Towards the Chemical Semantic Web. An introduction to RSS.

Peter Murray-Rust^a and Henry S. Rzepa^b

^aUnilever Centre for Molecular Informatics, University of Cambridge, UK, ^bDepartment of Chemistry, Imperial College London, SW7 2AY.

Keywords: RSS (RDF Site Summary), XML (eXtensible Markup language), CML (Chemical Markup language), Semantic Web, News alerts.

It is appropriate after almost exactly ten years of Web-based publishing to assess what this revolution has meant for the dissemination of scientific and chemical information. Well in excess of 2 billion documents, of which perhaps 5% have some scientific content, have been generated (authored is perhaps putting it a trifle too strongly). To this one must add a limitless number of "virtual" documents generated dynamically upon request. Even within a relatively specialised discipline such as chemistry, the effect upon a human can be overwhelming; keeping up with the literature has evolved from a weekly browse of the tables of content of perhaps ten key paper-based journals to having to visit a much larger number of computer-based Web sites. It might be fair to say that many humans are not coping too well! Ironically of course, computers are much more efficient than humans at scouring a large number of information sources in an error, boredom-free and periodic manner; the challenge is merely in specifying what it is they should be on the lookout for. The first generation Web (circa 1993-present) turns out in retrospect to have been rather unsuited to this task. This is not actually the fault of the original design, merely a facet of how it was (or more accurately was not) implemented. The key omission was meta-data, this being a concise and structured declaration of the content model of the document.

Before this last topic is elaborated, its worth noting the "Google phenomenon", an index of the entire Web which for most scientists renders it actually useful as a scientific information source.¹ Thus the erstwhile need to constantly "bookmark" key locations found by a possibly accidental and probably not reproducible path, is now productively replaced by being able to rely on Google to find information based on a few choice key words. Google augments this with additional meta-information derived from choices made by earlier searchers using the same key words; in effect a coarse but gradually accumulating peer review mechanism which naturally selects the survival of the fittest information.

The original and still current Web, Google not withstanding, in practice contains very little overtly structured information in the form of declared meta-data; this may be simply the title of a document and little else. The human authors of most documents have either cared little or knew little about the relatively arcane mechanisms the original designers of e.g HTML had put in place to capture meta-data. This is something of a chicken and the egg issue; the average chemist has hitherto had no compelling reason to learn these arcane methods, preferring instead to focus on chemical applications and doing what they do best! The consequence has been that the Web has not fulfilled its anticipated potential for "serendipity", or the art of accidental and fortuitous discovery of unexpected information. This is still an act expected to be achieved by humans exercising their perceptive skills and not by computers.

At this point in the discourse we introduce RDF-based Site Summary or RSS (acronyms can indeed be successful at capturing the world's imagination, viz HTML!). This is a simple but powerful XML-based implementation of meta-data of which (like HTML) the arcane aspects can be hidden behind user-friendly software referred to variously as a news reader, aggregator or RSS client. RSS grew out of an idiosyncratic phenomenon known as Web logging. Web Logs (now known as "Blogs") are essentially personalized Web servers containing chronologically organised items of information. Four items of meta-data are implicit and indeed mandatory in a Web Log; the identity of the author, the date of each item, a brief description of it and how to link to it. The "blogging" community perceived the need to summarise Blogs using such meta-data and then to broadcast these summaries in a manner which allowed "aggregation" of themes into larger "channels" of such information, and thence into syndication; a procedure not very different from e.g. newspapers or broadcast television. This process in due course merged with another and rather more formal vision of evolving the homogeneous unstructured Web into a Semantic Web,³ where in addition to meta-data, information is carried which allows machines to process documents without the necessity of human intervention. The resulting fusion has resulted in a formal XML-based specification known as RSS², which is supported by ready availability of RSS-aware software which can act upon the contents of such RSS files. The remainder of this article shows an example of how RSS is currently applied and used, and follows with a call for its application in chemistry and a scenario of how it might be applied in this subject in the future.

