Examining the Sentiment Expressed in Tweets Related to COVID-19 and the Omicron Variant Using Deep Learning Classifiers
Conference paper, Lecture Notes in Electrical Engineering, 2025, DOI Link
View abstract ⏷
This study employs advanced deep learning models, including convolutional neural networks (CNN), recurrent neural networks (RNN), hybrid architectures, bidirectional long short-term memory (BiLSTM) networks, and transformers, to analyze sentiment in COVID-19 and Omicron-related tweets. The goal is to explore the relationship between social media popularity and classification accuracy while addressing challenges associated with false information during the pandemic. The research aims to enhance accuracy in identifying misinformation, offering insights for public health, digital literacy, and crisis management. Comparative analysis of the models reveals their strengths and weaknesses, establishing a benchmark for future misinformation detection studies. While emphasizing the importance of accurate information during crises, the study acknowledges limitations such as a lack of multilingual analysis, Twitter-centric focus, and potential bias in sentiment analysis datasets. The difficulties in interpreting massive neural networks and the transformative impact of social media on information dissemination are also recognized. Results showcase accuracy metrics for different classifiers, highlighting variations in sentiment analysis performance across datasets. In conclusion, the study contributes to understanding misinformation complexities during the pandemic, providing a nuanced analysis of sentiment in social media. It establishes a foundation for future studies on misinformation detection, emphasizing the crucial role of accurate information in navigating global challenges. However, it falls short in detailing potential social and regulatory repercussions from social media restrictions.
MATSFT: User query-based multilingual abstractive text summarization for low resource Indian languages by fine-tuning mT5
Article, Alexandria Engineering Journal, 2025, DOI Link
View abstract ⏷
User query-based summarization is a challenging research area of natural language processing. However, the existing approaches struggle to effectively manage the intricate long-distance semantic relationships between user queries and input documents. This paper introduces a user query-based multilingual abstractive text summarization approach for the Indian low-resource languages by fine-tuning the multilingual pre-trained text-to-text (mT5) transformer model (MATSFT). The MATSFT employs a co-attention mechanism within a shared encoder–decoder architecture alongside the mT5 model to transfer knowledge across multiple low-resource languages. The Co-attention captures cross-lingual dependencies, which allows the model to understand the relationships and nuances between the different languages. Most multilingual summarization datasets focus on major global languages like English, French, and Spanish. To address the challenges in the LRLs, we created an Indian language dataset, comprising seven LRLs and the English language, by extracting data from the BBC news website. We evaluate the performance of the MATSFT using the ROUGE metric and a language-agnostic target summary evaluation metric. Experimental results show that MATSFT outperforms the monolingual transformer model, pre-trained MTM, mT5 model, NLI model, IndicBART, mBART25, and mBART50 on the IL dataset. The statistical paired t-test indicates that the MATSFT achieves a significant improvement with a p-value of ≤ 0.05 compared to other models.
Sentimental Analysis on Drug Reviews Using Fined Tuned Transformers
Garaga M.R., Paleti D., Syed G.A., Gudivaka B.R., Maddi K., Mudigonda K.S.P.
Conference paper, 2024 15th International Conference on Computing Communication and Networking Technologies, ICCCNT 2024, 2024, DOI Link
View abstract ⏷
Main goal of this work is to analysis the drug review by using sentimental analysis. Nowadays we can see media platforms has become a portion of everyone's lives, they share most of their views and interest in social media platforms like review sites, twitter etc. As social media has become an easier way to communicate, many individuals are posing their opinions in it which also include drug related reviews like providing useful remedies and providing valuable understandings which makes pharmacological companies more useful. In this work, we are mainly targeting on finding the sentiment score of drug reviews which are acquired from 'drugs.com' and 'drugslib.com' sites. Here we performed some preprocessing techniques on the data and then calculated the accuracy of each model LSTM, RNN, CNN and BERT, on comparing the accuracy we proposed that LSTM is giving the best accuracy when compared to other models with the accuracy of 97%.
Predictive Control Strategy for DC Microgrid Integrated EV Charging Stations in Oslo
Naik K.R., Kolhe M.L., Prasad M.K.S., Agundis G.D., Vasquez J.C., Guerrero J.M.
