Note: Descriptions are shown in the official language in which they were submitted.
METHOD OF AND SYSTEM FOR APPRAISING RISK
BACKGROUND OF THE INVENTION
Technical Field
[0001] The present invention relates to the field of risk management, and more
particularly to
a method of and a system for appraising risk.
Description of Related Art
[0002] With the development of the intemet technology, massive amount of
public sentiment
text information is generated everyday over the internet, and a great deal of
valuable
information relevant to risk states of enterprises is included therein. It is
a direction of
application extremely worthy of research as how to gather these pieces of
information
together, and to analyze and generate index data relevant to enterprise risks.
[0003] In the state of the art, public sentiment systems for using in
enterprise risk surveillance
are partly directed to collection over the whole intemet with respect to total
enterprises, but
colossal resources are occupied, the cost is not only high, but the
utilization rate of public
sentiment information is not much high; the other public sentiment systems
employ an
enterprise follow-up list to carry out total name collection, but this easily
leads to
overconcentration of public sentiment information of the enterprise early
warning systems,
and the actual utilization effect is far from being ideal. At the same time,
it is further
problematic in the state of the art that public sentiment texts are
inaccurately associated with
enterprise entities, or that the associating method is so strict that the
amount of public
1
Date Regue/Date Received 2022-07-27
sentiment collection is deficient and coverage is insufficient, or that the
associating method
is so over ambiguous that public sentiments as collected are inaccurately
associated, and that
great quantities of public sentiment noises are generated.
[0004] Accordingly, there is an urgent need to propose a method of appraising
risk capable of
enhancing public sentiment collection efficiency and achieving quantitative
appraisal of risks,
so as to solve the aforementioned technical problems pending in the state of
the art.
SUMMARY OF THE INVENTION
[0005] In order to deal with deficiencies prevalent in prior-art technology, a
main objective of
the present invention is to provide a method of and a system for recognizing
risk, so as to
solve the aforementioned technical problems pending in the state of the art.
[0006] To achieve the above objective, according to the first aspect, the
present invention
provides a method of appraising risk, and the method comprises:
[0007] obtaining, from a preset public sentiment text library, a target public
sentiment text to
which a target object corresponds and whose generating time is within a preset
time range,
according to a keyword library to which the target object corresponds;
[0008] predicting an event category to which each target public sentiment text
corresponds by
employing a trained classification model;
[0009] generating a public sentiment risk score to which the target public
sentiment text
corresponds according to the event category to which the target public
sentiment text
corresponds as obtained by prediction and according to preset negative hot
words included
in the target public sentiment text; and
2
Date Regue/Date Received 2022-07-27
[0010] determining whether to send out first risk early warning to which the
target object
corresponds according to the public sentiment risk score and a preset first
risk early warning
threshold.
[0011] In some embodiments, the method further comprises:
[0012] generating a first risk total score to which the target object
corresponds and determining
a scoring time to which the first risk total score corresponds according to
all public sentiment
risk scores to which the target object corresponds;
[0013] generating a second risk total score to which the target object
corresponds according to
the first risk total score whose scoring time is within a preset historical
time period and a time
decay item determined according to the scoring time; and
[0014] determining whether to send out second risk early warning to which the
target object
corresponds according to the second risk total score and a preset second risk
early warning
threshold.
[0015] In some embodiments, the target object includes enterprises, and, prior
to the step of
obtaining, from a preset public sentiment text library, a target public
sentiment text to which
a target object corresponds and whose generating time is within a preset time
range, according
to a keyword library to which the target object corresponds, the method
further comprises:
[0016] enquiring any associated enterprise to which the target object
corresponds from a preset
enterprise association map and obtaining enterprise information to which the
associated
enterprise and the target object respectively correspond;
[0017] processing the enterprise information to which the associated
enterprise and the target
object respectively correspond, and generating a keyword to which the target
object and the
associated enterprise correspond; and
[0018] determining the keyword library to which the target object corresponds
according to the
3
Date Regue/Date Received 2022-07-27
keyword to which the target object and the associated enterprise correspond.
[0019] In some embodiments, prior to the step of enquiring any associated
enterprise to which
the target object corresponds from a preset enterprise association map and
obtaining
enterprise information to which the associated enterprise and the target
object respectively
correspond, the method further comprises:
[0020] generating a node to which each enterprise corresponds according to an
enterprise
business database, wherein the node includes a node attribute and an
associated attribute, the
node attribute includes the enterprise information to which the enterprise
corresponds, and
the associated attribute is employed for enquiring any associated enterprise
to which the
enterprise corresponds; and
[0021] establishing the enterprise association map according to the associated
attribute to which
each node corresponds.
[0022] In some embodiments, the enterprise information includes an enterprise
full name, the
keyword includes an enterprise abbreviation, and the step of processing the
enterprise
information to which the associated enterprise and the target object
respectively correspond,
and generating a keyword to which the target object and the associated
enterprise correspond
includes:
[0023] recognizing abbreviation fields included in the enterprise full name by
employing a
preset regular expression; and
[0024] filtering the abbreviation fields according to a preset filtering rule,
and generating a
corresponding enterprise abbreviation.
