Let Us Guide You

Select Industry
Select Business Problem
Find Solutions

News Applied Technology: Phonetic Search Technology In Digital Archives

Applied Technology: Phonetic Search Technology In Digital Archives

By: Drew Lanham
Jun 25, 2015



High-definition broadcasting pioneer WRAL recently digitized its entire archive of nearly 20,000 videotapes — content dated back to 1976. The Raleigh, North Carolina-based station undertook that massive effort not only to ensure that those assets would be preserved, but to complete its transition to an entirely tapeless workflow using Sony’s Optical Disc Archive technology. The problem is that many of those assets lacked sufficient metadata to enable an effective search. The station found a solution. 

To solve that problem, WRAL integrated Nexidia Dialogue Search into its workflow. This software tool makes all of the media in a library searchable on any combination or words or phrases. Dialogue Search works across a wide variety of voices with varied accents. The probabilistic structure of the search algorithm means the tool can find results even in the presence of background noise. Unlike speech-to-text solutions, phonetic search does not rely on perfect spelling or dictionaries and does not have to be trained to recognize new words.

Dialogue Search is built on Nexidia’s patented phonetic search technology. Phonetic searching uses an algorithm comprising two phases — indexing and searching. In the first phase, which only occurs once, the software indexes all of the speech in a given library of media files. Once indexing is complete, the original media files are no longer required when searching the media. This means that the index can be generated on the highest quality media available for the highest accuracy, but then a proxy file can be used for subsequent playback.

The search engine returns a list of matching files, including the confidence score and location of each match within each file. These files can then be filtered on standard metadata such as creation date or source, previewed within the system, and added as time-coded markers to the MAM and/or video-editing applications. Results can be replayed in a standard media player.

The station relies on a Sony Optical Disc Archive. It accommodates multiple users and facilitates a variety of tasks such as archive, search, and content retrieval from the nearline, scalable library. 

The Nexidia Dialogue Search tool first indexed all the media in the archive. The tool allows staff to search independently of other metadata like logging, transcription, and/or captions. When metadata does exist, the tool can search and filter results using that metadata in combination with the dialog to refine and improve the search.Search results are collected into project bins that can be exported directly back to the MAM or into any NLE. This supports the flexibility to use a variety of editing solutions. Staff can search the 40,000+ hours of media in in only seconds. 

Peter Sockett, WRAL’s director of engineering and operations, said: “Nexidia Dialogue Search gives us back our archive. It’s as simple as that. Now we’ll actually be able to repurpose those assets easily and get more value out of them.”