# US 2019/0244122 A1 · Explainable artificial intelligence

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2019/0244122 A1
- Document ID: US20190244122A1
- Status: Published application
- Kind: A1
- Application number: 16/019,120
- Filed: 2018-06-26
- Published: 2019-08-08
- Earliest claimed: 2018-02-05
- Assignees: Accenture Global Solutions Limited
- Applicants: Accenture Global Solutions Limited
- Inventors: Chung-Sheng LI; Guanglei XIONG; Ashish JAIN; Emmanuel MUNGUIA TAPIA; Sukryool KANG; Benjamin Nathan GROSOF
- Web page: https://patentgrep.com/patent/US20190244122A1

## CPC

- G06N 5/048: Fuzzy inferencing
- G06N 5/022: Knowledge engineering; Knowledge acquisition
- G06N 5/042: {Backward inferencing}
- G06N 20/00: Machine learning
- G06N 5/045: Explanation of inference; Explainable artificial intelligence [XAI]; Interpretable artificial intelligence
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

Examples of artificial intelligence-based reasoning explanation are described. In an example implementation, a knowledge model having a plurality of ontologies and a plurality of inferencing rules is generated. Once the knowledge model is generated, based on a real-world problem, a knowledge model from amongst various knowledge models is selected to be used for resolving a real-world problem. The data procured from the real-world problem is clustered and classified into an ontology of the determined knowledge model. Inferencing rules to be used for deconstructing the real-world problem are identified, and a machine reasoning is generated to provide a hypothesis for the problem and an explanation to accompany the hypothesis.

## Claims

### Claim 1

An artificial intelligence (AI) system comprising:
a knowledge model constructor to generate a knowledge model, the knowledge model comprising a plurality of ontologies and a plurality of inferencing rules for generating an explanation for a decision made by the AI system, the knowledge model constructor comprising:
an ontology creator to create the plurality of ontologies based on data procured from real-world interactions of the knowledge model constructor, the data procured being real-world data, the real-world data comprising structured data, unstructured data, and semi-structured data; and
a rule constructor to construct the plurality of inferencing rules based on the ontology and the real-world data procured from the real-world interactions; and
an executor for decision-making when faced with a real-world problem, the executor comprising:
a knowledge model detector to select the knowledge model to be used for resolving the real-world problem, based on the real-world problem; and
an explainer to execute a decision-making process and to generate an explanation for the decision made, the explainer comprising:
a classifier and cluster-generator to classify and cluster data procured from the real-world problem into an ontology of the determined knowledge model;
a rule identifier to identify inferencing rules to be used for deconstructing the real-world problem; and
an inferencer and hypothesis generator to provide a machine-learning based reasoning to provide a hypothesis for the real-world problem and an explanation to accompany the hypothesis, wherein the inferencer and hypothesis generator is to capture a confidence level parameter indicating a strength of the explanation for the decision made.

### Claim 2 (depends on 1)

The AI system as claimed in claim 1, wherein the explainer is to generate a visual representation of the decision and the explanation using machine learning techniques.

### Claim 3 (depends on 1)

The AI system as claimed in claim 1, wherein the rule identifier is to parse the real-world problem to curate logical sentences in the real-world problem.

### Claim 4 (depends on 1)

The AI system as claimed in claim 1, wherein the rule identifier is to provide a human-readable form of the deconstruction of the real-world problem to preview a line of reasoning.

### Claim 5 (depends on 1)

The AI system as claimed in claim 1, wherein the identified inferencing rules are triggered using one of a forward chaining process and a backward chaining process.

