Tag Archives: Information Retrieval

Introduction to Findability, Cyril Doussin [WSG meetup 28/05/08]

Description unavailable

Image from Flickr

Here are my notes from tonight’s Web Standards Group London Meetup, the topic was Findability.

It’s a big subject, so he’s going to focus mostly on theoretical aspect. The topic of findability has been very well research by Peter Morville (semanticstudios.com) – Ambient Findability book

“Finding” and what it means.


  • Discover or perceive by chance o unexpectedly
  • Discover after a delibarate search
  • Succeed in obtaining
  • What is exposed to us (on purposes on inadvertently)
  • After search

We’re searching for: physical items. We’re also searching for digital information, mostly knowledge about: oneself, concepts (mining of something), more detailed information (e.g., products), entities in the same society (people, organizations, businesses).

Besides knowledge we’re trying to find opinions also. To validate feelings or judgments, in order to feel more conformable about data; we also search to establish trust relationships; complementary judgments (finding different POV on things).

Definition of information is a complex subject, but authors definition is:

  • Data: a string of identified but unevaluated symbols
  • Information: evaluated, validated or useful data
  • Knowledge: information in the context of understanding

Information is very closely tied to communication.

Examples are memes,  pieces of information that are transmitted from one mind to another. Either verbally or physical action. One example of this is “Rick Rolling”. If you are able to achieve this, you have “viral marketing”.

Multi-agent systems, systems composed of interacting intelligent agents. It’s a domain of AI. There are two types of agents:

  • Reactive agents (e.g., colonies of ants; they base their communication on very simple signals and they don’t really have a clue about the larger task)
  • Cognitive agents (e.g., closer to human beings; they are able to memorize things, they have beliefs and they have pretty complex ways of communication with each other)

Interesting base to study collective environment.

Findability referees to the quality of things we find.

Item level:

Evaluate to what degree a particular object is easy to discover or ..

System level:

How well a psychical or digital environment supports navigation and retrieval

Wayfinding: a complex events of what people to get from one place to go:

5 step process:

  • Knowing where you are
  • Knowing your destination
  • Following the best route
  • Being able to recognize your destination
  • Finding your way back

How do we make something findable?

Make sure that the item is easy to discover or locate

Have a well organized system which supports easy navigation and retrieval

“In your face” discovery principle; expose the item in places known to be frequented by the target audience. This is a case for advertising and commercial display. Advertisers have to understand how people navigate and use the world in which they live. Contextual example are airport related adds around airports since these are the people who usually drive around.

Hand-guided navigation:

  • Sorting/ordering
  • Sign-posting

Example: restaurant menus are sorted by the dish types and when you eat it in the process.

Describe and browse:

  • Similar to asking for directors
  • Similar to asking random questions
  • Get list of entry-points to pages

It’s also possible to mix things up. First example is from Google, direct links to custom web sites and inline search. Search assist for Yahoo proposes stuff to users other interesting things around this term.


  • Describe intent
  • Casual discussions
  • Advice
  • Past-experience

Essentially they are heavily based upon communication between peers..

Web is essentially a giant referral system. Anyone can add signs to entry-doors on your site. But this leads to need for relevancy system; someone seeing the signs don’t really know if that is the best way to go. One solution for this is PageRank, in order to put ranking on links; peer based example is Digg.

Relevance has two ways to measure effectivens the best as possible:

  • Precisions: how well a system retrieves only relevant documents
  • Recall: how well a system retrieves all relevant documents

Precisions = (number relevant and retrieved) / Total number retrieve

Recall = (number relevant and retrieved) / total number of relevant

When we talk about relevance, we need to identify the type of search that is being performed by the user:

  • Sample search: a small subset of documents are sufficient (most of the time; we often don’t look at the second page of results) (precision method)
  • Existence search: search for a specific document (precision method)
  • Exhaustive search: full set of documents needed (recall method)

Content Organization:

  • Taxonomy: organization through labeling
  • Ontology: taxonomy + inference rules (RDF, Dublin core)
  • Folksonomy: adding a social dimension

It’s increasingly important as the volume of information grows and information is shared. Very good base for search engines.

Measuring Findability on the Web:

  • Count the number of steps (less steps, better)
  • Many ways to get your data (search engines are predominate; peer-based lists and directories are also important)


  • Aim to strike a balance between sources
  • Know the path your audience will follow (do user testing)
  • Understand the type of search
  • Make advertising relevant (which is highly subjective, and a hard thing to do)
  • Make your content rich and relevant
  • Make your content structured