Text Classification for Deception Detection in Amazon Reviews

Pre-processed using Regex, trained using K fold cross validation and Linear SVC classifier, determined the precision-recall accuracy and f-score. A comparison was done by adding additional features and different stemming and lemmatization process. Finally, did an exploratory analysis on the dataset using seaborn and Matplotlib to explore some of the linguistic and stylistic traits of the reviews and compared the two classes

Sequence Labelling for Movie Queries

CRF Tagging of Movie. Tagging on the movie queries dataset first done using BIO tags and then improved by adding more features and using POS tags.

Identifying the genders of the characters in the Eastenders British Soap Opera.

The following are performed to experiment and optimize the classification: preprocessing, tokenization, different ngram models, line and sentence level features, cross-validation, and more.

Word Representation and Text Classification with Neural Networks

Sentiment Analysis of Twitter and IMDB Large Movie Review Own word embedding is obtained by training a skip-gram neural network model using negative sampling. Glove Embeddings were also used as pre-trained embedding. Various algorithms like LSTM and CNN were explored. Finally, all these are applied to real-world text classification tasks like the sentiment analysis on Twitter and the IMDB Movie reviews.

Neural Machine Translation and Neural Dialogue Systems

Implemented a version of the seq2seq NMT model, and enriched that with attention, using a pre-trained BERT as a classifier model, implementing a series of dialogue act taggers and created an end-to-end dialogue system.

Deep learning Course-works

Critical analysis and study performed on Generative Adversarial Nets and on super-resolution using Convolutional Neural Networks. These two were used on the MNIST dataset and the PyTorch framework is explored

Data Mining Course-works

Data Preprocessing, Data exploration and visualization (exploring sklearn, seaborn, pandas, numpy, and matplotlib libraries), Data Warehousing and On-line Analytical Processing (exploring cubes library), Classification, clustering, and Association Analysis (exploring mlxtend library), Outlier Detection, Web Mining (exploring BeautifulSoup library)

Big Data Processing project

Ethereum Analysis involving Time Analysis, Scam analysis, and comparative evaluation of the usage of Spark and MrJob. Thus, the following are performed: The number of transactions per month, average transactions per month, top ten most popular stories, top ten most active miners, Lucrative scams, Scam over time and correlation, Gas used over time.

Stellapps Technologies Private Ltd

Consultant• June 2022 - Present

Purple Beard Training, London, UK

Data Technical Trainer• Jan 2023 - Mar 2023

Quest Global Technologies Ltd

Consultant• Aug 2018 - July 2020

HSBC Software Development (India) Pvt. Ltd & HSBC UK

Software Engineer & Onsite Co-ordinator• Jan 2008 - Aug 2014