# US 2021/0241906 A1 · Machine-aided dialog system and medical condition inquiry apparatus and method

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

- Number: US 2021/0241906 A1
- Document ID: US20210241906A1
- Status: Published application
- Kind: A1
- Application number: 16/631,824
- Filed: 2019-02-20
- Published: 2021-08-05
- Earliest claimed: 2018-09-19
- Assignees: Boe Technology Group Co., Ltd.
- Applicants: Boe Technology Group Co., Ltd.
- Inventors: Zhenzhong ZHANG
- Web page: https://patentgrep.com/patent/US20210241906A1

## CPC

- G06N 5/04: Inference or reasoning models
- G16H 50/20: for computer-aided diagnosis, e.g. based on medical expert systems
- G06F 16/2379: {Updates performed during online database operations; commit processing}
- G06F 40/30: Semantic analysis
- G06F 40/35: Discourse or dialogue representation
- G06N 5/022: Knowledge engineering; Knowledge acquisition
- G16H 10/20: for electronic clinical trials or questionnaires
- G16H 15/00: ICT specially adapted for medical reports, e.g. generation or transmission thereof
- G16H 70/60: relating to pathologies
- G16H 80/00: ICT specially adapted for facilitating communication between medical practitioners or patients, e.g. for collaborative diagnosis, therapy or health monitoring
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A machine-aided dialog system as well as a medical condition inquiry method and apparatus employing the machine aided dialog system are disclosed. There is provided a machine-aided dialog system, comprising: a semantic framework for providing a dialog theme; a knowledge database for storing semantic knowledge, the semantic knowledge comprising at least one group of content-related language expressions manners; and a dialog manager for managing a dialog according to the dialog theme provided by the semantic framework, wherein the managing a the dialog comprises: generating a question expression of the dialog theme based on the semantic knowledge acquired from the knowledge database; and ending the dialog theme once upon determining that the dialog theme is completed in response to a dialog ending condition.

## Claims

### Claim 1

A machine-aided dialog system, comprising:
a semantic framework for providing a dialog theme;
a knowledge database for storing semantic knowledge, the semantic knowledge comprising at least one group of content-related expressions; and
a dialog manager for managing a dialog according to the dialog theme provided by the semantic framework, wherein managing the dialog comprises: generating a question expression of the dialog theme based on the semantic knowledge acquired from the knowledge database; and ending the dialog theme upon determining that the dialog theme is completed in response to a dialog ending condition.

### Claim 2 (depends on 1)

The machine-aided dialog system of claim 1, wherein managing the dialog further comprises: according to a received response expression, extracting a keyword from the received response expression based on a semantic context and/or syntactic structure in the semantic knowledge; and generating one or more question expressions according to the keyword in the response expression and a preset expression generation template.

### Claim 3 (depends on 1)

The machine-aided dialog system of claim 1, wherein the dialog manager comprises a mapping sub-unit for mapping a term in a response expression into a corresponding expression of the semantic knowledge.

### Claim 4 (depends on 3)

The machine-aided dialog system of claim 3, wherein the mapping sub-unit comprises:
a term vector calculator for calculating a term vector of each term in a term set, wherein both a set of predetermined expressions of the sematic knowledge and a set of terms in the response expression are proper subsets of the term set;
a similarity calculator for respectively calculating similarities between a term in the response expression and each expression in the set of predetermined expressions according to a term vector of the term and term vectors of all expressions in the set of predetermined expressions; and
a replacer for replacing the term in the response expression with the corresponding expression of the semantic knowledge in response to the condition that the calculated similarity meets a threshold condition.

### Claim 5 (depends on 4)

The machine-aided dialog system of claim 4, wherein the term vector calculator is further configured to indicate each term in the term set as a multi-dimensional Gaussian distribution and take a mean value of the multi-dimensional Gaussian distribution as a term vector of the each term.

