Language Processing Types

Since 1.9.0

Types for natural language processing (NLP) applications.

Overview

Language Processing Types provides type-level natural language processing utilities. It supports tokenization, parsing, semantic analysis, sentiment detection, translation, and text classification.

This module enables building type-safe NLP applications with proper constraints for tokens, sentences, embeddings, and language models.

Basic Usage

import type { Token, Sentence, Document, Embedding, SentimentResult } from 'uni-types'

// Define token type
type WordToken = Token<{ lemma: string, pos: PartOfSpeech }>

// Define document type
type TextDocument = Document<{ title: string, author: string }>

// Define embedding
type WordEmbedding = Embedding<{ word: string }>

Key Types

TokenType

Token type classification.

type TokenType =
  | 'word'
  | 'punctuation'
  | 'number'
  | 'whitespace'
  | 'symbol'
  | 'emoji'
  | 'url'
  | 'email'
  | 'mention'
  | 'hashtag'

Token

Token base interface.

interface Token<T = unknown> {
  id: number
  text: string
  type: TokenType
  start: number
  end: number
  lemma?: string
  pos?: PartOfSpeech
  tag?: string
  data?: T
}

Sentence

Sentence type with tokens.

interface Sentence<T extends Token = Token> {
  id: number
  text: string
  tokens: T[]
  start: number
  end: number
  parseTree?: ParseTree
  dependencies?: DependencyRelation[]
  sentiment?: SentimentResult
}

Document

Document type with paragraphs.

interface Document<P extends Paragraph = Paragraph> {
  id: string
  text: string
  paragraphs: P[]
  metadata?: DocumentMetadata
  language?: Locale
}

ParseTree

Parse tree structure.

interface ParseTree {
  root: ParseNode
  nodes: Map<string, ParseNode>
  depth: number
}

PartOfSpeech

Part of speech enumeration.

type PartOfSpeech =
  | 'noun'
  | 'verb'
  | 'adjective'
  | 'adverb'
  | 'pronoun'
  | 'preposition'
  | 'conjunction'
  | 'interjection'
  | 'determiner'
  | 'numeral'
  | 'particle'
  | 'auxiliary'
  | 'proper_noun'

Embedding

Embedding type with vector.

interface Embedding<T = unknown> {
  vector: EmbeddingVector
  dimension: number
  model: string
  data?: T
}

SentimentResult

Sentiment analysis result.

interface SentimentResult {
  score: SentimentScore
  label: SentimentLabel
  confidence: number
  aspectSentiments?: AspectSentiment[]
}

EntityType

Named entity recognition types.

type EntityType =
  | 'person'
  | 'organization'
  | 'location'
  | 'date'
  | 'time'
  | 'money'
  | 'percent'
  | 'quantity'
  | 'event'
  | 'product'
  | 'language'
  | 'nationality'

TranslationResult

Translation result type.

interface TranslationResult {
  text: string
  sourceLanguage: Locale
  targetLanguage: Locale
  alternatives?: string[]
  confidence?: number
  metadata?: TranslationMetadata
}

Locale

Locale type (ISO 639-1).

type Locale =
  | 'en' | 'zh' | 'es' | 'ja' | 'de' | 'fr' | 'ko' | 'ru'
  | 'pt' | 'it' | 'ar' | 'hi' | 'bn' | 'pa' | 'id' | 'ms'
  // ... more locales

Related

• AI/ML - ML model types
• Error Handling - Error types
• Functional Programming - Functional patterns