While still nascent, R&D lab at SRIT has been highly successful in developing some of the most path breaking innovations in our focus domains of Telecom, Healthcare and Enterprise. SRIT follows a unique support and incubation methodology to culture innovative talent. The initiative is personally overseen and driven by Dr. Madhu Nambiar, Founder CEO & Managing Director supported by other R&D scientists, some with more than 2 decades of core R&D experience.

Going forward, SRIT has concise concrete plans to architect and structure a unique practice, serving as an innovative bed to engage innovations on a larger scale, both laterally & vertically with a higher frequency sustained on a progressive basis.

Areas of Research
Recent Announcements

 AREAS OF RESEARCH

Advance Machine Learning, Cognitive Science and Intelligence

We pursue research on automated reasoning, analysis and applications of decision making and self-learning algorithms. Our research vision is focused on developing software that learn by themselves, without any training data unlike neural networks.

Currently, we are on the verge of developing a very advanced algorithm that learns by itself. It has the ability to classify information automatically without any human intervention or sample data. Given hundreds and thousands of electronic documents, the algorithm is intelligent enough to classify or cluster the related information. It then creates its own knowledge base that could be used in various applications.

1. Human-Computer Interaction
2. Search and Retrieval
3. Robotics
4. Contextual Data Integrity – Healthcare Data Exchange

Human-Computer Interaction

The fundamental ways we communicate with computers – the keyboard and monitor, will be replaced by what’s been the communications standard for centuries - voice and gestures.

In 1937, British philosopher Alan Turing devised a simple test to determine if a machine was intelligent. The Turing Test envisions a person sitting at a computer terminal in a room, exchanging messages with a computer in another room. If the first person cannot tell if he’s communicating with a computer instead of a human, the computer is intelligent.

The algorithm, which we have worked on is a concept based on Advance Machine Learning Techniques, which uses a novel mathematical approach to identify the relationship between the words in a set of given documents. This technique or the algorithm does not necessarily need training data to make decisions on matching the related words together but actually has the ability to do the classification by itself. All that is needed is to give the algorithm a set of natural documents. The algorithm is intelligent enough to classify the relationships automatically without any human guidance during the process. For e.g. if documents on animals are fed to the algorithm, the algorithm creates its own knowledge base after processing the content in the given documents. This knowledge base depicts the relationships between the animals. Say, the user types in “Lion” as the keyword; the algorithm would suggest “Tiger, Cheetah, Leopard, Big Cat” etc. These suggestions have a relationship with the word “Lion”.

If this algorithm were to be implemented in a robot that has visual abilities, the robot would be able to read through the documents with the help of adjacent video-to-text software and understand the relationships between the words. The benefits are obvious – a machine can act as a guide to humans. Here’s a simple example. If the user tells the machine that he/she would like to eat fruits, the machine would respond “Would you want an apple or a mango?” Another example would be if the user tells the machine that he/she is feeling sick, the machine or robot would respond “Why don’t you see a physician or why don’t you have a Crocin?” This is possible because the machine or robot has learnt from the documents that “sick, physician, crocin…” are all related. If additional documents are provided to the algorithm, it has the ability to update its knowledge-base based on the data in the given documents.

Note: The algorithm would depict the relationships only if enough data or words relevant to the keyword given by the user is available and also if the link between them is reasonable.

[Back]

Search and Retrieval

The next generation of search engines would be based on context search, wherein the search engine would know the context of what the user might be looking for and provides the best results and also related pertinent information offering the user an expert search. Hence, the user would get to know additional information, which would not have been perceived otherwise.

We have developed a Proof of Concept to prove the point by implementing the revolutionary algorithm on top of a “Search Engine”, making it an “Intelligent Search Engine”. There are various features that are derived from the algorithm that help power the search engine abilities to a very great extent.

In the context of the web, the web-crawler would act as a feeder to the algorithm; the algorithm categories the information automatically and the data gets indexed, hence, during a search if the user searches for a topic, say, “Tiger” the system would throw up information that exactly matches “Tiger” and the agents that have a relationship with “Tiger”. Hence, the system in this case would display “Lion, Leopard, Big Cat etc…” in the result set as a related topic, which makes retrieval very efficient.

[Back]

Robotics

If advance machine learning techniques were to be implemented on a robot, the robot would be able to read through the documents with the help of adjacent video to text software’s and understand the relationships between the words using LMai. The benefits are obvious. The machine can be programmed to act as a guidance system to human. A simple example would be if the user speaks to the robot and enquires about the Topic - “Diarrhea”, the robot uses the contextual relationship dictionary created by LMai and could reply back to the user by saying “Diarrhea” is related to “Dehydration”, “Coli Enteritis”, “Bacterial Gastroenteritis”, “Campylobacter Enteritis” etc., the robot could also narrow down the information on “Diarrhea” by asking the user if he/she needs information specific to “Diarrhea” like “Induced Diarrhea”, “Diarrhea Prevention”, “Diarrhea Diet” etc. Interesting?

