Blog Post

Nexidia Does Contextual Ads for Video

Last week, we saw a pretty exciting demo of a brand-new “AdSense for video“-enabling technology from speech recognition firm Nexidia. Sure, the demo depended on canned examples, but the product had only been available for three days at the point Drew Lanham, senior vice president of media for Nexidia, showed it to us.

NexidiaIn the example in this screenshot (larger version after the jump), just as a newscaster starts talking about a new Hewlett-Packard program, text ads pop up for three sites selling HP products. The ads would change throughout the newscast, with enough control allotted so stations so they could specify, for instance, that airline ticket and vacation offers would not appear next to a bit about a plane crash, said Lanham.

Atlanta-based Nexidia, which has $27 million in venture backing, uses a phonetic approach for speech analysis, meaning it compares the bits of sound in audio tracks to sounded-out versions of potential transcriptions rather than translating the audio tracks into text. The company contends this method is faster and more accurate than speech-to-text alternatives, and more informative than depending on metadata and tags.

Nexidia, like speech-to-text competitors Podscope and Podzinger, is entering the consumer space after years of success selling military applications of the same technology. In addition, Nexidia has built a significant business in call center monitoring. About a year ago the company set up a media division headed by Lanham to explore breaking into the exploding markets for digital media online.

Lanham has some interesting ideas for where such technology could be applied: video aggregators could screen uploads to see which files are mature and which files to run against a fingerprint engine; distributed media companies like Gannett could create searchable archives of all their assets; and media companies could get some independence from search engines by offering video and audio search as a proprietary index on their own sites.

The downside, as Lanham admitted, is Nexidia does not generate a text transcript, so it cannot be fully integrated into search.

Nexidia has only one media deployment so far, on, an Atlanta NBC affiliate owned by Gannett. A searchable video archive, when it comes to timely local news, is even better than posting station clips to YouTube.

Imagine being able to find the precise piece on a story you missed last night, or every mention of some public figure’s name, right from your browser. Now imagine why the station would probably like to be able to sell contextual ads based on that, since there’s really no more point in tuning in for the evening newscast besides kicking back with a beer.


9 Responses to “Nexidia Does Contextual Ads for Video”

  1. There are three important differences between Nexidia and other players in this space:

    1) Nexidia owns 100% of the voice technologies used to build these types (VideoNow, Adwords for Rich Media, etc..) of applications, as opposed to leveraging closed captioning text data or 3rd party speech to text engines.

    2) Nexidia’s phonetic approach is superior to sppech to text if you want broad coverage, open domain, unrestricted vocabulary, inexact spelling, user determined depth of search and practical processing speeds (Nexidia renders audio and video content searchable at 65 times real time vs. 1 times real time per processor). This efficiency (speed, processing power) and open vocabulary are critical elements for building AdWords for Rich Media and other media search related applications.

    3)Nexidia’s applications are simple enough for partners to deploy and maintain on their own sites for their own content.

    These are emerging categories of applications, so the ability for the partner to simply and efficiently deploy these services greatly enhances the partners ability to experiment as the business model developes.

  2. Liz Gannes

    Images should be working now…sorry, we had multiple server switches over the weekend.

    Ken, I thought I alluded to our previous coverage of the TVEyes (Podscope) announcement in the story. Thanks for the elaboration.

  3. TVEyes was first to announce this capability last October, in its news release dated Oct. 10.

    TVEyes Announces First Contextually Relevant Text Ad Solution for Video Sharing and Publishing Web Sites

    Text ads served based on content of audio instead of meta tags or untargeted video pre-roll; taps vast demand for advertising on for video sites in accepted form.

    Fairfield, Conn., Oct. 10 – Online video sharing, news and entertainment sites have gained enormous popularity but haven’t been able to monetize their millions of daily site visitors. Advertisers have also been frustrated by a lack of inventory on video sites and the general dislike site visitors have for pre-roll video ads. Now, by implementing TVEyes’ patent-pending ConText Ad Solutionsm, owners of popular Web sites can quickly begin presenting text ads, driving ad serving engines such as Google Adsense or Yahoo! Publisher Network based on words spoken in the video segment being viewed. A demonstration of ConText Ad Solution, using video from YouTube, may be seen at

    “We’ve made it possible for sites with millions of daily visitors to efficiently tap into the largest and most acceptable form of online advertising, and offer advertisers uniquely targeted access to millions of engaged site visitors,” said David J. Ives, president and chief executive officer of TVEyes, Inc. “Internet users dislike pre-roll advertising intensely, but are conditioned to click on text ads when the content is relevant. With the TVEyes ConText Ad Solution, which automatically indexes site content and provides an interface to an ad serving system, site owners can immediately monetize their millions of page views without adding staff to sell advertising.”

    TVEyes creates a Spoken Word Index™ for every word in a site’s library of video files which is used to dynamically drive the ad serving system each time the video is played. The solution is equally applicable to audio sites, such as radio station Web sites. It is available now and is in beta testing with a number of popular Web destinations.

    “Other than selling sponsorships or pre-roll video ads, both labor-intensive from sales and implementation perspectives, popular video sharing, news and entertainment properties have had limited means to monetize their audience,” continued Ives. “Contextually relevant text ads, based on the content being displayed, presents the most promising means to date for these sites to create a meaningful revenue stream.”

    About TVEyes Inc.
    TVEyes Inc., headquartered in Fairfield, Conn., provides online, real-time search and indexing for television and radio broadcasts, audio and video search infrastructure for search engines, and advertising solutions for video and audio sharing and publishing Web sites through its patent-pending ConText Ad Solution. Its Media Monitoring Suitetm is in use by corporations, professional sports teams, political campaigns, elected officials and the military to provide up-to-the-second intelligence on broadcast news, in multiple languages for US and international markets. More information on TVEyes can be found at


    Spoken Word Index and Spoken Word Search are trademarks of TVEyes Inc. ConText Ad Solution and Media Monitoring Suite are servicemarks of TVEyes Inc. All other trademarks referenced are property of their respective owners. ©2006 TVEyes Inc.