The Audio Mining is a technique used to automate the research and analysis within audio recordings. Phrases, keywords or even individual phonemes are searched in the audio-video files, and the search results are shown by highlighting the number, position within the file and the degree of confidence. The user who performed the research is able to verify the results proposed, and confirm whether or not the accuracy, enabling continuous improvement of the system performance. Some systems that make use of the Audio Mining:
- Multimedia Information Retrieval: integrated search of text, sound and images in multimedia databases.
- Speech Analytics: analysis of audio recordings to decompose information on the verbal and emotional content.
Thanks to the skills in speech recognition, Voisis has developed a solution for audio mining to search for words and phrases within multimedia archives.
Voisis Audio Mining Processor allows users to search , list and analyze the content of the media files in the manner typical of the search engine.
Thanks to the technology mix of phonetic similarity and lexical – semantic processing, Mining Voisis Audio Processor is able to create complex queries on large multimedia archives, enabling the application of semantic criteria such as the topic and the “sentiment”.
The solution requires the user to insert the keywords, those for which you are searching or those connected to them, and provides as output the files relevant to the research, their “structure” corresponding to the different speakers , the arguments in they contain, and the exact location of your search terms or those connected to them with the corresponding “score”. The Web platform makes it usable by the system from both the private corporate network and the Internet through authentication policies, and can be integrated with other Content Management System ( CMS) or digital archiving systems.