[0025] In some embodiments, the step of obtaining, from a preset public
sentiment text library,
a target public sentiment text to which a target object corresponds and whose
generating time
4
Date Regue/Date Received 2022-07-27
is within a preset time range, according to a keyword library to which the
target object
corresponds includes:
[0026] obtaining, from the preset public sentiment text library, a public
sentiment text that
contains one or more keyword(s) in the keyword library;
[0027] predicting an emotion category to which each public sentiment text
corresponds by
employing a preset emotion analyzing model; and
[0028] determining any public sentiment text whose emotion category is
negative from the
public sentiment texts to serve as the target public sentiment text.
[0029] In some embodiments, the step of generating a public sentiment risk
score to which the
target public sentiment text corresponds according to the event category to
which the target
public sentiment text corresponds as obtained by prediction and according to
preset negative
hot words included in the target public sentiment text includes:
[0030] determining an influence factor to which the target public sentiment
text corresponds
according to a preset corresponding relation between influence factors and
event categories;
[0031] determining a hot word risk value to which the target public sentiment
text corresponds
according to the number of preset negative hot words included in the target
public sentiment
text; and
[0032] determining the public sentiment risk score to which the target public
sentiment text
corresponds according to the influence factor to which the target public
sentiment text
corresponds and according to the hot word risk value.
[0033] In some embodiments, the method further comprises training the
classification model,
and a process of training the classification model includes:
[0034] training the classification model according to a preset training corpus
set;
[0035] verifying whether the classification model satisfies a preset condition
according to a
Date Regue/Date Received 2022-07-27
preset testing corpus set; and
[0036] generating the trained classification model when the classification
model satisfies a
preset training condition.
[0037] According to the second aspect, the present application provides a
system for appraising
risk, and the system comprises:
[0038] an obtaining module, for obtaining, from a preset public sentiment text
library, a target
public sentiment text to which a target object corresponds and whose
generating time is
within a preset time range, according to a keyword library to which the target
object
corresponds;
[0039] a predicting module, for predicting an event category to which each
target public
sentiment text corresponds by employing a trained classification model;
[0040] a generating module, for generating a public sentiment risk score to
which the target
public sentiment text corresponds according to the event category to which the
target public
sentiment text corresponds as obtained by prediction and according to preset
negative hot
words included in the target public sentiment text; and
[0041] a judging module, for determining whether to send out first risk early
warning to which
the target object corresponds according to the public sentiment risk score and
a preset first
risk early warning threshold.
[0042] According to the third aspect, the present application provides an
electronic equipment
that comprises:
[0043] one or more processor(s); and
[0044] a memory, associated with the one or more processor(s) and used for
storing a program
instruction. The program instruction performs the following operations when it
is read and
executed by the one or more processor(s):
6
Date Regue/Date Received 2022-07-27
[0045] obtaining, from a preset public sentiment text library, a target public
sentiment text to
which a target object corresponds and whose generating time is within a preset
time range,
according to a keyword library to which the target object corresponds;
[0046] predicting an event category to which each target public sentiment text
corresponds by
employing a trained classification model;
[0047] generating a public sentiment risk score to which the target public
sentiment text
corresponds according to the event category to which the target public
sentiment text
corresponds as obtained by prediction and according to preset negative hot
words included
in the target public sentiment text; and
[0048] determining whether to send out first risk early warning to which the
target object
corresponds according to the public sentiment risk score and a preset first
risk early warning
threshold.
[0049] The present application achieves the following advantageous effects.
[0050] The present application provides a method of appraising risk,
comprising obtaining,
from a preset public sentiment text library, a target public sentiment text to
which a target
object corresponds and whose generating time is within a preset time range,
according to a
keyword library to which the target object corresponds; predicting an event
category to which
each target public sentiment text corresponds by employing a trained
classification model;
generating a public sentiment risk score to which the target public sentiment
text corresponds
according to the event category to which the target public sentiment text
corresponds as
obtained by prediction and according to preset negative hot words included in
the target
public sentiment text; and determining whether to send out first risk early
warning to which
the target object corresponds according to the public sentiment risk score and
a preset first
risk early warning threshold. The public sentiment text to which a target
object whose risk is
required to be appraised corresponds can be extracted from the public
sentiment text library
7
Date Regue/Date Received 2022-07-27
through a preset keyword library to which each target object corresponds, and
risk appraisal
of the target object is realized on the basis of the public sentiment text to
which the target
object corresponds through recognition of the event category and negative hot
words of the
public sentiment text, whereby efficiency in collecting public sentiments
associated with
texts is enhanced, quantitative appraisal of public sentiment risks is
realized, and accuracy in
recognizing public sentiment risks is ensured.
[0051] Moreover, the present application further proposes enquiring any
associated enterprise
to which the target object corresponds from a preset enterprise association
map and obtaining
enterprise information to which the associated enterprise and the target
object respectively
correspond; processing the enterprise information to which the associated
enterprise and the
target object respectively correspond, and generating keywords to which the
target object and
the associated enterprise correspond; and determining the keyword library to
which the target
object corresponds according to the keywords to which the target object and
the associated
enterprise correspond, whereby the association between the obtained public
sentiment text
and the target object is enhanced, and collection coverage rate of public
sentiment texts of
target objects is ensured.
