Category: Aggregation

The Open Citation Index

Eugene Garfield, one of the founders of biliometrics and scientometrics, once claimed that “Citation indexes resolve semantic problems associated with traditional subject indexes by using citation symbology rather than words to describe the content of a document.” This statement led to the advent and a new dawn of Web-based measurements of citations, implemented as a way to describe the academic re-use of research.

However, Garfield had only reached a partial solution to a problem about measuring re-use, as one of the major problems with citation counts is that they are primarily contextless: they don’t tell us anything about why research is being re-used. Nonetheless, citation counts are now at the very heart of academic systems for two main reasons:

  • They are fundamental for grant, hiring and tenure decisions.
  • They form the core of how we currently assess academic impact and prestige.

Working out article-level citation counts is actually pretty complicated though, and depends on where you’re sourcing your information from. If you read the last blog post here, you’ll have seen that search results between Google Scholar, Web of Science, PubMed, and Scopus all vary to quite some degree. Well, it is the same for citations too, and it comes down to what’s being indexed by each. Scopus indexes 12,850 journals, which is the largest documented number at the moment. PubMed on the other hand has 6000 journals comprising mostly clinical content, and Web of Science offers broader coverage with 8700 journals. However, unless you pay for both Web of Science and Scopus, you won’t be allowed to know who’s re-using work or how much, and even if you are granted access, both services offer inconsistent results. Not too useful when these numbers matter for impact assessment criteria and your career.

Cartoonstock is the source of Hagen Cartoons’ Struggling scientists.

Google Scholar, however, offers a free citation indexing service, based, in theory, on all published journals, and possibly with a whole load of ‘grey literature’. For the majority of researchers now, Google Scholar is the go-to powerhouse search tool. Accompanying this power though is a whole web of secrecy: it is unknown who Google Scholar actually crawls, but you can bet they reach pretty far given by the amount of self-archived, and often illegally archived, content they return from searches. So the basis of their citation index is a bit of mystery and lacking any form of quality control, and confounded by the fact that it can include citations from non-peer-reviewed works, which will be an issue for some.

Academic citations represent the structured genealogy or network of an idea, and the association between themes or topics. I like to think that citation counts tell us how imperfect our knowledge is in a certain area, and how much researchers are working to change that. Researchers quite like citations; we like to know how many citations we’ve got, and who it is who’s citing and re-using our work. These two concepts are quite different: re-use can be reflected by a simple number, which is fine in a closed system. But to get a deeper context of how research is being re-used and to trace the genealogy of knowledge, you need openness.

At ScienceOpen, we have our own way to measure citations. We’ve recently implemented it, and are only just beginning to realise the importance of this metric. We’re calling it the Open Citation Index, and it represents a new way to measure the retrieval of scientific information.

But what is the Open Citation Index, and how is it calculated? The core of ScienceOpen is based on a huge corpus of open access articles drawn primarily from PubMed Central and arXiv. This forms about 2 million open access records, and each one comes with its own reference list. What we’ve done using a clever metadata extraction engine is to take each of these citations and create an article stub for them. These stubs, or metadata records, form the core of our citation network. The number of citations derived from this network are displayed on each article, and each item that cites another can be openly accessed from within our archive.

Visualising citation networks: pretty, but complex. (Source)

So the citation counts are based exclusively on open access publications, and therefore provide a pan-publisher, article-level measure of how ‘open’ your idea is. Based on the way these data are gathered, it also means that every article record has had at least one citation, and therefore we explicitly provide a level of cross-publisher content filtering. It is pertinent that we find ways to measure the effect of open access, and the Open Citation Index provides one way to do this. For researchers, the Open Citation Index is about gaining prestige in a system that is gradually, but inevitably and inexorably, moving towards ‘open’ as the default way of conducting research.

In the future, we will work with publishers to combine their content with our archives and enhance the Open Citation Index, developing a richer, increasingly transparent and more precise metric of how research is being re-used.

Moving beyond a journal-based filtering system

The amount of published scientific research is simply enormous. Current estimates are over 70 million individual research articles, with around 2 million more being published every year. We are in the midst of an information revolution, with the World Wide Web offering rapid, structured and practical distribution of knowledge. But for researchers, this creates the monolith task of manually finding relevant content to fuel their work, and begs the question, are we doing the best we can to leverage this knowledge?

There are already several well-established searchable archives, scientific databases representing warehouses for all of our knowledge and data. The most well-known include the Web of Science, Scopus, PubMed, and Google Scholar, which together are the de facto mode for current methods of information retrieval. The first two of these are paid services, and attempts to replicate searches between all platforms produce inconsistent results (e.g., Bakkalbasi et al., Kulkarni et al.), raising questions about each of their methods of procurement. The search algorithms for each are also fairly opaque, and the relative reliability of each is quite uncertain. Each of them, though, have their own benefits and pitfalls, which are far better discussed elsewhere (e.g. Falagas et al.).

So where does this leave discoverability for researchers in a world that is becoming more and more ‘open’?

Continue reading “Moving beyond a journal-based filtering system”  

Next page