…something very close to magic.”
Award Winning, Patented Dialogue Analysis
With over a dozen patents, Nexidia’s technology is robust and proven: We have created and power many award-winning products for media production, broadcast and online distribution. We’ve licensed our products to enterprise customers like Turner Broadcasting System, MSNBC.com, and Microsoft. And our technology is easy to integrate.
Nexidia’s technology is unique because it analyses the 30-40 individual sounds of a given language (phonemes) rather than trying to recognize the half-million or so words that make up that language. In other words, we convert text to sounds, then look for that string of easily recognized sounds in your media.
This approach is far more accurate than speech-to-text applications. Nexidia’s dialogue analysis easily handles slang and proper names, is highly tolerant of spelling errors and poor audio quality, and doesn’t require frequent dictionary updates. It also supports multiple languages and dialects.
Nexidia’s dialogue analysis technology is extremely fast. With current applications indexing audio at approximately 200x real-time per CPU core, a standard 12-core workstation can index 20 hours of media in less than a minute. Once analyzed, 100,000 hours of media can be searched in less than a second.
Results are consistently more accurate than speech-to-text transcriptions. Because Nexidia’s technology analyzes the individual sounds of words, it is not dependent on text-based dictionaries. The benefits are many: There is no need to constantly update dictionaries to search for new terms. Proper names, slang and new terms such as “Facebook” or “Blogosphere” are easily recognized. It’s also forgiving of spelling variations such as “Gadhafi” or “Qaddafi”, and highly tolerant of individual accents and poorly recorded audio.
In order to help the user find exactly the clip they need even more quickly, Nexidia’s search technology supports Boolean logic to search multiple terms. Additionally, terms spoken within a specific timeframe can be identified, such as “congress” and “ethics” spoken within 5 seconds of each other.
Multiple Language Support
To serve the increasingly global media markets, multiple languages and dialects are currently supported, including US and UK English, European French, Italian, German, Castilian and Latin American Spanish, Dutch, and Modern Standard Arabic.
Nexidia’s technology is straightforward to license and easy to integrate with any application or web interface using basic XML commands and our simple JAVA, .NET or REST-over-HTTP APIs. Most common audio and video file formats and codecs are supported natively.
To view the most common audio and video file formats and codecs, click here.