Accepted Papers

  • Approaches to Question Answering using LSTM and Memory Networks
    G Rohit and S Natarajan,PES University,India.

    Question Answering (QA) is a field of Natural Language Processing that deals with generating answers automatically to questions asked to a system. It can be categorized into two types - Open-domain and Closed-domain QA. Open-domain QA can deal with questions about anything, whereas closed-domain QA deals with Questions in a specific domain. In our work, we use the architectures of LSTM and Memory Networks to perform closed-domain Question Answering and compare the performances of the two. LSTMs are specialized RNNs that can forget irrelevant data and remember necessary data. Since data in QA consists of stories and questions based on them, this model seems appropriate, with the ability to handle long sequences. On the other hand, Memory Networks provide an architecture where there is a provision to store the information learnt by the system in an explicit memory component, rather than just as weight matrices. This also seems like an architecture well-suited to Question Answering. We implement each model and train it on the Facebook bAbi dataset. This dataset is specifically generated for the purposes of evaluating QA systems on the twenty prerequisite toy bAbi tasks. Each dataset corresponds to one task and checks whether the model is able to perform chaining, counting, answer with single and multiple supporting facts, understand relations, directions, etc. Based on the performances of each model on the bAbi tasks, we perform a comparative study of the two.

  • Scheduling job seekers using Genetic Algorithms
    Hussain M. Alfadhli

    This paper applies the heuristic algorithms to the problem of scheduling the job seekers over the available jobs in the public sector (government). As a way of providing equal chances for the graduates to win a job in government sector, the government of Kuwait uses a central scheduling system to nominate the graduates of the colleges and universities over the available job vacancies in the government. Usually the government uses a very classical scheduling system to nominate job seekers to the available jobs. The current scheduling method uses very rigid criteria which considers some parameters like Major, GPA, Graduation year, Marital status in a sequential order to evaluate the job seekers. We will use the genetic algorithms to get an optimal scheduling to job seekers over the available jobs. The aim of this project is to make the process of assigning the person for the future job as perfect and fair as possible. We will put in mind two main criteria. The first is to put the right person (in term of educational degree) in the right position (the job duties) depending on the list of approved certificates to hold the job for each job title. The second criterion is to put the person in a job he likes depending on a list of desired jobs selected by the person himself.

    Balasubramaniam Srinivasan,University of California, San Diego

    Audience engagement is key for broadcasting agencies, who purchase rights for huge sums of money for telecasting football. One way to do this, is to improve the overall quality of the pre-match and post-match analysis provided. Furthermore, analysis performed using sophisticated data driven approaches can also provide strategical advantages to teams. This work provides for novel techniques to predict and evaluate the performance of football teams and players based on the passing distribution statistics. The passing distributions data is mainly composed of how players successfully pass the ball around between each other which results in a goal or other events over the course of the game. In this work, similarity analysis across teams is also performed along with the observation of power-law. A low dimensional representation of the nodes of the football network also reveal a strong community structure.

  • Heart Problem Prediction System using Machine Learning
    Nimai Chand Das Adhikari, Rajat Garg, Arpana Alka

    Heart is the most important organ of a human body. It circulates oxygen and other vital nutrients through blood to different parts of the body and helps in the metabolic activities. Apart from this it also helps in removal of the metabolic wastes. Thus, even a minor problem in heart can affect the whole organism. Researchers are diverting a lot of data analysis work for assisting the doctors to predict the heart problem. So, an analysis of the data related to different health problems and its functioning can help in predicting with a particular probability of the wellness of this organ. In this paper we have analyzed the different prescribed data of 1094 patients from different parts of India. Using this data we have built a model which gets trained using this data and tries to predict whether a new out-of-sample data has a probability of having any heart attack or not. This model can help in decision making along with the doctor to treat the patient well and creating a transparency between the doctor and the patient. In the validation set of the data, its not only the accuracy that the model has, rather the True-Positive Rate and False-Negative Rate along with the AUC-ROC helps in building/fixing the algorithm inside the model.

    Doru CONSTANTIN, Costel BALCAU and Alina-Florentina ?TEFAN, University of Pitesti, Romania

    We present the comparative study of convergence for multiunit algorithms based on negentropy function for estimating the independent components.

    Deyvison de Paiva Penha and Adriana Rosa Garcez Castro,Federal University of Para, Brazil.

    This paper presents the proposal of a new methodology for the identification of residential equipment in non-intrusive load monitoring systems. The system is based on a Convolutional Neural Network to classify equipment. As inputs to the system, transient power signal data obtained at the time an equipment is connected in a residence is used. The methodology was developed using data from a public database (REED) that presents data collected at a low frequency (1/3 Hz). The results obtained in the test database indicate that the proposed system is able to carry out the identification task, and presented satisfactory results when compared with the results already presented in the literature for the problem in question.