Databases

MIT Software Greatly Simplifies Tough Database Querying

• By Dian Schaffhauser
• 08/05/16

New software from researchers at MIT could make even the most complex databases easier for non-experts to use. The program's home screen looks like a spreadsheet, and it allows users to create database queries and reports by combining functions familiar to spreadsheet users rather than forcing them to learn SQL or some other data scripting language.

The research project, underway in MIT's Computer Science and Artificial Intelligence Laboratory, lets people use drop-down menus to pull data into the tool from multiple sources; then they can sort and filter that data, perform integration and tuck away unneeded columns and rows. From those activities the program generates the database queries that deliver results.

The work is being led by Eirik Bakke, an MIT graduate student in electrical engineering and computer science, and his advisor, David Karger, an MIT professor of electrical engineering. Bakke's visual query system tool promises to help organizations tap into their data without having to wait for an expert to build specialized interfaces or forms.

As a research paper recently presented on the project explained, three requirements are "essential" to any alternative to standard SQL-based queries and tailored interfaces:

• Query specification can be done through direct manipulation of results, allowing the user to view results immediately upon each stage of the query;
• The user needs the ability to view and modify any part of the current query without departing from the direct manipulation interface; and
• The tool needs to be able to deliver SQL-like expressiveness, accommodating combinations of operations such as joins, calculations and aggregations.

Compared to Microsoft's Access database management system, the investigation found, Bakke's software provides far greater usability.

Bakke has named the system, "SIEUFERD" (pronounced soy-fird), which stands for "Schema-Independent End-User Front-End for Relational Databases." The tool's main drop-down menu has 17 entries, such as "hide," "sort" and "filter," that are rich enough to be able to reconstruct any database query that could be formulated with SQL-92, a common version of SQL that's taught in many database classes.

The researchers ran two user studies to test the viability of the program. In the first, 14 participants with a variety of technical and professional backgrounds were given "standardized tasks" aimed at assessing the initial learnability of the tool. No prior training was given; initial tasks served to train the users on subsequent tasks.

Some users were also given a chance to do more open-ended tasks on datasets, including some from their own organizations. In those cases they received demonstrations and further instructions for operating the program.

As one participant summed up, "I haven't used SQL that often in the recent past. Given [two] hours, I think I could make an accurate report in [SIEUFERD], allowing for mistakes, and fixing my mistakes. Take that same period in SQL, and I think I would still be at sea."

A second user study was intended to have people rate SIEUFERD against an existing database system, which is where Access came in. Again, 14 participants were involved, just a handful of which had any exposure to Access. Then participants rated their experiences using the System Usability Scale, a standard measure that allows the comparison of different types of software. The scores hovered around the 50th percentile for business software, which, the researchers noted, wasn't bad for an academic research project. Even then, that scoring was heads and shoulders above their experiences with Access.

"This supports the original hypothesis of our paper: database querying is hard, but can be made significantly easier using a direct manipulation interface," wrote Bakke.

At present, Bakke's tool enables query construction on an existing database, but it doesn't enable the direct entry or modification of data. That functionality is on the to-do list, along with additional features to be added and bugs to fix.

"It's almost ironic," said Karger, in a prepared statement. "Eirik's software is far more robust than just about everything that graduate students have built. But he's not satisfied with releasing it in its current form. He's aiming for something of commercial quality."