Hi Ian,

Thanks for your reply to this blog post. I think you’ve raised a really important point here. Your example is an interesting one, since it demonstrates a search which has different meaning to different people – this, I believe, is the crux of the issue. If I as an analytical chemist searched for caffeine, I might be interested in particular spectra or techniques. Similarly if I as a food scientist searched for caffeine, I may be interested in how caffeine is metabolised. A free text, or Boolean search does not understand this and will also include results for anything that mentions caffeine or fits my set of Boolean operators, e.g. shopping sites, coffee shop locations, blogs, health information, news articles, images etc.

Drilling down into a specific result from 59 million hits can be daunting for the user – however most users will eventually feel by tweaking their initial search they have reached an optimum conclusion and will feel no sense of loss for the other millions of results not considered in that particular search.

Your suggestion of meta-tagging the ‘type’ of article touches on a larger movement by search engines to semantic searches and entity mapping. This is to try and add ‘meaning’ to free text searches and hence to what individuals and audiences find valuable in their search results. A recent example of this kind of mark-up can be found at http://www.schema.org, this is a collaborative project created by Google, Yahoo and Yandex which uses rich snippets to add further information to documents to be read by the search engine. For example I could mark-up the article type and author using a predefined set of tags. An excellent example of this in action can be found on the White House – Veterans Job Bank (http://www.whitehouse.gov/blog/2011/11/07/open-innovation-heroes-introducing-veterans-job-bank), which uses a predefined schema to mark-up its job listings.

Extending a simple set of tags to incorporate information which could be of use to the scientific community could allow a search engine yield more valuable results from their search. This could be particularly useful when considering the standard formats for academic documents. The formal abstract, intro, references etc are rigorously structured, however the data within these sections, such as the reagents, methods, and results, are inherently unstructured - therefore one could highlight relevant information using a semantic mark-up. However there are a number of issues: How specific should these tags be (article type, scientific area, method, reagents)? What is the relative worth to the person performing the search? Who should be responsible for tagging the content? How should these tags be standardised?

At the RSC taken some small steps towards semantic search, such as the synonym function on ChemSpider. I think at the moment these concepts are very much in their infancy. There is an obvious balance which would need to be achieved by marking up data; such that relevant results become more discoverable to the individual and that the task isn’t completely time consuming. In addition to marking up documents effectively, it's also critical that the wider scientific and web community can see the value and therefore adopt a best practice - and ultimately that the search engines are also on board.

You mentioned in your reply that part of the solution rests with authors, I’d be interested to hear your thoughts on how we could use this mark-up, or any other methods to make results more valuable to users.

Jack
e-Publishing Specialist, RSC

Retrieving relevant information is a nightmare. Using your "caffeine" example as a keyword, I tried Google and found there were "about 59,000,000" strikes. Ideally, we would want a search engine that operates correctly on Boolean logic (Google does not) but I am not sure how we get that. The suggestion that the RSC optimizes output to promote its ranking is partially sound, however what happens when everybody else does the same thing? How are rankings promoted?

In my opinion, part of the problem, and part of the solution, rests with authors. What authors need is some readily available description of how tags (and similar search assistance terms) work, and how to use them more effectively. Besides tags (at least as I understand them, which is not all that well - I need help too) we need some sort of classification system by which authors can indicate what sort of an article it is. Thus epidemiologists do not need to be flooded with articles on chemical composition of various plants, etc. If there were some sort of acceptable classifier that set the type of article, or the type of information, and everybody agreed to use it, that should make searching easier.