Struct FastText 

pub struct FastText { /* private fields */ }

FastText model for learning word representations with character n-grams

Decomposes each word into character n-grams (subwords) and learns embeddings for both whole words and their constituent n-grams. This enables:

• Out-of-vocabulary word handling (any word can be represented via its n-grams)
• Morphological awareness (similar prefixes/suffixes produce similar vectors)
• Robustness to misspellings and rare word forms

Implementations

impl FastText

pub fn new() -> Self

Create a new FastText model with default configuration

pub fn with_config(config: FastTextConfig) -> Self

Create a new FastText model with custom configuration

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

Set a custom tokenizer

pub fn extract_ngrams(&self, word: &str) -> Vec<String>

Extract character n-grams from a word

Wraps the word with boundary markers < and > before extracting. For example, “fox” with min_n=3, max_n=4 produces: 3-grams: “<fo”, “fox”, “ox>” 4-grams: “<fox”, “fox>”, “<fox>”(if len allows)

pub fn build_vocabulary(&mut self, texts: &[&str]) -> Result<()>

Build vocabulary from texts

pub fn train(&mut self, texts: &[&str]) -> Result<()>

Train the FastText model

pub fn get_word_vector(&self, word: &str) -> Result<Array1<f64>>

Get the embedding vector for a word (handles OOV words via subwords)

For in-vocabulary words, returns the average of the word vector and its n-gram vectors. For OOV words, returns the average of matching n-gram vectors.

pub fn most_similar( &self, word: &str, top_n: usize, ) -> Result<Vec<(String, f64)>>

Find most similar words to a given word

pub fn most_similar_by_vector( &self, vector: &Array1<f64>, top_n: usize, exclude_words: &[&str], ) -> Result<Vec<(String, f64)>>

Find most similar words to a given vector

pub fn analogy( &self, a: &str, b: &str, c: &str, top_n: usize, ) -> Result<Vec<(String, f64)>>

Compute word analogy: a is to b as c is to ?

Uses vector arithmetic: result = b - a + c, then finds most similar words. Works with OOV words since FastText can compute vectors for any word.

pub fn word_similarity(&self, word1: &str, word2: &str) -> Result<f64>

Compute cosine similarity between two words

Both words can be OOV.

pub fn save<P: AsRef<Path>>(&self, path: P) -> Result<()>

Save the model to a file

Saves in a format that includes word vectors, n-gram info, and config. Uses a custom header format: Line 1: FASTTEXT <vocab_size> <vector_size> <min_n> <max_n> <bucket_size> Lines 2+: word vector_components…

pub fn load<P: AsRef<Path>>(path: P) -> Result<Self>

Load a FastText model from a file

pub fn vocabulary_size(&self) -> usize

Get the vocabulary size

pub fn vector_size(&self) -> usize

Get the vector size

pub fn ngram_range(&self) -> (usize, usize)

Get the n-gram configuration (min_n, max_n)

pub fn ngram_count(&self) -> usize

Get the number of unique n-grams discovered

pub fn contains(&self, word: &str) -> bool

Check if a word is in the vocabulary

pub fn can_represent(&self, word: &str) -> bool

Check if a word can be represented (either in vocab or has matching n-grams)

pub fn get_vocabulary_words(&self) -> Vec<String>

Get all words in the vocabulary

Trait Implementations

impl Clone for FastText

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for FastText

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for FastText

fn default() -> Self

Returns the “default value” for a type.

impl WordEmbedding for FastText

fn embedding(&self, word: &str) -> Result<Array1<f64>>

Get the embedding vector for a word

fn dimension(&self) -> usize

Get the dimensionality of the embedding vectors

fn find_similar(&self, word: &str, top_n: usize) -> Result<Vec<(String, f64)>>

Find the top-N most similar words to the query word

fn solve_analogy( &self, a: &str, b: &str, c: &str, top_n: usize, ) -> Result<Vec<(String, f64)>>

Solve the analogy: a is to b as c is to ?

fn vocab_size(&self) -> usize

Get the number of words in the vocabulary

fn similarity(&self, word1: &str, word2: &str) -> Result<f64>

Compute cosine similarity between two words