Video is Data

Video is an engaging medium. While it can be associated with entertainment, its use for training, executive town halls and other forms of corporate communications can’t be understated. However, it’s easy to overlook how valuable video content is longterm. Part of this is attributed to difficulties in utilizing larger archives. This is something that was brought to light by Wainhouse Research in the report Unlocking the Hidden Value of Business Video. Of the 1,801 executives surveyed as part of that report, an alarming 79% noted that: “One frustration of using on-demand video is not being able to quickly find the piece of information I am looking for when I need it.” As a result, traditionally people have approached video differently than they do a written report, which is seen as already searchable.

…but they don’t have to. Video can contain a rich depth of data and be searched against to discover this information. This article talks about approaching content libraries with the video is data mindset, traditional challenges in this methodology and technological advances that overcome them.

The hidden value of video

Ever been stuck in the situation where you know a particular item of importance was said in a meeting a few weeks ago, but can’t recall the exact details?

Inside video content can be a wealth of data. Being both a visual and audio based medium, a ton of information can be delivered to viewers. However, it can be seen as wasteful to develop a strategy based solely on live video for corporate communications. Excluding on-demand video would create a gap where you couldn’t reference previous statements or assist in on-boarding newer employees with already existing training.

Video assets can have longevity to them, presenting value to end users long after their initial creation. These benefits are realized by a large number of companies already. In fact, 55% of chief executives participating in the Unlocking the Hidden Value of Business Video report say they “strongly agree” that video archives are a valuable storehouse of institutional knowledge. For those not already taking advantage of video archives, this value can be unlocked from a methodology approach to content. That can include being disciplined and making sure important meetings, executive addresses, training and more is not only archived but in an accessible fashion to employees.

Video is data methodology

When adopting a video as data methodology, a variety of different considerations have to be made to make it successful.

The first is understanding that not all corporate communications may be seen as relevant longterm. A one-on-one meeting likely has far less importance than a department wide strategy summit. Identifying important content should be the first step in this process, and then to set it up so it’s streamed or recorded.

The second step is making sure this material is not only archived, but that you are working toward developing a centralized database of these assets. Pain points to avoid here include an overabundance of sources that might host this content, which could lead to end user frustrations in logging into a variety of different services on a quest to find a particular piece of content.

Third step is maintaining the archive. This includes actively unpublishing or deleting assets that are outdated, such as product specific training material that might talk to features that are no longer there. The goal here is to try and prevent users from stumbling across information in video content that is no longer accurate.

The final step is around discovery, aiding users in successfully finding assets when they need them. At one time this meant developing a structure and hierarchy where content could be navigated to from predetermined areas. However, this approach has issues with scalability, especially as video archives continue to grow at faster raters. In fact, from the survey data collected in Unlocking the Hidden Value of Business Video, 19% of businesses are adding at least 25 hours worth of video content to their archives each month. As a result, a better approach can be enabling search through typing out phrases or queries and having relevant content returned. For this to work, though, a lot of metadata is needed behind the scenes.

Challenges in using the hidden value of video

Once someone starts to approach video as a data source, the next question is around discovery. There is rich information found inside video content, but if it’s unstructured and untagged it can be difficult to utilize effectively.

One of the solutions can be to manually tag video assets. This generally encompasses someone watching the content and adding metadata, ranging from listing topics to speakers. This is a beneficial practice, making content easier to find or helping to organize material. However, it can be very time consuming, especially in face of how much content is being generated.

An alternative approach is to lean on artificial intelligence for this process. This can involve automating the creation of metadata that can be used for search. One of the more straight forward methods can be to utilize the spoken word in the content. This involves speech-to-text transcription, developing a literal transcript for that asset that’s searchable. This can be taken a step further, though, by allowing users to search inside of the videos themselves and giving them the ability to jump to these moments.

Video is Data: Player Search

For example, let’s say we are researching an executive at our company. We type in their name and are given a list of 19 relevant videos, which is already a good start. However, we notice that many of these are hour long sessions that involve multiple speakers. As a result, that’s still a daunting amount of content to sift through for a relatively small amount of information we might be looking for. Luckily, if we are able to search inside the video we can relatively quickly see these relevant moment and jump right to them. So let’s say the first video is a 80 minutes long and the executive doesn’t actually start talking until 30 minutes into it. Rather than having to try and skim through the video to find this, we are able to jump to potentially relevant moments.

This technology, from the AI-driven transcription to the ability to search inside the video player, is available from IBM’s enterprise video streaming solution.

Summary

It’s easy to overlook how valuable historic video content can be. If an organization commits to looking at video as a source of data, they can create an engaging knowledge database. A database that would allow stakeholders to quickly retrieve information as needed.

Curious on how the AI aspect of this process could work? Register for our Getting Started with IBM Watson Captioning webinar. This examines the technology for the generation of closed captions, an added benefit of the speech-to-text process, but also looks at using this for search and discovery of video assets as well.