公开(公告)号：US20180157744A1

公开(公告)日：2018-06-07

申请号：US15604677

申请日：2017-05-25

Applicant: INSTITUTE FOR INFORMATION INDUSTRY

Inventor： Ping-I CHEN ,  Tai-Ta KUO ,  Yen-Heng TSAO ,  Yu-Chuan YANG

IPC: G06F17/30 ,  G06F17/27

CPC classification number: G06F17/277 ,  G06F16/313 ,  G06F16/3344 ,  G06Q10/10

Abstract: A comparison table automatic generation method that includes the steps outlined below is provided. An interface is provided to set comparison topics, a basic article, a basic article object and marked paragraphs. Correlation between basic article words of the marked paragraphs is calculated to generate a marked main tag and marked enriched words to further retrieve collected article and a collected article object accordingly. Correlation between collected article words of the collected article paragraphs are calculated to generate main tag and enriched words of collected article to be compared with the marked main tag and the marked enriched words to calculate a similarity to further generate selected paragraphs accordingly. A comparison table that includes the comparison topics, the basic and collected article objects as the items of rows and columns therein is established such that the marked and the selected paragraphs are filled in entries of the comparison table.

公开(公告)号：US20220147835A1

公开(公告)日：2022-05-12

申请号：US17111499

申请日：2020-12-03

Applicant: Institute For Information Industry

Inventor： Hsin-Yi KUO ,  Wen-Nan WANG ,  Jia-Wei KAO ,  Wen-Fa HUANG ,  Po-Hsien CHIANG ,  Fu-Jheng JHENG ,  Yi-Hsiu LEE ,  Yu-Chuan YANG

IPC: G06N5/02 ,  G06N5/04 ,  G06F40/279 ,  G06K9/62 ,  G06N3/08

Abstract: A knowledge graph construction system and method are disclosed. The system generates a recommended subject entity, at least one recommended object entity, and at least one recommended relation for a piece of text data according to the text data and a plurality of triples. The system displays the recommended object entity and the recommended relation at a current paragraph of the text data according to the recommended subject entity for user to select. The system receives a confirmed message related to the recommended subject entity, a recommended object entity selected by user from the at least one recommended object entity, and a recommended relation selected by user from the at least one recommended relation. The system adds the recommended subject entity and the selected recommended object entity and recommended relation to the triples, and constructs a current knowledge graph by using the triples according to the confirmed message.

公开(公告)号：US20230135782A1

公开(公告)日：2023-05-04

申请号：US17456588

申请日：2021-11-25

Applicant: INSTITUTE FOR INFORMATION INDUSTRY

Inventor： Tai-Ta KUO ,  Yu-Chuan YANG ,  Jia Wei KAO ,  Fu-Jheng JHENG ,  Yi Hsiu LEE ,  Ping-I CHEN

IPC: G06Q40/08 ,  G06N5/02 ,  G16H10/60

Abstract: A disability level automatic judgment device is disclosed. The disability level automatic judgment device includes a processor and a memory. The processor is configured to create a diagnosis information graph according to a diagnosis content, to compare the diagnosis information graph and a standard disability graph, so as to determine a first disability level, and to generate a judgment result according to the first disability level. The memory is coupled to the processor, and the memory is configured to store the standard disability graph.

公开(公告)号：US20250139506A1

公开(公告)日：2025-05-01

申请号：US18515660

申请日：2023-11-21

Applicant: INSTITUTE FOR INFORMATION INDUSTRY

Inventor： Yu-Chuan YANG ,  Tai-Ta Kuo ,  Ping-I Chen

IPC: G06N20/00

Abstract: A training system and a training method for a domain-specific data model are provided. The training method includes configuring a computing device to perform the following processes: generating, by a training set generation module, a training data set based on a domain knowledge graph; updating the data model based on the training data set; generating, by the training set generation module, training input text corresponding to the domain knowledge graph; inputting the training input text into the data model to obtain training output text; evaluating and generating a score by an evaluation module based on a correlation between the training output text and the domain knowledge graph; and adjusting, by a reinforcement learning module, parameters of the data model according to the score and an optimization goal of the reward model until the score meets a training completion condition, taking the data model as the domain-specific data model.

公开(公告)号：US20220138508A1

公开(公告)日：2022-05-05

申请号：US17106171

申请日：2020-11-29

Applicant: INSTITUTE FOR INFORMATION INDUSTRY

Inventor： Chu-Chun HUANG ,  Yu-Chuan YANG ,  Yen-Heng TSAO ,  Tzu-Ying CHEN ,  Po-Hsien CHIANG ,  Fu-Jheng JHENG

IPC: G06K9/62 ,  G06F40/284

Abstract: A device and a method for detecting a purpose of an article are provided. The device is configured to divide the article into a plurality of sentences and input the sentences to a feature identification model to generate a contextualized word vector corresponding to each of the sentences. The device further inputs the representation to a specific purpose detecting model to generate a distributed representation similarity of the article. When the distributed representation similarity of the article is greater than a threshold, the device determines that the article conforms to a specific purpose.