Implementation of a Customized Named Entity Recognition (NER) Model in Document Categorization

In institutions, one of the problems that arise is related to accumulating a large amount of documentation about their processes and other important information; a recurring task to manage these documents occurs when classifying them. In addition to this, in current times, with the exploitation of applications using Artificial Intelligence (AI) and Natural Language Processing (NLP), they are allowing to provide solutions at a technological level. In this work, we propose a methodology for the development of a customized Named Entity Recognition (NER) model and implement it, for the task of classifying documents and determining and verifying if the classification performed corresponds to a manual classification. To evaluate the performance of the classifier, an automatic classification of 1049 documents from the corpus for knowledge management of an institution was used, using a customized NER model. The results allow us to determine that the presented model achieves 95% positive classifications; the methodology is developed with the purpose of being replicated and scaled, according to the different needs of organizations.