Conference paper, 2024 IEEE 4th International Conference on Sustainable Energy and Future Electric Transportation, SEFET 2024, 2024, DOI Link
View abstract ⏷
Norway holds highest per capita Electric vehicle (EV) shares globally. To meet the recharge demand from such large EVs fleet, fast charging infrastructure is essential. But, the fast charging stations are unable to fulfill the charging request of EV owners due to restricted grid power supply. Peak shaving has partially addressed this issue with a solar PV-Battery energy storage system (BESS) integrated Microgrid (MG) configuration. But, maintaining the reliability and stability of Microgrid (MG) against intermittent EV arrivals during peak load of the grid is a critical objective. To address this issue a Micro hydro generator-Solar PV-BESS integrated DC MG with iterative predictive control (ITPC) algorithm is proposed in this paper. With the prediction of EV load arrivals on iteration basis, the proposed algorithm operates the Micro hydro generator in such a way that EV load is supplied under the peak load condition with optimal dependency on BESS. The performance of the proposed strategy evaluated against the real time EV load data of Oslo as a case study. The proposed strategy achieves 9.3% improved fall in DC-link voltage and 67.735% reduced depth of discharge of BESS.
Extractive text summarization on medical insights using fine-tuned transformers
Prasad Mudigonda K.S., Lingineni N., Manisai Y., Pennada M., Gadde M., Sai Aluri R.
Article, International Journal of Computers and Applications, 2024, DOI Link
View abstract ⏷
Text summarization is a fundamental Natural language processing task that plays a crucial role in efficiently condensing large textual documents into concise and clear summaries for human comprehension. The amount of data being generated in the medical domain nowadays requires substantial application of the current deep learning approaches such as transformers. The main goal of this research is to extract relevant summaries from the abstracts of the research articles published related to cancer, blood cancer, tinnitus, and Alzheimer's. As the domain-related data requires special attention, our approach uses a fine-tuned transformer model, to guarantee that the summaries produced are not only brief but also accurate. Moreover, as a part of this research, we have effectively collected the information from PubMed and also prepared the data for analysis. A comparative analysis of the Bidirectional and Auto-Regressive Transformers (BART), Text-to-Text Transfer Transformer (T5), Textrank, and Lexrank models on the dataset is carried out in this study to understand the medical insights effectively. The fine-tuned transformer's performance in comparison with other models brings out a newer dimension for future studies.
Extractive Text Summarization of Clinical Text Using Deep Learning Models
Chandra Shekar G., Sai Teja K., Nithin Datta D., Geetha Sri Abhinay P., Krishna Siva Prasad M.
Conference paper, 2nd International Conference on Emerging Trends in Information Technology and Engineering, ic-ETITE 2024, 2024, DOI Link
View abstract ⏷
This project focused on using clinical text data from the PubMed dataset to train transformer models and deep learning models for text summarization. The primary goal was to develop a system capable of identifying and extracting meaningful information from large clinical texts. Using transformer models and deep learning techniques, the goal was to improve the search for information in the medical literature. The ROUGE score, a widely accepted metric for automated summary assessment, was used to analyze the performance of the trained models. This project involved not only training and optimizing transformer and deep learning models to obtain a comprehensive summary, but also comparing their ROUGE scores to determine which model outperformed the others. This comparative analysis was necessary to determine the most effective model for extracting important insights from clinical texts. The findings have the potential to significantly impact information in the clinical domain, providing researchers and healthcare professionals with faster access to critical information.
Capturing multiple emotions from conversational data using fine tuned transformers
Mudigonda K.S.P., Bulusu K., Sri Y., Damera A., Kode V.
Article, International Journal of Computers and Applications, 2024, DOI Link
View abstract ⏷
Emotion detection is one of the crucial topics of Natural Language Processing (NLP) in recent years, and now one of the biggest motivating factors in correct identification and interpretation of a wide range of emotional expressions in textual data. This research examines the use of Bidirectional Encoder Representations from Transformers (BERT), a trained transformer model for emotion detection in textual data. This analysis evaluates how good BERT is at identifying emotions such as surprise, anger, fear, happiness and sadness compared with ordinary machine learning as well as deep learning techniques. The addition of weighted emotions approach enhances the model performance and gives a deeper emotional context awareness making it more effective to deal with complicated emotional utterances. The purpose of the research is to develop a fine-tuned BERT model with a weighted emotion framework to enhance the accuracy of emotion classification in conversational text. The context is improving emotion recognition in scenarios where multiple emotions co-exist, addressing limitations in traditional models by capturing subtle and overlapping emotional expressions. In terms of training the methods multiple datasets are considered and also the research examines various models performance. The article further discusses possible application areas of BERT modified in light of NLP.
