Natarajan Subramanyam

Bio: Natarajan Subramanyam is an academic researcher from PES University. The author has contributed to research in topics: Tantra & Generalization. The author has an hindex of 3, co-authored 15 publications receiving 23 citations.

Approaches to Question Answering Using LSTM and Memory Networks.

Abstract: 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 remember necessary data and forget the irrelevant bits. Since data in QA consist 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 purpose 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.

Efficient fusion of cluster ensembles using inherent voting

Abstract: Discovering interesting, implicit knowledge and general relationships in geographic information databases is very important to understand and to use the spatial data. Spatial Clustering has been recognized as a primary data mining method for knowledge discovery in spatial databases. In this paper, we have analyzed an efficient method for the fusion of the outputs of the various clusterers, with less computing .We have discussed our proposed slice and dice cluster ensemble merging technique (SDEM) for spatial datasets and used it in our three-phase clustering combination technique in this paper. Voting procedure is normally used to assign labels for the clusters and resolving the correspondence problem, but we have eliminated by usage of Degree of Agreement Vector. Another common problem in any cluster ensembles is the computation of voting matrix which is in the order of n2, where n is the number of data points, which is very expensive with respect to spatial datasets. In our method, as we travel down the layered merge, we calculate degree of agreement (DOA) factor, based on the count of agreed clusterers. Using the updated DOA at every layer, the movement of unresolved, unsettled data elements will be handled at much reduced the computational cost. Added advantage of this approach is the reuse of the gained knowledge in previous layers, thereby yielding better cluster accuracy and robustness

Containing COVID-19 pandemic using community detection

Abstract: Diseases such as COVID-19 that quickly spread through social contact where Infections remain undetected for long pose new kinds of challenges Traditional epidemic models when applied to COVID-19 not only give highly alarmist predictions but also fail to indicate any way to target interventions The models also do not factor social structure/context/dynamics To address this challenge we propose a novel Social Infection Analysis Model explain the intuition behind it and a methodology that factors social structure while performing the analysis SIAM is particularly beneficial as standard epidemic models are dependent on wide-spread testing Our methodology includes (i) modelling society/localities using social network paradigm and analysis at macro level (ii) detecting communities and (iii) analysis of infection spread within a community over time To arrive at an outlook for infection spread within communities we make use of the NK Model for solving complex, long term problems coupled with agent based modeling for member behavior We make use of NetLogo to run simulations We intend to validate our research in Indian context © 2021 Institute of Physics Publishing All rights reserved

Robustness to Augmentations as a Generalization metric.

Abstract: Generalization is the ability of a model to predict on unseen domains and is a fundamental task in machine learning. Several generalization bounds, both theoretical and empirical have been proposed but they do not provide tight bounds .In this work, we propose a simple yet effective method to predict the generalization performance of a model by using the concept that models that are robust to augmentations are more generalizable than those which are not. We experiment with several augmentations and composition of augmentations to check the generalization capacity of a model. We also provide a detailed motivation behind the proposed method. The proposed generalization metric is calculated based on the change in the output of the model after augmenting the input. The proposed method was the first runner up solution for the NeurIPS competition on Predicting Generalization in Deep Learning.

Transforming India’s electoral democracy using Tantra framework

Abstract: India has a rich tradition where good governance is highly valued. The theme of good governance is emphasized by a lineage of thought leaders from Kautilya in 230 BCE (परजासख सख राजञः परजाना च हित हितम । नातमपरिय हित राजञः परजाना त परिय हितम ॥ अरथशासतरः १.१९.४३ ॥ Chapter 19 titled राजपरणिधिः duties of the king in Arthashastra of Kautilya. The happiness of the king is in the happiness of the people, his welfare is in the welfare of the people. The welfare of the king lies not in what he desires, but what his subjects desire) to Mahatma Gandhi [Mahatma Gandhi propounded the concept of “Su-raj” and according to him “good governance” has the following eight attributes, which link it to its citizens: (1) accountable, (2) transparent, (3) responsive, (4) equitable and inclusive, (5) effective and efficient, (6) follows the rule of law, (7) participatory and (8) consensus oriented] in recent times. However, achieving good governance in a country of size of India is hard. Good governance needs to address social, cultural, ethical and process dimensions. Governance not only needs to be good and efficient, but needs to human-centric. At times even well-intentioned and well-thought out processes of Government cause lot of trauma to common citizens. On the contrary doing away with safeguards is inviting disaster. To meet the objective of human-centric good governance, we have developed Tantra social information management framework. Tantra framework makes use of concepts from Zachman framework and unified foundational ontology. The framework itself is modeled as a social network (entity–entity network). Tantra framework interoperates with models such as balanced score card, theory of change and Bartels’ theory of separations. This paper describes the Tantra framework and how it can be applied to transform India’s electoral democracy.

