Struct HashingVectorizer 

pub struct HashingVectorizer { /* private fields */ }

Stateless hashing vectorizer for streaming/large-scale text.

Maps tokens to feature indices using a hash function. Does not store a vocabulary, making it memory-efficient but irreversible.

Examples

use aprender::text::vectorize::HashingVectorizer;
use aprender::text::tokenize::WhitespaceTokenizer;

let docs = vec!["hello world", "hello rust"];

let vectorizer = HashingVectorizer::new(1000)
    .with_tokenizer(Box::new(WhitespaceTokenizer::new()));

let matrix = vectorizer.transform(&docs).expect("transform should succeed");
assert_eq!(matrix.n_rows(), 2);
assert_eq!(matrix.n_cols(), 1000);

Implementations

impl HashingVectorizer

pub fn new(n_features: usize) -> Self

Create a new HashingVectorizer with specified number of features.

pub fn with_tokenizer(self, tokenizer: Box<dyn Tokenizer>) -> Self

Set the tokenizer.

pub fn with_lowercase(self, lowercase: bool) -> Self

Set lowercase.

pub fn with_ngram_range(self, min_n: usize, max_n: usize) -> Self

Set n-gram range.

pub fn with_stop_words_english(self) -> Self

Use English stop words.

pub fn transform<S: AsRef<str>>( &self, documents: &[S], ) -> Result<Matrix<f64>, AprenderError>

Transform documents to hashed feature matrix (stateless).