Yewno Q&A With Dr. Berthold Gilitzer Of The Bavarian State Library

Yewno Q&A with Dr. Berthold Gilitzer of the Bavarian State Library

1. What do you think is the greatest challenge facing librarians today?

The main challenge is this:   How do libraries fit into the digital change in our culture? This change is characterized by breakup of large information units and the domination of a stream of tiny information units. It is in direct opposition to the large forms of information libraries and librarians traditionally dealing with: books, articles, manuscripts and so on. If librarians don’t want to lose contact with this changing culture, they have to find a way to mediate between the fragmented modern digital culture and their main corpus of content in large units. Likewise, libraries must find ways to integrate the constantly generated smaller parts of fragmented information with traditional content.

2. Do you think that the current methods available for cataloguing digital information are sufficient?

Libraries have done a lot of things to make digital content retrievable and accessible for their patrons aside from traditional cataloguing.  However, even full text indexes and electronic full text search are not sufficient to bridge the gap between traditional content and the more fragmented, smaller pieces of information that we addressed in your previous question. All too often, the user gets lost in the huge results lists they receive using library catalogues or even conventional discovery systems. The presentation is not structured in a way to get an overview over the amount of hits, and even relevance ranking is often of only little help. There are serious problems for both users and the librarians with the current methods available.

3. How might these methods be improved and how do you think AI can help?

If anything really helps, then AI is the best candidate. We need more clever search engines which are a kind of a partner of the user. One first goal is to go deeper into the text which means: the search engine should know what the text is about and it should know that in considerable detail. This cannot be reached by human intellectual power and traditional subject cataloguing. The amount of information is by far too big to reach this target by manual work. Only if the machine is able to “read” the books and articles and to identify what they are about exists a prospect to be successful. The other goal is a dialogue between machine and user which presents the results in a much more intuitive way, so that the user can find the information he needs. The user must be able to have a dialogue with the machine about what he is really searching for and what the documents are really about.

4. Do you think that living in a digital age, with a growing trend toward shorter form and unstructured content is compromising the quality of University-level research?

Yes, this seems to be the case. I sometimes hear complaints from university teachers about a lower level quality of bibliographic research from their students, and about problems getting students to search for really relevant material.  The consequence – at least for humanities or social sciences – is that this affects the quality of the scientific work itself. However, improvement is possible. If we are able to make our search tools more efficient and more clever, users would be guided to more relevant books or articles that they otherwise would have no chance of finding. The hope is that they would even go beyond the scope of their original research to discover thematic connections and relations they otherwise would not have detected.

5. What role do you think that visualisation can play in helping people uncover connections?

Scrolling through results lists is not an efficient way to perform research.  People do not usually get beyond the first ten results listed. In order to better uncover useful information, AI should present a visually understandable overview of results to give users a choice of which information might be truly valuable by showing connections between concepts and letting the user decide which connections to pursue. The visualisation in the graphic display of interconnections as Yewno does is one proper way to solve this problem.

6. Why is original language text considered to be more valuable to researchers?

Original sources are often only available in original language. In Europe, the vast majority of text is still produced in a multitude of languages despite of the fact that English dominates the areas of information technology and medicine. While original texts can be digitized eventually, automatic translation is lacking, some things including emotion and semantics get “lost in translation” as the expression goes.  At least in the humanities and social sciences, original language text is indispensable and intelligent discoverable multilingual library collections will be crucial factor in facilitating original language research.

7. How does Yewno Discover and its recent introduction of Multilingual content make research easier?  More fun?

Detecting connections across the boundaries of languages is the crucial step towards making Yewno a serious tool in the context of a multilingual society, as I regard the community of scientists. Yewno has shown that it can work with very different languages by integrating German and Chinese, and we look forward to other languages following, specifically French, Italian, Spanish and Eastern European languages, as we have a very important special collection relating to this topic at our library. Yewno offers the possibility to discover hidden interdisciplinary connections within texts that could only be surfaced by machine learning. Adding multilingual capability adds another dimension to that value.

Jun Ge