As it relates to both knowledge retention and comprehension, it is well-documented that video is a more effective way to engage with employees than email, PowerPoint slides or even voice calling. But with Cisco predicting that increased adoption of video will cause enterprise network traffic to grow by a factor of 3X by 2021*, will video be its own worst enemy? Is it possible to create so many video assets that the information employees truly seek will be buried in an endless sea of video files, never to be found?
In today’s Ask the Experts video, Qumu President and CEO Vern Hanzlik explains how enterprise video platforms can solve the massive challenge facing large, video-intensive organizations today—increasing the value of enterprise video by making it easily searchable. Then, in the article below we break it down further, diving into the basics of video search technology today and trends for the future.
From Basic Content Search to Sophisticated Data Mining
Compared to traditional types of content like text documents, slide decks and spreadsheets, video is a very different animal. Take a typical PowerPoint presentation for example, where 20 megabytes is a massive file size and 50 megabytes is almost unheard of. Yet just 5 minutes of HD video can be 250 gigabytes! Multiply this by hundreds of hours of video, and you have an enormous amount of content that needs to be stored, indexed and easily searchable by anyone in the organization. That’s a lot of content to wade through to find that one video—not to mention the precise clip—you need. Of course, time is money, especially when you’re the compliance executive, customer service rep or field tech who needs an answer now.
“As our clients unleash video in an enterprise, we invariably see bandwidth and storage demand increase dramatically. As a result, we expect video repositories to grow 5- to 10-fold in the next few years. The gating factor will be the ability of the enterprise to manage it and get at the knowledge in real time.”
—Vern Hanzlik, Qumu President and CEO
So how do users quickly and easily cut through all of this video and find the one or two snippets they’re actually looking for? To answer that question, we will review three types of video search technologies which have emerged over time, and allow organizations to progress from finding content to mining that data to access rich knowledge resources.
Search Type #1: Metadata Search
Metadata, or “data about the data,” is a standard element of an enterprise video content management system. Think of it as a profile of the video: a title, description, key words, author, date, expiration date, etc. When you “ingest” a video asset into the video repository, it is automatically wrapped in a few very basic pieces of metadata. However, because most of the details are input manually by the user, the quality and quantity of information can vary considerably depending upon the skill level (and available time) of the person who is processing it.
Let’s say as as example, you are searching the repository for a video presentation in which the speaker discussed the concept of “digital dexterity.” Unless the individual who uploaded the video into the repository entered that phrase into the metadata, a standard metadata search might not result in finding the file. And if it did, you would likely still need to manually locate the section where “digital dexterity” is discussed. Clearly metadata is important, but given its inherent weaknesses and reliance on proper tagging by a human being, it cannot be the only video search type available for users.
Search Type #2: Speech or Phonetic Search
Speech search technology (also called phonetic search) overcomes the limitations of metadata, by adding the analysis of actual spoken content. Essentially, Speech Search recognizes the audible elements of speech (phonemes) and allows users to find spoken words or phrases—even with differences in speaker accents and varying recording quality—without manual tagging.
Speech Search allows a user to jump to exactly where a word or phrase is spoken, in a matter of seconds. And something called a “Confidence Score” helps users narrow a search to videos with the closest match to the selected terms or terms. Once users identify a promising video, they can see the precise locations in the video player navigation bar where the phrase is used, and simply preview that segment to see if it is what they were looking for.
Search Type #3: Speech to Text
As we progress in complexity, the third and still-emerging phase of video search is AI-Enabled Speech to Text capability, where the audio track of a video in the repository is automatically converted to searchable text. Speech to Text creates a data-rich environment for mining and search, which in some cases can be combined with AI functions to increase accuracy.
But Speech to Text goes much further than that. Imagine what could be learned about knowledge and collaboration trends across a global enterprise if a comprehensive analytics and dashboarding functionality were included as well. With the right analytics in place, organizations can now begin to reveal information around what and how video (or knowledge in general) is being consumed and shared. And with an open, service-based architecture, other applications will be able to integrate with Speech to Text services. In this way organizations can make greater use of their video content, while eliminating redundancy and out-dated information.
Video search is a robust technology that continues to advance, enabling large enterprises to scale up their use of video. If you need to take full advantage of the power of video, talk to DGvO about our Qx enterprise video platform and its industry-leading video search capabilities.
*Cisco Visual Networking Index, 09/2017