Word occurrence based extraction of work contributors from statements of responsibility

This paper addresses the identification of all contributors of an intellectual work, when they are recorded in bibliographic data but in unstructured form. National bibliographies are very reliable on representing the first author of a work; however, secondary contributors are frequently represented only in the statements of responsibility that are transcribed by the cataloguer from the book into the bibliographic records. The identification of work contributors mentioned in statements of responsibility is a typical motivation for the application of information extraction techniques. This paper presents an approach developed for the specific application scenario of the ARROW rights infrastructure being deployed in several European countries to assist in the determination of the copyright status of works that may not be under public domain. An evaluation of our approach was performed in catalogues of nine European national libraries of countries that are available in the ARROW rights infrastructure, which cover eight different languages. The evaluation has shown that it performs reliably across languages and bibliographic datasets. It achieved an overall precision of 98.7 % and recall of 96.7 %.