Spatial and Temporal Information Retrieval in Textual Corpora - Geographical Information Retrieval in Textual Corpora - page 31

1) the overlapping surface;

2) the measure of the distance between calendar references appearing,
respectively,inthequeryandinthedocuments(inotherwordsthetimelineseparating
two calendar periods whose overlapping is nil);

3) the measure of the proportion ratio from the view point of the range of the
entities;

4) the measure of the distance weighted by the range of the entities;

5) the measure of the overlapping ratio.

Alonso[ALO 08]proposesaformulathat,afterstandardizationoftherangeofthe
temporal entities and of the number of calendar expressions, calculates the number of
standardized expressions present at the same time in the query and in the document.

The majority of GIR projects listed in Table 2.1 are based on geographic
information systems (GISs) for the storage and parsing of indexes followed by the
calculations of spatial similarity. Similarly, they are generally based on database
management system (DBMSs) for the storage and parsing of indexes followed by the
calculations of temporal similarity. The approximation of spatial and temporal
entities with numeric representations is necessary for the implementation of such
calculations.

2.3.7. Evaluation campaigns, corpora and resources

An early evaluation of the different IR approaches was deemed necessary. Several
criteria have been defined by Cleverdon in [CLE 63]:

- the ease of use of the system;

- the access/storage cost;

- the presentation of results;

- the capacity of the system to select relevant documents.

Forthislastcriterion,Cleverdonhasproposedtwonow-recognizedmeasuresused
in most evaluations of IR methods. These are recall R and precision P:

R = Number of relevant documents selected
Number of relevant documents

= RS
R

P = Number of relevant documents selected
Number of documents selected

= RS
S