An Interview with Umberto Anderle, BibliU's Head of Product

Over the office’s coffee machine, I speak with Umberto Anderle, Head of Product at BibliU about the e-reader that has been built over the last couple of years. This conversation focuses on the intelligent search function that BibliU has created.

The Importance of Intelligent Search

Umberto "Umbi" Anderle joined the BibliU team in London three years ago from Perth, Australia. He works on developing the product daily. With a slight tone of bemusement he describes his understanding of the archaic way that universities currently approach hosting textbooks on their platform.
"Universities are spending mammoth amounts of money on books. They have access to so much content, all of it — high-quality and reliable," he says.
"This quality and reliability is not something that Google has right now. But still," He pauses for effect. "students battling against a late night deadline will always turn to Google for a quick solution to a query they have about Aerospace equations, say. As a student, I certainly did."

"Why does this happen? Because universities are often so focused on getting all the most up-to-date academic content that they forget to put in a system that really allows their students to sift through it."

"Yesterday, I spoke with an American university whose library spend ten million a year on new content. Ten million! After they have amassed all this content, however, they all too often dust off their hands and just hand it over to their students."

"That’s crazy!" Umbi gesticulates. "Who’s going to read it? Who’s going to find it? It will just disappear into a black hole.’"

Speaking with Umbi, it becomes clear that the intelligent search was born out of a frustration that one couldn’t nimbly sift through academic resources using a Google-like search tool. But neither could one rely on Google to offer up search results which had a stamp of high quality.
Umbi sums up the role of a platform like BibliU as he jokes: "We provide an intelligent, Google-like search for students without the cat videos."

Often No More Complicated Than Ctrl + F

Other content providers currently provide a frustratingly basic keyword search. "Our competitors know that modern students want to quickly search through course materials," Umbi tells me. "But their solution is just a basic keyword search that is then bolted on to their digitised textbook."

‍
"Keyword matching," Umbi almost shivers as he says it. "basically takes the search term a student has punched in and finds all literal matches in a single textbook. It is not really any more complicated than Ctrl + F on a PDF ... it’s analogue."

"At its heart, BibliU’s a technology company," Umbi tells me. "While our competitors often begin from the physical textbook and then look to digitise it, our approach is different."

Umbi begins to list a whole range of features using technological language and various acronyms to describe BibliU’s intelligent search. He pauses to explain them more slowly to me.

"Ok, well simply we are just doing obvious things, things that a student expects to have with a content platform without a second thought," he says. "We’ve built a fuzzy match search. So if you type in 'chemstry’ accidentally we can work out that you really want to see search results for 'chemistry.'"

It works to more thoroughly understand the textbook, rather than simply reproducing a version of the print textbook behind a computer screen.

Our Technology’s Attempts to Understand the Textbook We Host

‍
Umbi continues: "Our search is intelligent. It works to more thoroughly understand the textbook, rather than simply reproducing a version of the print textbook behind a computer screen."

‍
Like Google, all the material that BibliU hosts is indexed. The platform aims to understand the textbook it hosts by categorising its chunks of information. It understands what is a heading or a Q and A section within the textbook. As a result it is able to relevancy rank the information. A summary piece of knowledge about synthetic biology e.g. might be more easily found at the beginning of a chapter, a more granular explanation in the middle.

Umbi says: "If you search for 'Genetic Engineering,' the search function understands that term as a topic that it needs to find in the textbook, rather than just a word."

"Searching “Genetic Engineering” might result in a thousand different results in a single textbook, right. We would be able to narrow down this search," he explains. "By treating it as a topic, we might surface a block of text that the algorithm has identified as an introduction to genetic engineering as the first search result. Our second search result could be an in-depth exploration of the area."

"In short," he adds, "we actually try to understand what you’re querying and then return the most relevant content based on the search."

The Search Functions Attempts to Understand the User

As we continue talking, Umbi explains that BibliU’s intelligent search is made more efficient by its attempts to gain an understanding of the student who uses it.

At this point, Tom — Head of Growth — has joined us at the coffee machine. He jumps in: "Yeah, what we do in a very basic sense right now is personalization. It’s about a deeper understanding of the user."

Umbi nods and describes how you can work out the year of study of the student and the level of knowledge they have already attained based on the material that they have read and the search terms they enter.

"A third-year chemistry student wants a different set of results when they search nanocrystals, compared with a first year student who types in the same term. Why not harness this knowledge?"

Tom replies, "The cool thing about the search function we've got is that a university that has signed its students up for a number of years to receive textbooks through BibliU’s platform allows us to get to know the individual learner far better and hence tailor the learning experience directly to them."

He continues: "We’re going to end up with an in-depth profile of the student as they progress year by year. We will know what they have read, how likely they are to have retained it and we can then provide them with information tailored to this profile."

Umbi replies back: "You can ultimately personalize it for students not just on their activity … but on the activity of their peers. We can understand which students are linked by looking at the textbooks they have in common — we can assume that they share a class. We can then improve the learning experience of each individual by looking at how their classmates have studied."

"If five students have already searched “Amorphous Alloys” and have ignored the first result, but highlighted and noted the third one. By the time the sixth student searches for the term we can reorient our search results."

By now, Umbi is not directing his answers to me, in his excitement he speaks directly and rapidly to Tom. This is textbook learning done just as any Google-user would expect to have it.

"Really, I am just creating something that I would love to have had as a student," Umbi says as he collects his coffee and takes it back to his desk.