Please enable JavaScript.
Coggle requires JavaScript to display documents.
Natural Language Processing - Coggle Diagram
Natural Language Processing
Evaluation & Fairness
Model robustness & counterfactual tests
Representation bias in embeddings
Demographic Parity gap metric
Deep-Learning Architecture
Transformer & Self-Attention
BERT details
[CLS] pooled vector
Limitations & proposed improvements
Masking scheme (80-10-10)
Applications & Pipelines
Information Extraction
Rule / pattern vs supervised vs distant supervision
Neuro-symbolic hybrids
Entities, relations, events
Sentiment Analysis
Logistic-regression baseline
Fine-tuned BERT vs BoW
Lexicon-based rules (+ negation handling)
Hate-Speech Detection & Interpretability
Temporal Processing (Reichenbach tense model)
Fundamentals
Zipf's Law & long-tail vocab
Tokenisation
Language-specific hurdles (German compounds, CJK, Arabic morphology
Stop-word removal & over-normalisation trade-offs
Punkt sentence splitter
Why represent text as vectors?
Text Representation Techniques
TF-IDF (weights rare terms higher)
Bag-of-Words (counts)
Word embeddings
Word2Vec geometry & analogies
Bias & semantic drift issues
Contextual Embeddings
Sub-word tokenisers (BPE / WordPiece)
Transformer-based models (BERT, GPT)
Semantics
Semantic Role Labelling (who-did-what-to-whom)
Formal Compositionality
Multi-word expressions & compositionality score
Lambda-calculus
Frege's principle
Advanced & Future Directions
Foundation models & in-context / few shot learning
Model efficiency (sparse attention, pruning)
Domain-specific pre-training (BioBERT, FinBERT)
When regex still wins (structured text)