Skip to main content
Fig. 6 | Journal of Occupational Medicine and Toxicology

Fig. 6

From: Gendermetrics.NET: a novel software for analyzing the gender representation in scientific authoring

Fig. 6

A context-sensitive dictionary for learning the integration & grouping process. (Top) Structure and main functions exemplified by the dictionary of cities. Given a particular context, a key (consisting of only lowercase letters and spaces) is associated with a value in the sense of a partial function (key-value-pair). In the case of the city dictionary the context is given by the country of the city. By applying the dictionary during the data integration process corrections of incorrectly written (“correct”) or versatile names (“unify”) can be made. Furthermore, dictionaries allow learning associations across different data types improving the identification rate significantly. In this example the institution name ‘maximilians univ’ is correctly maped to the corresponding city name ‘Muenchen’ (“associate”). Finally, the city dictionary allows the subsuming of nearby cities (“group”). In this example the suburb ‘planegg’ is subsumed under the term ‘Muenchen’. Substitutions once learned can by expanded by function compositions. Here, through the intermediate step of ‘Muenchen’, ‘planegg’ is gradually replaced by ‘Munich’. Recursions are avoided by prohibiting cycles. For easy reproduction and correction, each entry within the dictionaries is provided with a time stamp. (Bottom) Grouping exemplified by the dictionary of institutions. Dictionaries may be used to group data. Thus, e.g. given a common context (here: a city name) an institution can be considered as representative for many other institutions. Here the given institutions with the common context “Boston” are subsumed under the term “Harvard University”. Generally, such dictionaries exist for cities, countries, journals and institutions. As a basic principle, each reclassification of a given entity during the working process can be stored by updating the corresponding dictionary, respectively. The consistent use of the learning function allows the complete algorithmic reintegration of formerly integrated data

Back to article page