NLP with Classification and Vector Spaces

Translation as linear transformation of embeddings

In this project, I use linear transformation of word embeddings to translate English words to French.

• Generate embedding matrices from a python dictionary
• Compute the gradient of loss function(square Frobenius) norm in respect to transform matrix R
• Find the optimal R with gradient descent with a fixed number of iterations
• Search for the translation embedding by K-NN(1-NN)
• Test our translation and compute its accuracy.

Finding the most similar tweets with LSH(locality sensitive hashing)

In this project, I use LSH instead of normal KNN to find the most similar tweet to a specific tweet.

• Create document(tweet) embeddings
• Calculate hash number for a vector
• Implement hash buckets
• Create hash tables

Logistic Regression Classifier-Sentiment Analysis

in this project, I use logistic Regression for sentiment analysis on tweets(nltk: twitter_samples). Given a tweet, will decided if it has a positive sentiment or a negative one.

• Process tweets: eliminate handles and URLs, tokenize the tweet into individual words, remove stop words and apply stemming.
• Extract features for logistic regression
• Implement logistic regression from scratch
• Apply logistic regression on a natural language processing task
• Test logistic regression and calculate the accuracy
• Predict on your own tweet

Naive Bayes Classifier-Sentiment Analysis

in this project, I use Naive Bayes for sentiment analysis on tweets(nltk: twitter_samples). Given a tweet, will decided if it has a positive sentiment or a negative one.

• Process tweets: eliminate handles and URLs, tokenize the tweet into individual words, remove stop words and apply stemming.
• Calculate logPrior and loglikelihood for each word
• Train a Naive Bayes model on a sentiment analysis task
• Test Naive Bayes and and calculate the accuracy
• Predict on your own tweet