Home > Articles > Data > SQL Server

SQL Server Reference Guide

Hosted by

Toggle Open Guide Table of ContentsGuide Contents

Close Table of ContentsGuide Contents

Close Table of Contents

English Query - Part 4

Last updated Mar 28, 2003.

In the last tutorial in this series I started our example in English Query using the Wizard to build a project. We'll return to that project today and flesh out the back-end solution.

On my test system I've opened the Windows Start menu and selected the All Programs | Microsoft SQL Server | English Query | Microsoft English Query item. I then clicked on the Existing tab and opened the Pubs folder and then the Pubs.eqp file. This file is the project that I created in the last tutorial. If you selected a different name or location for your sample project, open that file and you will see a screen similar to this one:

Other than the menu and icon bars at the top of this screen, there are four parts within this interface that we will be using. On the far left side is the document outline. In this section we'll have access to all the objects within the semantic model. In the center of the screen is the primary work area for the active object. Right now it's blank. At the bottom of that is a Task List, which shows the items left to complete the project. It's blank at the moment, but as we go along the system will populate some values and you can add and edit more at any time. If you work like I do, you might make an outline of the things you need to do so that you don't forget anything as you move along. The Task List is a good place to keep that outline. On the right-hand side of the screen are the Project Explorer which contains all of the files within this project, and the Data Explorer which shows the connections for the project.

In this graphic I've double-clicked the Pubs project in the Project Explorer and then double-clicked the Pubs.eqm file within it. Next I expanded the objects in the main work area, and then dragged the author Entity to the right hand side of that pane.

In the last tutorial I said that you can use the Wizard to create the objects and then you can edit them later. If you double-click the author Entity you can edit the Semantic and Physical or Database properties for it.

This is the main entrance for working with Entities. In the Words box you enter the names and synonyms for the Entity, and in the Entity type box you set whether the Entity is a who (people), where (location), when (time) or measure (numeric) noun. Setting the attribute of the Entity properly buys you all kinds of answers for "free." That means that the engine has extra intelligence built-in to extrapolate questions and answers for these kinds of nouns. I'll reiterate that taking your time with the design in English Query will make or break your project. Although it may seem tedious, you should be as thorough as you can.

On the right-hand side of this panel are the database options. In this section you tie out the Semantic object (nouns, in this case) to the Physical Database objects (a table or fields). If you find that your database design does not fit the Semantic model, you may have to create objects in your database and populate them using triggers or other logic. If you find yourself having to create many objects, you might want to consider creating a separate database for the English Query source, and populate it using Extract, Transform and Load (ETL) procedures similar to a Data Warehouse.

At the bottom of this panel are three buttons. The first shows the relationships that have already been defined on the Entity. In our case, the Wizard built the following relationships:

The next button allows you to create new relationships. If you click that now, you should receive a message that there are no new relationships possible. That's because we added all possible relationships for authors during the Wizard process.

The last button deals with advanced settings for the Entities. In our case these are unavailable, since the author Entity doesn't have these kinds of properties.

In this graphic, I've closed out the property panel and then dragged the author_names_are_the_names_of_authors relationship to the right-hand part of the work area:

If you study this screen you'll see that two Entities (author and author_name) are linked via this relationship. Once again you can double-click the objects in the right-hand side of the work area to bring up the properties for that item. I've double-clicked the relationship item to bring up the following panel:

In this example, the Phrasings area connects author and author_name with a Name/ID Phrasing type, as I explained in a previous tutorial. This means that authors have names, and names are held by authors. The reason for this two-way thought process is due to the English language and the way you can ask questions about things. Your users might ask what the name of an author is or they might ask which author belongs to this name. You'll see that relationship as we ask the system questions.

By clicking the Add button on this panel we can create more relationships. In this graphic I've done that and then highlighted the Subset Phrasing relationship:

This is useful in cases where things might be related, but not in every case. For instance, my mother doesn't have a middle name. If I wanted to include that kind of question (what are the authors' middle names?) I'd want to set this kind of relationship on the author and middle_name Entities.

I'll cancel out of that selection for now and click on the Database tab in the Relationship properties panel.

Here you can set the joins for the relationship. Since the pubs database is normalized, you don't have to take that step here. You can also set some Transact-SQL (T-SQL) that creates a condition for the relationship. If you're careful when you create your model, you can have some really interesting results using this one feature.

I've closed the relationship properties panel and rearranged my screen a little to show you another view. In this graphic I've dragged another relationship onto the work area, this time the authors_are_in_author_cities relationship.

You can actually double-click any of the objects in the left-hand side of the work area without first dragging them to the right-hand side, but placing them in this area allows you to see the relationships between the Entities. This comes in very handy when you're troubleshooting later on. Without some sort of graphical view it's difficult to see why a certain question returns no answer or the wrong one.

You've now seen most of the major parts of the screen. Let's see what our model can do so far. In the menu bar, I've selected Debug and then Start. After a moment, the system is ready to run an English Query sample application. As I mentioned at the top of this series, you have several options for writing a front end applications for this project. This is one that Microsoft includes in the product.

At the top of the panel it presents you'll see a Query box. In the query box, I've entered the text "Who are the authors?" and I've clicked the icon that looks similar to a table. Setting this option displays the results from the database. With that all set, I pressed Return.

The first box in this panel shows the command sent to the system. The second box shows the answer header the users get. The third shows the T-SQL commands sent to the database, and the final box (if you select the table icon) shows the results from the database.

Now you can really see the power of this system. Try other questions like "What are the author's names?" and "What are the names of the authors?" Keep up this process until you find a question the system can't answer, and note that question down. That becomes an action-item for you to work on later.

If you click on the Analysis tab you'll be able to tell which Entities and Relationships the system is using for the answers.

Although you can find out the ID numbers of the books (try "What are the books?"), you won't get an answer when you ask who wrote a particular title. That comes next week.