### Claim 6 (depends on 5)

The machine-aided dialog system of claim 5, wherein the term vector calculator calculates the term vector of each term in the term set in a following manner:
it is assumed that each term w in the term set represents a multi-dimensional Gaussian distribution f(w)˜N(μw, Σw), where μw and Σw respectively represent a mean value and a covariance of the multi-dimensional Gaussian distribution of the term w, and Σw is a diagonal matrix,
L(w, cp, cn)=max(0, 1−S(w, cp)+S(w, cn)) (1)
where cp and cn both are elements in the term set, cp is a term in context of the term w in all response expressions received by the machine-aided dialog system, cn is a term not appearing in the context of the term w in any response expressions received by the machine-aided dialog system, and function S(.,.) represents a similarity calculation function; and for given terms w1and w2, a similarity between w1 and w2 is calculated as follows:
S
⁡
(
w
1
,
w
2
)
=
⁢
∫
N
⁡
(
x
;
u
w
1
,
Σ
w
1
)
⁢
log
⁢
N
⁡
(
x
;
u
w
2
,
Σ
w
2
)
N
⁡
(
x
;
u
w
1
,
Σ
w
1
)
⁢
dx
=
⁢
1
2
⁢
{
t
⁢
r
⁡
(
Σ
w
1
-
1
⁢
Σ
w
2
)
+
(
u
w
1
-
u
w
2
)
T
⁢
Σ
w
1
-
1
⁢
(
u
w
1
-
u
w
2
)
-
⁢
log
⁢
det
⁡
(
Σ
w
2
)
det
⁡
(
Σ
w
1
)
-
d
}
(
2
)
where tr( ) represents calculating a trace of the matrix, Σ−1 represents an inversion of the matrix, det represents calculating a value of a determinant, and d represents a constant irrelevant to μ and Σ;
the mean value μw of the Gaussian distribution of the term w, the mean value μcp of the Gaussian distribution of cp, and the mean value μcn of the Gaussian distribution of cn when L is minimum are calculated by substituting the formula (2) to the formula (1), and the mean values are respectively taken as term vectors of the term w, cp and cn; and the above operations are repeated till the term vectors of all terms in the term set are obtained.

### Claim 7 (depends on 4)

The machine-aided dialog system of claim 4, wherein the term vector calculator is further configured to decompose a term-term matrix, at least partially map a term to a potential semantic space based on an Explicit Semantic Analysis (ESA) to obtain a vector of the term in the potential semantic space and take the vector as a term vector of the term.

### Claim 8 (depends on 7)

The machine-aided dialog system of claim 7, wherein the term vector calculator calculates the term vector of each term in the term set in a following manner:
⁢
M
≈
P
T
⁢
W
(
3
)
min
P
,
W
⁢
∑
i
=
1
n
⁢
∑
j
=
1
m
⁢
(
M
i
,
j
-
P
i
T
⁢
W
j
)
2
+
α
⁢
∑
i
=
1
n
⁢
∑
f
=
1
n
⁢
E
i
,
f
⁢

P
i
-
P
f

F
2
+
λ
⁡
(

P

F
2
+

W

F
2
)
(
4
)
where V={v1, v2, . . . , vm} represents the term set, vi and vj (1≤i, j≤m) each represents a single term, M∈Rm×m represents the term-term matrix, Mi,j is a Term Frequency-Inverse Document Frequency (tf-idf) weight of the term vj in the term vi, Ei,j is a similarity between the terms vi and Vj, ∥·∥2F is a Frobenius norm, α and λ are preset non-negative parameters, P∈Rk×m, W∈Rk×m, and R is a real number; and P is calculated under a condition in which the formula (4) is minimum according to the formula (3) and the formula (4), and a term vector Pi of the term vi and a term vector Pj of the term vj are obtained.

### Claim 9 (depends on 4)

The machine-aided dialog system of claim 4, wherein the condition that the similarity meets the threshold condition comprises: a cosine similarity between the term vector of the term in the response expression and a term vector of the corresponding expression of the semantic knowledge is greater than or equal to a predetermined threshold.

### Claim 10

A method of machine-aided dialog, comprising:
providing a semantic framework for providing a dialog theme;
providing a knowledge database for storing semantic knowledge, the semantic knowledge comprising at least one group of content-related expressions; and
providing a dialog manager for managing a dialog according to the dialog theme provided by the semantic framework, wherein managing the dialog comprises: generating a question expression of the dialog theme based on the semantic knowledge acquired from the knowledge database; and ending the dialog theme upon determining that the dialog theme is completed in response to a dialog ending condition.

### Claim 11 (depends on 10)

The method of claim 10, wherein managing the dialog further comprises:
according to a received response expression, extracting a keyword from the received response expression based on a semantic context and/or syntactic structure in the semantic knowledge; and generating one or more question expressions according to the keyword in the response expression and a preset expression generation template.

### Claim 12 (depends on 10)

The method of claim 10, wherein the dialog manager comprises a mapping sub-unit for mapping a term in a response expression into a corresponding expression of the semantic knowledge.

### Claim 13 (depends on 12)

The method of claim 12, wherein the mapping sub-unit comprises:
a term vector calculator for calculating a term vector of each term in a term set, wherein both a set of predetermined expressions of the sematic knowledge and a set of terms in the response expression are proper subsets of the term set;
a similarity calculator for respectively calculating similarities between a term in the response expression and each expression in the set of predetermined expressions according to a term vector of the term and term vectors of all expressions in the set of predetermined expressions; and
a replacer for replacing the term in the response expression with the corresponding expression of the semantic knowledge in response to the condition that the calculated similarity meets a threshold condition.

### Claim 14 (depends on 13)

The method of claim 13, wherein the term vector calculator is further configured to indicate each term in the term set as a multi-dimensional Gaussian distribution and take a mean value of the multi-dimensional Gaussian distribution as a term vector of the each term.