Creating RSS

One way of regarding RSS is as relating to the management and description of hyperlinks in the same way that the earlier HTML (and more recently XML) markup languages relate to the management and declaration of content and data. RSS therefore augments a conventional Web site (or the smaller scale Web log), and can be added to a site without the need for any additional infrastructures or software. Many major news sites already do this, but it is still rare for chemically related sites. At the server end, deploying RSS requires nothing more than the addition to the contents of a Web site of two entries;

1. the creation of a document (by default variously named index.rss or rss.xml) which defines meta-data for a channel of related documents and links. The example shown in Scheme 1 follows the W3C-supported specification known as RSS 1.0², which includes support for RDF (Resource Description Framework), the latter being described as a "lightweight ontology system for supporting the exchange of knowledge on the Web".² We note here a competing (and simpler) XML-based specification currently known as RSS 2.0 (here the acronym stands for Really Simple Syndication) ² which does not include the RDF components; both are recognised by most current RSS software (but not all the components are necessarily processed).² We recommend here the use of RSS 1.0, since this has the greater potential for extension into domains such as chemistry, but the choice is not critical since the two formats can always be easily interconverted in the future using e.g. XSLT stylesheets.
2. In order to assist the robot-based "auto-discovery" of this file, an entry can (optionally) be made to the root document of the Web server (i.e index.html or equivalent) identifying the location of the RSS document

   <link rel="alternate" type="application/rss+xml" title="RSS" href="http://.../index.rss" />
   

In fact, the RSS document need not even be real; it can (perhaps even should be) generated dynamically by appropriate query of any content management system which is used to generate documents for the Web site. Like HTML, RSS can also be generated using simple authoring tools; a good one is to be found at http://rssxpress.ukoln.ac.uk/. An example of an RSS file is shown in Scheme 1:

<?xml version="1.0"?>
<rdf:RDF 
  xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
  xmlns="http://purl.org/rss/1.0/"
  xmlns:dc="http://purl.org/dc/elements/1.1/"
  xmlns:cml="http://www.xml-cml.org/schema/cml2/core" >
<channel rdf:about="http://www.xml-cml.org/">
<title>Chemical Markup Language</title>
<link>http://www.xml-cml.org/index.html</link>
<description>CML Highlights</description>
<webMaster>admin@cmlconsulting.com</webMaster>
<image rdf:resource="http://ww.xml-cml.org/cml.gif" />
<items>
<rdf:Seq>
<rdf:li rdf:resource="http://www.xml-cml.org/rss/" />
</rdf:Seq>
</items>
</channel>
<item  rdf:about="http://www.xml-cml.org/rss/">
  <title>This RSS file contains an embedded Molecule in  CML</title>
  <link>http://www.xml-cml.org/cml.rss</link>
  <description>Currently,  RSS  clients are not capable of acting 
  upon  the  cml namespace in the RSS descriptor, and  so ignore this 
  information. In future, one can anticipate that a  CML-aware  RSS client 
  will not only extract the molecular information, but be capable of 
  searching/filtering/transforming it  into more useful forms (ultimately 
  for example being able to detect  particular sub structures or other 
  specific items of interest to the reader).
 </description>
 <dc:creator>Henry Rzepa</dc:creator>
 <dc:date>2003-04-06T10:00:00-05:00</dc:date>
 <cml:molecule id="a1" title="water">
 <cml:formula conciseForm="H 2 O 1"/>
 </cml:molecule>
</item>
</rdf:RDF>

Scheme 1. A representative RSS file using the RSS 1.0 specification. Note particularly that although the CML component is not currently processed by any standard RSS software, its presence does not disable processing of the other components by such clients.

Using RSS

1. The best way of viewing an RSS feed or channel is to install an appropriate client and add the URL of the RSS file on the remote server. Some RSS clients can also "auto-discover" such files. Invoking http://purl.org/net/syndication/subscribe/?rss=http://www.xml-cml.org/index.rss for example will initiate the following process.
   • The RSS file http://www.xml-cml.org/index.rss is passed to a syndication service.
   • This service checks for the existence of the file
   • The RSS file is then validated to ensure it conforms to an appropriate RSS specification.
   • A page listing a number of current RSS clients is presented. The user can then install an appropriate one⁴.
   • The same page allows the RSS channel to be subscribed to the just installed RSS client
2. Typically, the user will install a relatively controlled and limited list of RSS channels which they find are particularly relevant to their interests (Figure 1).
   
