UNT to turn plant specimens digital

by Chris Speight of North Texas Daily

UNT received a grant for more than $700,000 in September to investigate the digitizing of plant specimens, called herbarium sheets, found at the Botanical Research Institute of Texas.

UNT's Texas Center for Digital Knowledge is collaborating with the Botanical Research Institute on the project.

Herbarium sheets are the records of plant specimens. Each specimen is flattened and stored on a sheet of 11-inch by 17-inch, acid-free archival paper. Label data is included on each sheet. The data shows where and when a certain specimen was found, who found it and what type of botanical classification it is.

William Moen, the director and primary investigator for the Texas Center for Digital Knowledge, said the $738,075 National Leadership Grant from the Institute of Museum and Library Services was presented to NT because of the connections and mutual interests shared by the university and the institute.

Jason Best, information technology manger of the institute, said the goal of the project is to maximize the efficiency that people interface with when using the Atrium Biodiversity Information System, which is the official name of the software being used to scan the plant specimen sheets into the computer.

The Botanical Research Institute of Texas houses more than one million plant specimens and is the largest independent herbarium in the Southwestern United States. Some of its specimens date back to the 18th century.

Moen said researchers will scan millions of archival herbarium sheets that hold all the pertinent data about a particular plant specimen to get a digital image of the specimen and its label data.

Once a digital image of the whole sheet is captured, software called Optical Character Recognition translates the text from the specimen label into a form that can be read by a computer.

Best said that because there were no computers when many of the specimens were first collected, much of the label data was hand written or typed.

"Optical Character Recognition is only successful translating approximately 41 percent of the data from these handwritten sheets, leaving a lot of work for the software's human counterparts," Best said.

According to the project's outline, the goal is to figure out what interface method is the most effective in converting the data from the herbarium sheets into computer data.

The project starts Dec. 1 and is expected to take about two years to complete.

Moen said he and his colleagues are preparing to release applications that will lead to hiring three graduate students to work with Moen on the project.

"The knowledge and skill set will limit the selection process mainly to students in the College of Information and Library Science and Technologies, the College of Engineering and the College of Arts and Sciences with a biology discipline," Moen said. "Aptitudes and attitudes of the student we choose have to be successful."