National Institute of Technology Tiruchirappalli
Central Library
Institutional Digital Repository
Powered by National Digital Library of India (NDLI)
Communities in DSpace
Select a community to browse its collections.
Recent Submissions
Improving readability through extractive summarization for learners with reading difficulties
(Elsevier B.V., 2013-10-19) Nandhini, K; Balasundaram, S.R
In this paper, we describe the design and evaluation of extractive summarization approach to assist the learners with reading difficulties. As existing summarization approaches inherently assign more weights to the important sentences, our approach predicts the summary sentences that are important as well as readable to the target audience with good accuracy. We used supervised machine learning technique for summary extraction of science and social subjects in the educational text. Various independent features from the existing literature for predicting important sentences and proposed learner dependent features for predicting readable sentences are extracted from texts and are used for automatic classification. We performed both extrinsic and intrinsic evaluation on this approach and the intrinsic evaluation is carried out using F-measure and readability analysis. The extrinsic evaluation comprises of learner feedback using likert scale and the effect of assistive summary on improving readability for learners' with reading difficulty using ANOVA. The results show significant improvement in readability for the target audience using assistive summary.
Challenges in using a standard speech recognition engine in small vocabulary domain
(Blue Eyes Intelligence Engineering and Sciences Publication, 2019-07-22) Srinivasan, Narayanan; Balasundaram, S.R.
This paper discusses the challenges and proposes recommendations on using a standard speech recognition engine for a small vocabulary Air Traffic Controller Pilot communication domain. With the given challenges in transcribing the Air Traffic Communication due to the inherent radio issues in cockpit and the con-troller room, gathering the corpus for training the speech recognition model is another important problem. Taking advantage of the maturity of today’s speech recognition systems for the standard English words used in the communication, this paper focusses on the challenges in decoding the domain specific named entity words used in the communication.
Summarizing learning materials using graph based multi-document summarization
(IGI Global, 2021-10) Krishnaveni, P; Balasundaram, S.R.
The learners and teachers of the teaching-learning process highly depend on online learning systems such as E-learning, which contains huge volumes of electronic contents related to a course. The multi-document summarization (MDS) is useful for summarizing such electronic contents. This article applies the task of MDS in an E-learning context. The objective of this article is threefold: 1) design a generic graph based multi-document summarizer DSGA (Dynamic Summary Generation Algorithm) to produce a variable length (dynamic) summary of academic text based learning materials based on a learner’s request; 2) analyze the summary generation process; 3) perform content-based and task-based evaluations on the generated summary. The experimental results show that the DSGA summarizer performs better than the graph-based summarizers LexRank (LR) and Aggregate Similarity (AS). From the task-based evaluation, it is observed that the generated summary helps the learners to understand and comprehend the materials easily.
Affix-based distractor generation for tamil multiple choice questions using neural word embedding
(Aesthetics Media Services, 2021-05-12) Murugan, Shanthi; Balasundaram, S.R.
Assessment plays an important role in learning and Multiple Choice Questions (MCQs) are quite popular in large-scale evaluations. Technology enabled learning necessitates a smart assessment. Therefore, automatic MCQ generation became increasingly popular in the last two decades. Despite a large amount of research effort, system generated MCQs are not useful in real educational applications. This is because of the inability to produce the diverse and human alike distractors. Distractors are the wrong choices given along with the correct answer (key) to confuse the examinee. Especially, in educational domain (grammar learning) the MCQs deal with affix-based or morphologically transformed distractors. In this paper, we present a method for automatic generation of affix-based distractors for fill-in-the-blanks for learning Tamil Vocabulary. Affix-based distractor generation relies on certain regularities manifest in high dimensional spaces. We investigate the quality of distractors generated by a number of criteria, including Part-Of-Speech, difficulty level, spelling, word co-occurrence, semantic similarity and affixation. We evaluated our proposed method in grammar based Multiple Choice Questions (MCQs) dataset. The result shows that affix-based distractors, yield significantly more plausible outcomes in certain grammar based questions
Two-factor-based RSA key generation from fingerprint biometrics and password for secure communication
(Springer International Publishing, 2022-02-18) Suresh, K; Pal, Rajarshi; Balasundaram, S.R.
