By: Anthony Cresci, VP Business Development and Operations, Theta Lake
First a few interesting facts to set context.
- 85% of people want to see more video from companies (Source: Wyzowl)
- By 2019, 80% of global consumer internet traffic will be comprised of video (Source: Cisco Digital Transformation Survey)
- Enterprise video platform market is expected to grow at a 20.5% CAGR through 2023 (Source: Frost & Sullivan)
- 90% of consumers indicate product videos directly inform purchase decisions (Source: WebFX)
The growth of video within enterprises is staggering but very much expected as the benefits and results are becoming more easily demonstrated. As organizations continue to expand their use of live and on-demand video, their video libraries will grow exponentially and their need for tools to analyze video content and provide insights will become increasingly more important. Insights derived from video use data will enable organizations to be more effective in their usage of video, driving up an already demonstrable ROI.
Concurrent to the growth of video, the advancements in artificial intelligence (AI) and machine learning (ML) have enabled organizations to extract insights in a scalable and meaningful way on digital content. For example, video content management systems now enable more precise search across video content and are able to identify and provide the most relevant content to viewers. Technology vendors are providing solutions that enable sales reps and customer service agents during calls on what to say as well as providing post-feedback on performance. In addition, organizations such as Theta Lake are enabling highly regulated organizations to adopt and expand their video initiatives by analyzing video content for compliance risks to help them comply with industry regulations.
Better yet, these companies are extending beyond the use of machine learning and AI to just extract information. They are layering capabilities to go past extracting the data to turning that data into actionable insights presented in an automated way. For example, important AI and ML-based capabilities to extract data can include:
- Automated Speech Recognition – creating transcripts of the audio
- Optical Character Recognition – Converting images of text into machine-readable text
- Image Content Analysis – Understanding and labeling images
- Facial Recognition – Identifying or verifying a person from a video frame
- Neural Networks – Pattern recognition to extract insights in the data
That baseline extraction is great. But the next layer or steps to make that usable for a company is critical and can include:
- Natural Language Processing/Understanding – to augment transcription with relevance and context such as looking at common Word Error Rate (WER) mistakes specific to calls in financial services, banking, or insurance. These errors can be adjusted to most likely terms and that contextually normalized transcript can be served up to machine learning based classifiers that can tell if that set of words or phrases actually represent a potential regulatory risk.
- Supervised ML policies to take a labeled image and decide if it represents a risk such as the display of a weapon or a sensitive document.
- Taking an unsupervised detection of a face across videos, applying an identity, and suggesting a risk score based on risks in conversations associated to that identity.
- Additionally, AI can be used beyond adding these last mile cognitive insights. They can be used to automate manual tasks to free up human capacity to focus on more important, judgment-based work.
Global spending on cognitive and AI systems will reach $57.6 billion in 2021 according to IDC and US-based AI startups raised $9.3bn in 2018. This massive increase in spending an investment is driving unrivaled innovation within artificial intelligence, which is why PWC reported that 54% of executives say AI solutions implemented in their businesses have already increased productivity.
As AI and ML models continue to improve, along with the continuous growth in data sets such as video content, AI vendor capabilities and insights will evolve providing even more value to organizations. To expand on this example, 1 billion hours of content are watched on YouTube daily and 576,000 hours worth of content is uploaded daily. Google is investing in new content discovery features to make it easier to discover and engage with content, which can only be done at this scale by using artificial intelligence to analyze the content.
Furthermore, artificial intelligence and machine learning have increased the accessibility and usage of video within specific verticals that have been slower to adopt, such as Financial Services, Insurance, and Healthcare as they can now leverage AI capabilities to supervise the content. For example, Financial Service organizations are subject to a myriad of regulations around retention and supervision of electronic communication, including video. This requires firms to review all marketing and sales videos to make sure they do not contain misleading statements, inappropriate promotions, personal or sensitive information, and that they do contain the appropriate disclaimers. They also need the ability to recover video to provide support to regulators in the case of an audit review or for legal discovery.
Supervision of this content has historically been a manual process for compliance teams. However, the growing volume of video content has made this unscalable and unsustainable. Regulated firms can look to AI to drive efficiencies and efficacy in supervision, review, and eDiscovery of digital content. Theta Lake is using AI to help streamline supervision for FinServ firms in several use cases including:
- Visual medium detection and policy application extracting images and OCR of text shown or shared in video and then applying ML policies to determine risk.
- Transcript Relevance & Normalization (TranscripitonRN) – NLP and ML can provide relevance and normalization to transcripts. Instead of a word jumble, Theta Lake can take mis-transcribed versions of the words account, won, for, you, two, dirty, etc. and better infer that it is “account 31422” and that that should factor into a potential policy hit for PII.
- Risk Detection and Prioritization – Deep learning based policies identify regulatory, corporate compliance, and conduct risks within audio and visual content
- Compliance workflow automation – Based on common, repetitive human workflow actions to risk detections, the AI-assisted workflow learns those actions and automates them over time for human reviewers.
- True participant identification – Identify participants by face without any pre-registered face samples or identities, and identify risky and non-risky individuals while providing efficiencies in discovering individuals in the video when complying with GDPR data subject rights and right to be forgotten.
Vbrick and Theta Lake are partnering to explore how Artificial Intelligence (AI) and Machine Learning (ML) can build the IQ of video.
About Theta Lake
Theta Lake, a RegTech 100 company, provides cloud-based compliance AI for video, voice, chat, and other modern digital communications. Its patent-pending technology uses AI, deep learning, and seamlessly integrates with the leading enterprise video platforms, such as Vbrick. Using AI to also power insights and automation, Theta Lake provides directed workflow to add consistency, efficiency, and scale to the compliance review and supervision process, driving down the cost of compliance.