Conference Papers
Permanent URI for this collectionhttps://idr.nitk.ac.in/handle/123456789/28506
Browse
3 results
Search Results
Item Employing Differentiable Neural Computers for Image Captioning and Neural Machine Translation(Elsevier B.V., 2020) Sharma, R.; Kumar, A.; Meena, D.; Pushp, S.In the history of artificial neural networks, LSTMs have proved to be a high-performance architecture at sequential data learning. Although LSTMs are remarkable in learning sequential data but are limited in their ability to learn long-term dependencies and representation of certain data structures because of the lack of external memory. In this paper, we tackled two main tasks, one is language translation and other is image captioning. We approached the problem of language translation by leveraging the capabilities of the recently developed DNC architectures. Here we modified the DNC architecture by including dual neural controllers instead of one and an external memory module. Inside our controller, we employed a neural network with memory-augmentation which differs from the original differentiable neural computer, we implemented a dual controller's system in which one controller is for encoding the query sequence whereas another controller is for decoding the translated sequences. During the encoding cycle, new inputs are read and the memory is updated accordingly. In the decoding cycle, the memory is protected from any writing from the decoding controller. Thus, the decoder phase generates a translated sequence at a time step. Therefore, the proposed dual controller neural network with memory-augmentation is then trained and tested on the Europarl dataset. For the image captioning task, our architecture is inspired by an end-to-end image captioning model where CNN's output is passed to RNN as input only once and the RNN generates words depending on the input. We trained our DNC captioning model on 2015 MSCOCO dataset. In the end, we compared and shows the superiority of our architecture as compared to conventionally used LSTM and NTM architectures. © 2020 The Authors. Published by Elsevier B.V.Item Loss Optimised Video Captioning using Deep-LSTM, Attention Mechanism and Weighted Loss Metrices(Institute of Electrical and Electronics Engineers Inc., 2021) Yadav, N.; Naik, D.The aim of the video captioning task is to use multiple natural-language sentences to define video content. Photographic, graphical, and auditory data are all used in the videos. Our goal is to investigate and recognize the video's visual features, as well as to create a caption so that anyone can get the video's information within a second. Despite the fact, that phase encoder-decoder models have made significant progress, but it still needs many improvements. In the present work, we enhanced the top-down architecture using Bahdanau Attention, Deep-Long Short-Term Memory (Deep-LSTM) and weighted loss function. VGG16 is used to extract the features from the frames. To understand the actions in the video, Deep-LSTM is paired with an attention system. On the MSVD dataset, we analysed the efficiency of our model, which indicates a major improvement over the other state-of-art model. © 2021 IEEE.Item Analysis of Tweets for Cyberbullying Detection(Institute of Electrical and Electronics Engineers Inc., 2023) Mathur, S.A.; Isarka, S.; Dharmasivam, B.; Jaidhar, C.D.Cyberbullying takes place online on gadgets like smartphones and computers. Cyberbullying can occur through social media platforms. This paper presents a real-time cyber-bullying detection system for Twitter using Natural Language Processing (NLP) and Machine Learning (ML). The system is trained on a dataset of cyberbullying tweets using several ML algorithms and their performance is compared. Random Forest was found to provide the best results after tuning. To achieve real-time analysis, Selenium was used to scrape tweets from a given Twitter account and store the timestamp of the already checked tweets. Additionally, an image captioning model was employed to generate descriptions for images posted on the account and compare them with user-written captions to filter out spam tweets. The proposed work aims to prevent cyberbullying and provides a valuable tool for online platforms to detect and remove harmful content. The results of this study have shown that the selection of appropriate ML algorithms and preprocessing techniques significantly impact the performance of cyberbullying detection on Twitter. Our model sheds light on the appropriateness of different ML algorithms for the detection of cyberbullying. © 2023 IEEE.