[0052] Not all products according to the present application are required to
necessarily possess
all the aforementioned effects.
BRIEF DESCRIPTION OF THE DRAWINGS
[0053] To more clearly describe the technical solutions in the embodiments of
the present
application, drawings required to illustrate the embodiments will be briefly
introduced below.
Apparently, the drawings introduced below are merely directed to some
embodiments of the
present application, while persons ordinarily skilled in the art may further
acquire other
8
Date Regue/Date Received 2022-07-27
drawings on the basis of these drawings without spending creative effort in
the process.
[0054] Fig. 1 is a flowchart illustrating public sentiment system surveillance
provided by an
embodiment of the present application;
[0055] Fig. 2 is a flowchart illustrating the method provided by an embodiment
of the present
application;
[0056] Fig. 3 is a view illustrating the structure of the system provided by
an embodiment of
the present application; and
[0057] Fig. 4 is a view illustrating the structure of the electronic equipment
provided by an
embodiment of the present application.
DETAILED DESCRIPTION OF THE INVENTION
[0058] To make more lucid and clear the objectives, technical solutions and
advantages of the
present application, the technical solutions in the embodiments of the present
application will
be clearly and comprehensively described below with reference to the
accompanying
drawings in the embodiments of the present application. Apparently, the
embodiments as
described are merely partial, rather than the entire, embodiments of the
present application.
Any other embodiments makeable by persons ordinarily skilled in the art on the
basis of the
embodiments in the present application without creative effort shall all fall
within the
protection scope of the present application.
[0059] As noted in the Description of Related Art, parts of surveillance
systems in the prior-art
9
Date Regue/Date Received 2022-07-27
technology perform total collection and analysis on massive amount of public
sentiment texts,
which occupies large quantities of computational resources; another part
thereof carry out
total name collection based on the enterprise follow-up list, thereby the
collection effect is
rendered not ideal.
[0060] To solve the above technical problems, the present application provides
a method of and
a system for appraising risk for using in such target surveillance objects as
enterprises, the
public sentiment text to which a target object whose risk is required to be
appraised
corresponds can be extracted from the public sentiment text library through a
preset keyword
library to which each target object corresponds, and risk appraisal of the
target object is
realized on the basis of the public sentiment text to which the target object
corresponds
through recognition of the event category and negative hot words of the public
sentiment text,
whereby efficiency in collecting public sentiments associated with texts is
enhanced,
quantitative appraisal of public sentiment risks is realized, and accuracy in
recognizing public
sentiment risks is ensured.
[0061] Embodiment 1
[0062] To implement the method of appraising risk disclosed by the present
application, an
embodiment of the present application provides a public sentiment risk
surveillance system
that comprises a public sentiment collecting module, a public sentiment
associating module,
a public sentiment analyzing module, and a public sentiment early warning
module.
Specifically, as shown in Fig. 1, the process of using the public sentiment
risk surveillance
system disclosed by this embodiment to perform public sentiment collection and
surveillance
early warning includes the following.
[0063] S100 ¨ determining a to-be-monitored follow-up set according to a
preset enterprise
Date Regue/Date Received 2022-07-27
association map.
[0064] The enterprise association map is established in advance by the public
sentiment
collecting module according to an enterprise business database. The enterprise
business
database stores enterprise information and association information of
enterprises. The
association information includes investment relations of an enterprise,
including upstream
investing enterprises and downstream invested enterprises of the enterprise.
The enterprise
information includes such information associated with an enterprise as an
enterprise full
name, an enterprise unified information code, an enterprise abbreviation, an
enterprise
corporate name, projects or products, an enterprise legal person, core senior
executives, a
trademark, a telephone number, a website domain name, and an enterprise
mailbox, etc. The
enterprise business database can provide an output interface to the outside,
and an invoker
can send a query statement containing the enterprise full name and/or the
enterprise unified
credit code to the enterprise business database via the output interface and
receive association
information and enterprise information returned from the output interface.
[0065] An enterprise association map containing all enterprises can be
established according to
the enterprise business database. The enterprise association map includes
plural nodes, each
of which corresponds to one enterprise. Each node includes corresponding node
attribute and
associated attribute, of which the node attribute includes the enterprise
information of a given
enterprise, and the associated attribute includes external investment
relations of the enterprise.
Nodes can be associated according to the associated attribute of each node to
generate the
enterprise association map that contains all enterprises to which the nodes
correspons.