In an asymmetric-key cryptosystem, the secure storage of private keys is a challenging task. This paper proposes a novel approach for generating the same public and private key pair on a need basis. Hence, the need for secure storage of the private key is done away with. The proposed approach for generating the key pair is based on two factors: fingerprint biometrics and password. A stable binary string is generated from the distances among pairs of minutiae points in a fingerprint using a gray code-based method. Experiments show that gray code representation significantly reduces the number of inconsistencies between the generated bit strings from two instances of the same fingerprint as against the binary code representation. Hence, the Reed–Solomon error correction code successfully corrects errors due to variations in multiple instances of the same fingerprint to induce stability in the generated string. Hash of the stable string generated from the fingerprint and the string generated from hashed password are XORed to derive a stable seed value. The proposed approach uses this seed value to generate two large prime numbers. These prime numbers are used to generate the public and private key pair using the RSA key generation method. This seed value ensures the generation of the same key pair every time. The experimental results show that the proposed approach can ensure a stable generation of the key. It is not required to store either the fingerprint template or the password. Moreover, the generated private key is also not stored. It can be regenerated on a need basis.
Enhancing Command Recognition in Air Traffic Control Through Advanced Classification Techniques
(Sciendo, 2024-06-12) Srinivasan, Narayanan; Balasundaram, S.R.
This paper addresses the persistent challenges in speech processing within the Air Traffic Control (ATC) domain, a field where despite extensive research, issues such as handling noisy environments, accented speech, and the need for strict adherence to standard phraseology continue to undermine conventional language models. Our study employs a hybrid approach that integrates syntactic analysis with advanced machine learning classification algorithms - Logistic Regression, Lagrangian Support Vector Machine, and Naïve Bayes. By mixing and matching algorithms tailored for specific aspects of speech processing, our approach moves away from traditional reliance on a singular integrated system, illustrating through rigorous testing with the ATCOSIM dataset that such a multifaceted strategy markedly improves command recognition accuracy and adapts more effectively to the unique linguistic features of ATC speech. Results highlight the superior performance of Logistic Regression across various command recognition categories, pointing towards a promising direction for future advancements in ATC speech recognition technologies aimed at reducing human workload and increasing automation precision. This paper explores the complexities of the required analysis techniques and underscores the necessity of employing diverse algorithms in the processing pipeline to enhance overall system accuracy.
Ensemble learning with RAG model to reduce redundant question topics in auto-generated exam questions
(Springer Science and Business Media B.V., 2025-08-01) Sairaj, R. Tharaniya.; Balasundaram, S.R.
Reducing redundant question topics during automatic question generation (AQG) is essential for enhancing the quality of test sheets in assessment. Existing AQG models frequently generate repetitive questions due to insufficient named entity (Question Topic) diversity. This study aims to reduce redundancy in auto generated questions by improving diversity in question topics. The methodology is organised in three main phases: first, a fuzzy ontology mapping technique with ensemble learning is applied to generate expanded entity-relationship set generation from external Knowledge Graphs. Second, the generated entity-relationship set is integrated with Retrieval Augmented Generation (RAG) model for source text expansion via augmentation. Third, T5-shd, a pretrained AQG model is adopted to reduce repetition in generated questions. Comparison against baselines such as T5-e2e and T5-ppl shows substantial performance gain as well as reduction in redundancy. Experimental results on various datasets show that the proposed RAG + T5 model reduces redundant question topics along with an improvement in terms of ROUGE-2 metric (up to 5%) and BERTScore (up to 12%) over existing methods. The application of Ensemble Pruning, specifically with the m-EPIC algorithm, further enhances accuracy while reducing computational overhead (around 26%). These findings highlight the efficacy of combining ensemble learning with RAG-based transformers to refine AQG, ensuring improved question diversity and balanced relevance of generated questions. Additionally, this approach helps to reduce model complexity in Automatic Question Generation.
