Introduction to SemanticMeasure

In documents on the web and elsewhere, units of measure are embedded in text. The text does not store the information semantically, that is, according to its meaning. As far as a computer is concerned, "10 feet" does not mean the same thing as "3 meters" or "120 inches". That is where SemanticMeasure comes in. This software can scan text to infer just that kind of meaning, and it can then output the information in a variety of different ways.

The purpose of this application is to demonstrate my abilities with .NET development and with solving technical problems. All code was written in C# by Carl Sorenson.

The application

• See it in action!
• Manage the list of units of measure recognized by this application

Technical features

• Transform page
  • I use regular expressions to find instances of measurements in a sample of raw text. Any measurements in the input text are converted into XML.
  • The regular expression strings are constructed from the information in the user-editable DataSet.
  • Either of two XSLT stylesheets can be used to render the XML into HTML for customized viewing.
  • I implemented IComparer for sorting.
• Managing the list of units
  • I use an ADO.NET DataSet to store and retrieve the basic information about the units and their conversion factors. The default entries are retrieved from an XML file. The DataSet uses two tables that are connected by a DataRelation to enforce constraints and enable lookup. You can modify the entries in the DataSet, creating your own units and using them on the Unit Converter page. (The XSLT stylesheets will not recognize the units, however.)
  • The DataSet is displayed and edited in two GridViews. The connection between the GridViews and the DataSet is through two ObjectDataSource objects.
  • Custom validators make sure that duplicates are not added.
• Limitations and possible improvements
  • The XSLT stylesheets do not yet allow the option of displaying results in American spelling, British spelling, or symbols.
  • The XSLT stylesheet could be automatically created from on the information in the DataSet. This would allow the user's changes to be reflected in the transformation (they are currently only reflected first stage - conversion to XML).
  • The XML parser can't handle large numbers.  That means that while the regular expression will recognize "150 million km," it will show up as NaN when it is transformed.
  • The error handling and validation are not robust.

Source code

You can download the project to view the source code.

Copyright 2006 Carl Sorenson