IP Library Patent Application 17008714
Patent Application
App. No. 17/008,714

TRAINING APPARATUS AND NON-TRANSITORY COMPUTER READABLE MEDIUM

Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US None
App. No.
17/008,714
Abstract

A training apparatus includes an input unit that inputs multiple pairs of input and output, a processor, and an output unit. The processor is configured to, through execution of a program, generate the pairs of input and output as positive examples, and generate, as negative examples, pairs in which the combinations of input and output are changed. The processor is further configured to train a filter model by using the positive examples and the negative examples, and use the filter model to perform filtering by removing incorrect pairs from the pairs of input and output.

Claims (55)

1 . A training apparatus comprising:

an input unit that inputs a plurality of pairs of input and output;

a processor; and

an output unit,

wherein the processor is configured to, through execution of a program,

generate the plurality of pairs of input and output as positive examples, and generate, as negative examples, pairs in which the combinations of input and output are changed,

train a filter model by using the positive examples and the negative examples, and

use the filter model to perform filtering by removing incorrect pairs from the plurality of pairs of input and output.

2 . The training apparatus according to claim 1 ,

wherein the processor is further configured to

train a model by using the filtered pairs of input and output, the model obtaining the output in response to the input.

3 . The training apparatus according to claim 1 ,

wherein the processor is configured to

generate the negative examples by switching the plurality of pairs of input and output randomly.

4 . The training apparatus according to claim 2 ,

wherein the processor is configured to

generate the negative examples by switching the plurality of pairs of input and output randomly.

5 . The training apparatus according to claim 1 ,

wherein the processor is configured to

generate the negative examples on a basis of a degree of similarity between the input and the output.

6 . The training apparatus according to claim 2 ,

wherein the processor is configured to

generate the negative examples on a basis of a degree of similarity between the input and the output.

7 . The training apparatus according to claim 2 ,

wherein the processor is configured to

subject the filter model to reinforced training on a basis of an output result from the trained model obtaining the output in response to the input.

8 . The training apparatus according to claim 1 ,

wherein the filter model uses a discrimination probability indicating whether or not a pair of input and output is correct.

9 . The training apparatus according to claim 2 ,

wherein the filter model uses a discrimination probability indicating whether or not a pair of input and output is correct.

10 . The training apparatus according to claim 3 ,

wherein the filter model uses a discrimination probability indicating whether or not a pair of input and output is correct.

11 . The training apparatus according to claim 4 ,

wherein the filter model uses a discrimination probability indicating whether or not a pair of input and output is correct.

12 . The training apparatus according to claim 5 ,

wherein the filter model uses a discrimination probability indicating whether or not a pair of input and output is correct.

13 . The training apparatus according to claim 6 ,

wherein the filter model uses a discrimination probability indicating whether or not a pair of input and output is correct.

14 . The training apparatus according to claim 7 ,

wherein the filter model uses a discrimination probability indicating whether or not a pair of input and output is correct.

15 . The training apparatus according to claim 1 ,

wherein the filter model uses entropy calculated from a discrimination probability indicating whether or not a pair of input and output is correct.

16 . The training apparatus according to claim 2 ,

wherein the filter model uses entropy calculated from a discrimination probability indicating whether or not a pair of input and output is correct.

17 . The training apparatus according to claim 3 ,

wherein the filter model uses entropy calculated from a discrimination probability indicating whether or not a pair of input and output is correct.

18 . The training apparatus according to claim 1 ,

wherein the input is text data and the output is summary data of the text data.

19 . The training apparatus according to claim 1 ,

wherein the input is original-text data and the output is translation data of the original-text data.

20 . A non-transitory computer readable medium storing a program causing a computer to execute a process comprising:

inputting a plurality of pairs of input and output;

generating the plurality of pairs of input and output as positive examples, and generating, as negative examples, pairs in which the combinations of input and output are changed;

training a filter model by using the positive examples and the negative examples; and

using the filter model to perform filtering by removing incorrect pairs from the plurality of pairs of input and output.

Assignments (2)
CHANGE OF NAME Recorded Apr 28, 2021
From: FUJI XEROX CO., LTD.
To: FUJIFILM BUSINESS INNOVATION CORP.
Reel/Frame 056078/0098 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Sep 1, 2020
From: KANO, RYUJI; TANIGUCHI, TOMOKI; OHKUMA, TOMOKO
To: FUJI XEROX CO., LTD.
Reel/Frame 053653/0675 →