? A companion video is hosted at https://youtu.be/Ld1s6mEpA2Q.
Congratulations to the organisers for the huge success of the Bibliometricenhanced Information Retrieval conference series. Organising even a single workshop is di cult, and a series of ten is a major achievement. Workshops help to forge people into a community and de ne a eld, so this series has had a lasting impact on academic research.
Scholarly information retrieval is still a major challenge that still needs more work. I teach literature searching to maths and computing students and conduct frequent literature searches. Despite this experience, I frequently must remind myself that literature searching is more di cult than it seems.
The rst thing that I tell my students is that nding relevant research is di cult, even though Google Scholar (their main search tool) gives a lot of help. For a beginner, a key task is learning the terminology of a eld, which is often not straightforward and is essential for good literature searches. I give the example of Google. To conduct a literature search for the technology behind a search engine like Google, the keyword Google is not particularly useful, neither is the more general phrase \search engine." Instead, after reading into the topic, the students will discover that the name given to the eld producing the core research behind (the search part of) Google is information retrieval, so this is a good term to search for (although it gets too many results). I think that it is a major challenge to develop an information retrieval system that can accommodate beginners in this initial task: translating the words that they would use to describe a topic into the jargon used within the eld.
Using the wrong terminology has caught me out many times when exploring unfamiliar elds. Once I submitted what I thought was the rst article on a topic, but a referee pointed to a monograph and an edited volume of collected papers from a workshop, all devoted to the topic but using di erent terminology to me. My basic mistake is to deduce that there is no relevant research if I have used reasonable keywords to search for it and got no results.
Once a beginner has learned the jargon of a eld there are still major search challenges. Article titles can be misleading, important results can be in articles with a di erent focus, and there may be far too many relevant articles to read individually. I tell my students to look out for productive scholars in a eld and read their most cited work, to look for review articles, to nd highly cited articles, and to citation chain from core articles to nd more recent updates. These mostly exploit bibliometric information to reduce the amount of reading needed. One interesting challenge is that authors know the topic of their article so well that they may not be the best people to decide on a relevant title. Usually a title makes sense to me after a paper has been read but sometimes the content is not clear before then. In addition, whilst in some elds titles often directly describe the core outcome (\IL-1 receptor blockade restores autophagy and reduces in ammation in chronic granulomatous disease in mice and in humans"), in others titles might be funny (\Another one bites the dust: faecal silica levels in large herbivores correlate with high-crowned teeth"), speculative (\Why does unsupervised pre-training help deep learning?"), or include an illustrative quote (\'Do I really have to eat that?': A qualitative study of schoolchildren's food choices and preferences"). This makes the titles more interesting but adds to the human challenge of identifying relevant documents and adds greatly to the information retrieval challenge of nding and ranking documents matching the searcher's information need.