Bhavik Ardeshna

I see a purposeful and versatile vocation in natural language processing, yet what I love most is its fusion of a scientific core with a semantic understanding of languages and also aligning linguistic and graphical understanding with an abstraction of multi-modular tactics.

I am an author for Becoming Human: Artificial Intelligence Magazine featured on Google. My neoteric article [1] and [2] accentuates the outlook of the CQA and visiolinguistic learning. Additionally, I am opensource contributer at Ivy - The Unified Machine Learning Framework and worked on nest functional API #PR2002. I have also published 14 multilingual transformer models in HuggingFace🤗 for seven divered languages for QA downstream task.

During my time at DDU, I was engaged in research natural language, where I worked with Dr. Brijesh Bhatt and Prof. Harium Pandya on QA downstream task for low-resource languages and investigate the efficacy for cross-lingual transfer using parameter-efficient adapter. Also generated the GujaratiQASuite benchmark for Gujarati QA system.

Before that, I completed my SDE internship from Heliconia Solutions, worked with automation team for developing utility tools for their product. I am also a Kaggle 3x Expert, published many benchmarking datasets and notebooks which accentuate the concepts of Weights & Biases (wandb), Transformers, EDA, and various ensemble techniques..

Email  /  Resume  /  LinkedIn  /  GitHub /  Kaggle /  Medium /  HuggingFace🤗

I'm broadly interested in natural language processing, visiolinguistic representations zero-shot learning. I have worked on low-resource language models, transformers, adapters, cross-lingual transfer, image-text feature alignment. I have also explored the conversational question answering(CQA), open-domain learning.

Gujarati Question Answering Dataset (GujQA), which is the baseline QA benchmark for the Gujarati language. Additionally, we have offered thorough assessments of the GujQA using a variety of linguistic criteria and have seen encouraging outcomes. We believe that GujQA, GujAdapter, and Gujwiki will not only advance the study of the Gujarati language’s understudied QA but also provide a door to the cross-lingual study involving the languages of the typologically varied domain

We have investigated the efficacy of cascading adapters with transformer models to leverage high-resource language to improve the performance of low-resource languages on the question answering task. We trained four variants of adapter combina- tions for - Hindi, Arabic, German, Spanish, English, Vietnamese, and Simplified Chinese languages. We demonstrated that by using the transformer model with the multi-task adapters, the performance can be improved for the downstream task.

The researcher has been working to develop an array of intelligent dialogue system that not only matches or surpasses a human’s level in carrying out an interactive conversation but also answers questions on a variety of topics.

Vision-language tasks, such as image captioning, visual question answering, and visual commonsense reasoning, serve as rich test-beds for evaluating the reasoning capabilities of visually informed systems.

It emphasizes the featurization of image and question, feature fusion, and answer generation for the multimodal system for VisioLinguistic tasks, using different text and vision transformers to evaluate the efficacy of QA downstream task using a multimodular framework.

We trained four variants of adapter combinations for - Hindi, Arabic, German, Spanish, English, Vietnamese, and Simplified Chinese languages. We demonstrated that by using the transformer model with the multi-task adapters, the performance can be improved for the downstream task.

Features Provided By PDFConverter Tools

• Rotate Pdf [90°,180°,270°,360°]
• Merge Pdf (Merger all uploaded or droped pdfs)
• Split Pdf (Split the pdf as per provided page numbers)
• Convert Pdf to [JPEG, PNG, HTML, Text]
• Crop Pdf & Download the modified pdf or zip

Creatively.io provides features to create their own canvas and with the help of its beautifully designed UI user can share to the world.

InstaBook was smartly created using Django and other tools such that it supports all features which are provided by Instagram. InstaBook is created to change social media communication and many new users would get attracted and use InstaBook and be socialize.

VQA is a multimodal task wherein, given an image and a natural language question related to the image, the objective is to produce a natural language answer correctly as output. VQA entails a wide range of sub-problems in both CV and NLP (such as object detection and recognition, scene classification, counting, and so on). Thus, it is considered an AI-complete task.

The Yahoo! Answers topic classification dataset is constructed using the 10 largest main categories. Each class contains 140,000 training samples and 6,000 testing samples. Therefore, the total number of training samples is 1,400,000, and testing samples are 60,000 in this dataset. From all the answers and other meta-information, we only used the best answer content and the main category information.

The Amazon reviews dataset consists of reviews from amazon. The data span a period of 18 years, including ~35 million reviews up to March 2013. Reviews include product and user information, ratings, and a plaintext review. It supports text classification and sentiment-classification: The dataset is mainly used for text classification: given the content and the title, predict the correct star rating.

Website template taken from Jon Barron.