The word 'mashup' may bring to mind references to the popular television show, GLEE, in which two or more songs or rhythms are combined to create a more interesting, catchier tune. In fact, this same concept is being applied to the web, where thousands of end users are remixing data sources to create more interesting and relevant data for use by themselves and their communities.

Imagine reading news from the New York Times, CNN.com, CBS, and MSNBC, where the articles only interest you, and you never have to switch to a different website. Or imagine searching for a new house using only one website that aggregates data from every realtor in your area onto one map. That's the power of a mashup.

Mashup programming environments are popping up to help end users to create mashups that tailor and individualize data streams. This means that the power of creation is in the hands of the end user. However, the mashups created by end users are often littered with errors and deficiencies that can make them error-prone and hard to understand. Further, users often 'reinvent the wheel' by creating mashups that have the same functionality as mashups created by other users.

We have designed techniques to simplify the structures of mashups and make them easier to understand, as can be shown by the mashups in the figures below, which were built in the Yahoo! Pipes environment. A more understandable and simpler structure, such as the mashup shown on the right, can lead to increased reuse and protect against common errors from external components, such as websites that no longer exist.

We have designed a tool prototype (grey bar at the bottom of the images) that identifies common errors in the mashup being viewed or created by a user and automatically restructures the mashup to address the issues. Our techniques can be applied to many environments that have the structure of a 'pipe-like mashup' where there are many data sources that create an output. Our study results show that the new mashups are easier for users to understand, easier to maintain, and more preferable among end users.

For more information, including more screenshots and sample code, contact Sebastian Elbaum or Kathryn Stolee at the University of Nebraska-Lincoln.

When we say "spreadsheet", we often think of users doing calculations. But, in fact, many people just use spreadsheets as a place to organize and store textual data like phone numbers and university room numbers. These same kinds of data show up in many end-user programs, including web applications created through Visual Studio Express Edition or web macros created through CoScripter.

These textual strings often contain typos and other errors. For example, "KED 1148" isn't a valid university room number that should be "KEC 1148". Excel doesn't even recognize that phone numbers never end with "11". Since Excel and other tools don't know about these domain-specific kinds of data, they can't find the errors.

So we decided to teach Excel some new tricks--or, more specifically, we provided a new data model that ordinary end users can use to teach Excel about domain-specific data.

This new data model is called "topes", and it not only describes how to find typos, but it also enables end users to automatically reformat data. For example, given a tope for university room numbers, Excel could quickly reformat thousands of strings like "KEC 1148" so that they look like "Kelley Engineering Center 1148".

Once a user has taught Excel how to recognize a new kind of data, the model can be saved and loaded into other end-user programming tools so that they, too, can "speak the user's language". For example, we have worked with IBM to integrate topes into a new programming tool called Vegemite that enables users to rapidly download and manipulate data from the web. We are now working with two other companies who want to integrate topes into decision-support software in 2010.

For more information, including more screenshots and sample code, contact Chris Scaffidi at Oregon State University.