Explore GEN family of websites, social media experiences, blogs and more.

Unleashing ideas.
Strengthening startups.

Back to search results
    Date submitted
  • 23-Jan-2019

Vera

Abstract

Vera is a smart AI assistant that will revolutionize the way meetings are conducted. Vera will accurately transcribe all the conversations, intelligently summarize them and makes them accessible in a variety of ways. Thus, removing the redundancies in the task and making the employees more productive.

Video

Original Vimeo URL: Open

Introduction Video

Additional Questions

Who is your customer?

American Legal industry is our first market that we will target. We are primarily focusing on eliminating the redundancy in a stenographer’s job, who transcribes speech in shorthand, by making it more seamless and efficient. Vera can effortlessly perform all the tasks of a stenographer with a touch of a button. Stenographers have a market of $50 billion in US ($53,000 (avg. Stenographer salary) times 938,000 ( number of Stenographer in US, 2018) ). This is a huge untapped market which our product can easily capture, if executed correctly. We will be specifically targeting law firms and private practitioners in Pittsburgh in our first year of business. To start with, we will be reaching out to potential customers in Pittsburgh to sell them our platform. Eventually, we will advertise our product on all the local legal and law-related social boards. Since lawyers are a close-knit community, word-of-mouth will be our most valuable marketing philosophy.

What problem does this idea/product solve or what market need does it serve?

Office meetings have been conducted in exactly the same way since the dawn of time. There has been little to no innovation in improving the meetings and that has significantly decreased the impact of meetings. The current system of conducting meetings is very inefficient and unproductive. The work required to prepare, conduct and summarize meetings seems redundant since it comes with an unnecessary cost of assigning specific people to take notes, summarizing them, doing follow up tasks and scheduling extra meetings to recall previous meetings etc. This is an industry wide problem which we are aiming to solve one day, but we will start by targeting the American Legal Industry first. We specifically chose the same due its constant need of accurate transcription and summarization. Legal industry is still catching up with the latest technology so it will be faster to make an impact and get the low hanging fruits.

What attributes will make this idea/product successful? Why do you believe that those features will create success?

Vera is a smart AI assistant that will revolutionize the way meetings are conducted. We offer an array of services but what differentiates us from the others is: Real-time accurate transcription: We are building a custom model for transcribing the conversations to replace the third party Speech to text cloud service that we are currently using. It will give us an edge over competitors since we can fine tune our model according to the jargon used in the industry where our service is deployed. This will give us great performance boost and save 45% of our costs. Human level Summarization: Vera summarizes the conversations in a way a human would do, if he/she were to summarize it. Search through text: This is an additional feature which is solely offered by us and it makes the summary and the transcription accessible in a variety of ways. It gives our users the ability to search any historical transcription in real time. For instance, asking Vera, “What did Jim mention at the beginning of the meeting?” or searching through keywords “Give me the details where ‘privacy’ was mentioned”.

Explain how you (your team) will execute to make this idea/product successful? What gives you (your team) an advantage over others already in the market or new to this market?

The first feature, Real-time accurate transcription is a very hard problem since the industry we are targeting, Legal Industry, expects a very high recall. Not only that, when multiple speakers are engaged in a conversation, correctly and accurately identifying who the speakers are and specifically understanding each of their dialects is particularly hard. We are advised in this area by Prof. Rita Singh, who specializes on Automatic Speech Recognition and on Speech processing. Abstractive summarization (human level text summarization) is still an unsolved problem and it is our main technical challenge. It is something that even humans cannot do accurately since summarization involves a lot of subjectivity, i.e the summary may differ from person to person. The best approach to solve this problem is to use sequence to sequence models, though they give state of the art performance, even these models are not that accurate and we are constantly working on improving them and also trying new architectures. We are advised in this area by Prof. Bhiksha Raj, who specializes in Speech recognition, Audio processing, Neural networks and Privacy/Security for voice processing. The third feature is not that hard to implement but is quite useful since it makes the summary and the transcription accessible in a variety of ways as mentioned in the use case above. We are a group of researchers who have experience in Artificial Intelligence and Machine Learning, along with the expertise of experienced professors, Prof. Bhiksha Raj and Prof. Rita Singh, professors at the Language Technologies Institute, School of Computer Science, Carnegie Mellon University. We come with 25 years of experience in Natural Language Processing, Machine Learning and Artificial Intelligence and we believe that we have the ability to bring our product, Vera to the market.