[0066] On receiving a public sentiment risk surveillance request of a new
target enterprise, the
public sentiment risk surveillance system can obtain from the enterprise
association map the
target enterprise and the node attribute of any associated enterprise of the
target enterprise as
11
Date Regue/Date Received 2022-07-27
determined according to the associated attribute, and establish a follow-up
set to which the
target enterprise corresponds. Taking the target enterprise is enterprise A as
an example, the
associated enterprises of enterprise A are enterprise Al and enterprise A2,
then the follow-up
set to which the target enterprise corresponds is:
[0067] {A: (attribute 1, attribute 2, .................................... );
Al: (attribute 1, attribute 2, ); enterprise A2:
(attribute 1, attribute 2, .. )1
[0068] Any directly associated enterprise in which the target enterprise
directly invests and any
indirectly associated enterprise in which the target enterprise indirectly
invests through the
directly associated enterprise can be determined according to the associated
attribute, and it
is possible to designate any associated enterprise to which the target
enterprise corresponds
as a directly associated enterprise, an indirectly associated enterprise, or
as directly associated
enterprise and indirectly associated enterprise according to business
requirements, to which
no restriction is made in the present application.
[0069] The to-be-monitored follow-up set can be determined according to a
follow-up set to
which the new target enterprise corresponds and the follow-up set to which the
target
enterprise originally existent in the public sentiment collecting module
corresponds.
[0070] S200 ¨obtaining a target public sentiment text to which each target
enterprise
corresponds by employing the public sentiment associating module according to
the to-be-
monitored follow-up set and the public sentiment text library.
[0071] The public sentiment text library stores a massive amount of public
sentiment texts
collected from such channels as the internet.
12
Date Regue/Date Received 2022-07-27
[0072] The process of obtaining the target public sentiment text can
specifically include:
[0073] S210 ¨ determining a keyword library to which each target enterprise
corresponds
according to the to-be-monitored follow-up set.
[0074] All enterprise information to which the target corresponds can be
obtained according to
the to-be-monitored follow-up set, the target enterprise, and node attributes
of associated
enterprises of the target enterprise. A keyword library to which each target
enterprise
corresponds can be generated by further analyzing and processing the
enterprise information.
[0075] Specifically, keywords included in the keyword library can include
original enterprise
information to which the target enterprise corresponds, and such information
as enterprise
abbreviation and enterprise domain name generated after analyzing and
processing.
[0076] Specifically, when the original enterprise information does not contain
any enterprise
abbreviation, the process of generating enterprise abbreviation includes the
following.
[0077] S211 ¨ recognizing abbreviation fields contained in an enterprise full
name according
to a preset regular expression.
[0078] The regular expression can be summarized and obtained in advance
according to a
naming rule of enterprise full names. For instance, as the common enterprise
naming mode
is placename + corporate name + nature of the enterprise, it is then possible
to eliminate such
fields irrelevant to abbreviation as the placename and the nature of the
enterprise that are
contained in the enterprise full name through a regular expression, and the
fields that remain
are precisely abbreviation fields contained in the enterprise full name.
13
Date Regue/Date Received 2022-07-27
[0079] S212 ¨ filtering any preset keyword to be filtered contained in the
abbreviation fields
according to a preset keyword filtering dictionary, and generating a
corresponding enterprise
abbreviation.
[0080] The keyword filtering dictionary stores preset keywords to be filtered,
and is employed
for recognizing such useless keywords to be filtered as "new energy" that is
common in
enterprise full names. After the keyword to be filtered contained in the
abbreviation fields
has been filtered, the fields that are retained in the enterprise full name
are precisely an
enterprise abbreviation of the corresponding target enterprise.
[0081] When the original enterprise information does not contain any website
domain name of
the enterprise, the process of generating website domain name includes:
[0082] S221 ¨ recognizing a corresponding website domain name according to
enterprise
website URLs contained in the enterprise information.
[0083] A keyword library to which the target enterprise corresponds can be
generated according
to keywords to which the target enterprise and associated enterprises of the
target enterprise
correspond.
[0084] Taking the target enterprise is an X Di Food Co., Ltd. as an example,
its associated
enterprises include Y Fresh Milk Co., Ltd. and Z Green Rice Development Co.,
Ltd., and the
keyword library to which the X Di Food Co., Ltd. corresponds is as shown in
Table 1
according to enterprise information to which the three enterprises
respectively correspond as
obtained from the enterprise association map.
14
Date Regue/Date Received 2022-07-27
[0085] Table 1
Enterprise Monitored Surveillance Keywords
X Di Food Co., Ltd. X Di Food, X Di, ZHANG San,
xdifood.com,
Y Fresh Milk Co., Ltd. Y Fresh Milk, Y Fresh, LI Si,
yfreshmilk.com,
Z Green Rice Development Co., Ltd. Z Green Rice, Z Green,
zgreenrice.com,
[0086] Corresponding key-value pairs can be generated to serve as index
according to the target
enterprise and the keyword library to which the target enterprise corresponds.
A
corresponding index set can be generated according to indexes to which all
target enterprises
correspond, so as to facilitate quick query from the key-value pairs to obtain
the
corresponding keyword library according to the target enterprise during
subsequent search,
wherein the keys in the key-value pairs can store such information for
recognizing target
enterprises as enterprise full names and enterprise unified information codes
of the target
enterprises, and the values can store corresponding keyword libraries.
[0087] S220 ¨ enquiring and obtaining, from a preset public sentiment text
library, a target
public sentiment text to which a target enterprise corresponds and whose
generating time is
within a preset time range, according to the keyword library.
