File size: 7,072 Bytes
a5ae1ac | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 | """
Tests for the Encoder.
Verifies against HLD spec:
- Word-level: direct lookup in embedding table
- Sentence-level: weighted average of word vectors
- OOV: zero vector (honest — "I don't have this word")
- Vectors are normalized (for cosine similarity)
- No training — just lookup
Uses synthetic vocabulary to avoid downloading GloVe in tests.
"""
import sys
from pathlib import Path
import numpy as np
sys.path.insert(0, str(Path(__file__).parent.parent / "src"))
from encoder import Encoder
DIM = 300
def make_vocab():
"""Small synthetic vocabulary for testing."""
rng = np.random.RandomState(42)
words = {
"shakespeare": rng.randn(DIM).astype(np.float32),
"hamlet": rng.randn(DIM).astype(np.float32),
"wrote": rng.randn(DIM).astype(np.float32),
"playwright": rng.randn(DIM).astype(np.float32),
"english": rng.randn(DIM).astype(np.float32),
"cat": rng.randn(DIM).astype(np.float32),
"dog": rng.randn(DIM).astype(np.float32),
"the": rng.randn(DIM).astype(np.float32),
"who": rng.randn(DIM).astype(np.float32),
"is": rng.randn(DIM).astype(np.float32),
}
return words
def make_encoder():
enc = Encoder(data_dir="/tmp/test_encoder", dim=DIM)
enc.load_from_dict(make_vocab())
return enc
class TestEncoderWord:
def test_known_word_returns_vector(self):
enc = make_encoder()
v = enc.encode_word("shakespeare")
assert v.shape == (DIM,)
assert np.linalg.norm(v) > 0
def test_known_word_is_normalized(self):
enc = make_encoder()
v = enc.encode_word("hamlet")
assert abs(np.linalg.norm(v) - 1.0) < 1e-5
def test_oov_returns_zero(self):
"""OOV = zero vector. Honest about not knowing."""
enc = make_encoder()
v = enc.encode_word("glorpnax")
assert np.all(v == 0)
def test_case_insensitive(self):
enc = make_encoder()
v1 = enc.encode_word("Shakespeare")
v2 = enc.encode_word("shakespeare")
np.testing.assert_array_equal(v1, v2)
def test_strips_whitespace(self):
enc = make_encoder()
v1 = enc.encode_word(" hamlet ")
v2 = enc.encode_word("hamlet")
np.testing.assert_array_equal(v1, v2)
def test_has_word(self):
enc = make_encoder()
assert enc.has_word("cat") is True
assert enc.has_word("glorpnax") is False
def test_returns_copy_not_reference(self):
"""Modifying returned vector should not change the vocab."""
enc = make_encoder()
v1 = enc.encode_word("cat")
v1[0] = 999.0
v2 = enc.encode_word("cat")
assert v2[0] != 999.0
class TestEncoderSentence:
def test_single_word_sentence(self):
enc = make_encoder()
v_word = enc.encode_word("hamlet")
v_sent = enc.encode_sentence("hamlet")
# Single word sentence should equal the word vector (both normalized)
np.testing.assert_array_almost_equal(v_word, v_sent, decimal=5)
def test_sentence_is_normalized(self):
enc = make_encoder()
v = enc.encode_sentence("who wrote hamlet")
norm = np.linalg.norm(v)
if norm > 0:
assert abs(norm - 1.0) < 1e-5
def test_all_oov_returns_zero(self):
"""Sentence with no known words → zero vector. Honest abstention."""
enc = make_encoder()
v = enc.encode_sentence("glorpnax zibble fweep")
assert np.all(v == 0)
def test_empty_string_returns_zero(self):
enc = make_encoder()
v = enc.encode_sentence("")
assert np.all(v == 0)
def test_mixed_known_oov(self):
"""OOV words are skipped, known words contribute."""
enc = make_encoder()
v_pure = enc.encode_sentence("hamlet")
v_mixed = enc.encode_sentence("glorpnax hamlet zibble")
# Should be close to "hamlet" since it's the only known word
sim = float(np.dot(v_pure, v_mixed))
assert sim > 0.99
def test_different_sentences_different_vectors(self):
enc = make_encoder()
v1 = enc.encode_sentence("who wrote hamlet")
v2 = enc.encode_sentence("the cat is english")
# Different sentences should produce different vectors
sim = float(np.dot(v1, v2))
assert sim < 0.99
def test_word_order_matters(self):
"""Position weighting means order changes the vector."""
enc = make_encoder()
v1 = enc.encode_sentence("cat dog")
v2 = enc.encode_sentence("dog cat")
# Should be similar but not identical
sim = float(np.dot(v1, v2))
assert sim > 0.9 # mostly the same words
assert sim < 1.0 # but order differs
def test_punctuation_stripped(self):
enc = make_encoder()
v1 = enc.encode_sentence("who wrote hamlet?")
v2 = enc.encode_sentence("who wrote hamlet")
np.testing.assert_array_almost_equal(v1, v2, decimal=5)
def test_tokenization(self):
enc = make_encoder()
tokens = enc._tokenize("Who wrote Hamlet?")
assert tokens == ["who", "wrote", "hamlet"]
def test_tokenization_special_chars(self):
enc = make_encoder()
tokens = enc._tokenize("cat's dog-eared, the.")
assert tokens == ["cat", "s", "dog", "eared", "the"]
class TestEncoderNearestWords:
def test_nearest_to_itself(self):
"""A word's vector should be nearest to itself."""
enc = make_encoder()
v = enc.encode_word("hamlet")
nearest = enc.nearest_words(v, k=1)
assert len(nearest) == 1
assert nearest[0][0] == "hamlet"
assert abs(nearest[0][1] - 1.0) < 1e-5
def test_nearest_returns_k(self):
enc = make_encoder()
v = enc.encode_word("cat")
nearest = enc.nearest_words(v, k=3)
assert len(nearest) == 3
def test_nearest_sorted_by_similarity(self):
enc = make_encoder()
v = enc.encode_word("cat")
nearest = enc.nearest_words(v, k=5)
sims = [s for _, s in nearest]
assert sims == sorted(sims, reverse=True)
def test_nearest_zero_vector(self):
"""Zero vector = OOV. No nearest words."""
enc = make_encoder()
nearest = enc.nearest_words(np.zeros(DIM, dtype=np.float32))
assert nearest == []
class TestEncoderVocab:
def test_vocab_size(self):
enc = make_encoder()
assert enc.vocab_size == 10
def test_load_from_dict(self):
enc = Encoder(data_dir="/tmp/test", dim=DIM)
assert enc.vocab_size == 0
enc.load_from_dict({"hello": np.ones(DIM)})
assert enc.vocab_size == 1
def test_vectors_normalized_on_load(self):
"""All loaded vectors should be unit-length."""
enc = Encoder(data_dir="/tmp/test", dim=DIM)
enc.load_from_dict({"word": np.ones(DIM) * 5.0})
v = enc.encode_word("word")
assert abs(np.linalg.norm(v) - 1.0) < 1e-5
if __name__ == "__main__":
import pytest
pytest.main([__file__, "-v"])
|