Data processing method and device, electronic equipment and computer readable storage medium
A technology for data processing and computer programs, applied in the fields of data processing, devices, electronic equipment, and computer-readable storage media, can solve problems such as time-consuming, slow training speed, and inability to meet practical needs, so as to reduce the amount of calculation and improve The effect of data processing efficiency
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment approach
[0113] In this alternative embodiment, the method further includes:
[0114] According to the maximum character length of each sample text in the training sample, each sample text in the training sample is paired by the character length to obtain the character length, which is obtained after the pre-treatment of training sample set, where the pre-processed training sample set each sample The text is consistent with the length of the text and matches the maximum character length;
[0115] The training sample is set to the initial neural network model, including:
[0116] Enter the pre-processed training sample set to the initial neural network model.
[0117] Among them, the maximum length of the character length of each sample text is determined to be trained as the maximum character length of each sample text pair (hereinafter also referred to as the minimum character length corresponding to the training sample set) of each sample text. .
[0118] For each training sample set, ea...
specific Embodiment approach
[0145] Alternatively, the loss of the training sample set is determined according to the predicted results corresponding to the blocking text fragment of each sample text, respectively, respectively, respectively, respectively, the label, respectively, the loss, including:
[0146]According to the respective prediction results corresponding to the occlusion text fragment of each sample text, respectively correspond to the real text fragment of the segments of each sample text, respectively, the losses of each occlusion text fragment are determined;
[0147] Determine the loss of training samples based on the loss of each occlusion text fragment.
[0148] In this implementation, the loss of each occlusion text fragment refers to the corresponding prediction result of the blocking text fragment and the loss value between the segment of the shutdown section corresponding to the segment.
[0149] When the loss of the training sample set is calculated, the loss of the training sample s...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More