[0088] The surveillance system can extract corresponding public sentiment
texts from such
channels as the internet periodically according to a preset time period and
store the same in
the preset public sentiment text library. During appraisal, it is possible to
enquire from the
index set and obtain a keyword library to which the target enterprise
corresponds, to
recognize the public sentiment text that contains one or more keyword(s) in
the keyword
library, and to determine the public sentiment text as the public sentiment
text to which the
target enterprise corresponds.
Date Regue/Date Received 2022-07-27
[0089] In order to facilitate query and to reduce storage space occupied by
text storage, the
public sentiment text can be preprocessed before it is stored in the database
to generate a
structured public sentiment text.
[0090] The title of the public sentiment text and any entity name contained in
the text proper
can be recognized through a preset recognizing rule, and the entity name is
matched with
keywords in the keyword library of the target enterprise in the index library,
if there is a
corresponding keyword, it can then be determined that the public sentiment
text is the public
sentiment text to which the target enterprise corresponds.
[0091] Table 2 illustrates an exemplary structured public sentiment text.
Entity names contained
therein can be extracted through the preset recognizing rule: Y DI Milk,
Shenzhen Stock
Exchange, Green Eco-Rice. Through the matching results in the index library,
the target
enterprises to which the public sentiment text corresponds are: Y Fresh Milk
Co., Ltd. and Z
Green Rice Development Co., Ltd., that is to say, the public sentiment text is
a public
sentiment text to which Y Fresh Milk Co., Ltd. and Z Green Rice Development
Co., Ltd.
correspond.
16
Date Regue/Date Received 2022-07-27
[0092] Table 2
Public Sentiment Field Field Contents
Title Whether financial liquidity is depleted,
Y Di Milk with a net profit loss of 0.12
billion was consecutively thrice asked
by Shenzhen Stock Exchange
Time 2021-05-17 12:35:25
Source sina.com
Text Over the recent days, as shown by the
data of an annual report, Y Di Milk got an
operating income of 0.5 billion in the
year 2020, but suffered a net profit loss
of 0.12 billion. Faced with the problem of
abrupt decreases in the operating income
and the profit, Y Di Milk again received
inquiry letters concerning three big
problems from Shenzhen Stock
Exchange in the annual report 2020.
What is the reason counting for the cliff
descent of achievements? As Y Di Milk
considers, there will be no problem for its
operation in the future 12 months.
Shenzhen Stock Exchange demanded
that Y Di Milk to explain the specific
reason for the abrupt decreases in
business income, net profit, net profit
after deduction of non-recurring gains
and losses, and net cash flow within the
duration of the report in conjunction with
such factors as income, cost, period
expense, gross profit, receipt and
payment, and capital chain.
Its brother company, Green Eco-Rice,
has also been recently complained by
consumers as being suspected of fraud...
...
[0093] S300 ¨ the public sentiment analyzing module predicting an emotion
category to which
17
Date Regue/Date Received 2022-07-27
the public sentiment text corresponding to each target enterprise corresponds
by employing
a trained emotion analyzing model, and to determine any public sentiment text
whose
emotion category is negative as the target public sentiment text to which the
target enterprise
corresponds.
[0094] The text and the title contained in the public sentiment text can be
input in the emotion
analyzing model, and the emotion analyzing model can predict to obtain its
corresponding
emotion category. Specifically, emotion categories can at least include three
categories as
positive, neutral, and negative.
[0095] The emotion analyzing model can be trained according to public
sentiment texts and
artificially marked emotion categories, when a prediction precision rate of
the model exceeds
a preset threshold, it can be determined that the model is a trained emotion
analyzing model
and is used to predict emotion categories of target public sentiment texts.
The text can be an
originally existent public sentiment text or any such random text information
as public corpus
collected over the internet.
[0096] The aforementioned emotion analyzing model can be any of such models
trained to have
emotion category predicting capabilities as a trained fasttext (shallow
network) model, a
CBOW (Continuous Bag of Words Model, continuous vocabulary), a Skip-gram
(Continuous
Skip-gram Model) model, a bidirectional long-term and short-term memory
network
(BiLSTM), an ALBERT (A Lite BERT) model, to which no restriction is made in
this context.
[0097] New training corpora can be periodically collected and artificially
marked, and the
emotion analyzing model is thereafter updated and trained again, so as to
ensure precision of
model prediction.
18
Date Regue/Date Received 2022-07-27
[0098] S310 ¨ the public sentiment analyzing model predicting by employing a
trained
classification mode a corresponding event category according to the target
public sentiment
text whose emotion category is negative.
[0099] Specifically, the event category includes, but is not limited to, such
randomly preset
negative events as bankruptcy and termination of business, mortgage and pawn,
deficit,
default and negative news, illegal fund raising, infringement and counterfeit,
contract dispute,
violation of regulations and laws, product statuses, and personnel changes,
etc.
[0100] Before the corresponding event category is predicted, the process of
training the
classification model includes the following.
[0101] S311 ¨ obtaining a training corpus set.
