The "INDIAai" National AI Portal of India offers weekly articles that showcase the AI research contributions of universities across the country. Each article provides an in-depth report on the work of a specific university, allowing researchers and students to present concise explanations of their research. 

BVRIT HYDERABAD College of Engineering for Women (BVRITH), established in 2012 by Sri Vishnu Educational Society (SVES) under the leadership of Chairman Sri K V Vishnu Raju, aims to empower women to pursue better careers. Initially admitting 240 students across four branches—EEE, ECE, CSE, and IT—the college now accommodates 540 students and offers new programs in emerging technologies, including UG courses in CSE (AI&ML) and a PG course in M.Tech (Data Science). BVRITH, affiliated with Jawaharlal Nehru Technological University Hyderabad and approved by AICTE and the Government of Telangana, was granted UGC autonomous status starting from the academic year 2023-24. The institution is accredited by NAAC with an 'A' Grade, and its UG programs in EEE, ECE, CSE, and IT are accredited by NBA.

This week, the portal highlights the top AI research contributions from BVRIT Hyderabad College of Engineering for Women, Hyderabad.

Applied-behavioural analysis therapy for autism spectrum disorder students through virtual reality

Authors:

• T. Subetha, Department of Information Technology, BVRIT HYDERABAD College of Engineering for Women, Hyderabad, India
• Kayal Padmanandam, Department of Information Technology, BVRIT HYDERABAD College of Engineering for Women, Hyderabad, India
• L. Lakshmi, Department of AI/ML, BVRIT HYDERABAD College of Engineering for Women, Hyderabad, India
• S.L. Aruna Rao, Department of Information Technology, BVRIT HYDERABAD College of Engineering for Women, Hyderabad, India

Autism spectrum disorder (ASD) affects social interactions and understanding of social cues. Interactive virtual reality (VR) technologies offer a promising solution to help individuals with ASD learn social behaviors. This study aims to develop VR-based training using applied behavior analysis therapy to teach essential social-communication skills. The system includes two sessions: a training session with VR content to demonstrate social skills, and a practice session where students apply what they learned. A multimodal gesture recognition system and deep neural networks identify student gestures and speech. Successful interactions are compiled into a video using automatic video self-modelling (VSM), helping learners improve specific social behaviors. A comparative study shows that students using this system have better learning and communication skills in the real world, fulfilling the hopes of their parents and families.

This study leverages AI technology to help individuals with Autism Spectrum Disorder (ASD) improve social interactions. It uses interactive virtual reality (VR) to create a training program based on applied behavior analysis therapy. The program includes a training session to teach social skills and a practice session for students to apply what they've learned. AI technologies, such as multimodal gesture recognition systems and deep neural networks, are used to identify student gestures and speech.

Effective Automated Transformer Model based Sarcasm Detection Using Multilingual Data

Authors:

• Vidyullatha Sukhavasi, Department of CSE, BVRIT HYDERABAD College of Engineering for Women, Bachupally, Telangana, 500090, India
• Venkatesulu Dondeti, Department of Advanced CSE, Vignan’s Foundation for Science, Technology & Research, Guntur, Andhra Pradesh, 522213, India

Sarcasm detection on social media is important for understanding the true meaning behind posts. Simply analyzing text isn't enough for modern social networks; emojis also play a crucial role. This study introduces a model that analyzes both text and emojis in Hindi and English to detect sarcasm. Traditional transformer models haven't performed well enough, so this new model uses an attention-based transformer for better results. The process begins with pre-processing steps like stop word removal and tokenization to clean the data. Text features are then extracted using the ATF-IDF method, and a Gated Temporal Bidirectional Convolution Network (GT-BiCNet) creates the text model. An emoji-to-vector model (E-VM) converts emojis into feature vectors. These features are combined using deep feature fusion and classified using an Attention LSTM model with ALABerT. The Enhanced Pelican Optimization Algorithm (EpoA) reduces network losses, and a softmax layer classifies the data as sarcastic or non-sarcastic. The model outperforms existing methods, achieving 99.1% accuracy on English Twitter data and high accuracy, recall, precision, and F-measure on Hindi Twitter data.

This study presents an AI model for detecting sarcasm on social media by analyzing both text and emojis in Hindi and English. Traditional transformer models were insufficient, so an attention-based transformer was developed for better performance.

Artificial rabbits optimization based reconfiguration and distributed generation allotment in the distribution network

Authors:

• Ganney Poorna Chandra Rao, Department of Electrical and Electronics Engineering, Vignana Bharathi Institute of Technology, Hyderabad, India 
• Pallikonda Ravi Babu, Department of Electrical and Electronics Engineering, Sreenidhi Institute of Science and Technology, Hyderabad, India
• Mailugundla Rupesh, Department of Electrical and Electronics Engineering, BVRIT Hyderabad College of Engineering for Women, Hyderabad, India
• Puvvula Venkata Rama Krishna, Department of Electrical Electronics and Communication Engineering, GITAM (Deemed to be University), Hyderabad, India

This study uses AI to reduce power losses and improve voltage stability in power networks. Distributed generation (DG) produces affordable energy close to where it is used. The method combines network reconfiguration with DG distribution to enhance efficiency. The loss sensitivity factor (LSF) helps find the best spot for DG units, and the artificial rabbits optimization (ARO) technique identifies the ideal network setup and DG size. Testing on two power systems showed that combining reconfiguration with DG distribution is more effective than reconfiguration alone.

