NLP_Portfolio

Summary

Hi! My name is Ashwin Somasundaram. I am currently a senior studying CS at UT Dallas. This portfolio contains a comprehensive overiew of my experience with NLP. Natural Language Processing (NLP) has become one of the hottest fields in machine learning due to its diversity in applications. As teh demand for automation is constantly increasing, NLP has started to play a crucial role in analyzing vast amounts of text data to help create meaningful functions/conversations. From chatbots to sentiment analysis, machine translation, and speech recognition, NLP has transformed the way we interact with machines and computers. Through my undergradyate Human Language Technologies course, I have learned a great deal about this topic. I gained knowledge on numerous topics such as sentiment analysis, text generation, technologeis such as WordNet, Spacy, NLTK, etc. In the future I would love to collaborate on any NLP project with like-minded individuals. Please take a look at my portfolio below to understand the work I have done and where my abilities would be best suited.

Skills Gained:

• Numpy, Spacy, NLTK and various other python libraries.
• Neural Nets(CNNs,RNNs,LSTMs, etc) using SciKitLearn.
• General process of creating, training and building machine learning and deep learning models.
• Ability to work ony code indepently and on a team.
• Time management skills developed through semester projects.
• Acute Analytical thinking skills.

All my assignments/project from my NLP coursework

Overview of NLP

This document provides a high level overview of NLP, it's use cases and my reason for interest in it. It also goes on to describe the three approaches to NLP ranging from the 1960s to 2010s.

Porfolio Assignment 1

• This program takes in an input csv file of the details of people, processes the details, and re-organizes it in the form of objects to be respresented clearly.
• To run this program, execute python axs180183_portfolio1.py data/data.csv in terminal for a Unix/Linux based system.
• To run this program, execute python axs180183_portfolio1.py data/data.csv in powershell for a Windows based system.
• Please ensure that you have the proper libraries downloaded such as re, sys, pickle. Also ensure that python works on your operating system. (data/ folder should be on the same level as the program file)
• Python is generally a strong choice for text processing as it has many inbuilt methods for processing. libraries such as 're' are useful as well. One downside would be that it might be a little slower than other compiled languages for runtime but this is usually overcome by maintaining good coding style.
• In this assignment, I was able to review and learn many concepts such as file IO, regex in python as well as certain inbuilt functions in python. Especially through regex, i learned some new ways of making more code more efficient for matching strings given in input to what I would require.

Porfolio Assignment 2

• This program takes in an input txt file of a chapter in anatomy and performs some basic text processing operations on it. The latter half of the program also inlcudes a word guessing game built off of the most frequent nouns appearing from the text processed.
• To run this program, execute python axs180183_portfolio2.py anat19.txt in terminal for a Unix/Linux based system.
• To run this program, execute python axs180183_portfolio1.py anat19.txt in powershell for a Windows based system.
• Please ensure that you have the proper libraries downloaded:these are included at the top of the axs180183_portfolio2.py. Also ensure that python works on your operating system. (data/ folder should be on the same level as the program file)

Porfolio Assignment 3

• This python notebook explores all that WordNet and SentiWordNet have to offer, along with explorign some word similarity algorithms such as Wu-Palmer and Lesk.

Porfolio Assignment 4

• This python program utilizes ngrams on 3 trxt files to create 3 language models. Using these language models, the program then attempts to classify sentences in a test file to a particular language based on laplacian smoothing probabilities of the bigrams and unigrams.
• To run part1.py, execute python part1.py data/LangId.train.English data/LangId.train.French data/LangId.train.Italian in terminal for a Unix/Linux based system.
• To run part1.py, execute python part1.py data/LangId.train.English data/LangId.train.French data/LangId.train.Italian in powershell for a Windows based system.
• To run part2.py, execute python part2.py data/LangId.test in terminal for a Unix/Linux based system.
• To run part2.py, execute python part2.py data/LangId.test in powershell for a Windows based system.
• Please ensure that you have the proper libraries downloaded:these are included at the top of the part1.py and part2.py. Also ensure that python works on your operating system. (data/ folder should be on the same level as the program file)

Porfolio Assignment 5, Sentence Parsing

• This pdf document explores three different methods for sentence parsing, PSG parsing, dependency parsing, and SRL parsing.

Porfolio Assignment 6, Web Crawling

• This program performs web crawling to obtain 15 urls, for which it scrapes data and then does some text processing. In particular, the program creates a knowledge base from 10 most frequent words occuring in the text data, this can be utilized as the first step to building a chatbot.
• To run this program, execute python wc.py in terminal for a Unix/Linux based system.
• To run this program, execute python wc.py in powershell for a Windows based system.
• Please ensure that you have the proper libraries downloaded:these are included at the top of the wc.py. Also ensure that python works on your operating system.

Porfolio Assignment 7, Text Classification with machine learning

• This python notebook explores Text Classification utilizgin three learning models: Naive Bayes, Logistics Regression, and MLP Neural Networks. IT consists of a comprehensive report on the performance analysis of each model and the tradeoffs.

Porfolio Assignment 8, ACL paper summary

• This summary report contains the summary about a research paper regarding simulated bandit learning from user feedback for QA systems.

Porfolio Assignment 9, Chatbot

• This is a chatbot built using numerous NLP techniques such as NER, POS tagging, Tf meaures, along with OPENAI's API to create a comprehensive chatbot about all thing related to Star Wars.

Porfolio Assignment 7, Text Classification 2 with deep learning

• This python notebook explores Text Classification utilizgin three deep learning models (Sequential: RNN, LSTM, and CNNs along with the effect of the different approaches of embedding. It consists of a comprehensive report on the performance analysis of each model and the tradeoffs.