English Query stores a large amount of information about the vocabulary and semantics of a speech database when you build an English Query interface in a format called Semantic Modeling Format (SMF). SMF is the native persistence format of English Query 7.5, a product that will ship in beta as part of the July drop of the Microsoft Developer Network, as well as with SQL Server 7.5. For users of the existing English Query 7.0 product, a conversion utility called eqp2smf.exe is available to convert English Query 7.0 projects (.eqp and .eqm files) into SMF format. You can download it from http://www.microsoft.com/sql/70/gen/smf.asp.

    SMF is an open XML grammar that stores all of the entities (such as customer, product, or store) in a database. Many other items are also stored, like words and synonyms that apply to those entities (such as purchaser, item, or outlet), instances of those entities ( John Doe, widget, or Seattle), relationships between entities (between customers and products, between stores and cities) and phrasings ("customers buy products," or "stores are in cities") for those relationships.

    All of this information can be used by speech recognition software to bias its recognition to the words present in the SMF, as well as by using the phrasing information to help syntax-based analysis. The specification for the XML grammar implementation of SMF is also available at the URL listed previously.

Lernout and Hauspie Voice Express Pro 4.0

      Lernout and Hauspie will provide support for training from SMF files in Voice Express Pro 4.0, a new product that's being announced at the May 1999 TechEd conference. You can use SMF to train Voice Express Pro by executing the following steps.

    If you're using English Query 7.0, convert the English Query project's .eqp and .eqm files into SMF by running the command:

     EQP2SMF <basename of your project>

    If using the English Query 7.5 beta from Visual Studio, invoke Project | Export SMF File for the project from which you wish to train Voice Express Pro.

    Start up Voice Express Pro and select Vocabulary Extender. Click Documents | Add Documents, then select SMF Documents from the dropdown list of available document types. Point to the location where the SMF file you created is located.

    Voice Express Pro will extend the vocabulary of a user's profile based upon the words and phrases available in the SMF. Questions entered via speech input should operate much more accurately after this vocabulary extension has occurred.

    When you deploy your English Query application with Voice Express Pro for speech recognition, you should plan to distribute the SMF file with it. Since the vocabulary extension only applies to the profile of the user who executes this operation, you will need to make the SMF file for your English Query project available to all users. By running the Voice Express Vocabulary Extender on this file, users can get more accurate English Query question performance.

    Voice Express Pro is a great example of how English Query can be used not just to provide plain English database searching when you type in a question, but also when speech recognition is added to your application. By utilizing an XML-based SMF database in your English Query-enhanced application, you'll be able to respond more accurately to your users' spoken input.