Lecture 9 : Intro to NLP Flashcards
What is NLP?
Natural Language Processing –> automatic processing of written texts
1.Natural Language Understanding–>input=text
2.Natural Language Generation–>input=text
What is Speech processing?
Automatic processing of speech
1.Speech recognition -> input=acoustic signal
2.Speech synthesis–>output=acoustic signal
What is rule-based NLP(1950-mid 1980)?
-Cognitive approach
-Rules are developed by hand in collaboration with linguists
What is statistical NLP(mid 1980-2010)?
-Engineering approach
-Rules are developed automatically with ML
-linguistic features are hand-engineering and fed to ML model
What is deep language processing (2010-now)?
-Rules are developed automatically with ML
-Linguistic features are found automatically
What are the 2 limits of the Bag-of-Words model(BOW)?
-word order is ignored–> meaning of text is lost
-n-grams take a bit of the word order into account
What is the n-gram model(2)?
-Probability distribution over sequences of events
-Models the order of events
When is the n-gram model used?
To predict the next event in a sequence of event
What’s a language model?
Its a n-gram model over word/character sequences
What are examples of applications of language models(4)?
-Speech recognition
-Statistical machine translation
-Language identification
-Spelling correction
What are 2 issues with n-grams?
-Natural language is not linear
-There may be long-distance dependencies : syntactic, semantic, world knowledge
What are the 3 stages of NLU?
1.Parsing (syntax)
2.Semantic interpretation
3.Contextual/world knowledge interpretation
What is parsing?
-To assign syntactic structures to a sentence
-To determine how words are put together to form correct sentences
What is the goal of semantic interpretation?
-Map sentences to some representation of its meaning
What are the 2 types of semantics?
1.Lexical semantics–> meaning of individual words
2.Compositional semantics–>meaning of combination of words
