Validation
The term validation in ELRA is used in reference to the activity of checking the adherence to standards, and the quality control of the LR product. The contribution of ELRA can be seen as a validation of existing and newly developed resources and documentation of the results in the catalogue.
ELRA has established a validation centre (VC) for Spoken Language Resources (SLR) and one for Written Language Resources (WLR) which are being controlled and co-ordinated by the Vcom and the Board. The procedures of validation are supported on the one hand by validation committees linked to the Board and on the other hand by the VCs themselves.
Before distribution can proceed the LR must be subject to quality control and validation. In the first place, ELRA must provide a manual for validation, and persuade producers to adopt it as a means of adding to the marketability of their products. Where necessary, ELRA must ensure that validation is carried out.
The work of ELRA related to validation is applied to each of the two areas of activity:
Aiming to fulfil its objectives regarding the production of validation manuals, the ELRA VC for Written Language Resource (WLR) is working in close co-operation with highly recognised research centres in order to produce such manuals. The work being carried out capitalises on previous projects including, but not limited to, EAGLES, Parole, Simple, Multext, and previous work done by ELRA. The validation of the WLR available at ELRA is carried out by CST, Center for Sprogteknologi, (Denmark), the ELRA WLR Validation Centre. CST cooperates with a network of expert centres in Europe.
Speech Language Resource (SLR) validation concerns the quality evaluation of a database against a checklist of relevant criteria. These criteria are typically the specifications of the databases, together with some tolerance margins for deviations.
An article dealing with SLR validation has been published in the ELRA newsletter, volume 5 n° 4: the Art of Validation.
The validation of the SLR available at ELRA is carried out by SPEX, the Speech Processing EXpertise centre (The Netherlands), ELRA Validation centre for SLR.