zyLab_logo_2009.jpg A quick look back at our coverage of e-Discovery in recent times will show one glaring hole. That hole is audio e-Discovery, which hasn’t been covered because a solution has not been available. Until now, that is, with the release of Zylab’s Audio Search Bundle.

Zylab (news, site) doesn’t need any introduction, but for the record, it is one of the bigger players in the e-Discovery and information management markets, and, unlike some, has stuck firmly within this space.

Audio Search Use

The result is a number of established products, as well as some lesser-known ones. The Zylab Audio Search Bundle, if it works as Zylab says it does, should take off for two reasons.

  1. The first and most obvious one is that, according to the release notes, the software will cut the time it takes to go through audio files for information relevant to e-Discovery requests.
  2. The second thing is -- and putting it in a wider enterprise context -- it can help identify and recognize unstructured audio content that is coming into enterprises through a variety of sources

In this respect, think of all the audio tools even a small business uses, and then think of all the untapped information that is coming through those tools . Even at the very lowest level, those tools come from fixed-line telephone, VOIP, mobile and specialized platforms such as Skype or MSN Live.

Audio Information Analysis

And this will be where Zylab's new product is going to come into play. It is conceived as software for anyone involved in legal disputes to use by finding information, reviewing it and analyzing it in the same way that they might take any other electronic file.

However, there doesn’t appear to be any reason why it should be limited to law enforcement. As an off-the-cuff observation, it seems ideal for CRM, or business intelligence systems where an enterprise is looking to wring as much value as possible out of customer contacts and would probably welcome any possibility of identifying relevant audio information that can be pulled into other enterprise system.

Audio Search Bundle

That’s all just speculation for the moment, as there is no indication of integration capacities with third-party vendors.

Already, though, it is looking useful. According to Zylabs, what took the FBI three months to transcribe and analyze in the aftermath of the Enron case, where 2,800 hours of audio had to be checked for key phrases, would now take five minutes.

With the Zylab Audio Search Bundle available today, they could perform those same searches directly on the audio files – not a speech-to-text transcript – within about five minutes and instantly replay the segments to verify their relevance,” says Johannes C. Scholtes, chairman and chief strategy officer for Zylab.

And there are other benefits too. These include:

  • File-based tagging to enable audio record classification
  • Dataset filtering based on tags to enable iterative reduction of audio records
  • No dictionary dependence, meaning that the terms used for search face no restrictions (places, people, products, brands and jargon are all usable search terms)
  • Supports multiple search terms and the combination of terms
  • Flexible XML export plug-in framework to create custom data
  • While speech-to-text technology is limited to dictionary entries, the software from Zylabs transforms the audio recordings into phonetic representations to indicate how words were produced and used, giving users a sense of the context of the audio as well as the word-for-word meaning

It also offers a number of search options, such as Boolean and wildcard, giving greater accuracy and relevance of results.

It supports industry-standard audio formats, including G711, GSM6.10, MP3 and WMA, as well as the audio component of video files and uses a fraction of the hardware required by traditional solutions. Given the potential for audio information retrieval for use in other systems, it will be interesting to see where Zylab goes with this.