Title:
SENTENCE-LEVEL CONVOLUTION LSTM TRAINING METHOD, AND DEVICE AND READABLE MEDIUM
Document Type and Number:
WIPO Patent Application WO/2021/174824
Kind Code:
A1
Abstract:
The present invention relates to a sentence-level convolution LSTM training method, comprising: aggregating, in a one-dimensional convolution mode, hidden states of the current word and an adjacent word at the previous moment, and using the hidden states as a sentence vector input; using the sub-state of the current word of a sentence at the current moment as a word vector input; sending the sentence vector input, the word vector input, and a cell state of the current word at the previous moment into a logic gate to obtain a cell state of the current word at the current moment; and sending the sentence vector input, the word vector input, and the cell state of the current word at the current moment into an output gate to obtain and output the hidden state of the current word at the current moment. The present invention further relates to a computer device and a readable storage medium. According to the present invention, an integral sentence is regarded as a single state, containing a set of word-level sub-states in a sentence length, and local information around each word is aggregated by using one-dimensional convolution, so that the parallel computing capability is greatly improved, and the time and capital costs are saved.
More Like This:
Inventors:
ZHANG KAI (CN)
Application Number:
PCT/CN2020/118341
Publication Date:
September 10, 2021
Filing Date:
September 28, 2020
Export Citation:
Assignee:
SUZHOU INSPUR INTELLIGENT TECH CO LTD (CN)
International Classes:
G06F40/211
Foreign References:
| CN111597792A | 2020-08-28 | |||
| CN108304911A | 2018-07-20 | |||
| CN108363753A | 2018-08-03 | |||
| AU2018100320A4 | 2018-04-26 | |||
| CN109783817A | 2019-05-21 |
Attorney, Agent or Firm:
LIAN & LIEN IP ATTORNEYS (CN)
Download PDF: