Listen to this article
* – This article has been archived and is no longer updated by our editorial team –
Coseer is an enterprise grade AI software that understands and processes natural language just like humans do. They call this Deep Language Understanding. They use this capability to automate complex processes which are text based. The software configures and trains specifically to every situation, and gives highly accurate results. Below is our recent interview with Praful Krishna, CEO at Coseer:
Q: Could you tell us more about your products?
A: Coseer’s software has dozens different modules and capabilities. All of them come together to do one thing – understand texts in natural language and take decisions based on that. Think of this as 1,000 interns or entry level associates in your team – interns that are very fast, that have a very high attention to detail and can work 24/7 on very large problems.
Use cases vary – from assembling the right reports from scores of disparate sources as if someone went through every single piece of information to do it; to finding the one nugget of relevant information from millions of documents. From virtual assistants to solutions that can identify security risks and proactively block them – it’s all one code which can configure and train for specific workflows.
Recommended: Nobul Aims To Improve The Real Estate Buying And Selling Experience For Both Consumers And Industry Professionals
Q: Can you provide some real word use-cases that we can expect to see?
A: Sure.
Pharma companies use Coseer to bring drugs faster to market. To develop drugs scientists have to deal with knowledge that resides with different teams across organization or is buried in millions of documents. Coseer helps them find this information – the specific knowledge, not just the document – using natural language search. It is as if someone went read each piece of paper, understood the question, and reported back with the best material.
Companies in finance and other sectors use Coseer to analyze, draft and manage contracts. If you are a financial institution that deals with a very large number and variety of contracts, you need something that can judge their strength without worrying about how the different clauses are written. You need to find precedences, and may be you need to extract all the key data points from the contract to put into a table. Coseer does all of that.
The last one I will say relates to GDPR. Personal information is easy to manage when it comes in databases, but what about personal information residing in document repositories, archived emails, or other unstructured documents. Especially when it is not in prescribed format. Coseer’s technology is able to handle these cases very easily.
Our website has tons of use cases by industries.
Q: How is Deep Language Understanding different than Deep Learning?
A: Deep Learning is an machine learning algorithm that has been around for decades. It gives amazing results when you use it for structured data or data that can be neatly put into tables. You need a lot of it, and you need properly labelled. A lot of companies have used Deep Learning to get amazing results.
However, Deep Learning does not work for unstructured data. Labeling the data itself is so expensive that data scientists cannot iterate on their AI models, and accuracy is low.
Deep Language Understanding is the discipline of emulating human thought process when dealing with natural language. We are trying to teach a computer how to process language like a human does. We use and algo called Calibrated Quantum Mesh to implement DLU. This approach does not need any labeled data. The amount of training data is minimal to begin with.
Recommended: Let’s Talk Security! How To Protect Yourself With BitBay’s Multi-Level Cryptocurrency Wallets?
Q: What are your plans for the future?
A: We are very excited about future – near term and long term.
Near term the mission is simple – help more and more enterprises with their workflows. This means rolling out to more languages and adding more capabilities. For example, we are trying to develop some abstract thinking on questions like “what is bribery”, “what is an adverse pharmacovigilance event”, etc. We are also working to reduce the time it takes to configure and train our machine from weeks to days.
Longer term – we want to understand all that is being said over the internet, in real time. We think that will be the truly intelligent machine. Everything we do is a building block towards that ambitious goal.
