snipsco/snips-nlu
Snips Nlu
snips-nlu este o bibliotecă Python și un motor de înțelegere a limbajului natural (NLU) conceput pentru a converti textul nestructurat în date structurate. Identifică intențiile utilizatorului și extrage entitățile asociate din propoziții în limbaj natural pentru a permite procesarea comenzilor lizibile de către mașină.
Motorul funcționează ca un parser multilingv capabil să proceseze text în mai multe limbi. Acesta mapează entitățile identificate la valori canonice sau formate ISO standardizate, cum ar fi marcajele temporale, pentru a asigura consistența datelor.
Proiectul acoperă clasificarea intențiilor și recunoașterea entităților numite, utilizând etichetarea secvențelor și tokenizarea pentru a identifica obiectivele utilizatorului și sloturile de date specifice.
Features
- Multilingual Natural Language Understanding - Functions as a multilingual NLU engine that converts unstructured text into structured intents and entities.
- Natural Language Intent Processing - Converts unstructured natural language sentences into structured machine-readable intents and probability scores.
- Intent Classification Pipelines - Offers a Python library for mapping natural language input to predefined intents and entities.
- Slot Extraction Systems - Provides an engine that simultaneously predicts user intent and extracts associated entity slot values.
- Multilingual Model Loading - Supports multilingual processing by loading language-specific model weights and configuration files.
- Named Entity Recognition - Identifies and classifies specific parameters within text, mapping them to canonical or standardized values.
- Natural Language Classification - Identifies the underlying action of a natural language sentence and assigns a probability score to the match.
- Slot Normalization - Identifies variables in sentences and converts them into standardized formats like ISO datetimes.
- Sequence Labeling - Implements sequence labeling using Conditional Random Fields to predict entity labels for tokens in a sentence.
- Linear Classifiers - Uses a linear model with a softmax activation function to predict user intent from feature vectors.
- Canonical Entity Mapping - Provides a mechanism to map diverse entity synonyms to a single standardized reference value using a dictionary.
- Text Feature Extraction - Transforms raw text into numerical feature vectors using linguistic patterns and character-level n-grams.
- User-Defined Entities - Allows the definition of custom entity types using reference values and synonyms.
- ISO Date Normalization - Converts raw entity strings into standardized ISO 8601 date and time formats.
- Entity Format Standardizers - Standardizes strongly typed entities like dates and times into ISO timestamps.
- Chatbots and QA - Library for parsing natural language into structured information.
- Natural Language Processing - Library for extracting meaning from text.
- Python NLP Libraries - Production-ready library for intent parsing and slot filling.