[0102] The training corpus set includes public sentiment texts in which
corresponding event
categories are artificially marked.
[0103] S312 ¨training the classification model by employing the training
corpus set.
[0104] S313 ¨verifying whether the prediction precision rate of the
classification model
satisfies a preset condition by employing the training corpus set.
[0105] A corresponding precision rate threshold can be preset. When the
prediction precision
rate of the classification model exceeds the precision rate threshold, it can
be determined that
the prediction precision rate of the classification model satisfies the preset
condition and
19
Date Regue/Date Received 2022-07-27
determined that the classification model is a trained classification model.
[0106] The classification model can be any model that is trained to have text
classifying
capability, to which no restriction is made in the present application.
[0107] S320 ¨ the public sentiment analyzing module recognizing any preset
negative hot word
contained in the target public sentiment text according to a preset negative
hot word library.
[0108] The negative hot word library stores preset negative hot words, and it
can recognize
whether the target public sentiment text contains any preset negative hot word
according to
the negative hot word library.
[0109] S400 ¨ the public sentiment early warning module generating a public
sentiment risk
score to which the public sentiment text corresponds within a preset time
range according to
the event category to which the target public sentiment text corresponds as
obtained by
prediction, an influence factor to which the preset event category
corresponds, and preset
negative hot words contained in the preset public sentiment text.
[0110] Specifically, the process of generating the public sentiment risk score
includes the
following.
[0111] S410 ¨ determining the influence factor to which the target public
sentiment text
corresponds according to preset corresponding relations between influence
factors and event
categories, and the event category to which the target public sentiment text
corresponds as
obtained by prediction.
Date Regue/Date Received 2022-07-27
[0112] Influence factors to which each event category corresponds can be
preset in the
surveillance system, to facilitate obtainment as required.
[0113] S411 - determining a hot word risk value to which the target public
sentiment text
corresponds according to the number of preset negative hot words included in
the target
public sentiment text.
[0114] Specifically, corresponding relations between the numbers and the hot
word risk values
can be preset, so as to determine the corresponding hot word risk value when
the number of
preset negative hot words as included is recognized.
[0115] S412¨ generating a public sentiment risk score to which the target
public sentiment text
corresponds according to the corresponding influence factor and hot word risk
value.
[0116] S413 ¨ generating a first risk total score according to the public
sentiment risk score,
and judging whether to send out first risk early warning according to the
first risk total score
and a first preset risk early warning threshold.
[0117] After the public sentiment risk score has been generated, it is further
possible for the
public sentiment early warning module to generate a first risk total score
corresponding to
the target enterprise and to determine the corresponding scoring time
according to public
sentiment risk scores of all target public sentiment texts whose generating
times are within
the preset time range. When the first risk total score exceeds the first risk
early warning
threshold, the first risk early warning can be sent out.
[0118] The first risk total score Ro can be expressed as: first risk total
score = influence factor
21
Date Regue/Date Received 2022-07-27
(1 + hot word risk value), namely:
[0119] Ro = No (Xj(1 + Yi))
[0120] where X, represents the influence factor, Y, represents the hot word
risk value, and No
represents the total number of target public sentiment texts whose emotion
categories are
negative and whose generating times are within the preset time range and to
which the target
enterprise corresponds.
[0121] Specifically, taking the target enterprise is Y Fresh Milk Co. Ltd.,
the generating time to
which the currently appraised target public sentiment text corresponds is 24
hours before the
start of the appraisal, and the first risk early warning threshold Ho=5 as an
example, there are
three collected target public sentiment texts whose emotion categories are
negative, preset
influence factors of the corresponding event categories are 3, 2, 4 and none
of which contains
any preset negative hot word, then the first risk total score Ro to which the
target enterprise
corresponds is:
[0122] R0= No(X(1+ Yi)) = 3 * (1 + 0) + 2 * (1 + 0) + 4 * (1 + 0) = 9
[0123] Since the first risk total score Ro is greater than the first risk
early warning threshold
Ho, the surveillance system will send out the first risk early warning.
[0124] The public sentiment early warning module can store the first risk
total score and the
corresponding scoring time. According to the first risk total score and the
corresponding
scoring time, the public sentiment early warning module can further generate a
corresponding
second risk total score and judge whether to send out second risk early
warning. This process
includes the following:
22
Date Regue/Date Received 2022-07-27
[0125] S420 - generating a second risk total score to which the target object
corresponds
according to the first risk total score whose scoring time is within a preset
historical time
period and a time decay item determined according to the scoring time.
[0126] Specifically, taking a preset time period is one week before the start
of the current
appraisal as an example, the second risk total score fi can be expressed as:
fi =
Ess:8 Rse¨ciss , where s represents a difference between the scoring date and
the current time,
Rs represents the corresponding first risk total score, e¨ciss represents the
corresponding
time decay item, and this time decay item is an exponential function with
natural base number
e as base and with -0.5s as exponent.
[0127] S430 - determining whether to send out second risk early warning
according to the
second risk total score and a second risk early warning threshold.
