# US 12,001,518 B2 · Method for predicting matching degree between resume and post, and related device

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

- Number: US 12,001,518 B2
- Document ID: US12001518B2
- Status: Granted
- Kind: B2
- Application number: 18/083,211
- Filed: 2022-12-16
- Published: 2023-06-22
- Granted: 2024-06-04
- Earliest claimed: 2021-12-17
- Assignees: National University of Defense Technology
- Applicants: National University of Defense Technology
- Inventors: Honghui Chen; Taihua Shao; Chengyu Song; Miao Jiang; Mengru Wang; Xin Zhang; Fei Cai; Dengfeng Liu; Siyuan Wang
- Web page: https://patentgrep.com/patent/US12001518B2

## CPC

- G06F 18/241: relating to the classification model, e.g. parametric or non-parametric approaches
- G06F 17/16: Matrix or vector computation {, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization (matrix transposition G06F7/78)}
- G06N 3/044: Recurrent networks, e.g. Hopfield networks
- G06N 3/045: Combinations of networks
- G06N 3/047: Probabilistic or stochastic networks
- G06N 3/08: Learning methods
- G06N 3/084: Backpropagation, e.g. using gradient descent
- G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G06Q 10/1053: Employment or hiring
- G06N 20/10: using kernel methods, e.g. support vector machines [SVM]
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR

## Abstract

A method for predicting matching degree between a resume and a post, and a related device are provided in this disclosure. In the method for predicting the matching degree between the resume and the post, and the related device according to this disclosure, firstly the semi-structured keys and values in post information and resume information and their source are obtained. Then, the matching degree between the resume information and the post information is predicted by a prediction model including a cascaded pre-trained language model, a Transformer encoder and a single label classification model, based on the keys and values of a respective post information and resume information attribute, and corresponding source representations. Thus, by comprehensively searching internal interaction and external interaction of semi-structured multivariate attributes in person-post matching, the matching result is more accurate.

## Claims

### Claim 1

A method for predicting matching degree between resume information and post information, implemented via a processor, comprising:
receiving, via input interface, the resume information and the post information from storage;
obtaining, via a processor, a first key and a first value of a respective semi-structured post attribute in the post information and a second key and a second value of a respective semi-structured resume attribute in the resume information, the first key, the first value, the second key and the second value being all expressed in text data; and
predicting, via the processor, the matching degree between the resume information and the post information by a prediction model including a cascaded pre-trained language model, a Transformer encoder and a single label classification model, based on the first key and the first value of the respective post attribute, a first source representation corresponding to the post information, the second key and the second value of the respective resume attribute, and a second source representation corresponding to the resume information;
wherein predicting the matching degree between the resume information and the post information comprises:
for the first key and the first value of the respective post attribute, respectively encoding the first key and the first value into a semantic space through the pre-trained language model so as to obtain a first key embedding and a first value embedding, and fusing the first key embedding and the first value embedding so as to obtain a first fused embedding of the post attribute;
encoding the first source representation into the semantic space through the pre-trained language model so as to obtain a first source embedding;
for the second key and the second value of the respective resume attribute, respectively encoding the second key and the second value into the semantic space through the pre-trained language model so as to obtain a second key embedding and a second value embedding, and fusing the second key embedding and the second value embedding so as to obtain a second fused embedding of the resume attribute;
encoding, via the processor, the second source representation into the semantic space through the pre-trained language model so as to obtain a second source embedding; and
performing, via the processor, internal interaction of a first matrix including the first fusion embedding of the respective post attribute so as to obtain a first internal-interaction-attribute embedding matrix and performing internal interaction of a second matrix including the second fusion embedding of the respective resume attribute so as to obtain a second internal-interaction-attribute embedding matrix, with the Transformer encoder;
wherein the fusing and embedding operations are performed on all attributes in and to obtain matrix representations of and :
XJ=[j1a;j2a; . . . ;jma],
XR=[r1a;r2a; . . . ;rna],
where J is the post information, is the resume information, jia, rja∈din are fused and embedded expressions of an i-th attribute in and a j-th attribute in , respectively, i∈{1,2, . . . , m}, j∈{1,2, . . . , n}, Xj∈m×din is a first matrix and the matrix representation of , XR∈n×din is a second matrix and the matrix representation of , m and n are corresponding numbers of attributes in and , and din is a vector dimension after an attribute value and an attribute key are fused;
wherein a multi-head self-attention matrix representation of , MJ, and a multi-head self-attention matrix representation of , MR, are calculated according to:
M
J
h
=
softmax
(
(
X
J
·
Q
J
h
)
⁢
(
X
J
·
K
J
h
)
T
d
i
⁢
n
)
⁢
X
J
·
V
J
h
,
M
R
h
=
softmax
(
(
X
R
·
Q
R
h
)
⁢
(
X
R
·
K
R
h
)
T
d
i
⁢
n
)
⁢
X
R
·
V
R
h
,
M
1
=
concat
(
M
J
1
,
M
J
2
,
…
,
M
J
H
)
·
O
J
,
M
R
=
concat
(
M
R
1
,
M
R
2
,
…
,
M
R
H
)
·
O
R
,
where QJh, QRh∈din×dq, KJh, KRh∈din×dk, VJh, VRh∈din×dvk and OJ, OR∈H·dv×din are trainable network parameters; h∈{1,2, . . . , H}, H represents a number of headers of the Transformer encoder, dq=dk=dv=din/H, MJh and MRh are self-attention matrix representations of and at an h-th head;
MJ and MR are input into a feedforward layer of the Transformer encoder to obtain internal interaction representations of and as follows:
M
J
=
feedforward
(
M
J
)
,
=
[
j
1
a
′
;
j
2
a
′
;
…
;
j
m
a
′
]
,
M
R
=
feedforward
(
M
R
)
,
=
[
r
1
a
′
;
r
2
a
′
;
…
;
r
n
a
′
]
,
where jia′, rja′∈din are internal-interaction-attribute embeddings of the i-th attribute in and the j-th attribute in , i∈{1,2, . . . , m} and j∈{1,2, . . . , n}, respectively; and
wherein the prediction model is trained by minimizing a binary cross entropy loss according to:
=−1/NΣi=1N(i log(′i)+(1−i)log(1−′i)),
where i is a true matching degree of an i-th training instance; ′i is a predicted matching degree of the i-th training instance generated by the model; N is a total number of training instances,
thus predicting an accurate matching degree between the resume information and the post information to obtain rich representations of the resume information and the post information; and
presenting, via output interface, rich representations of the resume information and the post information, which comprising the resume information, the post information and the accurate matching degree, wherein when the accurate matching degree is 1 indicates match, and when the accurate matching degree is 0 indicates no match.

