The examples in the previous subsection illustrate not only the kind of knowledge to be collected, but also some of the methods to be used to capture the knowledge. Additional methods for knowledge capture are illustrated in this subsection.
James prepares a lengthy intelligence summary containing the sentence ``Thus, there is considerable evidence that Russia is covertly shipping enriched nuclear material disguised as raw-material containers from transshipment points in China.'' Because the report contains many instances of noun phrases identifying objects of frequent interest, and because a large number of those noun phrases appear in the given sentence, HAWK proposes to James that the given sentence is what our high-school teachers called a topic sentence, and proposes to use a version of that sentence as START-accessible annotation.
James prepares a shorter report, one for which HAWK cannot determine a topic sentence. HAWK asks James to indicate what sort of questions the report provides answers to, adding less than a few percent to the cost of producing the report. HAWK volunteers annotations that were attached to recent reports prepared by James, particularly if HAWK notices key word correspondence. James can ignore the suggestions or modify them.
At this point, James needs to know the expected reaction to overt US action in this region. When asked, START answers that PUBLIC REACTION is likely to be HIGH, NEGATIVE, and LONG-LASTING, using information from a relational database prepared by political scientists and converted into START accessible form by our knowledge recapturing system.
Over time, James's life becomes easier as journalists all over the world begin to annotate their magazine and newspaper articles so as to attract START queries.