The NSSN, like the APIB gateway, is a service for obtaining standards information over the Internet. When completed, the NSSN will provide an electronic infrastructure linking numerous databases of standards resources throughout the United States. An NSSN testbed implementation has been set up for developing and testing the concepts and technologies to be used in actual NSSN production environment. The NSSN testbed currently allows users to search collections of sample documents using either the popular WAIS search engine or PRISE, a full text retrieval system being developed at NIST that uses fast and space-efficient indexing and searching algorithms[ROGERS]. Traditional text query languages have a syntax based on Boolean logic. WAIS and PRISE, on the other hand, allow unstructured natural language queries. The PRISE search engine uses a statistical ranking technique to retrieve the documents in the data set that best match the query. Studies have shown that users often get satisfactory results faster with an information retrieval system supporting natural language queries with statistical ranking than than they would using a Boolean-based system[HAR]. Efforts are underway to extend the NSSN testbed implementation to automatically determine from a query which standards database to search. The testbed's developers have also implemented a ``web crawler'' tool for maintaining and indexing hyperlinks to standards resources on the web.
The NSSN and the APIB gateway are similar in that they both provide a means for users to quickly obtain useful standards information. They are different in their scopes and in their levels of granularity. The APIB's scope is limited to STEP while the NSSN is general in nature. Unlike the APIB gateway, whose architecture is tied to the STEP DTDs, the NSSN's architecture can accommodate any electronically available standards resource. However, the NSSN's generality comes at a price. Because it does not assume any particular predefined structure for its standards documents, its architecture cannot easily support ``STEP-centric'' search interfaces such as the APIB Browser and IR Query interface discussed in Section 3.
Full text search engines such as the PRISE system used in the NSSN prototype could be useful in the APIB. PRISE's support for natural language queries can potentially provide an attractive alternative to the SGML-structured queries supported by Pat. For example, the APIB gateway allows users to search for EXPRESS entities (or other schema objects) by specifying a substring of the object's name. It would be desirable if the APIB gateway also allowed users to search for EXPRESS entities by specifying a natural language description of the entity. PRISE could match this query against each entity's ENTITY.DESCRIPTION element (Figure 10 shows the ENTITY.DESCRIPTION element for the entity dimension_curve_directed_callout).