Transcript Smart Search

Smart Search
Brett Chavis
Mike Giles
Randy Murdock
Leslie Jacobson
Patrick Tollison
Background
Just to Recap:
Our project’s purpose was to prove that a search
engine could be implemented using natural English
sentences and questions as the query.
We were to build a system that
accepted these natural English
queries and manipulated them to
access a database, containing a
limited number of sites that satisfy
a given request.
In The Beginning
• Began with SQL 7.0
• Attempted for weeks to get SQL running
• Consulted numerous SQL books
• Consulted with local system administrators
• Viewed SQL instructional videos
• Visited IT Professionals in Winston Salem for
help
• Still did not work!
Our Solution
• Realizing that SQL with English Query is the
optimum solution for this problem, but that we no
longer had the time to pursue it for our project we
decided to implement the project exclusively in Perl.
• Next, we identified the major components needed:
- User Interface
- Parser/Sentence Analyzer
- Database
- Maintenance Capabilities
User Interface
• Web page consists initially of a single input box
and a submit button
• Accepts Plain English Queries or keyword
searches.
• The system returns 5 or less links to quality sites
that match the user’s request.
• The system also provides links to Yahoo!, Alta
Vista, and AskJeeves already formatted to submit
directly to the engine, eliminating retyping for the
user.
Parser
• Creates a keyword search string from the
inputted query.
• Eliminates certain grammatical constructs and
all punctuation.
• Modal verbs
• Auxiliary verbs
• Linking verbs
• Pronouns
• Prepositions
• etc.
Database
key1,key2,key3
.
.
.
siteName1,siteName2
key4,key5
siteName2,siteName3
siteName1
URL,description
siteName2
URL,description
siteName3
.
.
.
URL,description
Maintenance
• The maintenance code we have written is to
be used to administer the database.
• It provides the following capabilities:
• Add information to support a new query
• Add a new web site to an existing query
• Delete a query
• Delete a web site from a query
• Modify a web site entry
• List the keywords for all currently supported
queries
In Progress
• Maintenance needs to be converted from
command-line to web interface
• Continued research of grammar for the parser
• Database Development
• Final System Test
• Documentation: Technical Manual
Pitfalls
• Did not receive SQL Server software until the
week preceding Spring Break.
• No resources/documentation on English Query
• Content in directory was deleted TWICE!
• While pursuing SQL, progress was slowed due to
limited access to the machine on which it was
installed.
• No previous experience with databases
• Limited experience with Perl among the majority
of the team
Knowing What We Know Now
• SQL 7.0 with English Query is a very
powerful utility and would serve best to
accomplish this project.
• However, given the time constraints, we
would have never pursued SQL with English
Query from the beginning had we known the
complexity involved.
• The complexity of composing an English
lexer is unreasonable for a semester-long
course.
Team Successes
• Regrouped and completed a working
system in limited amount of time.
• Fully appreciate team experience of
working through major problems and being
proud of our accomplishments.
• Learned to find resources outside of the
University.