1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123

use crate::{NGram, NGramType, Tokenizer};

#[derive(buffalo::Read, buffalo::Write)]
#[buffalo(size = "static", value_size = 8)]
pub enum FeatureGroup {
	#[buffalo(id = 0)]
	Identity(IdentityFeatureGroup),
	#[buffalo(id = 1)]
	Normalized(NormalizedFeatureGroup),
	#[buffalo(id = 2)]
	OneHotEncoded(OneHotEncodedFeatureGroup),
	#[buffalo(id = 3)]
	BagOfWords(BagOfWordsFeatureGroup),
	#[buffalo(id = 4)]
	WordEmbedding(WordEmbeddingFeatureGroup),
	#[buffalo(id = 5)]
	BagOfWordsCosineSimilarity(BagOfWordsCosineSimilarityFeatureGroup),
}

#[derive(buffalo::Read, buffalo::Write)]
#[buffalo(size = "dynamic")]
pub struct IdentityFeatureGroup {
	#[buffalo(id = 0, required)]
	pub source_column_name: String,
}

#[derive(buffalo::Read, buffalo::Write)]
#[buffalo(size = "dynamic")]
pub struct NormalizedFeatureGroup {
	#[buffalo(id = 0, required)]
	pub source_column_name: String,
	#[buffalo(id = 1, required)]
	pub mean: f32,
	#[buffalo(id = 2, required)]
	pub variance: f32,
}

#[derive(buffalo::Read, buffalo::Write)]
#[buffalo(size = "dynamic")]
pub struct OneHotEncodedFeatureGroup {
	#[buffalo(id = 0, required)]
	pub source_column_name: String,
	#[buffalo(id = 1, required)]
	pub variants: Vec<String>,
}

#[derive(buffalo::Read, buffalo::Write)]
#[buffalo(size = "dynamic")]
pub struct BagOfWordsFeatureGroup {
	#[buffalo(id = 0, required)]
	pub source_column_name: String,
	#[buffalo(id = 1, required)]
	pub tokenizer: Tokenizer,
	#[buffalo(id = 2, required)]
	pub strategy: BagOfWordsFeatureGroupStrategy,
	#[buffalo(id = 3, required)]
	pub ngram_types: Vec<NGramType>,
	#[buffalo(id = 4, required)]
	pub ngrams: Vec<(NGram, BagOfWordsFeatureGroupNGramEntry)>,
}

#[derive(buffalo::Read, buffalo::Write)]
#[buffalo(size = "dynamic")]
pub struct BagOfWordsCosineSimilarityFeatureGroup {
	#[buffalo(id = 0, required)]
	pub source_column_name_a: String,
	#[buffalo(id = 1, required)]
	pub source_column_name_b: String,
	#[buffalo(id = 2, required)]
	pub tokenizer: Tokenizer,
	#[buffalo(id = 3, required)]
	pub strategy: BagOfWordsFeatureGroupStrategy,
	#[buffalo(id = 4, required)]
	pub ngram_types: Vec<NGramType>,
	#[buffalo(id = 5, required)]
	pub ngrams: Vec<(NGram, BagOfWordsFeatureGroupNGramEntry)>,
}

#[derive(buffalo::Read, buffalo::Write)]
#[buffalo(size = "static", value_size = 0)]
pub enum BagOfWordsFeatureGroupStrategy {
	/// The feature values will be 1 if the token is present in the source column value and 0 otherwise.
	#[buffalo(id = 0)]
	Present,
	/// The feature values will be equal to the number of occurrences of the token in the source column value.
	#[buffalo(id = 1)]
	Count,
	/// The feature values will be equal to the number of occurrences of the token in the source column value multiplied by the token's IDF.
	#[buffalo(id = 2)]
	TfIdf,
}

#[derive(buffalo::Read, buffalo::Write)]
#[buffalo(size = "dynamic")]
pub struct BagOfWordsFeatureGroupNGramEntry {
	#[buffalo(id = 0, required)]
	pub idf: f32,
}

#[derive(buffalo::Read, buffalo::Write)]
#[buffalo(size = "dynamic")]
pub struct WordEmbeddingFeatureGroup {
	#[buffalo(id = 0, required)]
	pub source_column_name: String,
	#[buffalo(id = 1, required)]
	pub tokenizer: Tokenizer,
	#[buffalo(id = 2, required)]
	pub model: WordEmbeddingModel,
}

#[derive(buffalo::Read, buffalo::Write)]
#[buffalo(size = "dynamic")]
pub struct WordEmbeddingModel {
	/// This is the number of values for each word.
	#[buffalo(id = 0, required)]
	pub size: u64,
	/// This is a map from words to row indexes in `values`.
	#[buffalo(id = 1, required)]
	pub words: Vec<(String, u64)>,
	/// This holds the word embedding values, stored as a 2d array of shape (n_words, dimension) in row major order.
	#[buffalo(id = 2, required)]
	pub values: Vec<f32>,
}