SP500207
NIST Special Publication 500-207: The First Text REtrieval Conference (TREC-1)
Application of the Automatic Message Router to the TIPSTER Collection
chapter
R. Jones
S. Leung
D.L. Pape
National Institute of Standards and Technology
Donna K. Harman
Centre for Electronic Document Reeearch
TREC Routing Experiments CPGHC and CPGCN
Svstems Summarv and Timing
I Construction of Indexes
The software does not invert the text. It inverts the queries (or filters) and passes the text
through the combined index formed from the queries.
Query Construction
D Automatically Built Queries (Routing)
1 Concept field used
2 Time to build query <5 seconds
3 (a) Terms selected from topic
(b) Terms weighted with weights based on terms from documents with
relevance judgements, and dynamically modified through the training set
and the test set.
c) Phrases extracted from topics
j) Automatic addition of Boolean connectors and proximity operators from
topics.
E Manually constructed queries (routing)
1 All topic fields used
2 Average time to build query 30 minutes
3 Query builder system expert
4 Data used to build query from topic
5 The creation of the query uses Boolean operators, and proximity operators.
249