[0128] Specifically, taking the aforementioned Y Fresh Milk Co., Ltd. is the
target enterprise as
an example, the first risk total score Rs and the corresponding time decay
value e¨ .ss of
the Y Fresh Milk Co., Ltd. in the past week are as shown in Table 3, and the
second risk early
warning threshold Fi = 10.
[0129] Table 3
Date Rs e¨o.ss __
past 1 day 5 0.61
past 2 days 2 0.36
past 3 days 1 0.22
past 4 days 2 0.14
past 5 days 3 0.08
past 6 days 4 0.05
23
Date Regue/Date Received 2022-07-27
past 7 days 1
[0130] Then the second risk total score to which the target enterprise
corresponds is:
[0131] /7? = Ess:8Rse¨'s = 9 x 1 + 5 x 0.61 + 2 x 0.36 + lx 0.22 + 2 x 0.14 +
3 x
0.08 + 4 x 0.05 = 13.71
[0132] Since the second risk total score is greater than the second risk early
warning threshold,
the second risk early warning is hence sent out.
[0133] Based on the method of appraising risk disclosed by this embodiment of
the present
application, the surveillance system can realize public sentiment surveillance
of focused
target enterprises, avoid wastage of resources due to total analysis of public
sentiment texts,
provide a quantitatively calculating method for public sentiment, and supply a
theoretic basis
for risk recognition.
[0134] Embodiment 2
[0135] Corresponding to the above embodiment, the present application further
provides a
method of appraising risk, as shown in Fig. 2, the method comprises the
following steps.
[0136] 2100 - obtaining, from a preset public sentiment text library, a target
public sentiment
text to which a target object corresponds and whose generating time is within
a preset time
range, according to a keyword library to which the target object corresponds.
[0137] Preferably, the target object includes enterprises, and, prior to the
step of obtaining, from
a preset public sentiment text library, a target public sentiment text to
which a target object
24
Date Regue/Date Received 2022-07-27
corresponds and whose generating time is within a preset time range, according
to a keyword
library to which the target object corresponds, the method further comprises:
[0138] 2110 - enquiring any associated enterprise to which the target object
corresponds from
a preset enterprise association map and obtaining enterprise information to
which the
associated enterprise and the target object respectively correspond;
[0139] 2111 - processing the enterprise information to which the associated
enterprise and the
target object respectively correspond, and generating a keyword to which the
target object
and the associated enterprise correspond; and
[0140] 2112 - determining the keyword library to which the target object
corresponds according
to the keyword to which the target object and the associated enterprise
correspond.
[0141] Preferably, prior to the step of enquiring any associated enterprise to
which the target
object corresponds from a preset enterprise association map and obtaining
enterprise
information to which the associated enterprise and the target object
respectively correspond,
the method further comprises:
[0142] 2120- generating anode to which each enterprise corresponds according
to an enterprise
business database, wherein the node includes a node attribute and an
associated attribute, the
node attribute includes the enterprise information to which the enterprise
corresponds, and
the associated attribute is employed for enquiring any associated enterprise
to which the
enterprise corresponds; and
[0143] 2121 - establishing the enterprise association map according to the
associated attribute
to which each node corresponds.
[0144] Preferably, generating a keyword to which the target object and the
associated enterprise
correspond, the method comprises:
[0145] 2130 - recognizing abbreviation fields included in the enterprise full
name by employing
Date Regue/Date Received 2022-07-27
a preset regular expression; and
[0146] 2131 - filtering the abbreviation fields according to a preset
filtering rule, and generating
a corresponding enterprise abbreviation.
[0147] Preferably, the step of obtaining, from a preset public sentiment text
library, a target
public sentiment text to which a target object corresponds and whose
generating time is
within a preset time range, according to a keyword library to which the target
object
corresponds includes:
[0148] 2140 - obtaining, from the preset public sentiment text library, a
public sentiment text
that contains one or more keyword(s) in the keyword library;
[0149] 2141 - predicting an emotion category to which each public sentiment
text corresponds
by employing a preset emotion analyzing model; and
[0150] 2142 - determining any public sentiment text whose emotion category is
negative from
the public sentiment texts to serve as the target public sentiment text.
[0151] 2200 - employing a trained classification model to predict an event
category to which
each target public sentiment text corresponds.
[0152] Preferably, a process of training the classification model includes:
[0153] 2210 - training the classification model according to a preset training
corpus set;
[0154] 2211 - verifying whether the classification model satisfies a preset
condition according
to a preset testing corpus set; and
[0155] 2212 - generating the trained classification model when the
classification model satisfies
a preset training condition.
26
Date Regue/Date Received 2022-07-27
[0156] 2300 - generating a public sentiment risk score to which the target
public sentiment text
corresponds according to the event category to which the target public
sentiment text
corresponds as obtained by prediction and according to preset negative hot
words included
in the target public sentiment text.
