28 Jul Syntactic And Semantic Analyses
The neural network technology behind 8vance Matching Technology requires an extensive data model to represent the hierarchical knowledge (called a taxonomy) the system has about people’s skills, knowledge, job titles, education etc.
The construction of such taxonomies initially requires a fair amount of manual work, not only in designing its structure but also filling the taxonomy with data. To speed up this process for one of the taxonomies that contains knowledge about job titles we have designed a module that will process the vast amount of real world data, extract new (previously unknown) job titles and some other relevant data and tries to automatically determine the position of each job title in the taxonomy.
In order to do this, a second taxonomy has been designed that analyses input in terms components and modifiers. The original job title taxonomy is analysed into tables. The same analysis is run on new data containing job titles. The resulting components and modifiers are looked up in the tables and all matches are assigned a probability score depending on the kind and occurrence frequency of both components and modifiers. The results are subsequently sorted on score, best score first. This process allows us to analyse in a short time a vast amount of data and update our models.