Technology.
Source : The Economic Times
courtsey.sapost.blogspot.in
