FOS-Fresh Online Services: by Shyam Kapur
Unstructured/opinion-rich data are growing fast across the Net. Several search engine meant for mining this real-time data are coming up. TipTop Technologies is one such service. In this podcast Shyam Kapur, President and CEO, of TipTop search engine speaks about the features of his service.
Shyam has been working with computers for over three decades He has already accomplished twenty plus years of outstanding research in data mining, natural language processing, machine learning and information retrieval. Shyam has also put his ideas and vision into practice at several major Internet companies (e.g., Yahoo!, Alta Vista, Infoseek) and has been involved in the founding or early stage development of successful start-ups such as Adchemy (funded in early stages by August Capital & Mayfield Ventures) and MetaLINCS (acquired by Seagate.).
Shyam obtained his Bachelor of Technology Degree from Indian Institute of Technology, Kanpur and his Ph.D. also in Computer Science from Cornell University. Prior to moving to the Silicon Valley, Shyam was a postdoc at University of Pennsylvania and an Assistant Professor at James Cook University in Australia.
Text of the Podcast
Welcome to J. Murali’s podcast
TipTop Technologies is a start-up company that builds programs to extract meaning and patterns from natural language and other unstructured data. Our first consumer product on the Internet can be found at FeelTipTop.com. It is an Insight Engine that helps users find, in real-time, insights from other users. These insights are mined through the application of highly sophisticated proprietary semantic technologies.
Human life is a series of experiences. Our past experiences provide us with the requisite know-how to deal with situations we come across every day. Before the computer age, learning from past experiences was mostly recorded only within the human mind and that is what humans relied on to make the best decisions in their day to day lives. As computers and then the Internet became more widespread, people started to rely on services like search to augment the knowledge within their own head. Today, the tens of billions of queries users type into search boxes each day are a testament to the fact that users recognize the value of having instant access to records of their and other’s past experiences.
In the last couple of years, human behavior began to shift towards not only leveraging one’s own past experiences but also other’s past experiences on a very wide scale. The “other” in this regard can be a friend or influencer or even a complete stranger. Each type of person – both as an individual and as part of a group in the aggregate - was recognized to have value. Web sites like Amazon, TripAdvisor and Yelp became successful in attracting users to enter records of their day-to-day experiences. In turn, consumers flocked to these sites to help find the best records of other’s past experiences in order to improve their own future experiences. Then, even more recently, social networks like Facebook and Twitter have become mainstream. On these networks, users make known their observations of their world – a record of many of their online and offline experiences – several times a day. These status updates are a mix of facts – what happened – and opinion – what was it like. In turn, what users read and how they behave is nowadays often influenced by what is promoted within such social media.
While just some years ago, even the presence of such data was considered a novelty, today, the sheer volume of relevant data of this nature is simply overwhelming. No human being can read and absorb the sometimes hundreds of thousands of user reviews & user status updates that may all be very relevant to a decision that they need to make.
Most of the data that is a record of past experiences is in natural language because that is the means for recording & communication that humans find natural. The numerous insights trapped within this kind of data have to be extracted using technology so that everyone can benefit the most from this data. At TipTop, we view the sea of unstructured information surrounding people as containing (among other things) ‘tips’ that, correctly identified and organized, can be useful to other people when similar considerations are at the ‘top’ of their mind. Hence, the name of the company TipTop. By successfully mining the right tips – those related to other peoples’ experiences in situations similar to ours – TipTop is in a position to help anyone cut through the fog, noise and chaos of the Web. It helps them discover the most useful, relevant and actionable information out there for any problem at hand, as well as connect with other interesting people engaged in similar pursuits.
Information that is relevant to any situation a user faces could be from the very recent past as well as from the less recent past. As the world moves more and more towards real-time communication, it is natural that users would expect search services to also provide the most up-to-date information. Knowing that the road two miles ahead is blocked is critical to know at the moment one is stuck in traffic not an hour later.
Traditional search services like Google are only slowly coming to grips with the changing nature of available information and rising user expectations. Information retrieval paradigms from 40 years ago on which the current crop of search engines are based suddenly feel hopelessly outdated. Users now expect search engines to not just show the familiar 10 blue links but rather powerful insights extracted from the data. Users want the machine to be smarter at recognizing the intrinsic worth of each piece of data – no matter how freshly generated it is – taking into account a number of relevant relationships not just simply the link structure which is now clearly no longer very useful to derive cues for relevance from within this new kind of content. Users want to also communicate in real time with other users who might have additional information that could be relevant to their current situation.
Information in real-time, social, unstructured text has a texture very different from that of static Web pages, just like a classroom lecture is different from a text book. So, the methods of gleaning this information must also be different. The approach we have taken at TipTop reflects our view, based on many years of reflection, research and experience, that language is a social creation which solves a definite human need for relevant information, and that it is structured accordingly. Algorithms can be designed, therefore, that take apart the everyday instances of language use (i.e. the sentences spoken and written by common people) in a way that reveals the semantic content within the unstructured text usefully.
The unique TipTop way is a combination of linguistic, statistical and heuristic techniques to build lightweight engines that illuminate the richness of information in the data that flows through them as they are used. Our engines read each and every word in their input streams, just like a human would.
Because our approach mimics the cognitive process itself, the data ‘takes care of itself’ so to speak. By using the full potential of the self-learning and data mining techniques built into our semantic engine, we can easily develop extensions of our algorithms geared towards specialized content such as financial news or medical research.
TipTop’s consumer product at FeelTipTop.com is way more than just a search engine. It is an Insight engine. No matter what the user’s situation might be, TipTop is able to uncover insights relevant to the user. Some of these insights are mined in real time from relevant social data. TipTop also provides a variety of new ways to help users communicate with other users, including with complete strangers. These ways have now become established as powerful ways to enrich social interactions.
In summary then, TipTop is redefining what search means in this new decade of the 21st century. The glimpses of the future of search that we are showing you today will only grow brighter as this year and this decade progresses. Visit TipTop at FeelTipTop.com to search tomorrow today.
Standard Podcast [11:40m]: Play Now
| Play in Popup