Supercharge Your Innovation With Domain-Expert AI Agents!

Language sequence model decoding method

A decoding method and sequence technology, applied in computing models, natural language data processing, special data processing applications, etc., can solve problems such as the inability to obtain global optimal solutions, reduce the number of generated vertices and the amount of calculation, speed up the solution process, The effect of reducing computational complexity

Pending Publication Date: 2020-08-25
EISOO SOFTWARE
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0021] The purpose of the present invention is to provide a language sequence model decoding method in order to overcome the defects in the above-mentioned prior art, so as to solve the problem that the prior art cannot obtain the global optimal solution

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Language sequence model decoding method
  • Language sequence model decoding method
  • Language sequence model decoding method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0081] Such as figure 1 Shown, a language sequence model decoding method comprises the following steps:

[0082] S1. Initialization: Based on the sequence model, use the greedy algorithm to generate the initial language sequence, and construct the directed graph to obtain the current shortest path length from the starting point to the end point and the shortest path length from the current vertex to the starting point in the initial language sequence;

[0083] S2. Extension: Input the current vertex information in the sequence model to generate a language sequence, and filter out temporary vertices according to the conditional probability of each word element in the language sequence;

[0084] S3, clipping: according to the existence of temporary vertices, filter to obtain ordinary vertices;

[0085] S4. Selection: Select a new current vertex from common vertices. If the word element corresponding to the new current vertex is the end word element, the language sequence corres...

Embodiment 2

[0145] In order to further verify the effectiveness of the present invention, this embodiment uses the language sequence model that has been trained, respectively uses the decoding algorithm based on the greedy algorithm to decode and compares the result sequence obtained by decoding based on the Dijkstra algorithm proposed by the present invention, as shown in Table 12- The comparison results shown in Table 14:

[0146] Table 12

[0147]

[0148]

[0149] Table 13

[0150]

[0151]

[0152] Table 14

[0153]

[0154] From the comparison results of Table 12 to Table 14, it can be seen that the decoding based on the greedy algorithm can only obtain a sequence of local optimal solutions, while the overall probability value of the sequence obtained by decoding based on the idea of ​​Dijkstra algorithm in the present invention is obtained based on the decoding of the greedy algorithm Several times or even ten times of the sequence probabil...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention relates to a language sequence model decoding method, which comprises the following steps of: initializing: based on a sequence model, generating an initial language sequence by utilizing a greedy algorithm, and respectively obtaining the current shortest path length from a starting point to an end point and the shortest path length from a current vertex to the starting point in theinitial language sequence by constructing a directed graph; expanding: inputting current vertex information into the sequence model, generating a language sequence, and performing screening to obtaina temporary vertex according to the conditional probability of each word element in the language sequence; cutting: performing screening to obtain common vertexes according to whether temporary vertexes exist or not; selection: selecting a new current vertex from the common vertexes, if the word element corresponding to the new current vertex is the word element of the terminal point, regarding that the language sequence corresponding to the shortest path between the starting point and the new current vertex is the global maximum occurrence probability sequence, and otherwise, returining to the extension stage to start a new round of solution. Compared with the prior art, the method has the advantage that the global optimal solution can be quickly and accurately solved in the decoding process.

Description

technical field [0001] The invention relates to the technical field of natural language processing, in particular to a language sequence model decoding method. Background technique [0002] Nowadays, the use of machine learning for natural language processing has become a mainstream practical method. Since the sequence model can learn and predict a series of ordered data queues with continuous relationships, such as voice data, text data, and video data, such The model has a wide range of applications in practice. Common sequence model applications include: speech recognition, image summarization, music generator, machine translation, DNA sequence analysis, named entity recognition, etc. Among them, for speech recognition, its input data and output data are both sequence data, input data X is an audio segment played in time sequence, and output Y is a word sequence; for image summarization, only output data is sequence data, That is, the input data X is a picture data, and...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
IPC IPC(8): G06F40/216G06F16/901G06N20/00
CPCG06F40/216G06F16/9024G06N20/00Y02D10/00
Inventor 肖强马祥祥
Owner EISOO SOFTWARE
Features
  • R&D
  • Intellectual Property
  • Life Sciences
  • Materials
  • Tech Scout
Why Patsnap Eureka
  • Unparalleled Data Quality
  • Higher Quality Content
  • 60% Fewer Hallucinations
Social media
Patsnap Eureka Blog
Learn More

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2025 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com