Enhanced Prediction of Intensive Care Unit Length of Stay using a Stack Ensemble of Machine Learning Models
(Dr D. Pylarinos, 2024-11-20) Tella, Ashok Kumar; Balasundaram, S.R
The Length of Stay (LoS) refers to the time between a patient's hospital admission and discharge. LoS is considered to increase as the complexity of the disease increases. A prolonged stay in the Intensive Care Unit (ICU) can consume clinical resources and be labor intensive. Models that correctly predict LoS are needed to help medical experts make better decisions. To define an ideal process system, healthcare models must consider the patient's condition, availability of beds, resources, etc. These predictions can also help insurance companies manage their budgets. Existing models deploy machines and deep learning techniques to predict LoS. However, there is a need for improvement, considering the features associated with the process. This study presents machine learning algorithms, such as SVM and a stack ensemble, with improved accuracy over existing models. Experiments were carried out on a benchmark dataset, MIMIC-III, specific to ICU patients. The SVM model achieved an accuracy of 93.88%, while the stack ensemble model showed an improved accuracy of 94.70%. The results show that combining machine learning models achieves better prediction rates, which helps healthcare professionals make better decisions.
A new green prospective of non-orthogonal multiple access (NOMA) for 5G
(MDPI AG, 2020-02-07) Basnayake, Vishaka; Jayakody, Dushantha Nalin K.; Sharma, Vishal; Sharma, Nikhil; Muthuchidambaranathan, P; Mabed, Hakim
Energy efficiency is a major concern in the emerging mobile cellular wireless networks since massive connectivity is to be expected with high energy requirements from the network operators. Non-orthogonal multiple access (NOMA) being the frontier multiple access scheme for 5G, there exists numerous research attempts on enhancing the energy efficiency of NOMA enabled wireless networks while maintaining its outstanding performance metrics such as high throughput, data rates and capacity maximized optimally. The concept of green NOMA is introduced in a generalized manner to identify the energy efficient NOMA schemes. These schemes will result in an optimal scenario in which the energy generated for communication is managed sustainably. Hence, the effect on the environment, economy, living beings, etc is minimized. The recent research developments are classified for a better understanding of areas which are lacking attention and needs further improvement. Also, the performance comparison of energy efficient, NOMA schemes against conventional NOMA is presented. Finally, challenges and emerging research trends, for energy efficient NOMA are discussed.
Hardware impaired self-energized bidirectional sensor networks over complex fading channels
(MDPI AG, 2020-09-29) Panic, Stefan R.; Jayakody, Dushantha Nalin K.; Affes, Sofiene; Muthuchidambaranathan, Palanivelu
Rapid emergence of wireless sensor networks (WSN) faces significant challenges due to limited battery life capacity of composing sensor nodes. It is substantial to construct efficient techniques to prolong the battery life of the connected sensors in order to derive their full potential in the future Internet of Things (IoT) paradigm. For that purpose, different energy harvesting (EH) schemes are relying on a wide array of sources. Following the same objective, in this work, we have observed a time-switching EH for half-duplex (HD) bidirectional WSN, which performed in-between relaying over Hoyt fading channels. For its comprehensive performance analysis, rapidly converging infinite-series expressions have been provided with focus on the outage probability (OP) and achievable throughput of the hardware-impaired system. Additionally, asymptotic behavior of these performance measures has also been provided. Further, an approach to the symbol-error probability (SEP) analysis is also presented in the context of the observed system. Finally, we consider the shadowing influence along the WSN propagation path. Performance analysis of observed EH system operating over Rician-shadowed fading channels has been carried out, with deriving exact corresponding infinite-series expressions for outage probability (OP) and achievable throughput of the system under the hardware impairment conditions. In addition, bidirectional relaying in a mixed fading environment has been considered.