A deep dive of deep learning models to Emotion Detection using weighted emotions
Mudigonda K.S.P., Bulusu V.K., Sri V.Y., Damera A., Kode V.
Conference paper, 2024 15th International Conference on Computing Communication and Networking Technologies, ICCCNT 2024, 2024, DOI Link
View abstract ⏷
This research paper explores different types of deep learning architectures for emotion detection through the use of Long Short-Term Memory (LSTM) networks. The main focus in this analysis is LSTM, a recurrent neural network (RNN) which can be used to understand cultural context due to its ability of capturing time dependencies in sequences. This study also looks into Bidirectional LSTM (BiLSTM) with Convolutional Neural Networks (CNNs), CNNs and RNNs one after another, independent CNNs and RNNs, and CNNs integrated with LSTM layers. Special attention here is given to the highly flexible and effective LSTM networks that incredibly capture even the most subtle emotional parameters as well as contextual information essential for improving accuracy in detecting emotions.
MMSFT: Multilingual Multimodal Summarization by Fine-Tuning Transformers
Article, IEEE Access, 2024, DOI Link
View abstract ⏷
Multilingual multimodal (MM) summarization, involving the processing of multimodal input (MI) data across multiple languages to generate corresponding multimodal summaries (MS) using a single model, has been under explored. MI data consists of text and associated images, while MS incorporates text alongside relevant images aligned with the MI context. In this paper, we propose an MM summarization model by fine-tuning transformers (MMSFT), focusing on low-resource languages (LRLs) such as the Indian languages. MMSFT comprises multilingual learning for encoder training, incorporating multilingual attention with a forget gate mechanism, followed by MS generation using a decoder. In the proposed approach, we use publicly available multilingual multimodal summarization dataset (M3LS). Evaluation utilizing ROUGE metrics and the language-agnostic target summary metric (LaTM) illustrates MMSFT's significant enhancement over existing MM summarization models like mT5 and VG-mT5. Furthermore, MMSFT yields better or equivalent summaries compared to existing MM summarization models trained separately for each language. Human and statistical evaluation reveal MMSFT's significant improvement over existing models, with a p-value ≤ 0.05 in paired t-tests.
Revitalizing the single batch environment: a ‘Quest’ to achieve fairness and efficiency
Article, International Journal of Computers and Applications, 2024, DOI Link
View abstract ⏷
In the realm of computer systems, efficient utilization of the CPU (Central Processing Unit) has always been a paramount concern. Researchers and engineers have long sought ways to optimize process execution on the CPU, leading to the emergence of CPU scheduling as a field of study. In this research, we have analyzed the single offline batch processing and investigated other sophisticated paradigms such as time-sharing operating systems and wildly used algorithms, and their shortcomings. Our work is directed toward two fundamental aspects of scheduling: efficiency and fairness. We propose a novel algorithm for batch processing that operates on a preemptive model, dynamically assigning priorities based on a robust ratio, employing a dynamic time slice, and utilizing periodic sorting to achieve fairness. By engineering this responsive and fair model, the proposed algorithm strikes a delicate balance between efficiency and fairness, providing an optimized solution for batch scheduling while ensuring system responsiveness.