PenBOT—Make Transcribing Easy with an AI Scribe

Authors:

• N. M. Sai Krishna, Department of Electronics and Communication Engineering, BVRIT Hyderabad College of Engineering for Women, Hyderabad, Telangana, India
• R. Priyakanth, Department of Electronics and Communication Engineering, BVRIT Hyderabad College of Engineering for Women, Hyderabad, Telangana, India
• C. Srinika Sharma, Department of Computer Science and Engineering—Artificial Intelligence and Machine Learning, BVRIT Hyderabad College of Engineering for Women, Hyderabad, Telangana, India
• Chithra Bhanu Aalla, Department of Computer Science and Engineering—Artificial Intelligence and Machine Learning, BVRIT Hyderabad College of Engineering for Women, Hyderabad, Telangana, India
• Sudiksha Kolluru, Department of Computer Science and Engineering—Artificial Intelligence and Machine Learning, BVRIT Hyderabad College of Engineering for Women, Hyderabad, Telangana, India
• Grahya Yalavarthy, Department of Electronics and Communication Engineering, BVRIT Hyderabad College of Engineering for Women, Hyderabad, Telangana, India
• K. Sai Uma Maheswari, Department of Electronics and Communication Engineering, BVRIT Hyderabad College of Engineering for Women, Hyderabad, Telangana, India

In the past, personal medical scribes were a dependable alternative for physicians who had an excessive number of patients. Frequently, medical students who were candidates for this position departed abruptly, resulting in a loss of time and money for providers in the recruitment process. However, contemporary AI-powered scribe software can alleviate physicians' duties by reducing data entry and EHR organization, thereby enabling them to concentrate on patient care. This restores the enjoyment of practicing medicine to physicians. Not to be burdened with documentation, physicians are intended to provide care to those in need.

PenBOT employs speech recognition technology, machine learning, and natural language processing to manage multiple speakers, filter minor conversations, and navigate interruptions. Physicians will be able to concentrate on delivering the highest quality of patient care and alleviate the stress associated with tedious documentation duties, all while ensuring the safety of patients at all times, using this technology. This technology helps doctors maintain high-quality patient care while minimizing the stress and burden of paperwork.

Performance Analysis of Cycle GAN in Photo to Portrait Transfiguration Using Deep Learning Optimizers

Authors:

• L. Lakshmi, Department of DS and AI, Faculty of Science and Technology (IcfaiTech), ICFAI Foundation for Higher Education, Hyderabad, India
• A. Naga Kalyani, Department of CSE (AI and ML), BVRIT Hyderabad College of Engineering for Women, Hyderabad, India
• D. Krishna Madhuri, Department of DS and AI, Faculty of Science and Technology (IcfaiTech), ICFAI Foundation for Higher Education, Hyderabad, India
• Sirisha Potluri, Department of CSE, Faculty of Science and Technology (IcfaiTech), ICFAI Foundation for Higher Education, Hyderabad, India
• Geetika Silakari Pandey, Department of CSE (AI and ML), BVRIT Hyderabad College of Engineering for Women, Hyderabad, India
• Shahid Ali, Dept. of Electronics Engineering, Peking University, Beijing, China
• Muhammad Ijaz Khan, Dept. of Mathematics and Statistics, Riphah International University, Islamabad, Pakistan
• Fuad A. Awwad, Dept. of Quantitative Analysis, College of Business Administration, King Saud University, Riyadh, Saudi Arabia
• Emad A. A. Ismail, Dept. of Quantitative Analysis, College of Business Administration, King Saud University, Riyadh, Saudi Arabia

In computer vision, image transformations are crucial for various applications like healthcare, image enhancement, and art analysis. Traditionally, training models requires lots of annotated images. Cycle-GAN is a powerful tool that allows for training with fewer images by using unsupervised learning.

This study presents a system that creates Monet-style paintings from realistic photos using Cycle-GAN. Due to the limited number of Monet paintings available, the system uses generator and discriminator neural networks to generate new Monet-style images. The model is trained with a dataset that includes 300 Monet paintings and 7,028 natural photos, using deep learning optimizers like RMSprop, ADAM, and SGD. The system performs well, especially with the SGD optimizer, showing good results in creating Monet-style artwork.

AI in computer vision, particularly with Cycle-GAN, enables the creation of Monet-style paintings from realistic photos by using unsupervised learning. This approach overcomes the challenge of needing large annotated image datasets by generating new artworks with fewer samples. The system uses generator and discriminator neural networks trained with a mix of Monet paintings and natural photos, showing impressive results, especially with the SGD optimizer.