Virtualization Technology News and Information
VMblog's Expert Interviews: Archive360 Says No Longer Blue-Skying - Achieving Artificial Intelligence (AI) in the Cloud, with Microsoft Azure Cognitive Services


Today, I am speaking with Bill Tolson, Vice President of Marketing for Archive360 on the topic of machine learning, information management, predicative information governance (PIG) and auto-categorization to automate the management of electronically stored information (ESI). 

VMblog:  One of the challenges the content/information management industry continues to face is how to extend machine learning to audio and video content.  Is this a challenge you are hearing from customers?

Bill Tolson:  Increasingly!  There are very few organizations today that do not have mounting stores of audio and video content.  It is everywhere - voicemail, WebEx presentations, conference calls, surveillance/security video, website content, podcasts and the list goes on.  Regardless of the customer size, industry, public, private or government - they all have audio and/or video content (most have both).  And, the majority find it nearly impossible to index and search on site - forget about it if its sitting in the cloud.  Regardless, this is a task that organizations are increasingly faced with whether it is for business, legal or regulations compliance purposes. 

VMblog:  There are some off-the-shelf solutions for this, but the users I have spoken with advise they can be complicated to deploy and manage, frustratingly imperfect and ridiculously expensive.  For the majority that do not have the time nor the budget to buy these 3rd party solutions, how are they handling search and index of their audio and video content?

Tolson:  Historically, whoever needed to find something within an audio file or video - whether it was a business line manager to the legal department, had to spend untold hours watching or listening to the video or audio file.  Looking for something specific - say a key-word, was the equivalent of looking for a needle in a haystack.  Then, they would manually classify, decide how long they should be saved, where they should be saved, how they should be protected, etc...

With the advent of Azure Cognitive Services, Azure-based information management of non-text based content became possible, and at a much lower price than the current "roll your own" independent audio/video applications.  With the addition of Azure Cognitive Services to a native Azure information management solution, audio and video data can now be managed just like text-based files with full indexing, search, review, tagging, policy management and analytics.  Cognitive Services for audio and video data also provides automatic transcription and translation, speech authentication, sentiment analysis with text analytics, video face tracking, image analysis, and the ability to OCR words in images, to name just a few.

VMblog:  I would imagine this delivers regulatory advantages?

Tolson:  You nailed it!  A number of governmental regulatory retention requirements include audio and video content that must be also kept and managed. Under the SEC 17a-4 Broker Dealer regulations, voicemail recordings with customers wishing to buy and sell securities are considered covered communications and must be kept and made quickly producible when asked by the government agency.

Another well-known government regulation is the EU based Markets in Financial Instruments Directive or MiFID II, which comes into effect in early 2018. MiFID II specifically requires the recording, capture, and management of all communications with customers including telephone conversations. In fact, Article 16(7) states:

"Records shall include the recording of telephone conversations or electronic communications relating to, at least, transactions concluded when dealing on own account and the provision of client order services that relate to the reception, transmission and execution of client orders."

Recital 57 of the MiFID II Directive states:

"Such records should ensure that there is evidence to detect any behavior that may have relevance in terms of market abuse, including when firms deal on own account."

These requirements highlight an obvious need; audio (and video) files must be captured, stored in the appropriate repository (WORM), indexed, and tagged with the appropriate metadata and keywords. Additionally, because regulatory authorities expect to be able to find suspect content quickly, the ability to search for and find the content quickly is a necessity.

VMblog:  Likewise, legal eDiscovery and litigation benefits?

Tolson:  Absolutely!  Audio and video analysis is an even greater requirement in the eDiscovery process. 70% to 80% of the cost of eDiscovery is consumed during the review phase and can be even higher if there is large amounts of audio and video content to review. Because audio/video files are rarely described or categorized before storage, upon eDiscovery, these files need to be reviewed if there is even a slight chance they could be responsive to the case. This means that thousands of additional review hours could be consumed driving up the cost of eDiscovery even more.

VMblog:  So, if you had a crystal ball, what's next? 

Tolson:  Today, Azure Cognitive Services already helps organizations to be more responsive to customers as it enables businesses to do things like analyze call center records and other customer communications to find and analyze those communications, say from dissatisfied customers, using audio and textual sentiment analysis.  We have all heard the recording leading up to a call with customer service that says something like, "This call may be recorded..."

In the future, wouldn't it be cool if an automatic alert would be sent to financial services management when a broker promises a client specific returns during a recorded phone call?  Or, a non-video/audio example would be an automatic notification - or to take it a step further, an automatic block - if an email is sent containing certain data/key words. 

The advantage of next generation machine learning capability built into a public cloud repository like Azure is huge. You will get the best of both worlds - powerful machine learning capability combined with the lowest possible price over that of individual, on premise solutions. 

VMblog:  Any parting advice?

Tolson:  IT, business, legal and compliance professionals should as always - do their homework.  It can also be beneficial to partner with a solutions provider that is expert in this area.  There are already solutions today that are positioned to take advantage of Azure Cognitive Services' current and emerging capabilities, as well as layer on additional key capabilities to take search, index, storage, management and alerting to the next level. 


Published Monday, October 02, 2017 7:33 AM by David Marshall
Filed under: ,
There are no comments for this post.
To post a comment, you must be a registered user. Registration is free and easy! Sign up now!
<October 2017>