Audio Mining and Speech Analytics

Tom Keating : VoIP & Gadgets Blog
Tom Keating
| VoIP & Gadgets blog - Latest news in VoIP & gadgets, wireless, mobile phones, reviews, & opinions

Audio Mining and Speech Analytics

I've always been fascinated with word spotting analysis. -- You know, the ability for the NSA to overhear you say, "Let's kill the president" on the telephone and then mysteriously three hours later the Secret Service shows up at your door. Not that that has ever happened to me. : )

Ok, maybe that scenario is a bit mythical, but the NSA & CIA do have some interesting technology to eavesdrop on conversations and perform wordspotting.

Reading this email and press release below about a private firm's wordspotting technology has me wonding how many years ahead the NSA is technology-wise. Definitely food for thought...

Here's the email/release.

As you know, call centers record hundreds, sometimes thousands of calls each day. There is a great treasure trove of information contained within these calls that can potentially impact every thing from customer service to the company's overall business strategy. Unfortunately, companies often struggle to extract the specific information they are looking for in a cost-efficient manner, and this knowledge often remains untapped.

Yesterday, an Atlanta-based company called Nexidia launched the latest version of its audio search product that helps address this very issue. As background, most audio search technologies operate by first converting spoken words to text (similar to closed captioned programming) and then searching the transcripted text for the desired information. What makes Nexidia's technology unique is that its search is based on phonetics. That means searched terms need not match letter-for-letter, but simply sound-for-sound. In this way, a typed search for 'stewart', ‘stuart' or 'stooart' will bring back the same results with the Nexidia search engine.

Nexidia's new 4.0 product is able to search audio at 100,000x faster than real-time playback and is more scalable and more accurate than ever before -- whether it's searching broadcast-quality audio or a cell phone transmission. It's also browser-based with expanded analytical tools to help slice and dice search results.

As you can imagine, the ability for call centers to automate this search process focusing on key words and phrases can completely reshape a process which, for most organizations today, is manually-intensive. If you have any questions about this product or would like to see a short demo, please let me know.


Expanded Offering Provides Direct Access to "Voice of the Customer"

Dallas, TX - Call Center Demo and Conference - January 31, 2005 - Nexidia, the only provider of a phonetics-based solution for audio mining and speech analytics, announced today the general availability of its NEXminer Enterprise 4.0 product. The browser-based solution delivers the most scalable enterprise audio search capability in the market with audio indexed at 40 times real-time. This makes NEXminer Enterprise the fastest solution to render any quality of audio searchable in multiple languages.

"Today many enterprises, such as contact centers, record the ‘Voice of the Customer' in order to gain insight into both their customers and business. Data gleaned from these recorded interactions guides decisions impacting the customer experience, revenue opportunity identification and lower operational costs. Traditionally, this analysis has been a manual process as technology failed to deliver an enterprise-wide speech analytics solution. NEXminer Enterprise 4.0 has been architected to render vast quantities of audio searchable at 40 times real time, and to give enterprise users the tools to analyze all of their audio data for key metrics, trends and data vital to the organization," said John Willcutts, President & CEO of Nexidia. "In addition, we have focused on ensuring a level of accuracy unparalleled in the industry. As a result, NEXminer Enterprise delivers a true, enterprise speech analytic solution allowing companies to benefit from a lower total cost of ownership, without limiting their ability to search and analyze audio-video data."

NEXminer Enterprise 4.0 features include:

Unparalleled Speed and Accuracy
NEXminer Enterprise uses an open-systems vocabulary retrieval system that can conduct searches against large collections of recorded speech at speeds up to 100,000 times faster than real-time playback. Additionally, its phonetics-based engine does not require extensive training or large dictionaries thereby improving the accuracy of searches on proper names, inexact spellings, industry terms and jargons, slang and colloquialisms.

Investigation Tools
The Investigation Tools available with NEXminer Enterprise allow users to easily view large volumes of search results in order to monitor key metrics, identify anomalies that require investigation, track trends, correlate results to support root cause analysis, and drill down to specific audio. Results can also be exported to external systems, providing users access through a variety of other applications.

NEXminer Enterprise can be used for specific term or phrase searches, as well as complex rules-based queries. The powerful phonetic search engine is also capable of delivering results from both poor quality audio such as cell phones and high quality sources such as broadcast audio.

Distributed, Open Systems Platform
The NEXminer Enterprise solution is designed on an open systems framework, supporting standard I/Os and interfaces, and also can be deployed on various platforms and servers. Integration to an existing framework, such as legacy systems and recorders, enables NEXminer Enterprise to access and mine all audio data.

NEXminer currently supports the following languages: North American English, Latin American Spanish, German, Mandarin Chinese, Korean, Japanese, Modern Standard Arabic, Gulf Standard Arabic, Farsi and Iraqi. Additional languages can be easily added without cumbersome resource-demanding dictionaries.

About Nexidia
Nexidia is a leading provider of highly scalable, highly accurate audio mining and speech analytics software. By transforming audio data into actionable intelligence, Nexidia allows both government agencies and commercial enterprises to leverage untapped information previously locked away in audio files. Based on years of research and development, NEXminer's phonetic engine is the only tool that allows the user to search on proper names, places, industry terms and jargon without extensive training and cumbersome dictionaries. And unlike the speech-to-text and word-spotting methodologies, Nexidia's phonetic approach makes it the fastest, most accurate, and highly scalable tool for audio analytics and search on the market today.

Featured Events