Toward Automatic Tagging of Cultural Heritage Documents

Abstract

The presented system provides intelligent suggestions of terms for tagging cultural heritage documents based on accompanying protocols. This system was made as an improvement of a previously designed system for organizing and manually tagging cultural heritage multimedia collection. The collection consists mainly of documents in the form of audio and video interviews with accompanying textual protocols. The labels for tagging are metadata (date and location of the record, participants in the conversation, language, ethnicity, and religious affiliation of the interlocutors), as well as the topics discussed. Another system was developed using natural language processing techniques for automatic context-based extraction of metadata and topics from those textual protocols. The system presented here combines these two systems to intelligently offer possible solutions for labels, as a help for a user who used to enter these terms manually. The system preselects appropriate choices, gives an option for additional corrections if necessary, and associates final tagging with the current document. This provides semi-automatic document tagging, because it uses automatic intelligent suggestions, but the user has to validate choices with a possibility for corrections.