   

   Figure 1. Typical User interface to an RSS client.

   This can be done either by selecting from pre-determined lists provided with the RSS client (although "science" as a category is included with most clients, this currently contains little of overt chemical interest), by discovery at visited web sites (RSS feeds are normally indicated by displaying the following icon: ) or by searches at sites specializing in RSS discovery⁵. We emphasize that numerous RSS channels already exist; it might be expected that in due course, syndication sites specializing in chemical content will arise.

3. The RSS client will then periodically perform various automatic tasks.
   • Foremost, it will periodically revisit each RSS site to check if any new information or items have been added. An interesting variation is an RSS client that converts any new items to an alerting email.⁶
   • The retrieved meta-data will normally be indexed by the RSS client to enable facile local searches of this content to be performed. Some clients can also archive it to the user's hard disk for longer term availability.
   • The items retrieved will be aggregated, sorted by various meta-data attributes, such as author, date, subject etc and presented in a condensed form for rapid browsing by the user (Figure 1).
   • Although not yet a reality, several RSS clients promise more sophisticated filtering of aggregated items. Thus the user may indicate that only items by a particular author are displayed, or those that contain a particular keyword or subject. This process can be arbitrarily sophisticated. Thus (as exemplified in the RSS example shown in the box) the meta-data could consist of molecular definitions carried using the appropriate XML descriptor known as CML,⁷ and the user could specify that a chemical sub-structure comparison be carried out on the RSS feed such that only items that reference e.g. a heteroaromatic ring containing two nitrogen atoms be displayed. Importantly, this action is now performed by software, and need not rely on the perceptive capabilities of a trained human chemist. We note that it is not possible to perform such checks using a conventional Web browser, and conventional HTML!
4. It is not necessary that the RSS file be processed by a visually-oriented RSS client. It is entirely feasible that other types of automatic software actions can be set up. A relatively simple one which illustrates the potential of this approach is invoked by http://rssxpress.ukoln.ac.uk/view.cgi?rss_url=http://www.ch.ic.ac.uk/index.rss This passes the contents of the RSS file to (in this instance) an XML process which applies a (XSLT) stylesheet to transform the RSS content to a different form (in this case a convention HTML-based web page, but arbitrarily any other XML form). A variation on this procedure is to include an expansion of an RSS channel within an existing document, to in effect update the document automatically with relevant information, e.g. the addition of <script src="http://rssxpress.ukoln.ac.uk/lite/viewer/?rss=http://www.ch.ic.ac.uk/index.rss"></script> to an HTML document.

   The results of such requests could then be re-used in a different context, as for example automated entry to a database, use by a synthesis robot, entry in a personal diary etc. One might imagine an example in the future where RSS channels from primary chemical journals will contain explicit chemical information encoded in e.g. CML⁷ which are automatically screened by the user's software for the presence of say a particular molecular structure, or molecules with particular properties. Any found matching these criteria could be automatically submitted for e.g. quantum mechanical calculation of further properties.⁸ The human comes to their (probably virtual) desk in the morning to find not only that overnight the system has discovered a molecule of interest to them, but has arranged calculation of its properties, or even its synthesis in the laboratory!

Current applications of RSS.

1. MDB - The Metalloprotein Database and Browser at the Scripps Institute.¹⁰ The URL for the RSS file takes the form: http://metallo.scripps.edu/services/api.php?func=metallopdb&metal=zn&count=12&format=rss

   This in fact is a dynamically generated RSS feed, resulting for a query fed to the (MySQL) database of metalloproteins, with the output being formatted in RSS. Installing this will alert the user to any new (in this case Zinc containing) metalloproteins recently added.

2. The National Cancer Institutes support an RSS feed¹¹ of the form: http://dtp.nci.nih.gov/dtpstandard/servlet/chemRSS.

   This is an feed generated using a Java servlet application to again query a database and return an appropriately formatted RSS file.

Adoption within Chemistry of RSS and the Creation of a Chemical Semantic Web.

