Title:  Highlighting 		                                Page 1 of  
Date Prepared:  November 21, 1995          Date Needed: 
		April 3, 1996
		May 28, 1996


Paragraphs Affected:    Section 7.1.4 Document and Query Indexes

References: RFC Viewing RetrievalQueries & RoutingQueries

Change Required:
---------------
	
	We propose 2 new operations to the Tipster Architecture.
	These operations will identify matching query terms and
	passages in Documents.  One MatchQueryTerms() operation will be
	added to the DocumentCollectionIndex and the other to the
	QueryCollectionIndex.

	Documentation shall be included to stress that these new
	operations will add Annotations to the Documents.  In
	addition, the documentation shall stress that these
	query term Annotations are not critical and need not be
	maintained over the life of the Document.


Specific Recommendation:
-----------------------
Section 7.1.4 [Document & Query Indexes]

	Class DocumentCollectionIndex
	...
	Operations
	   MatchQueryTerms(sequence of DocumentCollectionIndex, 
			   RetrievalQuery, 
		     	   Document)

	       Document is a Document which has been returned in the
	       Collection created by a RetrieveDocuments() operation.

	       RetrievalQuery is the query evaluated during the
	       RetrieveDocuments() operation.

	       sequence of DocumentCollectionIndex is the set of
	       indices against which the RetrievalQuery was evaluated.

	       The MatchQueryTerms() operation will add Annotations to the
	       Document to indicate terms and passages in the Document
	       which match the input RetrievalQuery. 

		    annotation type RelevantPassage 
		         {rank : String}
		    annotation type RelevantTerm    
		         {rank : String}


Title:  Highlighting		                              Page 2 of


	       The query term Annotations indicate where the RawData
	       component of a particular Document matches the
	       RetrievalQuery.  Because these Annotations are specific
	       to an instance of a RetrievalQuery they are not
	       intended to be a permanent part of the Document. 
	       Therefore, users may wish to delete these query term
	       Annotations when they are finished reviewing their
	       RetrieveDocuments results.

	Class QueryCollectionIndex
	...
	Operations
	   MatchQueryTerms(sequence of QueryCollectionIndex, RoutingQuery, 
		     	   Document)

	       Document is a Document which was compared to a set of
	       QueryCollectionIndexes in a RetrieveQueries()
	       operation.  

	       sequence of QueryCollectionIndex is the set of QCIs
	       used to compare the Document during the
	       RetrieveQueries() operation.

	       RoutingQuery is the current RoutingQuery representing a
	       DetectionNeed returned by the RetrieveQueries()
	       operation. 

	       Will add Annotations to the Document to indicate terms
	       and passages in the Document which match the input
	       RoutingQuery. 

		     annotation type RelevantPassage
		         {rank : String}
		     annotation type RelevantTerm    
		         {rank : String}

	       The query term Annotations indicate where a
	       particular Document matches a RoutingQuery.  Because
	       these Annotations are specific to an instance of a
	       RoutingQuery they are not intended to be a permanent
	       part of the Document.  Therefore, users may wish to
	       delete these query term Annotations when they are
	       finished reviewing their RetrieveQueries results.

	   GetQuery(QueryCollectionIndex, DetectionNeed): RoutingQuery or nil
		
	       GetQuery returns the RoutingQuery in QueryCollectionIndex
	       whose DetectionNeed component is identical to the
	       DetectionNeed given as an argument.  If no such
	       RoutingQuery exists in the QueryCollectionIndex, nil is
	       returned. 
	 	[see Viewing RetrievalQueries and RoutingQueries RFC]

Title:  Highlighting		                               Page 3 of


Reason for Proposed Change:
--------------------------
	Indicating which Document terms and passages matched the
	RetrievalQuery or RoutingQuery are typical highlighting functions
	performed by Detection systems.   Section 7 of the Tipster
	architecture defines the functionality of the Detection system
	yet does not mention these highlighting tasks as operations
	for of any of the Detection objects. 

	By adding a MatchQueryTerms() operation to both the
	DocumentCollectionIndex and the QueryCollectionIndex objects,
	the Tipster Architecture provides the expected query matching
	information users expect from an IR system.

	An additional operation GetQuery() has been added to the
	QueryCollectionIndex object.  A QueryCollectionIndex consists
	of RoutingQueries where  each RoutingQuery is a representation of a
	DetectionNeed. 

	The RetrieveQueries() operation returns only the Collection of
	DetectionNeeds matching a particular Document.  To provide
	highlighting information for that <Document, DetectionNeed>
	pair, the Search Engine requires the RoutingQuery
	representation of the DetectionNeed which caused the Document
	to be matched.  The GetQuery() operation allows the
	RoutingQuery matching the DetectionNeed to be retrieved from
	the QueryCollectionIndex and used later in the
	MatchQueryTerms() operation.

Change Requested By:
    Organization:     	University of Massachusetts, CIIR
    Name:  		Kathleen S DiBella
    Phone Number:    	(413)545-9781
    Date:	 	October 19, 1995