Please enable JavaScript.
Coggle requires JavaScript to display documents.
Natural Language Processing (NLP) - Coggle Diagram
Natural Language Processing (NLP)
Understanding Language
What is Language?
Formal Language: Artificial, unambiguous, rule-based (e.g., programming languages).
Natural Language: Used by humans, ambiguous, constantly evolving.
Challenges of NLP:
Ambiguity in natural language.
Dependence on world knowledge and common sense.
Constantly changing nature of language.
Levels of Language Analysis
Phonetics & Phonology:
Study of speech sounds and their organization.
Morphology:
Structure and formation of words from morphemes (meaningful units).
Lexicon:
Vocabulary and meanings of words in context.
Idioms: Phrases with non-literal meanings.
Syntax:
Grammar and arrangement of words in sentences.
Semantics:
Meaning of words, phrases, and sentences.
Discourse Analysis:
Relationships between sentences in a larger text.
Speech Acts: Actions performed through language (requests, promises).
Pragmatics:
Meaning in context, considering world knowledge and speaker intentions.
History of NLP
1950s:
Turing Test (measuring machine intelligence through language).
Georgetown Experiment (early attempt at machine translation).
1960s:
ELIZA (rule-based chatbot simulating a therapist).
SHRDLU (understanding language in a limited "blocks world").
1970s:
Conceptual ontologies (formal representations of knowledge).
Early chatterbots (e.g., Jabberwacky).
1980s:
Shift from rule-based systems to machine learning approaches.
Machine Learning in NLP
Advantages:
Handles large datasets and complex patterns.
Learns from data, reducing need for explicit rules.
Adapts to new language data.
Applications:
Text classification.
Machine translation.
Sentiment analysis.
NLP Tasks and Progress
Mostly Solved:
Optical Character Recognition (OCR).
Named Entity Recognition (NER) (identifying names, locations, etc.).
Spam detection.
Part-of-Speech (POS) tagging (for English).
Good Progress:
Sentiment analysis (understanding emotions and opinions).
Co-reference resolution (linking pronouns to their referents).
Word Sense Disambiguation (WSD).
Parsing (grammatical analysis).
Information Extraction (IE).
Really Hard:
Question answering (QA).
Summarization.
Paraphrasing.
Dialogue systems (natural conversation).
Challenges and Open Problems
World Knowledge and Common Sense: Enabling computers to reason like humans.
Handling Ambiguity: Disambiguating word senses, sentence structures, and meanings.
Natural Language Generation: Producing fluent and coherent text.
Natural Language Understanding: Extracting meaning and intentions from text.
Common Sense Reasoning
Importance: Essential for true language understanding.
Challenges: Implicit, context-dependent, difficult to formalize.
Examples: Logical puzzles, understanding implied information.
Approaches: Knowledge bases (e.g., ConceptNet), reasoning with defaults.