An obvious and immediate application might be as an alerting service for the primary chemical journals. Consider the current somewhat haphazard procedure most chemists have come to adopt since the majority of science journals have gone on-line. One periodically finds a browser "bookmark" to a favorite journal (and anticipating it is still functional) a visit to the "latest issue" section of the table of contents will reveal titles of the latest articles, and possibly an abstract that may impart more explicit chemical information (but only visually since it is almost certainly a graphical image which requires a human for perception). Further analysis will require the download of e.g. an Acrobat file, which normally arrives on the user's disk in the "downloads" directory. It still requires much action by the user to organise this reprint into a bibliographic database (for example Endnote), and the process has to be repeated for each journal, with the added difficulty that each publishers' "user-interface" is different and has again to be learnt. Most of us probably realise that this procedure does not "scale", and are wondering how we will cope with this in the future. This procedure could be replaced by a much more structured one based on RSS-derived metadata, derived (in the future) by automated processing of the original full (XML-based) article. In order to discover new and potentially interesting articles, the user subscribes to the RSS feeds of relevant publishers, and can e.g. simply search the latest items that appear automatically for key words of interest. The article download is still necessary, although it may be possible for the RSS client to automatically invoke e.g bibliographic software (or alternatively such software could support RSS directly). When primary scientific publications become available directly in XML (rather than e.g. Acrobat) the possibilities for their re-use increase enormously; no longer is one limited merely to printing the article!

Another immediate application of RSS is as an alerting services for new additions to chemical data bases (although here the sheer volume of new additions might require immediate mechanisms for filtering this down to a manageable quantity). In addition to the two examples noted above, we have easily implemented a simple extension to our php/MySQL-based ChemStock inventory system¹² to alert users to e.g the last five added entries at any given time. It is also apparent that the strength of the system is that separate alert streams from different communities (say chemistry and bio-informatics) might be semantically combined to create connections that are simply not being currently made due to the sheer overload of information. Perhaps the most significant aspect of an increasing deployment of RSS is that it could serve as a focal point for increasing awareness of the importance of creating properly structured information, which includes well defined meta-data, and of managing the links between such information (the hyperlinks) in a manner which allows software as well as humans to utilise these connections in a semantically meaningful manner. RSS does seem to be a tool which is bringing the ultimate vision of a chemical semantic web one step closer.³

References

1. Google: http://www.google.com/
2. For the formal specification, see http://web.resource.org/rss/1.0/. For the RSS 2.0 specification, see http://backend.userland.com/rss An RSS based channel for tracking development of W3C recommendations is http://www.w3.org/2000/08/w3c-synd/home.rss. More information can be found at http://www.w3.org/2001/10/glance/doc/howto#rss
3. P. Murray-Rust, and H. S. Rzepa "Towards the Chemical Semantic Web", Proc. 2002 International Chemical Information Conference, ed H. Collier, (Infonortics), 2002, pp 127-139; T. Berners-Lee and J. Hendler, "Publishing on the semantic web", Nature, 2001, 410(6832), 1023-4. &#0;
4. At the time of writing, we found http://www.awasu.com/ for Windows operating systems, or http://ranchero.com/netnewswire/ for MacOSX work effectively.
5. RSS syndication services include http://www.syndic8.com/ or http://rssxpress.ukoln.ac.uk/ for the UK.
6. For RSS2email conversions, see http://www.w3.org/2002/09/rss2email/
7. P. Murray-Rust and H. S. Rzepa, "Chemical Markup, XML and the Worldwide Web. Part 4. CML Schema", J. Chem. Inf. Comp. Sci., 2003, 43(3), in press and references therein.
8. For further discussion of extended CML families such as CMLComp, see P. Murray-Rust and H. S. Rzepa, chapter in "Handbook on Chemoinformatics", Ed J. Gasteiger and T. Engel, Wiley, 2003, in press.
9. See http://www.rss-search.com/
10. J. M. Castagnetto, S. W. Hennessy, E. D. Getzoff, J. A. Tainer, M. E. Pique and E. Michael, "Data sharing in Bioinformatics: Using the MDB's web services.", Abstracts of Papers, 225th ACS National Meeting, New Orleans, LA, United States, March 23-27, 2003, COMP-141.
11. D. Zaharevitz, personal communication.
12. H. S. Rzepa and M. J. Williamson, Internet J. Chem., 2002, article 6.