Using intelligent multimedia pattern recognition algorithms, audio mining automatically generates a wide range of metadata for media files, converting spoken words into searchable text.
With the audio mining system of Fraunhofer IAIS, audio and video tracks can be searched specifically for original sounds, and speaker recognition allows people to be found and targeted in the file.
© Photo Fraunhofer IAIS
Automatic preparation of audio media stocks
With automatic speech recognition (“speech-to-text”), audio data can be prepared for searching and automatically tagged. It also recognizes different speakers and distinguishes speech from other audio data (music, sounds). The metadata of the audio files can be enriched accordingly to support existing search functions.
Benefit & value
Speech recognition not only helps improve the search function, it can also be used for further optimization: Based on the spoken words, the content is enriched with automatically generated keywords and related to similar content. In this way, users can be given recommendations that point them to further content that is of interest to them. The user’s dwell time is thus extended and even older, no longer popular content is still accessed.
Flexibility and usability
Thanks to its service-oriented architecture and message-based communication, the audio mining system offers a high degree of flexibility and the possibility to tailor the range of functions to your individual needs. This allows the system to be integrated into an existing media archive and used, for example, as a metadata enrichment service, or to function as a stand-alone media archive.
According to your requirements
For your version of the audio mining system, we can use existing workflows, e.g. for text mining or audio transcription, or we can develop new individual workflows for you. In close cooperation with your team, customer-specific AI models can be trained, new analysis services can be developed or additionally existing services can be connected.