### Claim 6

An AI system comprising:
a knowledge model constructor to generate a knowledge model, the knowledge model comprising a plurality of ontologies and a plurality of inferencing rules for generating an explanation to decision made by the AI system, the knowledge model constructor comprising:
an ontology creator to create the plurality of ontologies based on data procured from real-world interactions of the knowledge model constructor, the data procured being real-world data, the real-world data comprising structured data, unstructured data, and semi-structured data, wherein the ontology creator is to utilize inductive reasoning techniques for creating the ontology; and
a rule constructor to construct the plurality of inferencing rules based on the ontology and the real-world data procured from the real-world interactions, wherein the rule constructor is to utilize deductive reasoning techniques for creating the plurality of inferencing rules; and
an executor for decision-making when faced with a real-world problem, the executor comprising:
a knowledge model detector to select the knowledge model to be used for resolving the real-world problem, based on the real-world problem; and
an explainer to execute a decision-making process and to generate an explanation for the decision made, the explainer comprising:
a classifier and cluster-generator to classify and cluster data procured from the real-world problem into an ontology of the determined knowledge model, wherein the classifier and cluster-generator is to use the inductive reasoning techniques for classifying and clustering data;
a rule identifier to identify inferencing rules to be used for deconstructing the real-world problem, wherein the rule identifier is to utilize the deductive reasoning techniques for deconstructing the real-world problem; and
an inferencer and hypothesis generator to provide a machine-generated reasoning to provide a hypothesis for the problem and an explanation to accompany the hypothesis, wherein the inferencer and hypothesis generator is to utilize abductive reasoning techniques for providing the hypothesis and the explanation.

### Claim 7 (depends on 6)

The AI system as claimed in claim 6, wherein the rule constructor is to build a strength of the explanation into the inferencing rules.

### Claim 8 (depends on 6)

The AI system as claimed in claim 6, wherein the inferencer and hypothesis generator is to determine a confidence parameter, indicative of a strength, associated with a decision and the associated explanation.

### Claim 9 (depends on 6)

The AI system as claimed in claim 6, wherein the rule identifier is to parse the real-world problem to curate logical sentences in the real-world problem.

### Claim 10 (depends on 6)

The AI system as claimed in claim 6, wherein the rule identifier is to provide a human-readable form of the deconstruction of the real-world problem to preview a line of reasoning.

### Claim 11 (depends on 6)

The AI system as claimed in claim 6, wherein the identified inferencing rules are triggered using one of a forward chaining process and a backward chaining process.

### Claim 12 (depends on 6)

The AI system as claimed in claim 6, wherein the inductive reasoning techniques comprise one of supervised clustering techniques and unsupervised clustering techniques.

### Claim 13 (depends on 6)

The AI system as claimed in claim 6, wherein the abductive reasoning techniques comprise evidence-based reasoning techniques.

### Claim 14

A non-transitory computer readable medium including machine readable instructions that are executable by a processor to:
generate a knowledge model, the knowledge model comprising a plurality of ontologies and a plurality of inferencing rules for generating an explanation to decision made by an AI system, wherein generating the ontology comprises creating the plurality of ontologies based on data procured from real-world interactions using inductive reasoning techniques for creating the ontology, and creating the plurality of inferencing rules based on the ontology and the data procured from the real-world interactions using deductive reasoning techniques for creating the plurality of inferencing rules;
determine the knowledge model to be used for resolving a real-world problem, based on the real-world problem;
classify and cluster data procured from the real-world problem into an ontology of the determined knowledge model using the inductive reasoning techniques for classifying and clustering data;
identify inferencing rules to be used for deconstructing the real-world problem using deductive reasoning techniques for deconstructing the real-world problem; and
provide a machine-learning based reasoning to provide a hypothesis for the problem and an explanation to accompany the hypothesis using abductive reasoning techniques for providing the hypothesis and the explanation.

### Claim 15 (depends on 14)

The non-transitory computer readable medium as claimed in claim 14, wherein to provide the machine-learning based reasoning, the processor is to generate a visual representation of the decision and the explanation using machine learning techniques.

### Claim 16 (depends on 14)

The non-transitory computer readable medium as claimed in claim 14, wherein to identify the inferencing rule, the processor is to parse the real-world problem to curate logical sentences in the real-world problem.

### Claim 17 (depends on 14)

The non-transitory computer readable medium as claimed in claim 14, wherein to identify the inferencing rule, the processor is to provide a human-readable form of the deconstruction of the real-world problem to preview of a line of reasoning.

### Claim 18 (depends on 14)

The non-transitory computer readable medium as claimed in claim 14, wherein the identified inferencing rules are triggered using one of a forward chaining process and a backward chaining process.

### Claim 19 (depends on 14)

The non-transitory computer readable medium as claimed in claim 14, wherein the inductive reasoning techniques comprise one of supervised clustering techniques and unsupervised clustering techniques.

### Claim 20 (depends on 14)

The non-transitory computer readable medium as claimed in claim 14, wherein the abductive reasoning techniques comprise evidence-based reasoning techniques.
