Next: Precisionrecall and specificity
Up: The KMS scenario
Previous: Information retrieval
In information retrieval, we distinguish three kinds of search units:
- Individual documents: the focus of the search is on a
specific document and not on a list of possible documents.
- Any document: the focus is on any document containing
a specific character string defined in a search inquiry.
- Information on a specific subject: the focus is on documents
detailing the given subject (concept).
For each kind of search we can identify typical user applications. In
case 1, the user knows that there is one single document which
contains the key information he is looking for. In case 2, the user
wants to study (analyse) the contexts of character strings, but not
the information which may be communicated by the retrieved documents;
typical users are terminologists and linguists. In case 3, the user
does not want a unique document: the information he wants must belong
to the selected subject about which he hopes to improve his knowledge.
Thus, the application areas for IR are very diverse. We may
distinguish between the following search methods, based on the
kind of search:
- Reference text search (bibliographic IR):
standard data and fact retrieval, this would include specialised
applications, such as IR for specific storage media (e.g. compact
disks) or mixed media (e.g. image retrieval), and IR combined with
problem solving (e.g. expert systems).
- Non-reference text search (full text systems):
- This would
include IR for on-line product information, reports, journals,
etc., and for short lived information (e.g. electronic mail) that could
be stored in personal databases or in a company's archives.
- Specialised text search (texts generated for electronic use
- This would include IR for encyclopaedic information
(short technical texts) and for hypertext (non-linear structured text).