mindspore.ops.EditDistance

class mindspore.ops.EditDistance(*args, **kwargs)[source]

Computes the Levenshtein Edit Distance. It is used to measure the similarity of two sequences. The inputs are variable-length sequences provided by SparseTensors (hypothesis_indices, hypothesis_values, hypothesis_shape) and (truth_indices, truth_values, truth_shape).

Parameters

normalize (bool) – If true, edit distances are normalized by length of truth. Default: True.

Inputs:

  • hypothesis_indices (Tensor) - The indices of the hypothesis list SparseTensor. With int64 data type. The shape of tensor is \((N, R)\).

  • hypothesis_values (Tensor) - The values of the hypothesis list SparseTensor. Must be 1-D vector with length of N.

  • hypothesis_shape (Tensor) - The shape of the hypothesis list SparseTensor. Must be R-length vector with int64 data type. Only constant value is allowed.

  • truth_indices (Tensor) - The indices of the truth list SparseTensor. With int64 data type. The shape of tensor is \((M, R)\).

  • truth_values (Tensor) - The values of the truth list SparseTensor. Must be 1-D vector with length of M.

  • truth_shape (Tensor) - The shape of the truth list SparseTensor. Must be R-length vector with int64 data type. Only constant value is allowed.

Outputs:

Tensor, a dense tensor with rank R-1 and float32 data type.

Raises

TypeError – If normalize is not a bool.

Supported Platforms:

Ascend

Examples

>>> import numpy as np
>>> from mindspore import context
>>> from mindspore import Tensor
>>> import mindspore.nn as nn
>>> import mindspore.ops.operations as ops
>>> class EditDistance(nn.Cell):
...     def __init__(self, hypothesis_shape, truth_shape, normalize=True):
...         super(EditDistance, self).__init__()
...         self.edit_distance = ops.EditDistance(normalize)
...         self.hypothesis_shape = hypothesis_shape
...         self.truth_shape = truth_shape
...
...     def construct(self, hypothesis_indices, hypothesis_values, truth_indices, truth_values):
...         return self.edit_distance(hypothesis_indices, hypothesis_values, self.hypothesis_shape,
...                                   truth_indices, truth_values, self.truth_shape)
...
>>> hypothesis_indices = Tensor(np.array([[0, 0, 0], [1, 0, 1], [1, 1, 1]]).astype(np.int64))
>>> hypothesis_values = Tensor(np.array([1, 2, 3]).astype(np.float32))
>>> hypothesis_shape = Tensor(np.array([1, 1, 2]).astype(np.int64))
>>> truth_indices = Tensor(np.array([[0, 1, 0], [0, 0, 1], [1, 1, 0], [1, 0, 1]]).astype(np.int64))
>>> truth_values = Tensor(np.array([1, 3, 2, 1]).astype(np.float32))
>>> truth_shape = Tensor(np.array([2, 2, 2]).astype(np.int64))
>>> edit_distance = EditDistance(hypothesis_shape, truth_shape)
>>> output = edit_distance(hypothesis_indices, hypothesis_values, truth_indices, truth_values)
>>> print(output)
[[1. 1.]
 [1. 1.]]