Natural Language Processing Models for Named Entity Recognition in Digital TV Audio

This study investigates the application of natural language processing (NLP) models for automatically classifying named entity recognition (NER) elements in digital TV audio transcriptions, aiming to characterize and understand (DTV) program contents. Additionally, a database consisting of 568 TV programs was created by collecting and manually labeling them with NER entities. Automatic results from three chosen NLP models were evaluated and compared with this manual labeling approach to determine their performance. The ultimate goal was to identify relevant words in transcriptions to recommend content related to what a user is watching, thus providing a more enriching and interactive viewing experience.