BERT+vnKG: Using Deep Learning and Knowledge Graph to Improve Vietnamese Question Answering System

Abstract: A question answering (QA) system based on natural language processing and deep learning is a prominent area and is being researched widely. The Long Short-Term Memory (LSTM) model that is a variety of Recurrent Neural Network (RNN) used to be popular in machine translation, and question answering system. However, that model still has certainly limited capabilities, so a new model named Bidirectional Encoder Representation from Transformer (BERT) emerged to solve these restrictions. BERT has more advanced features than LSTM and shows state-of-the-art results in many tasks, especially in multilingual question answering system over the past few years. Nevertheless, we tried applying multilingual BERT model for a Vietnamese QA system and found that BERT model still has certainly limitation in term of time and precision to return a Vietnamese answer. The purpose of this study is to propose a method that solved above restriction of multilingual BERT and applied for question answering system about tourism in Vietnam. Our method combined BERT and knowledge graph to enhance accurately and find quickly for an answer. We experimented our crafted QA data about Vietnam tourism on three models such as LSTM, BERT fine-tuned multilingual for QA (BERT for QA), and BERT+vnKG. As a result, our model outperformed two previous models in terms of accuracy and time. This research can also be applied to other fields such as finance, e-commerce, and so on.

Investigating Topic Modeling Techniques to Extract Meaningful Insights in Italian Long COVID Narration

Abstract: Through an adequate survey of the history of the disease, Narrative Medicine (NM) aims to allow the definition and implementation of an effective, appropriate, and shared treatment path. In the present study different topic modeling techniques are compared, as Latent Dirichlet Allocation (LDA) and topic modeling based on BERT transformer, to extract meaningful insights in the Italian narration of COVID-19 pandemic. In particular, the main focus was the characterization of Post-acute Sequelae of COVID-19, (i.e., PASC) writings as opposed to writings by health professionals and general reflections on COVID-19, (i.e., non-PASC) writings, modeled as a semi-supervised task. The results show that the BERTopic-based approach outperforms the LDA-base approach by grouping in the same cluster the 97.26% of analyzed documents, and reaching an overall accuracy of 91.97%.

Bengali Question Answering System Using Seq2Seq Learning Based on General Knowledge Dataset

Abstract: Context-based QA system is a leading research area in NLP. An automatic QA system that can respond to the answer, and which is related to the given context. A deep learning-based model provides a more factual result for today's QA system. Here we introduce a deep learning-based Seq2Seq model for Bengali context-based QA system using general knowledge dataset. Where context and question are part of the encoder and related answer is part of the decoder. All automatic system can discern any language by machine translation. Sequence wise learning is a good solution for those types of automatic system learning. Input tokens are encoded by the encoder and output tokens are decoded by the decoder. Each sequence is stored in LSTM cell that maintains the sequence of input and output. Most of the AI system is developed in different languages. Compared with other languages the Bengali language needs to expand research this field. The major perspective of this research to develop an AI based QA system for the Bengali language. For experiment total, two thousand Bengali general knowledge data is used that also provides a dataset for Bengali QA system. In the dataset, the context contains the main feature for the question. After training our model it gives 99 % accuracy for this dataset and 89% accuracy for validation. The trained model gives a good response in answer prediction.

Containing COVID-19 pandemic using community detection

Abstract: Diseases such as COVID-19 that quickly spread through social contact where Infections remain undetected for long pose new kinds of challenges Traditional epidemic models when applied to COVID-19 not only give highly alarmist predictions but also fail to indicate any way to target interventions The models also do not factor social structure/context/dynamics To address this challenge we propose a novel Social Infection Analysis Model explain the intuition behind it and a methodology that factors social structure while performing the analysis SIAM is particularly beneficial as standard epidemic models are dependent on wide-spread testing Our methodology includes (i) modelling society/localities using social network paradigm and analysis at macro level (ii) detecting communities and (iii) analysis of infection spread within a community over time To arrive at an outlook for infection spread within communities we make use of the NK Model for solving complex, long term problems coupled with agent based modeling for member behavior We make use of NetLogo to run simulations We intend to validate our research in Indian context © 2021 Institute of Physics Publishing All rights reserved

Global Experiences of Community Responses to COVID-19: A Systematic Literature Review

Abstract: Objective This study aimed to conduct a systematic review of the global experiences of community responses to the COVID-19 epidemic. Method Five electronic databases (PubMed, Embase, CINAHL, ScienceDirect, and Web of Science) were searched for peer-reviewed articles published in English, from inception to October 10, 2021. Two reviewers independently reviewed titles, abstracts, and full texts. A systematic review (with a scientific strategy for literature search and selection in the electronic databases applied to data collection) was used to investigate the experiences of community responses to the COVID-19 pandemic. Results This review reported that community responses to COVID-19 consisted mainly of five ways. On the one hand, community-based screening and testing for Coronavirus was performed; on the other hand, the possible sources of transmission in communities were identified and cut off. In addition, communities provided medical aid for patients with mild cases of COVID-19. Moreover, social support for community residents, including material and psychosocial support, was provided to balance epidemic control and prevention and its impact on residents' lives. Last and most importantly, special care was provided to vulnerable residents during the epidemic. Conclusion This study systematically reviewed how communities to respond to COVID-19. The findings presented some practical and useful tips for communities still overwhelmed by COVID-19 to deal with the epidemic. Also, some community-based practices reported in this review could provide valuable experiences for community responses to future epidemics.