Multiple Granularity Context Representation based Deep Learning Model for Disaster Tweet Identification
Conference paper, 2024 5th International Conference on Innovative Trends in Information Technology, ICITIIT 2024, 2024, DOI Link
View abstract ⏷
Twitter has evolved into a pivotal platform for information exchange, particularly during emergencies. However, amidst the vast array of data, identifying tweets relevant to damage assessment remains a significant challenge. In response to this challenge, this study presents a novel approach designed to identify tweets related to damage assessment in times of crises. The challenge lies in sifting through an immense volume of data to isolate tweets pertinent to the specific event. Recent studies suggest that employing contextual word embedding approaches, such as transformers, rather than traditional context-free methods, can enhance the accuracy of disaster detection models. This study leverages multiple granularity level context representation at the character and word levels to bolster the efficiency of deep neural network techniques in distinguishing between disaster-related tweets and unrelated ones. Specifically, the weighted character representation, generated with the self-attention layer, is utilized to discern important information at the fine character level. Concurrently, Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) algorithms are employed in the word-level embedding to capture global context representation. The effectiveness of the proposed learning model is assessed by comparing it with existing models, utilizing evaluation measures viz., accuracy, F1 score, precision, and recall. The results demonstrate the effectiveness of our model compared to existing methods.
Blockchain-enabled SDN in resource constrained devices
Book chapter, Blockchain-based Cyber Security: Applications and Paradigms, 2024,
A Deep Learning Based Approach In The Prediction Of Tinnitus Disease For Large Population Data
Mudigonda K.S.P., Nallapuneni V., Thindi A., Popuri H., Koka A., Batchu N.
Conference paper, 2023 14th International Conference on Computing Communication and Networking Technologies, ICCCNT 2023, 2023, DOI Link
View abstract ⏷
Tinnitus is a frequent sensory disorder that puts a lot of strain on the patient. Usually, tinnitus results from disturbances occurring to the sensory systems, such as the peripheral seldom central, the somatosensory system, the head and neck, or a mix of the two. This can be found in people with high stress, anxiety, depression, and hearing disorders. Although there is progress in the medical domain using artificial intelligence (AI), research related to tinnitus using AI is limited. This work aims to bridge the gap using deep-learning techniques for evaluating the patient record by examining various parameters. The proposed research also aims to target the same to understand the severity and possible recommendations for tinnitus disease. Our findings forecast how patients will react to tinnitus treatments. From the patients' electroencephalography (EEG) data, predictive EEG variables are extracted, and later feature selection approaches are used to determine the prominent features. The patient's EEG features are supplemented by AI algorithms for training and forecasting treatment outcomes. Higher accuracy levels of the proposed model using AI help the practitioners suggest the proper diagnosis for the patients and also check the patient's recovery over a period of time.
Path and information content based semantic similarity measure to estimate word-pair similarity
Mudigonda K.S.P., Dasari C.M., Pikle N.C.K., Shinde S.B.
Conference paper, AIP Conference Proceedings, 2023, DOI Link
View abstract ⏷
Extracting semantic features of text in natural language processing activities are important for many applications. Measuring semantic similarity of text can be carried out by various methods. Given two concepts or two short texts, the similarity between them can be carried out by similarity measures like corpus based and knowledge based measures. Measures which are corpus based are application specific and this paper focuses on measuring semantic similarity using knowledge based measures. Existing knowledge based measures use either information content or path length between the concepts to evaluate the similarity. Hence, in this paper an approach which uses both information content and path length is designed to evaluate the similarity between the concepts and a thorough analysis is done on the benchmark datasets and with results it is shown that the proposed measure is more efficient than all the existing measures.
Exploring intrinsic information content models for addressing the issues of traditional semantic measures to evaluate verb similarity
Article, Computer Speech and Language, 2022, DOI Link
View abstract ⏷
Semantic similarity measures play an important role in many natural language processing and information retrieval activities. It is highly challenging to measure semantic similarity with higher accuracy. A notable branch of semantic similarity evaluation based on information content (IC) is popular in this aspect. Intrinsic information content (IIC) models are another wing of IC based evaluation. Both IC based and IIC based approaches majorly handled similarity evaluation of nouns. Research related to semantic similarity assessment of verb pairs are rarely discussed. To bridge this gap, this work examines various IC based, IIC based approaches on verb pairs. A detailed discussion of the existing measures and their drawbacks are mentioned in this work. Strategies based on information content, length and depth of the concepts are discussed and tested on benchmark datasets. Existing intrinsic information content models are enhanced by addressing various issues like (a) dealing concepts with no path in WordNet and (b) handling the synonym sets of verb concepts. Measures based on path length, intrinsic information content, combined strategies and non-linear strategies for verb pairs are thoroughly inspected. This paper also presents novel strategies to understand novel aspects that are not addressed before. The strategies are experimented by generating the synonym sets of required parts-of-speech which proved very effective in improving the correlation with human judgment. Results on benchmark datasets specify that the proposed approaches for verb similarity will be a guiding factor for understanding the natural language processing tasks.
