|TensorFlow 1 version|
Generates a word rank-based probabilistic sampling table.
Compat aliases for migration
See Migration guide for more details.
tf.keras.preprocessing.sequence.make_sampling_table( size, sampling_factor=1e-05 )
Used for generating the
sampling_table argument for
sampling_table[i] is the probability of sampling
the word i-th most common word in a dataset
(more common words should be sampled less frequently, for balance).
The sampling probabilities are generated according to the sampling distribution used in word2vec:
p(word) = (min(1, sqrt(word_frequency / sampling_factor) / (word_frequency / sampling_factor)))
We assume that the word frequencies follow Zipf's law (s=1) to derive a numerical approximation of frequency(rank):
frequency(rank) ~ 1/(rank * (log(rank) + gamma) + 1/2 - 1/(12*rank))
gamma is the Euler-Mascheroni constant.
size: Int, number of possible words to sample. sampling_factor: The sampling factor in the word2vec formula.
A 1D Numpy array of length `size` where the ith entry is the probability that a word of rank i should be sampled.