Research

Publications

My research focuses on enhancing the reliability and interpretability of language understanding systems and improving the efficiency of deep learning models. I am particularly interested in the evaluation of sentence acceptability and how language models can be guided by structured representations and human judgments. I also focus on unsupervised methods for information extraction and summarization in real-world applications like e-commerce. Another key area of interest is the integration of external signals, such as weather data, to improve predictive modeling. Overall, my work aims to bridge linguistic insights with practical, scalable machine learning solutions.

Papers

1. Vijay Daultani, Hector Javier Vazquez Martinez, Naoaki Okazaki, Acceptability Evaluation of Naturally Written Sentences, IPSJ Transactions on Databases (IPSJ TOD), 2024.
   [DOI]
2. Vijay Daultani, Naoaki Okazaki, Improving Automatic Evaluation of Acceptability Based on Language Models with a Coarse Sentence Representation, Pacific Asia Conference on Language, Information and Computation (PACLIC), 2022.
   [URL]
3. Jeremiah Anderson, Vijay Daultani, Tariq Muman, Mohamed Batran, The Importance of Weather to E-Commerce Orders Forecasting, IEEE International Conference on E-Business Engineering (EBEE), 2019.
   [DOI]
4. Vijay Daultani, Lasguido Nio, Young-Joo Chung, Unsupervised Extractive Summarization for Product Description using Coverage Maximization with Attribute Concept, IEEE International Conference on Semantic Computing (ICSC), 2019.
   [DOI]
5. Vijay Daultani, Ohno Yoshiyuki, Ishizaka Kazuhisa, Sparse Direct Convolutional Neural Network, International Symposium on Neural Networks (ISNN), 2017.
   [DOI] [URL]
6. Vijay Daultani, Subhajit Chaudhury, Ishizaka Kazuhisa, Convolutional Neural Network Layer Reordering for Acceleration, Synthesis And System Integration of Mixed Information Technologies (SASIMI), 2016.
   [URL]

Patents

1. Summary Generating Device, Summary Generating Method, and Information Storage Medium: Vijay Daultani, Lasguido Nio, Youngjoo Chung. Filed: 2019, Patent No: US/2020/0134011
   [Link]
2. Search System, Search Method, and Program: Vijay Daultani. Filed: 2018, Patent No: WO/2019/176102
   [Link]
3. Information Processing System, Information Processing Method, and Program: Robin Swezey, Jing Mi, Vijay Daultani. Filed: 2018, Patent No: WO/2020/031232
   [Link]
4. Improved Sparse Convolutional Neural Network: Vijay Daultani. Filed: October 2016, Application No: PCT/JP2016/081973
   [Link]
5. Information Processing Method and Device for Neural Network: Vijay Daultani. Filed: June 2016, Application No: PCT/JP2016/068741
   [Link]
6. Information Processing Apparatus, Information Processing Method and Storage Medium Storing Program: Vijay Daultani. Filed: February 2016, Application No: PCT/JP2016/002686
   [Link]

Presentations

1. Bringing India at the Forefront of AI Research
   Venue: Sage University
   Date: Apr. 16 2025
   [Presentation]
2. Imminent AI – Generative Adversarial Networks
   Venue: Rakuten Technology Conference
   Date: Nov. 9 2019
   [Video]
3. Recurrent Neural Networks
   Venue: Rakuten Technology Conference
   Date: Oct. 28 2017
   [Presentation] [Video]
4. Association Rules Venue: Internal Presentation at Rakuten
   Date: March 2017
   [Presentation]

White Papers

1. Bringing India at the Forefront of AI Research
   [Link]

Personal Research Projects

1. Data Visualization using Seaborn
   I explored and visualized the Abalone dataset using Seaborn to uncover patterns in physical measurements and age classification. My work involved generating insightful plots such as pair plots, box plots, and heatmaps to analyze correlations and distributions across attributes. These visualizations aided in identifying relationships between features like shell dimensions and the number of rings (a proxy for age), supporting further predictive modeling efforts.
   [Code]

2. Parallel Computation using MPI
   As part of my coursework in parallel computation, I developed two high-performance computing assignments using MPI. In the first, I implemented a matrix averaging algorithm using block-cyclic distribution over a 2D process grid to optimize load balancing. In the second, I designed an efficient broadcast mechanism over a 2D torus network using multiple non-contending spanning trees to reduce network congestion.
   [Code]

3. Pintos
   This project involves extending the Pintos operating system, a teaching OS for the 80x86 architecture developed at Stanford. As part of a course on Resource Management in Computer Science, key modules were implemented: user program loading, virtual memory management, and file system support. The work enhances Pintos’ capability to handle user-space programs, manage memory efficiently, and perform file operations.

4. Energy Breakdown of Static and Dynamic Non-Uniform Cache Access
   This project focuses on analyzing energy consumption in Static and Dynamic Non-Uniform Cache Access (NUCA) architectures. It investigates power usage, heat generation patterns, and inter-bank data access behaviors. The study includes experiments comparing cache mapping policies, identifying high power-consuming areas, and evaluating hop counts under different write policies.
   [Presentation]