Reiner Kraft - April 16, 2007
Title: Contextual Shortcuts - A Platform for Entity Detection and Content Syndication
Abstract: Accessing and finding information on the Web is
becoming increasingly difficult due to its growth. Search engines and
hyperlinks help us to navigate and find information, yet often users
may want to perform additional tasks with items in web pages, blog
entries or emails. Simple tasks such as storing an address, looking up
a map for a location, or finding relevant information for a particular
item of interest are inadequately supported. Current hyperlinks do not
provide this desired functionality and may be disruptive, typically
leaving users to cut and paste portions of text into a tool external
to the current page. Furthermore, this requires users to make the
appropriate tool choice to get the most relevant content or or
service.
We introduce a platform for delivering "Contextual
Shortcuts", intelligent hyperlinks to provide simple single-click
access to relevant information and services. The platform
automatically detects "entities" of interest within unstructured
content, extracts their features, and annotates the original document
with meta-data derived from those features. Additionally, it provides
a framework for content syndicators and publishers to easily deliver
relevant information and services at the "point of inspiration".
The Contextual Shortcuts platform bridges the gap between
entity detection and seamless interaction with such entities. The
platform's high performance detection framework produces detection
accuracy on par with state-of-the-art techniques, and is equally
suited to a variety of view-time or publish-time annotation scenarios.
Bio: Reiner is a senior engineering manager at Yahoo!,
Sunnyvale, CA. He manages the contextual search team within web
search.
He invented Y!Q - Contextual Search, and turned his idea within a
short time period into a product that launched early last year in beta
at http://yq.search.yahoo.com. Y!Q allows users to search at the
point of inspiration, and uses contextual information to augment a
user's query and therefore provide more relevant results.
Reiner earned a Ph.D. in Computer Science from the University of
California at Santa Cruz. His advisor was Prof. Raymie Stata. He
performs research in the fields of web information retrieval, Internet
search technologies, machine learning, hypertext, and Internet
information systems. His recent research focus in on contextual search
and search personalization.
Reiner has been nominated with the TR100 award from MIT Technology
Review. The TR100 represents a young group of 100 innovators, all
under age 35 as of Jan. 1, 2002. Their innovations will have a deep
impact on how we live, work, and think in the century to come.
In the past 10 years he has filed over 100 patent applications for
IBM and Yahoo!, contributing very valuable and high quality ideas to
their patent portfolio. In 2003 he was nominated as an Master Inventor
within IBM's research division.
His PhD thesis focused on genre search and doc-type classification
for iterative, filtering meta-search. He designed and implemented the
buying guide finder (BGF). BGF is an intelligent web search agent (or
"web carnivore") that takes as input a specialized information need
and returns as output search results that satisfy this information
need. It obtains raw data from search engines, rather than directly
crawling and indexing data on the Web. The proposed technique makes it
economical to build very powerful, comprehensive search engines for
very small, specialized purposes.
Sami Rollins