JAWAHARLAL NEHRU TECHNOLOGICAL UNIVERSITY HYDERABAD B.Tech III NATURAL LANGUAGE PROCESSING Previous Question Paper (February - 2022 - R18)

Early NLP systems focused on rule-based approaches, symbolic logic, and pattern matching to process natural language. Key examples include:

1. ELIZA (1966): Developed by Joseph Weizenbaum, it was one of the first chatterbots. It used simple pattern matching and substitution to simulate a Rogerian psychotherapist, creating an illusion of understanding by reflecting user input back as questions.
2. SHRDLU (1970): Created by Terry Winograd, this system operated in a restricted 'blocks world'. It could understand natural language commands to manipulate objects, demonstrating early integration of syntax, semantics, and reasoning within a limited domain.
3. LUNAR (1971): Designed by William Woods, this was a question-answering system that allowed geologists to query a database about moon rock samples. It utilized an Augmented Transition Network (ATN) for parsing, marking a significant advancement in formal grammar processing.
4. PARRY (1972): Created by Kenneth Colby, it modeled a paranoid schizophrenic. Unlike ELIZA, it maintained a more complex internal state and belief system, representing an early attempt at modeling human personality and emotional states.

NLP systems face significant challenges due to the inherent complexity and fluidity of human language:

• Ambiguity: Language is highly ambiguous at multiple levels. Lexical ambiguity occurs when words have multiple meanings (e.g., 'bank'), while structural ambiguity arises when a sentence can be parsed in different ways (e.g., 'I saw the man with the telescope').
• Context Dependency: The meaning of words and phrases often depends heavily on the surrounding discourse, social context, or shared world knowledge, which is difficult to encode computationally.
• Non-Standard Language: Processing informal text, such as social media posts, slang, dialects, misspellings, and grammatical errors, remains a major hurdle for models trained on formal corpora.
• Sparsity and Data Scarcity: Many linguistic phenomena are rare. Training models to handle long-tail events or languages with limited digital resources (low-resource languages) is challenging.
• Pragmatics and Discourse: Understanding the intent behind an utterance (speech acts) and maintaining coherence across long conversations or documents requires deep reasoning beyond simple syntactic analysis.
• Multimodality and Cultural Nuance: Language is deeply intertwined with cultural norms and non-verbal cues. Capturing sarcasm, irony, and cultural metaphors requires more than just statistical pattern recognition.