### Claim 2 (depends on 1)

The method according to claim 1, wherein predicting the matching degree between the resume information and the post information further comprises:
fusing each element in the first internal-interaction-attribute embedding matrix respectively with the first source embedding so as to obtain a first updated embedding matrix;
fusing each element in the second internal-interaction-attribute embedding matrix respectively with the second source embedding so as to obtain a second updated embedding matrix;
performing external interaction between the first updated embedding matrix and the second updated embedding matrix with the Transformer encoder so as to obtain a first external-interaction-attribute embedding matrix and a second external-interaction-attribute embedding matrix; and
predicting the matching degree between the resume information and the post information through the single label classification model according to the first external-interaction-attribute embedding matrix and the second external-interaction-attribute embedding matrix.

### Claim 3 (depends on 2)

The method according to claim 2, wherein fusing the first key embedding and the first value embedding, fusing the second key embedding and the second value embedding, fusing each element in the first internal-interaction-attribute embedding matrix respectively with the first source embedding, and fusing each element in the second internal-interaction-attribute embedding matrix respectively with the second source embedding all comprises an additive fusing operation.

### Claim 4 (depends on 2)

The method according to claim 2, wherein fusing the first key embedding and the first value embedding, fusing the second key embedding and the second value embedding, fusing each element in the first internal-interaction-attribute embedding matrix respectively with the first source embedding, and fusing each element in the second internal-interaction-attribute embedding matrix respectively with the second source embedding all comprises a concatenate fusing operation.

### Claim 5 (depends on 2)

The method according to claim 2, wherein performing the internal interaction of the first matrix so as to obtain the first internal-interaction-attribute embedding matrix comprises: obtaining a first multi-head self-attention matrix using a multi-head self-attention mechanism in the Transformer encoder according to the first matrix; and obtaining the first internal-interaction-attribute embedding matrix according to the first multi-head self-attention matrix through a first fully connected feedforward layer in the Transformer encoder; and
performing the internal interaction of the second matrix so as to obtain the second internal-interaction-attribute embedding matrix comprises: obtaining a second multi-head self-attention matrix using a multi-head self-attention mechanism in the Transformer encoder according to the second matrix; and obtaining the second internal-interaction-attribute embedding matrix according to the second multi-head self-attention matrix through a second fully connected feedforward layer in the Transformer encoder.

### Claim 6 (depends on 5)

The method according to claim 5, wherein performing the external interaction so as to obtain the first external-interaction-attribute embedding matrix and the second external-interaction-attribute embedding matrix comprises:
concatenating the first updated embedding matrix and the second updated embedding matrix to obtain a joint matrix;
obtaining a multi-head self-attention joint matrix according to the joint matrix by using the multi-head self-attention mechanism in the Transformer encoder; and
obtaining the first external-interaction-attribute embedding matrix and the second external-interaction-attribute embedding matrix according to the multi-head self-attention joint matrix through a third fully connected feedforward layer in the Transformer encoder.

### Claim 7 (depends on 2)

The method according to claim 2, wherein predicting the matching degree according to the first external-interaction-attribute embedding matrix and the second external-interaction-attribute embedding matrix comprises:
respectively pooling the first external-interaction-attribute embedding matrix and the second external-interaction-attribute embedding matrix through a first pooling layer and a second pooling layer in the single label classification model so as to obtain a post vector and a resume vector;
aggregating the post vector and the resume vector to obtain an aggregated vector;
performing semantic fusion on the aggregated vector through a multilayer perceptron in the single label classification model, and obtaining a matching score by using a single neuron according to the aggregated vector subjected to the semantic fusion; and
obtaining the matching degree of the resume information and the post information using an activation function according to the matching score.

### Claim 8 (depends on 7)

The method according to claim 7, wherein the pooling is performed by using an adaptive average pooling function provided by a machine learning framework PyTorch.

### Claim 9 (depends on 1)

An electronic device, comprising a memory, the processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method in claim 1 when executing the computer program.

### Claim 10 (depends on 9)

The electronic device of claim 9, wherein predicting the matching degree between the resume information and the post information further comprises:
fusing each element in the first internal-interaction-attribute embedding matrix respectively with the first source embedding so as to obtain a first updated embedding matrix;
fusing each element in the second internal-interaction-attribute embedding matrix respectively with the second source embedding so as to obtain a second updated embedding matrix;
performing external interaction between the first updated embedding matrix and the second updated embedding matrix with the Transformer encoder so as to obtain a first external-interaction-attribute embedding matrix and a second external-interaction-attribute embedding matrix; and
predicting the matching degree between the resume information and the post information through the single label classification model according to the first external-interaction-attribute embedding matrix and the second external-interaction-attribute embedding matrix.

### Claim 11 (depends on 10)

The electronic device of claim 10, wherein fusing the first key embedding and the first value embedding, fusing the second key embedding and the second value embedding, fusing each element in the first internal-interaction-attribute embedding matrix respectively with the first source embedding, and fusing each element in the second internal-interaction-attribute embedding matrix respectively with the second source embedding all comprises an additive fusing operation.

### Claim 12 (depends on 10)

The electronic device of claim 10, wherein fusing the first key embedding and the first value embedding, fusing the second key embedding and the second value embedding, fusing each element in the first internal-interaction-attribute embedding matrix respectively with the first source embedding, and fusing each element in the second internal-interaction-attribute embedding matrix respectively with the second source embedding all comprises a concatenate fusing operation.

### Claim 13 (depends on 10)

The electronic device of claim 10, wherein performing the internal interaction of the first matrix so as to obtain the first internal-interaction-attribute embedding matrix comprises: obtaining a first multi-head self-attention matrix using a multi-head self-attention mechanism in the Transformer encoder according to the first matrix; and obtaining the first internal-interaction-attribute embedding matrix according to the first multi-head self-attention matrix through a first fully connected feedforward layer in the Transformer encoder; and
performing the internal interaction of the second matrix so as to obtain the second internal-interaction-attribute embedding matrix comprises: obtaining a second multi-head self-attention matrix using a multi-head self-attention mechanism in the Transformer encoder according to the second matrix; and obtaining the second internal-interaction-attribute embedding matrix according to the second multi-head self-attention matrix through a second fully connected feedforward layer in the Transformer encoder.

### Claim 14 (depends on 13)

The electronic device of claim 13, wherein performing the external interaction so as to obtain the first external-interaction-attribute embedding matrix and the second external-interaction-attribute embedding matrix comprises:
concatenating the first updated embedding matrix and the second updated embedding matrix to obtain a joint matrix;
obtaining a multi-head self-attention joint matrix according to the joint matrix by using the multi-head self-attention mechanism in the Transformer encoder; and
obtaining the first external-interaction-attribute embedding matrix and the second external-interaction-attribute embedding matrix according to the multi-head self-attention joint matrix through a third fully connected feedforward layer in the Transformer encoder.

### Claim 15 (depends on 10)

The electronic device of claim 10, wherein predicting the matching degree according to the first external-interaction-attribute embedding matrix and the second external-interaction-attribute embedding matrix comprises:
respectively pooling the first external-interaction-attribute embedding matrix and the second external-interaction-attribute embedding matrix through a first pooling layer and a second pooling layer in the single label classification model so as to obtain a post vector and a resume vector;
aggregating the post vector and the resume vector to obtain an aggregated vector;
performing semantic fusion on the aggregated vector through a multilayer perceptron in the single label classification model, and obtaining a matching score by using a single neuron according to the aggregated vector subjected to the semantic fusion; and
obtaining the matching degree of the resume information and the post information using an activation function according to the matching score.

### Claim 16 (depends on 15)

The electronic device of claim 15, wherein the pooling is performed by using an adaptive average pooling function provided by a machine learning framework PyTorch.