News

New Google –like search technology is curbing course dropouts

By Meris Stansbury, Managing Editor, @eSN_Meris
September 1st, 2015

Brilliant minds at the New Jersey Institute of Technology (NJIT) are developing an intuitive course-based multimedia search engine for students and faculty; say personalized search capability on the near horizon.

search-multimedia-UCSIn the wake of the multimedia resources boom in campus courses across the world, sifting through hour-long recorded presentations or hundreds of pages of online text to find specific information can be like trying to find the keyword needle in a massive academic haystack. But that’s about to change.

In an effort to help both students and faculty better categorize and access information-rich multimedia resources—with the ultimate aim of improving learning—computer science and software development experts at NJIT are in beta for their creation called Ultimate Course Search (UCS): an open source Google-esque search tool that sifts through course-generated multimedia to find specific keywords.

“We know how to search fundamental data, and we do this on a daily basis. But how can someone look up an image? Or a specific slide in a online presentation? Currently, the only way to search these multimedia objects are within a folder,” explained Vincent Oria, associate professor and chair of the computer science online program at NJIT. “We wanted to see if we could build a system like Google locally for students and faculty.”

Oria said UCS, which is funded by the National Science Foundation (NSF), aims to not only improve learning overall by giving students a better way to access information, but improve the online learning experience as well.

“Right now, thousands of students are registering for online learning courses, but the dropout rate is incredibly high. If a student could, for example, type in a search term and pull up a specific reference within that online course, as well as any linked material within that online course that specifically mentions that keyword, perhaps students would get more out of the material.”

Using Ultimate Course Search

Using UCS is much like using Google Search, only targeted for students.

First, students are given a web link to UCS, and presented with an authentication screen for login. Once the system has verified the student’s authenticity, the student is asked to select their institution (since UCS debuted last year and is still considered a prototype, only students at Montclair State University part of the beta study have access). Once an institution is selected, the student then chooses their course.

For example, say a student chose his/her course on Career Counseling. The student would then type in their desired keyword—for instance, “community college”—then select either the “Slides/Video” tab or “Textbook” tab to narrow the multimedia search.

If the student chose “Textbook,” search results would present a scroll-down list of hyperlinked references in one book or multiple books where the keyword was mentioned. Clicking on one of the hyperlinks on the left-hand side of the page, the student would then have a scroll-able box to the right, showing the scanned online pages of the book where the reference is made.

UCS1

Search results under Textbook. (Click for a larger view)

“Because of copyright agreements with textbook publishers, we can currently only provide a scroll-through of around 5 pages before and after where the specific reference is made within the text,” said Hardik Dasadia, software development intern at ADP NJIT and NJIT MS graduate student. “The text pages are also watermarked for security and inhibit copy and pasting. But there is a print option for students.”

If, however, a student chooses the “Slides/Video” tab after entering the keyword, presentation slides, as well as any video presentations mentioning the keyword, are listed. Presentation slides are also listed in order of relevancy.

UCS2

Search results under Slides/Video. (Click for a larger view)

“We determine relevancy of information by how many times the keyword term or phrase is mentioned, as well as where the term is mentioned,” noted Dasadia. “For example, if the term is within the title, it counts for more in the system. It’s a numerical statistic called TF-IDF, or Term Frequency-Inverse Document Frequency, and it’s the same statistic used by Google.”

Is it Helping Learning?

So far, NJIT has conducted two beta studies—one in the 2014 fall semester and one this past spring—to determine how students use the search tool, and whether or not it has any impact on learning outcomes.

According to the studies conducted by Dr. Edina Renfro-Michel, LPC, ACS, associate professor of Counseling, Dept. of Counseling and Educational Leadership at Montclair State University, the difference in attrition rate for students who used the search tool was noticeable.

“In the fall beta, we asked students in our experimental group to use UCS. Those in the control group did not have access,” she explained. “Though both classes are fully face-to-face, they have many multimedia resources. The difference in the attrition rate was this: Only 2-3 students dropped out in the class that had the tool, while 50 percent dropped out of the class that didn’t have access to UCS. To me, that’s staggering.”

What’s even more interesting is that in the second beta conducted this past spring, the study results were skewed because students started sharing their UCS login with the control class.

“We had to move to a hybrid study model for the spring beta because the students who had access to UCS started giving their login information to help the students in the control group who didn’t have access! So, our data for our research is a bit skewed at this point, but it’s a good indication that students really appreciate UCS,” she exclaimed.

She also noted that UCS is not an extra time-suck for faculty. “We didn’t have to change the way we uploaded our multimedia materials,” she said. “We just uploaded into the system and UCS took care of it.”

Renfro-Michel aims to have a paper published on data collected from the beta studies within the next few months.

Big Implications for the Future

Though Dasadia says the next version of UCS will allow students to search by all courses at all participating institutions, the UCS team is also working on finding search terms via text-to-speech.

“This means that a student could enter a search term and the system could locate the exact moment in a video or verbal presentation where the term was said aloud,” he explained.

Oria says that though UCS is robust (doesn’t break when used frequently), the team must now figure out how to turn the software into a marketable product.

In the meantime, building off of the technology used in UCS, a team composed of NJIT and CUNY faculty (Dr. Soon Ae Chun, associate professor, CUNY-Staten Island; Dr. James Geller, professor, NJIT; and Dr. Reza Curtmola, associate professor, NJIT) is working on a different search tool that could be considered an added functionality of UCS once it’s further developed and tested. The tool is working on search via ontology, or the taxonomy of ideas.

“Think of it as an umbrella of terms,” said Dasadia. “If the user searches for “denial of service” and can’t find the exact phrase or term match, the tool will run through an ontology of descending terms; meaning that the tool will skip down to the next closest idea to the term or phrase and provide resources, continuing to descend down the taxonomy list until something is found for the user.”

Search via ontology. (Click for a larger view)

Search via ontology. (Click for a larger view)

But even providing ontology-based search functionality is merely scratching the tip of the search technology iceberg, said Oria.

“The future of search, just like learning overall, is heading toward personalization,” he emphasized. “By this, I envision a student posing a question into the search system and the system giving a personalized answer. The answer will be personalized not just based on relevancy of material, but on the system recognizing the student’s unique learning preferences and tailoring the search results to that individual student; in other words, presenting the information to the specific student in the modality and format that works best for him or her.”

He concluded that this personalized search technology may be seen in classrooms within the next five years.


Add your opinion to the discussion.