Similarity of Sentences With Contradiction Using Semantic Similarity Measures
Article, Computer Journal, 2022, DOI Link
View abstract ⏷
Short text or sentence similarity is crucial in various natural language processing activities. Traditional measures for sentence similarity consider word order, semantic features and role annotations of text to derive the similarity. These measures do not suit short texts or sentences with negation. Hence, this paper proposes an approach to determine the semantic similarity of sentences and also presents an algorithm to handle negation. In sentence similarity, word pair similarity plays a significant role. Hence, this paper also discusses the similarity between word pairs. Existing semantic similarity measures do not handle antonyms accurately. Hence, this paper proposes an algorithm to handle antonyms. This paper also presents an antonym dataset with 111-word pairs and corresponding expert ratings. The existing semantic similarity measures are tested on the dataset. The results of the correlation proved that the expert ratings are in order with the correlation obtained from the semantic similarity measures. The sentence similarity is handled by proposing two algorithms. The first algorithm deals with the typical sentences, and the second algorithm deals with contradiction in the sentences. SICK dataset, which has sentences with negation, is considered for handling the sentence similarity. The algorithm helped in improving the results of sentence similarity.
Multi-sense Embeddings Using Synonym Sets and Hypernym Information from Wordnet
Article, International Journal of Interactive Multimedia and Artificial Intelligence, 2020, DOI Link
View abstract ⏷
Word embedding approaches increased the efficiency of natural language processing (NLP) tasks. Traditional word embeddings though robust for many NLP activities, do not handle polysemy of words. The tasks of semantic similarity between concepts need to understand relations like hypernymy and synonym sets to produce efficient word embeddings. The outcomes of any expert system are affected by the text representation. Systems that understand senses, context, and definitions of concepts while deriving vector representations handle the drawbacks of single vector representations. This paper presents a novel idea for handling polysemy by generating Multi-Sense Embeddings using synonym sets and hypernyms information of words. This paper derives embeddings of a word by understanding the information of a word at different levels, starting from sense to context and definitions. Proposed sense embeddings of words obtained prominent results when tested on word similarity tasks. The proposed approach is tested on nine benchmark datasets, which outperformed several state-of-the-art systems.
Modified Path Measure to Assess Sentence Similarity
Conference paper, 2018 Conference on Information and Communication Technology, CICT 2018, 2018, DOI Link
View abstract ⏷
Sentence similarity can be calculated by various measures, but the measures that can use semantic information between the words perform better compared to others. Out of these, the measures which are related to a corpus, or which are knowledge related are more significant in the sentence similarity domains. The applications which use knowledge based measures tend to give more accurate results and are very much in coincidence with human similarity. These measures use path length between the concepts or information content between the words to derive the similarity between the words. Some of the semantic similarity measures generate synonym sets of the words to evaluate similarity, but these measures focus mainly on generating noun or verb synonym sets which can be enhanced by generating all the synonym sets. In this paper, a metric PathM is proposed to calculate word-pair similarity by generating all the synonym sets of the words and it is enhanced to calculate the sentence similarity. The measure is compared with path measure and the results obtained are better in comparison with other measures.
Combining Common Words and Semantic Features for Sentence Similarity
Conference paper, 2018 9th International Conference on Computing, Communication and Networking Technologies, ICCCNT 2018, 2018, DOI Link
View abstract ⏷
Assessing the similarity of short texts or sentences is an important phase in many natural processing activities. The paper describes the importance and role of sentence similarity in various domains and also the paper provides a mechanism to calculate the similarity between short texts. The main idea in this paper is to extract the syntactic and semantic features between the sentences, to calculate the similarity between them. The syntactic features are evaluated by finding the common words between the sentences, whereas the semantic features are evaluated using the information content between the concepts of the sentences. For obtaining the information content between the concepts knowledge based measures are used. Three information content based measures are compared in this paper over bench mark sentence similarity dataset. The results show that the integration of syntactic and semantic features increases the performance of the system.
