Semantic Web
How OneIS makes sense of information
When the web became more than just static web pages, commentators called it Web 2.0, recognising a dramatic shift in its capabilities. Similarly, the Semantic Web has been called Web 3.0, with the potential to start another revolution in the online world.
The core concept of the Semantic Web is to turn hyperlinks in web pages into something which describes relationships between real things. This then allows computers to ‘understand’ these relationships and answer search results precisely, rather than guessing from ambiguous text on the pages.
For example, if there was a web page describing the author “William Shakespeare”, and other web pages describing books by him, links from the books to the author would be annotated with “written by” in the Semantic Web. Then, a computer could very easily answer the query “which books were written by William Shakespeare?” by finding those links. The answer would be precise and unambiguous, compared to current techniques of looking for keywords within pages — a rather inaccurate way of determining authorship.
Practical applications of the Semantic Web have been slow to take off in the wider internet, as it requires everyone to rewrite their web sites. But the concepts have been used to great effect within smaller self-contained applications, such as OneIS.
The Semantic Web and OneIS
When designing OneIS we looked in detail at the kinds of information stored and used in the real world. We noticed that no piece of information stood alone; everything was related to something else. The key observation of the Semantic Web is that those relationships can be used to enable really accurate searching, as well as navigation through the information.
So we borrowed some ideas from the Semantic Web and adapted them to work within a small organisation. Of course, none of the complexity of the underlying structure is shown to the user — when entering information it’s as simple as filling in a form.
Everything in OneIS is represented by an Object, described in more detail here. These can be ‘things’, such as People and Organisations, or ‘concepts’, such as terms in a hierarchical taxonomy like ‘intellectual property’, which are used to classify actual things.
An Object can be linked to other objects using the link data type. For example, a Person object could be linked to an Organisation in the ‘works for’ attribute.
This is exactly how the Semantic Web specifies relationships between things with annotated links. The links between Objects in a OneIS are the equivalent of an annotated hyperlink between web pages.
Of course, in real life, there are lots of links between objects. Here’s a simplified example of some objects and the links between them which might be found in a OneIS:
Using Semantic Web links
OneIS uses these links in three main ways:
1. Hyperlinks
When an object is displayed, the links are shown as hyperlinks to the linked objects. This allows easy navigation between objects, and discovery of related information.
2. Associated objects
Under every object, OneIS displays a list of objects linked to it. So viewing a client shows a list of everything related to it, like staff, documents and so on.
This is taken one step further for some types of information. When viewing a Person, the details of the organisations they work for are included on the page. Similarly, when viewing an organisation, a clickable list of the people who work for it are shown. So, the information you’re likely to need is shown, and the displayed information is kept up to date automatically.
3. Searching
The most important use of the links is in searching. When OneIS finds an object, it traverses the links to find other information which is relevant.
Here’s an example where the user searches for “acme consulting”, the name of a client.
The object representing the Acme Consulting organisation is found. Then, OneIS uses the highlighted links to bring in other objects which are also relevant. The relevancy of the results is set so the organisation will appear near the top, followed by the other objects.
This technique yields very natural search results. But, because OneIS is using the relationships between objects as well as just keywords, the results and ordering will highlight useful things which are unlikely to be found by a traditional system. In this example, the people involved with the client will show up near the top, reflecting their importance to the user.
In a more complex example, the user searches for “research and development”.
OneIS uses Objects to represent everything, including abstract concepts. Since it isn’t intuitive to include concepts in search results, OneIS is configured to remove the Subject object from the search results, replacing it with the Objects linked to it. In this case, OneIS finds a Subject object representing the concept “research and development”, then includes all the objects classified by it so the search behaves as the user expects.
As well as direct links to the subject object, OneIS can follow hierarchical links. So objects linked to the “product development” subject are also included, since OneIS considers links to any object linked as a broader subject to be an equivalent link to “research and development”. This allows us to naturally and efficiently implement taxonomies.
You can configure the path a search takes through the links, specifying:
- what should be done for each type of object found
- whether the original object should be removed from the search
- the path to take (as deep as required)
- how relevant the linked object is
- which attributes the links must be in
- whether the link should be hierarchical, matching things linked to the found object’s children
In addition to the basic configuration, advanced users can explicitly specify a path through the links in a search query. For example, a query could find everyone who works for an organisation which operates in a particular market, or all organisations for which a particular staff member had ever written a document.
OneIS is ready to go out of the box
You can customise OneIS completely, modeling your organisation’s information with the Semantic Web features. But you don’t need to.
OneIS has a default schema, suitable for small organisations. You can extend this without knowing anything about the Semantic Web. But if you’d like to take things further, OneIS has the power you’ll need to make a significant impact with only a little bit of work.