The main focus of my research has been to study the growth and structural evolution of semantic networks and collaboration networks. Since Albert and Barabasi’s 1998 seminal paper on statistical properties of large scale networks the research interest in studying the dynamics and statistics of networks increased phenomenally. Networks like social networks, citation networks, collaboration networks, semantic networks; biological networks like protein interaction networks or gene mapping networks; physical networks like power grids or road networks, and many more, can be used to represent data of many types. The impact of this research on sociology, biology, physics, economics, psychology, etc. is immense as networks first time offer a concrete means to study phenomenon and processes that cannot be explained by raw statistics. This is the reason I chose to study semantic networks and social networks.
My research can broadly be divided in 2 areas
- Study of semantic networks; their representation, statistical analyses and growth models and applications
My research in this area can be divided into 2 phases
Phase 1 – Semantic networks as representations of course concepts
I proposed a unique way to represent course knowledge to be represented as a semantic network of concepts. I created a repository of such semantic networks for core computer science undergraduate courses such as Computer Communication Networks, Operating Systems and Algorithms and made them publicly available. Researchers in semantic networks at that time were busy in generating generalized semantic networks but did not concentrate of creating specialized semantic networks catered for specific applications. I was the first to create a semantic network for building and analyzing course materials.
Later I also created applications to use the semantic network. The first application called as “intelligent course composer” generated new course materials from a random walk model of the course semantic network based on a few constraints. The next application called as the “course material analyzer” qualitatively analyzed a given course material object for various parameters.
The project was funded by NSF Digital Library initiative.
Phase 2 – Study of statistical properties of semantic networks; a growth model and applications of the theory
Semantic networks can be formed by connecting adjacent words in a book, or from synonyms in a thesaurus, or by playing word association games and there are many more ways to form semantic networks or language networks. However there were surprisingly few research efforts statistically analyzing semantic networks and proposing processes on semantic networks. Specifically I am the first person to propose a growth model for semantic networks in the present of constraints. I propose that like any other physical network the semantic networks stored in the human short term memory are subject to constraints and therefore cannot follow the “general” standard accepted models of networks growth. I give a mathematical model that adheres to realities of constraints in the growth of a semantic network and use it to predict how an individual’s semantic networks grows during the process of text comprehension.
For decades psychologists have proposed theories of human text comprehension, however have failed to provide concrete mathematical representations of these theories which can be translated into machine coded algorithms. Similarly, computer scientists for some years have been working on the problem of machine text comprehension by statistical natural language processing however have failed to take advantage of the theories of psychologists. In my research I have given a mathematical representation of theories of text comprehension. I have proposed a new growth model for semantic networks which explains how people comprehend text during the process of reading. This model can help in creating applications which understand text in a “human-like” manner. This is an alternative to well-known natural language understanding approaches.
It has been known for a few years that the semantic networks of people who suffer from various ailments of brain function like Alzheimer’s, Autism Spectrum Disorder, Epilepsy, Dementia, and many more, show “structural holes” or anomalies as compared to control subjects. My research helps in explaining how the semantic networks for these patients form. This research can help in creating mental exercise applications or specially catered texts for patients with mental ailments that are cognizant of their impairment.
- Statistical analysis of collaborative networks and co-operative work
Large scale collaboration has been an important research area in the recent past. With the advent of applications such as Amazon Mechanical Turk, Kickstarter, Indiegogo, Twitter, etc. the power of crowdsourcing has come to the fore. “Wisdom of the crowds” is a term that is often used to indicate this process of leveraging the cumulative knowledge of the crowd to accomplish tasks of estimation and innovation.
The underlying structure in large scale crowdsourcing is a social network of people collaborating and communicating with each other. In my research I study properties of these collaborative networks and the processes that run on these networks. Specifically the problem of collaborative networks specifically geared to support enterprise innovation has not be visited often. In my research I am trying to set a gold standard for metrics of collaborative networks that support collaborative innovation. The idea of modifying network structure to change certain metric values to facilitate certain processes is novel and unique.
I have proposed a method of detect “gaming” behavior and detect the phenomenon called as “herd behavior” by analyzing the votes cast by members of a crowd network. It is currently successfully being used in series of innovation management products.
I am also the first person to analyze innovation networks i.e. the structure of networks of people collaborating with each other specifically geared towards the task of co-operative innovation. There are no other works to my knowledge that specifically analyzes these networks. I have collected metrics about the networks structure of a typical innovation network which are used to benchmark networks of these types.