Conference Papers
Permanent URI for this collectionhttps://idr.nitk.ac.in/handle/123456789/28506
Browse
5 results
Search Results
Item Unsupervised KeyPhrase Extraction using Graph and Embedding Techniques(Institute of Electrical and Electronics Engineers Inc., 2023) Kumar S, J.K.; Anand Kumar, M.The process of extracting keyphrases from a document automatically, without any supervision, is referred to as Unsupervised Keyphrase Extraction. This method aims to produce a brief summary of the main content of the document. Embedding-based methods comprise computing similarity between candidate keyphrases and documents embeddings. In this paper, we find that filtering candidate keyphrases using graph-based techniques enriches frequent candidates which are reranked using embeddings. On comparing the proposed model to the current state-of-the-art unsupervised Keyphrase Extraction approaches across three KPE benchmarks, it was found that the proposed model outperformed them. © 2023 IEEE.Item Creation and Classification of Kannada Meme Dataset: Exploring Domain and Troll Categories(Springer Science and Business Media Deutschland GmbH, 2024) Kundargi, S.Y.; N, N.; Anand Kumar, M.; Chakravarthi, B.R.In this pioneering research, the first-ever Kannada memes dataset is established, marking a groundbreaking contribution. This dataset encompasses 2002 memes, spanning various categories such as movies, politics, sports, trolls, and non-troll memes. The classification models have been meticulously fine-tuned for memes, incorporating image-based models using DenseNet169 and text-based models with BERT for text encoding. An innovative multimodal approach combines insights from images and text, acknowledging the comprehensive nature of meme content. Throughout the study, model strengths and weaknesses are assessed, emphasizing their reliance on cutting-edge technologies like Deep Learning and Natural Language Processing. Valuable improvements are recommended, such as the implementation of oversampling techniques and regular dataset updates to enhance relevance and accuracy. This work extends beyond immediate research, contributing to the development of adaptive meme classification systems, particularly for Kannada-speaking audiences within the evolving meme culture landscape. Notably, the results indicate that multimodal models achieved the best scores for domain classification, while image-based models excelled in troll meme classification, further highlighting the significance of this approach within the field. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.Item End-to-End Space-Efficient Pipeline for Natural Language Query based Spacecraft Health Data Analytics using Large Language Model (LLM)(Institute of Electrical and Electronics Engineers Inc., 2024) Ram, G.V.R.; Ashinee, K.; Anand Kumar, M.There is a requirement of automated Space-craft Health monitoring and mission maintenance System which is able to process Natural-Language Query and revert back in required format for which size of space database is a hurdle. Hence, we propose an end-to-end customizable real-time pipeline for space mission health monitoring, utilizing LLM that addresses issue of very large databases by extracting only relevant columns in initial stages of pipeline itself leveraginf BERT for NER, LLM for fetching schema and PandasAI to execute these queries on large datasets efficiently, producing user-friendly outputs. The pipeline is robust, space-efficient, and customizable, offering features such as cross-table referencing and handling same feature names in multiple tables. We achieved 70% realtime accuracy. © 2024 IEEE.Item SCaLAR NITK at Touché: Comparative Analysis of Machine Learning Models for Human Value Identification(CEUR-WS, 2024) Praveen, K.; Darshan, R.K.; Reddy, C.T.; Anand Kumar, M.This study delves into task of detecting human values in textual data by making use of Natural Language Processing (NLP) techniques. With the increasing use of social media and other platforms, there is an abundance in data that is generated. Finding human values in these text data will help us to understand and analyze human behavior in a better way, because these values are the core principle that influence human behavior. Analyzing these human values will help not only in research but also for practical applications such as sentiment evaluation, market analysis and personalized recommendation systems. The study tries to evaluate the performance of different existing models along with proposing novel techniques. Models used in this study range from simple machine learning model like SVM, KNN and Random Forest algorithms for classification using embeddings obtained from BERT till transformer models like BERT and RoBERTa for text classification and Large Language Models like Mistral-7b. The task that has be performed is a multilabel, multitask classification. QLoRA quantization method is used for reducing the size of weights of the model which makes it computationally less expensive for training and Supervised Fine Tuning (SFT) trainer is used for fine tuning LLMs for this specific task. It was found that LLMs performed better compared to all other models. © 2024 Copyright for this paper by its authors.Item Sexism Identification Using Annotator Ranking in Memes: A Multimodal Approach Using Transformers(CEUR-WS, 2025) Jha, D.K.; Mandal, M.K.; Anand Kumar, M.Memes are a popular medium for sharing information on social media, often embedding humor and interactive content. However, they can also propagate sexism, targeting specific genders, particularly females. This paper presents a multimodal approach to detect sexism in memes and classify the intent of sexist memes and sexism categorization. We leverage BERT for textual analysis, BLIP for multimodal processing, and Vision Transformers (ViT) for image feature extraction. Our model achieves approximately 68.49% accuracy in identifying sexist memes and 68.52% accuracy in determining the source intention and 49.31% accuracy in Sexism Categorization. This work contributes to creating safer digital spaces by automating the detection of biased content on social media. © 2025 Copyright for this paper by its authors.