Consistency as an indicator of the indexing reliability (rightness): the cases of LISA, PASCAL and VOCED

Abstract

The possibilities of the consistency as an indexing quality indicator are analyzed through a double procedure. First of all, the consistency found within the LISA and PASCAL databases after applying both the Hooper’s and Rolling equations is stated. This theory demonstrates that, if the context in which the document is described is not taken into account, consistency may not necessarily show the indexing quality of the text. Next, the link between consistency and indexing quality in the VOCED database is also stated, by studying the relationship between consistency itself and the remaining factors that build indexing: correction, exhaustivity and specificity. In order to reach to that conclusion, the VOCED indexing is compared to a so called “model” or “golden pattern” indexing, which enables us to add the Kappa index to the two already mentioned formulas (Hooper and Rolling). This procedure therefore constitutes a model that includes all the factors (semantic, pragmatic and environmental factors) that may affect the correct indexing process from a quality indicator. The method states that consistency itself is not a quality indicator, but a required element to complete it. It also defines the reasons of consistency ratio variability and finds the mistakes or possible omissions in the indexing process