También disponible en Español


The digital magazine of InfoVis.net

Knowledge Crystallisation
by Juan C. Dürsteler [message nº 31]

Knowledge crystallisation is the process that allows you to transform data into understandable information.

It's a process that we perform daily almost inadvertently. For example, when we gather the data about the monthly sales, grouping it by product type and demarcation and entering it in a spreadsheet to get a bar chart, we are performing an operation of knowledge crystallisation. This operation culminates when we really extract the knowledge that underlies all this data: the sales of a particular product have grown by 15% since the beginning of the advertising campaign.

Knowledge crystallisation is a concept popularised by Stuart Card and other researchers at the Xerox PARC (Palo Alto Research Center). They define it as the process through which "one person gathers information for some purpose, makes sense of it by building a representational schema, and then packages it into some form of communication or action". (See the book Readings in information Visualization by Card et al. Readings in Information Visualization page 10).

The phases that this process comprises of can be summarised as:

  • Information foraging. Gathering data related with the problem we intend to solve. Typically this happens inside Internet databases or in the company's database. For example collecting information on last month sales.
  • Searching for a schema. We need to identify the variables or attributes that characterise the information that we want to extract. In our example, the monthly sales by demarcation and product.
  • Instantiation of the schema with data. Once the schema has been built taking into account the variables of interest for our problem, we must populate it with the data already gathered. In many cases we get some residual data that doesn't fits well in the schema. If this data is a majority part, it's obvious that the schema is insufficient for our purpose, it's badly constructed or the data is too unrelated. In that case we need to refine or change it. In our example we would fill in the spreadsheet with the data and group it by demarcation and product. Incidences or comments could constitute the residual data.
  • Problem solving. Using the instantiation of the schema we can extract the patterns of information that are the objective of our work. In our example, the bar chart allow us to quickly see the sales trend and whether the advertising campaign is being successful or not.
  • Packaging the result in the form of communication or action. All the previous operations lead finally to the effective use of the knowledge obtained. This can be done through the communication of it or by taking appropriate action. In the example, if the advertising campaign hadn’t promoted the sales properly, we would have to study the reason for that further, modify the sales forecast and / or communicate it to the management.

Information visualisation contributes to the creation of the schema and to the problem solving phase. If the schema is already well established, knowledge crystallisation reduces to information retrieval. The sales statistics are not re-invented each month. We just fill it in with the new data that then allow us to see how the pattern has changed since the last month. Anyway the bar chart is still what reveals the trend every month.

The idea underlying the concept of knowledge crystallisation is simple, yet it reveals the usual process of digestion of information and, hence, it's a powerful starting point in order to improve the way in which we extract knowledge out of data.

Links of this issue:

© Copyright InfoVis.net 2000-2018