Please enable JavaScript.

Coggle requires JavaScript to display documents.

Cognitive Science - Coggle Diagram

- - - - History of NLP
        
        Chomsky: Build grammar and ontologies
        
        70-80s: AI cold
        
        80-10s: Many algorithms developed for npl for specific tasks, by learning from data
        
        10s: Neural Networks dominate
        
        20s: Apparition of transformers
      - Applications
        
        Sentiment Analysis: The evaluation of whether the words expressed by a person are conveyed in a positive or negative way
      - Subfields
        
        Text Mining
        
        Text Vectorization
        
        Binary
        
        Bag of words representation
        
        Document Terms Matrix
        
        3 more items...
        
        Term Frequency
        
        Term Frequency Matrix
        
        Components
        
        3 more items...
        
        Processing
        
        2 more items...
        
        Similarity between documents
        
        Jaccard Coefficient: Nr of words in common / Nr of words in documents
        
        Cosine Similarity: dot product of both documents / (sqrt(dot product d1) * sqrt(dotproduct d2))
        
        Done for Term Frequency Matrix after Normalization
        
        Preprocessing
        
        abbreviations
        
        Regular Expressions
        
        Kleene Expressions
        
        3 more items...
        
        Negation
        
        1 more item...
        
        Disjunction
        
        1 more item...
        
        Substitution
        
        1 more item...
        
        Creative expressions, text
        
        slang
        
        emoticons
        
        character flooding
        
        capitalization
        
        concatenations
        
        typos
        
        Normalization
        
        Lemmatization: Was becomes be
        
        Stemming (eliminating plural): Cats becomes Cat
        
        Case folding (Eliminate capital letters)
        
        Levenshtein Edit Distance: To fix misspellings
        
        Encodings
        
        Byte pair Encoding (didnt understand)
        
        WordPiece Encoding (didnt understand)
        
        n gram encoding (didnt understand)
      - Model Evaluation
        
        Confusion Matrix
        
        Precision
        
        Recall
        
        Fb score
        
        MRR
        
        AP
        
        Humans annotate labels
        
        Cohen's Kappa
      - Classification Models
        
        N-gram language model
        
        how likely is word given context
        
        how likely is context given word
        
        Sequence Probabilities
        
        P(survival|Im going to pass) = Count(Im going to pass survival) / Count(Im going to pass)
        
        Smoothing Laplace
        
        Count(Im going to pass survival) + 1 / Count(Im going to pass) + length of the vocabulary
        
        Naive Bayes
    - - Uses of Psycholinguistics in NLP
        
        Black boxes: Creativity and linguistic questions are needed to diagnose NLP models
        
        Bias: Human set or models have bias, which can lead to wrong inferences
        
        Imperfection: Expertise is required for debugging or making relevant, complete, and appropiate test sets
  - - - Why is language difficult?
        
        The surface form of language is sentences
        
        Syntax: The structure of a sentence
        
        Semantics: Meaning of the word
        
        These all are used and done differently by different people, which causes ambiguity
        
        Different interpretations of a sentence
        
        These make language hard to learn . Interpreting, inferring, understanding and representing language is complex
        
        1 more item...
        
        Context
        
        Sarcasm
        
        Negation
        
        Dependencies
        
        Part of Speech: Adverbs, Nouns, Verbs, Articles, Determinants, Adjectives, prepostions, conjunctions
        
        Discourse: Adds information to a sentence