Speech-based metadata generation for web map search

Abstract

Metadata is indispensable for data discoverability and interoperability. Most datasets utilize automatic techniques to create metadata; nevertheless, metadata creation still requires manual interventions and editions, yet manually metadata creation is a tedious task. The study proposes a prototype that introduces speech recognition in the metadata creation process. Users can generate content by speaking. Afterward, the prototype transforms it into metadata with JSON-LD format, a popular metadata format and utilized by mainstream search engines. A user study was conducted to understand the impact of speech-based interaction on user performance and user satisfaction. The result showed no signi cant performance di erence between speech-based and typebased by the e ciency, slip rate, and di culty rating evaluation. In the user experience evaluation, participants consider the type-based metadata creation is pragmatic, and speech-based metadata creation is hedonic. It suggests that the mix-mode can complement mutually with the advantages of each and optimize the user experience.