Słownik AI
Kompletny słownik sztucznej inteligencji
Named Entity Recognition (NER)
Natural language processing technique that identifies and classifies predefined entities in unstructured text into categories such as persons, organizations, locations, or dates.
Entity Linking
Process of connecting named entities detected in text to their unique identifiers in a structured knowledge base like Wikipedia or DBpedia.
IOB Tagging
Tagging system using Inside (I), Outside (O), and Beginning (B) prefixes to mark the beginning and continuity of entities in a token sequence.
Conditional Random Fields (CRF)
Discriminative statistical model used for sequence labeling, particularly effective for named entity recognition by capturing contextual dependencies between adjacent observations.
Bidirectional LSTM
Recurrent neural network architecture that processes sequences in both temporal directions, enabling better context understanding for named entity recognition.
BERT
Pre-trained transformer-based language model that uses a bidirectional attention mechanism to understand the deep context of words in named entity recognition.
Tokenization
Fundamental process of segmenting text into basic lexical units (tokens), constituting the first essential step of named entity recognition.
Chunking
Syntactic segmentation technique that groups adjacent words into coherent phrases, facilitating the identification of complex named entities.
Nested NER
Variant of named entity recognition capable of identifying entities nested within each other, such as 'University of Paris' containing 'Paris' as a nested entity.
Fine-grained NER
Approach to named entity recognition using very specific and numerous categories (over 100 types) for detailed entity classification.
Zero-shot NER
Technique for recognizing entity types never seen during training by using textual descriptions or prototypes as guidance.
CoNLL-2003
Standard benchmark dataset for evaluating named entity recognition systems, containing news texts annotated with four main entity types.
Entity Disambiguation
Process of resolving ambiguity between homonymous entities by using context to determine the correct entity referenced in the text.
spaCy
Open-source natural language processing library optimized for production, including powerful pre-trained models for named entity recognition.
Transfer Learning for NER
Approach consisting of reusing knowledge acquired from large text corpora to improve named entity recognition performance on specific domains with limited data.
Active Learning for NER
Semi-automatic annotation methodology where the model intelligently selects the most informative examples to annotate, optimizing human effort for NER training.
Weak Supervision for NER
Learning paradigm using imperfect annotation sources such as heuristics, dictionaries, or less accurate models to automatically generate NER training data.
Stanford NER
Named entity recognition system developed by Stanford University, using CRF models and offering custom learning capabilities.
Part-of-Speech Tagging
Process of grammatically tagging words that serves as an important feature for named entity recognition models, helping to distinguish proper nouns from common nouns.
BiLSTM-CRF
Hybrid architecture combining bidirectional LSTM networks for feature extraction and CRFs for modeling sequential dependencies in named entity recognition.