Creating a Search Engine
Tuesday 19 December 2006 8:10:00 pm
Any website could benefit from a specialized local search engine that uses datatype and semantic information. However this is especially important in an enterprise environment, because information requirements and data structure tend to be more interdependent. First we will look at the special data situation in enterprises. While some or all of these characteristics may apply to other websites as well, they are almost always present in an enterprise context:
- The data is mostly well structured (for example invoices, product information, technical documentation, etc). In general, structured data is the most valuable for users.
- On an enterprise website, the link structure between the documents is much less manual and ad hoc than on the web. Often links are primarily implemented through a navigation menu based on the document hierarchy. Thus the link relationships that indicate site popularity on the web do not apply in a local context.
Beyond the analysis of data, a more important question is: What is the difference in the information needs of enterprise users compared to users performing normal web searches? On the web there are many similar relevant documents - users are searching for the best documents. In an enterprise environment, users are most often searching for the right answer. [1] In that case, only a few documents are relevant. The popularity of a document does not help here - therefore, fortunately, the lack of link relationships is not a problem in the local context.
Often users are looking for a particular document that they have already seen and for which they remember some characteristics. For example, a user might remember the author, or when the document was published, or some other special information. Because the content in an enterprise environment is more structured, these data characteristics are available for searching.
A good search interface enables users to search for both structure and content related conditions.