Technology: Python, Google Collab.
• Programmed the bigram model with and without add-one Laplace smoothing for a given corpus.
• Computed positive pointwise mutual information for terms and context word with/out add-2 smoothing.
• Implemented HMM and the Viterbi algorithm to assign POS tags to given text.
• Trained and evaluated a feed-forward neural model on Reuters corpus using Google Collab and varying the parameters for best efficiency and accuracy.
• Implemented the CKY Parser for CNF Grammar on given text.
• Executed the constituency parser with a self-attentive encoder (Kitaev & Klein 2018) from Github.
• Implemented manual and automatic Semantic Role Labeling using ProbBank definitions and neural SRL code from Github on the given text.
• Implemented Simple LESK WSD algorithm to disambiguate the given text.
• Identified named entities, IOB notations, temporal expressions with normalization and relation extraction for given text.
HOMEWORK 1
- Regular Expressions
- N-Grams
- Vector Semantics
- Part-of-Speech Tagging
HOMEWORK 2
- CKY PARSER
- Statistical Parsing
- Semantic Role Labeling
HOMEWORK 3
- Simple LESK Word Sense Disambiguation
- Information Extraction