### Claim 15 (depends on 14)

The method of claim 14, wherein the term vector calculator calculates the term vector of each term in the term set in a following manner:
it is assumed that each term w in the term set represents a multi-dimensional Gaussian distribution f(w)˜N(μw, Σw), where μw and Σw respectively represent a mean value and a covariance of the multi-dimensional Gaussian distribution of the term w, and w is a diagonal matrix,
L(w, cp, cn)=max(0, 1−S(w, cp)+S(w, cn)) (1)
where cp and cn both are elements in the term set, cp is a term in context of the term w in all response expressions received by the machine-aided dialog system, cn is a term not appearing in the context of the term w in any response expressions received by the machine-aided dialog system, and function S(.,.) represents a similarity calculation function; and for given terms w1 and w2, a similarity between w1 and w2 is calculated as follows:
S
⁡
(
w
1
,
w
2
)
=
⁢
∫
N
⁡
(
x
;
u
w
1
,
Σ
w
1
)
⁢
log
⁢
N
⁡
(
x
;
u
w
2
,
Σ
w
2
)
N
⁡
(
x
;
u
w
1
,
Σ
w
1
)
⁢
dx
=
⁢
1
2
⁢
{
t
⁢
r
⁡
(
Σ
w
1
-
1
⁢
Σ
w
2
)
+
(
u
w
1
-
u
w
2
)
T
⁢
Σ
w
1
-
1
⁢
(
u
w
1
-
u
w
2
)
-
⁢
log
⁢
det
⁡
(
Σ
w
2
)
det
⁡
(
Σ
w
1
)
-
d
}
(
2
)
where tr( ) represents calculating a trace of the matrix, Σ−1 represents an inversion of the matrix, det represents calculating a value of a determinant, and d represents a constant irrelevant to μ and Σ;
the mean value μw of the Gaussian distribution of the term w, the mean value μcp of the Gaussian distribution of cp, and the mean value μcn of the Gaussian distribution of cn when L is minimum are calculated by substituting the formula (2) to the formula (1), and the mean values are respectively taken as term vectors of the term w, cp and cn; and the above operations are repeated till the term vectors of all terms in the term set are obtained.

### Claim 16 (depends on 13)

The method of claim 13, wherein the term vector calculator is further configured to decompose a term-term matrix, at least partially map a term to a potential semantic space based on an Explicit Semantic Analysis (ESA) to obtain a vector of the term in the potential semantic space and take the vector as a term vector of the term.

### Claim 17 (depends on 16)

The method of claim 16, wherein the term vector calculator calculates the term vector of each term in the term set in a following manner:
⁢
M
≈
P
T
⁢
W
(
3
)
min
P
,
W
⁢
∑
i
=
1
n
⁢
∑
j
=
1
m
⁢
(
M
i
,
j
-
P
i
T
⁢
W
j
)
2
+
α
⁢
∑
i
=
1
n
⁢
∑
f
=
1
n
⁢
E
i
,
f
⁢

P
i
-
P
f

F
2
+
λ
⁡
(

P

F
2
+

W

F
2
)
(
4
)
where V={v1, v2, . . . , vm} represents the term set, vi and vj (1≤i, j≤m) each represents a single term, M∈Rm×m represents the term-term matrix, Mi,j is a Term Frequency-Inverse Document Frequency (tf-idf) weight of the term vj in the term vi, Ei,j is a similarity between the terms vi and Vj, ∥·∥2F is a Frobenius norm, α and λ are preset non-negative parameters, P∈Rk×m, W∈Rk×m, and R is a real number; and P is calculated under a condition in which the formula (4) is minimum according to the formula (3) and the formula (4), and a term vector Pi of the term vi and a term vector Pj of the term vj are obtained.

### Claim 18 (depends on 13)

The method of claim 13, wherein the condition that the similarity meets the threshold condition comprises: a cosine similarity between the term vector of the term in the response expression and a term vector of the corresponding expression of the semantic knowledge is greater than or equal to a predetermined threshold.

### Claim 19

A machine-aided dialog device, comprising:
an I/O interface for receiving an input from a user and providing an output to the user;
a processor;
a memory for storing executable instructions, wherein the executable instructions upon execution by the processor causing the device to perform any one of the method of claims.

### Claim 20

A medical condition inquiry apparatus, comprising:
a user interactive interface for providing a dialog with a patient, including inquiring a patient's health status and receiving a response from the patient;
a semantic analyzer, for extracting medical information of the patient based on the response from the patient, wherein the response from the patient is a first expression in a first expression category;
a medical knowledge database storing a second expression of a medical condition corresponding to the medical information, wherein the second expression is in a second expression category; and
a report generator, for generating a medical condition inquiry report having the medical information with the second expression.

### Claim at position 21

21-33. (canceled)