Natural Language Processing (BIT471CO)
Speech and language processing, morphological parsing with FSTs, N-grams, HMM POS tagging, feature unification, WordNet lexical semantics, and discourse pragmatics.
Unit 1: Introduction to NLP6 hrs
- Definition, issues, and strategies in speech and language processing
- Application domains and software tools for NLP
- Linguistic organization of NLP, Natural Language Processing vs Programming Language Processing
- Word classes, review of Regular Expressions, Context-Free Grammars (CFG), and parsing techniques
Unit 2: Morphology and Phonology7 hrs
- Inflectional and derivational morphology
- Morphological parsing with Finite State Transducers (FSTs) and combinational rules
- Phonology: Speech sounds, phonetic transcription (IPA), phoneme definitions and phonological rules
- Optimality theory and machine learning of phonological rules
- Phonological aspects of prosody and speech synthesis (TTS)
Unit 3: Pronunciation, Spelling and N-grams7 hrs
- Spelling error detection and correction using probabilistic noisy channel models
- Pronunciation variation: lexical, allophonic, and dialectal variations
- Decision tree models for pronunciation
- Counting words in corpora and simple N-gram language models
- Smoothing techniques: Add-One (Laplace), Witten-Bell, Good-Turing discounting
- N-grams for spelling correction and pronunciation modeling
Unit 4: Syntax and Part-of-Speech Tagging6 hrs
- Penn Treebank tagsets and word categories
- Concept of Hidden Markov Model (HMM) taggers
- Rule-based vs stochastic POS tagging
- Viterbi algorithm for HMM decoding and tagging
- Transformation-Based Learning (Brill) tagger
Unit 5: Sentence Level Construction & Unification Semantics7 hrs
- Noun phrase structures, co-ordination, and sub-categorization
- Concept of feature structures and unification
- Representing Meaning: Unambiguous representation, canonical form, expressiveness, meaning structure of language
- Basics of First-Order Predicate Calculus (FOPC) in semantic interpretation
- Syntax-driven semantic analysis, attachment, integration, and robustness
Unit 6: Lexical Semantics6 hrs
- Lexemes and semantic relationships: homonymy, polysemy, synonymy, hyponymy
- WordNet taxonomy and relational database structure
- Internal structure of words, metaphors, and metonymy with computational approaches
- Word Sense Disambiguation (WSD): Selectional restriction-based, machine learning-based, and dictionary-based (Lesk algorithm) approaches
Unit 7: Pragmatics and Discourse Structure6 hrs
- Discourse reference resolution and referential phenomena
- Syntactic and semantic constraints on co-reference
- Pronoun resolution algorithms (Hobbs algorithm, centering theory)
- Text coherence and discourse rhetorical structure
- Dialogues: Turns and utterances, grounding, dialogue acts and conversational structures
- Natural Language Generation (NLG): introduction to language generation architecture and discourse planning
Reference books
- Jurafsky, Daniel and James H. Martin, Speech and Language Processing: An Introduction to Natural Language Processing, Computational Linguistics, and Speech Recognition, Pearson Education.
- Allen, James, Natural Language Understanding, Benjamin/Cummings.
- Bharati, Akshar, Vineet Chaitanya, and Rajeev Sangal, Natural Language Processing: A Paninian Perspective, Prentice Hall of India.
- Charniak, Eugene, Statistical Language Learning, MIT Press.
- Manning, Christopher D. and Hinrich Schutze, Foundations of Statistical Natural Language Processing, MIT Press.