[Back]

Contextual Data Integrity - Healthcare Data Exchange

Contextual Data Integrity is about relative privacy, contingent to situation and context of use. It is founded on the logic that complete and absolute privacy is not real and meaningful.  That people share information both public and private all the time so long as the recipient is authorized access to the information forms the basis of this approach. That privacy is happily shared with others as long as certain social norms are met. E.g. all health consultation info provided to a doctor need not be shared with a lawyer or a colleague. It is when these norms are contravened - for example, when your doctor tells the personnel department all about your consultation - has your privacy been invaded.

There is little guidance on how to resolve the conflicts created by all the personal data now available with more technology available to only gather, store, manipulate and share this information. We live in a world where the ability to handle data is rapidly outpacing agreement about how that ability should be used.

[Back]

RECENT ANNOUCEMENTS

Innovation:  LMai^TM - Latent Metonymical Analysis and Indexing
Innovator:  Syed Yasin

Webcast: View

Latent Metonymical Analysis and Indexing (LMai^TM) is an algorithm that uses mathematical techniques in the realm of unsupervised or advance machine learning to derive results no other contemporary technology has been able to achieve till date.  We believe it is all set to power the next generation of internet search experience, where the search engine, besides searching for the keywords, would help in defining the very keyword(s) for the search. LMai’s algorithm works with the user from the start in incrementally defining the scope and the results thereof, helping a user who is not acquainted with the exact terminology for the search to start from any related word and quickly identify the desired area of interest.

For example, if a user were to type an ‘X’ brand of motorcycle, it would throw up pages which were directly related to the keyword. LMai on the other hand would also give contextual results like 2-Stroke Motorcycle Engines, 4-Stoke Motorcycle engines, V-twin Engines, Safety while riding, Helmets and so on. LMai defines this kind of relationship automatically without any set of training data enabling a machine or a system to act like an expert.

Sobha Renaissance Information Technology (SRIT) announced this revolutionary ‘Artificial Intelligence’ technology for immediate application in deriving context-based internet search results on May 07^th, 2007 through a high power media round table in Bangalore. SRIT is actively exploring all possible ways to commercialize LMai, reaching it to all domains and applications that can benefit from its capabilities.


Some domains of high interest are:

Innovation:  DynaRate^TM
Innovator:  Phani kanth

Dyna-Rate^TM is a telecom operator’s software system component which will allow for subscribers to decide their internet bandwidth on a real time and dynamic basis. It does this by provisioning (i.e. provide the requested bandwidth on availability) and rating (i.e. calculate the usage for billing purpose, in post-paid scenarios) or balancing (i.e. balance provisioning of bandwidth with the charge available in pre-paid scenarios) subscriber requests through operators service portals on an immediate reactive basis. The key differentiator is the real time provisioning and rating of bandwidth which till today is not possible.

‘On-demand’ services offered by operators today need a subscriber to send a request (either in print or online) for enhancement or reduction in the bandwidth. The request is handled manually by the customer care department which includes sending the request internally to the technical team for them to manually make those needed bandwidth allocation changes. Typical turns around times are between 2-3 days for most operators globally. The system is naturally fraught with human errors and long turn-around times making it redundant to the very concept of ‘on-demand’.

It supports reversal of charge for undelivered services. This happens through intelligence embedded in the system which tracks download from the operators content delivery platform (CDP). With Dyna-Rate^TM the user will pay ONLY for what he uses and receives. The user can initiate service requests for gaining access to protected services through service selection portal. High service definitions granularity allows services to be classified as Yahoo chat, email, browsing TV channel-1 and so on. The system requires little or no manual intervention for provisioning, operator revenue- sharing, and subscriber control. Dyna-Rate allows for real-time subscriber billing as per the service delivered to subscriber at the time of delivery.

With the bandwidth-on-demand engine, it’s not only the retail subscribers but the corporate customers of the Operators and Service Providers who will derive enormous benefits. With this software engine, a service provider will be able to offer flexibility to its corporate customers to define different levels of Internet access to their internal-users (employees). That’s not all. These employees can be given access to their corporate subscription from any place, where the Service Provider has coverage. Only the corporate customer gets billed, not the employee. The Software engine provides central-control mechanism to handle high level of complexities for a national level data operator.