class Cadmium::Glove::Model

• Cadmium::Glove::Model
• Reference
• Object

Defined in:

Constant Summary

BIAS_FILE = "word_biases.json"

COOC_FILE = "cooc_matrix.json"

CORPUS_FILE = "corpus.json"

VEC_FILE = "word_vectors.json"

Constructors

• .new(num_components : Int32 = 30, epochs : Int32 = 25, threads : Int32 = 4, learning_rate : Float64 = 0.05, alpha : Float64 = 0.75, max_count : Int32 = 100)

  Creates a new Glove::Model instance.

Class Method Summary

• .load(dir, corpus_file = CORPUS_FILE, cooc_file = COOC_FILE, vec_file = VEC_FILE, bias_file = BIAS_FILE, **options)

  Create a new model from an existing dataset.

Instance Method Summary

• #alpha : Float64
• #alpha=(alpha : Float64)
• #analogy_words(word1, word2, target, num = 3, accuracy = 1e-4)

  Get a word that relates to target like word1 relates to word2.

• #cooc_matrix : Apatite::Matrix(Float64)
• #cooc_matrix=(cooc_matrix : Apatite::Matrix(Float64))
• #corpus
• #corpus=(corpus : Corpus | Nil)
• #epochs : Int32
• #epochs=(epochs : Int32)
• #fit(text, **options)

  Fit a String or Glove::Corpus instance and build a co-occurrence matrix.

• #learning_rate : Float64
• #learning_rate=(learning_rate : Float64)
• #load(dir, corpus_file = CORPUS_FILE, cooc_file = COOC_FILE, vec_file = VEC_FILE, bias_file = BIAS_FILE)

  Loads training data from already existing files.

• #max_count : Int32
• #max_count=(max_count : Int32)
• #most_similar(word, num = 3)

  Get most similar words to word.

• #num_components : Int32
• #num_components=(num_components : Int32)
• #save(outdir, corpus_file = CORPUS_FILE, cooc_file = COOC_FILE, vec_file = VEC_FILE, bias_file = BIAS_FILE)

  Save trained data to files

• #threads : Int32
• #threads=(threads : Int32)
• #token_index : Hash(String, Int32)
• #token_index=(token_index : Hash(String, Int32))
• #token_pairs : Array(TokenPair)
• #token_pairs=(token_pairs : Array(TokenPair))
• #train

  Train the model.

• #vector(word)

  Find the vector row of @word_vec for a given word.

• #vector_distance(word : String | Apatite::Vector)

  Calculates the cosine distance of all the words in the vocabulary against a given word.

• #visualize

  TODO Generate a graph of the word vector matrix

• #word_biases : Array(Float64)
• #word_biases=(word_biases : Array(Float64))
• #word_vec : Apatite::Matrix(Float64)
• #word_vec=(word_vec : Apatite::Matrix(Float64))

Constructor Detail

Class Method Detail

Instance Method Detail

model.analogy_words("quantum", "physics", "atom")
# => [{"electron", 0.98583}, {"energi", 0.98151}, {"photon",0.96650}]