Supriya N. Pal

Phase Space Point Disribution Parameter for Speech Recognition

This paper presents the method for extracting the Phase Space Point Distribution parameter for improving the accuracy of speech recognition systems. By utilizing nonlinear or chaotic signal processing techniques to extract time domain based phase space features, a method is suggested for speech recognition. It is experimentally proved the accuracy of speech recognition system can be improved by appending the time domain based PSPD parameters along with the conventional frequency domain parameters. The presented parameter is also proved invariant to the speaking style say prosody of speech.

Design and Implementation of Pessimistic Commit Protocols in Mobile Environments

The Pessimistic commit protocol specifies set of rules which guarantee that every single transaction in a mobile database environment is executed to its completion or none of its operations are performed. To show the effectiveness of pessimistic commit protocols, a generic simulator is designed and implemented to demonstrate how the transactions are committed and how the data consistency is maintained when the transactions are executed concurrently. Further the constraints imposed by mobile database environments like mobility, blocking of data items etc. are effectively handled. The simulator is tested for efficiency for all timeout based strategies proposed in the literature.

A Hybrid Recommender System Using Weighted Ensemble Similarity Metrics and Digital Filters

Recommender Systems aim to predict the rating or preference of a user given to an item and provide suggestions of further resources that are likely to be of interest. The critical part of the recommender algorithm is finding the similarity metrics, which yield predictions with different accuracies and varieties. In the case of high dimensional feature space, the recommender systems using traditional similarity metrics suffer from several problems such as cold-start problem, scalability, over-specialization and data imbalance. In this paper, a recommender system using weighted ensemble hybrid similarity metric model is proposed by combining two or more traditional similarity metrics such as Pearson correlation, log-likelihood and Tanimoto coefficient. The digital filter is extended and adapted in order to handle the posterior intractability and spatial smoothing of high dimensional recommender space. The proposed recommender system is implemented using Apache Mahout. The evaluation of the model has been done using three large MovieLens datasets consist of 100 thousand ratings, 1 Million ratings and 10 Million ratings. We provide a quantitative and a qualitative evaluation. Interesting conclusions were extracted from the real-time execution of the proposed system on the above-mentioned data sets. The comparison with the results obtained from the traditional recommender systems shows that the proposed system achieved significant improvement in the accuracy for large data sets.

Graph Mining Framework for Finding and Visualizing Substructures Using Graph Database

In the scientific and commercial domains, graph as a data structure has become increasingly important for modeling sophisticated structures especially the interactions within them. Mining the knowledge from graph data has become a major research topic in recent data mining studies. Researchers have designed several efficient algorithms for mining various substructures (subgraphs) within the graph. Several graph visualization tools and techniques exist. But there is a need to define a unified framework for finding and visualizing substructures from graph. In this paper we propose a graph mining framework that captures entities and relations between entities from different data sources. The framework further models this data as a graph and facilitates the dense substructure extraction and frequent substructure discovery in order to find substructures. It also supports knowledge visualization using graphs.

A Big Data Analytics Framework for Enterprise Service Ecosystems in an e-Governance Scenario

In the recent times we have been seeing a fundamental shift from Enterprise Applications towards large scale Enterprise Service Ecosystems. Enterprise Service Ecosystems are developed by modularizing and bundling of individual business rules and functions in the form of services. These services are loosely coupled, distributed and heterogeneous components which orchestrate amongst themselves in a seamless manner. Ecosystem components record the events that are related to the activities performed by them. These components could span across Data Centre, Cloud Infrastructure and Internet of Things. Aadhaar Authentication Ecosystem and e-Governance Service Exchange are examples of Enterprise Service Ecosystems which recently emerged in national e-Governance scenario. A Big Data Analytics Framework for comprehensive mining and analyzing event data of Enterprise Service Ecosystems is proposed in this paper. The offered framework facilitates interesting real time analytics (e.g. Process Conformance Checking, Bottleneck Detection) as well as performing offline analytics (e.g. Process Discovery). The application of the proposed framework for real time analytics is explained using Aadhaar (Unique Identity) Authentication Ecosystem case study.

Implementation of learning analytics framework for MOOCs using state-of-the-art in-memory computing

MOOC aims at delivering online courses to tens of thousands to millions of heterogeneous learners at the same time, with minimal or no charge. It provides an alternate way to disseminate quality education to the section of people who cannot reach premier institutions. It has great potential to overcome the barriers of traditional learning systems. However, there are several challenges in MOOCs such as huge drop-out rates, improper automated assessments, varied student engagement, and attention etc. Learning Analytics helps us to contain such issues. Learning analytics, with the help of Big Data Technologies, helps us to interpret humongous MOOCs data to assess progress, predict performance and identify problems. To perform analytics, we developed a workflow using Apache Spark, a scalable inmemory computing framework. The data from edX platform has been used for experiments. It contains the information of more than 2 Lakh students from 39 courses. Initially, detailed statistical analysis has been carried out to understand the learning patterns and the behavior of online learners. Later, we have developed drop-out prediction models using various machine learning algorithms such as Random Forest, Gradient Boost, and Logistic Regression. A stacked ensemble model is developed and performance comparison with baseline models is carried out. It outperformed all other models with an accuracy of 91.2%.

NLP@Desktop: a service oriented architecture for integrating NLP services in desktop clients

Research and development in Natural Language Processing (NLP) has made significant progress over the last decade. Many robust NLP systems have been developed for handling machine translation, question-answering, summarization, topic detection, cluster analysis, information extraction, named entity recognition (NER), etc. Despite this advancement in NLP research, the results are still not accessible for common desktop users. In the current scenario, it is difficult to integrate a new NLP tool with the existing text processing applications. To overcome this, we have implemented a service deployer framework based on DBus[1] for the seamless integration of NLP applications with desktop clients like email clients, word processors, browsers etc. This framework enables NLP researchers to make their products equipped for the desktop clients with minimal efforts.

Pruthak: mining and analyzing graph substructures

In many scientific and commercial domains, graph as a data structure has become increasingly important for modeling of sophisticated structures. In the past few years, there has been sharp increase in research on mining graph data. We had proposed a unified framework for graph mining and analysis of extracted substructures, which was then an unattended task. Pruthak, a graph mining tool is developed based on this proposed framework. The tool provides preprocessing, frequent substructure discovery, dense substructure extraction and visualization techniques for graph representation of data. In this paper we discuss the approach taken in design and implementation of Pruthak. We then talk about our study on the Digital Bibliography & Library Project (DBLP) dataset for mining and analyzing substructures using this tool. The study results have demonstrated the intended correctness and usability of the tool.

Performing operations on graph through multimodal interface: An agent based architecture

In this paper we discuss about agent based multimodal interface architecture. To illustrate this concept, we present an application to mine information from a graph data structure. This application can be used as a user friendly interface for mining information where graph can be used as data structure for information storage. The application recognizes the gestures, text based input, speech and combination of all of these. Here the various input modalities are implemented as software agents which are independent of each other. There exist MONITOR agents which will synchronize various inputs from the user and act according to it. Users response will be mapped to system vocabulary space with help of a semantic analyzer and results will be passed to MONITOR for further processing.