[0157] Preferably, the step of generating a public sentiment risk score to
which the target public
sentiment text corresponds according to the event category to which the target
public
sentiment text corresponds as obtained by prediction and according to preset
negative hot
words included in the target public sentiment text includes:
[0158] 2310 - determining an influence factor to which the target public
sentiment text
corresponds according to a preset corresponding relation between influence
factors and event
categories;
[0159] 2311 - determining a hot word risk value to which the target public
sentiment text
corresponds according to the number of preset negative hot words included in
the target
public sentiment text; and
[0160] 2312 - determining the public sentiment risk score to which the target
public sentiment
text corresponds according to the influence factor to which the target public
sentiment text
corresponds and according to the hot word risk value.
[0161] 2400 - determining whether to send out first risk early warning to
which the target object
corresponds according to the public sentiment risk score and a preset first
risk early warning
threshold.
[0162] Preferably, the method further comprises:
[0163] 2410 - generating a first risk total score to which the target object
corresponds and
determining a scoring time to which the first risk total score corresponds
according to all
27
Date Regue/Date Received 2022-07-27
public sentiment risk scores to which the target object corresponds;
[0164] 2411 - generating a second risk total score to which the target object
corresponds
according to the first risk total score whose scoring time is within a preset
historical time
period and a time decay item determined according to the scoring time; and
[0165] 2412 - determining whether to send out second risk early warning to
which the target
object corresponds according to the second risk total score and a preset
second risk early
warning threshold.
[0166] Embodiment 3
[0167] Corresponding to Embodiment 1 and Embodiment 2, the present application
further
provides a system for appraising risk, as shown in Fig. 3, the system
comprises:
[0168] an obtaining module 310, for obtaining, from a preset public sentiment
text library, a
target public sentiment text to which a target object corresponds and whose
generating time
is within a preset time range, according to a keyword library to which the
target object
corresponds;
[0169] a predicting module 320, for predicting an event category to which each
target public
sentiment text corresponds by employing a trained classification model;
[0170] a generating module 330, for generating a public sentiment risk score
to which the target
public sentiment text corresponds according to the event category to which the
target public
sentiment text corresponds as obtained by prediction and according to preset
negative hot
words included in the target public sentiment text; and
[0171] a judging module 340, for determining whether to send out first risk
early warning to
which the target object corresponds according to the public sentiment risk
score and a preset
first risk early warning threshold.
28
Date Regue/Date Received 2022-07-27
[0172] Preferably, the generating module 330 can be further employed for
generating a first risk
total score to which the target object corresponds and determining a scoring
time to which
the first risk total score corresponds according to all public sentiment risk
scores to which the
target object corresponds; and generating a second risk total score to which
the target object
corresponds according to the first risk total score whose scoring time is
within a preset
historical time period and a time decay item determined according to the
scoring time; the
judging module 340 can be further employed for determining whether to send out
second risk
early warning to which the target object corresponds according to the second
risk total score
and a preset second risk early warning threshold.
[0173] Preferably, the target object includes enterprises, and the obtaining
module 310 can be
further employed for enquiring any associated enterprise to which the target
object
corresponds from a preset enterprise association map and obtaining enterprise
information to
which the associated enterprise and the target object respectively correspond;
the generating
module 330 can be further employed for processing the enterprise information
to which the
associated enterprise and the target object respectively correspond, and
generating a keyword
to which the target object and the associated enterprise correspond; and
determining the
keyword library to which the target object corresponds according to the
keyword to which
the target object and the associated enterprise correspond.
[0174] Preferably, the generating module 330 can be further employed for
generating a node to
which each enterprise corresponds according to an enterprise business
database, wherein the
node includes a node attribute and an associated attribute, the node attribute
includes the
enterprise information to which the enterprise corresponds, and the associated
attribute is
employed for enquiring any associated enterprise to which the enterprise
corresponds; and
establishing the enterprise association map according to the associated
attribute to which each
node corresponds.
29
Date Regue/Date Received 2022-07-27
[0175] Preferably, the enterprise information includes an enterprise full
name, the keyword
includes an enterprise abbreviation, and the generating module 330 can be
further employed
for employing a preset regular expression to recognize abbreviation fields
included in the
enterprise full name; and filtering the abbreviation fields according to a
preset filtering rule,
and generating a corresponding enterprise abbreviation.
[0176] Preferably, the obtaining module 310 can be further employed for
obtaining, from the
preset public sentiment text library, a public sentiment text that contains
one or more
keyword(s) in the keyword library; the predicting module 320 can be further
employed for
employing a preset emotion analyzing model to predict an emotion category to
which each
public sentiment text corresponds; and determining any public sentiment text
whose emotion
category is negative from the public sentiment texts to serve as the target
public sentiment
text.
[0177] Preferably, the generating module 330 can be further employed for
determining an
influence factor to which the target public sentiment text corresponds
according to a preset
corresponding relation between influence factors and event categories;
determining a hot
word risk value to which the target public sentiment text corresponds
according to the number
of preset negative hot words included in the target public sentiment text; and
determining the
public sentiment risk score to which the target public sentiment text
corresponds according
to the influence factor to which the target public sentiment text corresponds
and according to
the hot word risk value.
[0178] Preferably, the device further comprises a training module for training
the classification
model according to a preset training corpus set; verifying whether the
classification model
satisfies a preset condition according to a preset testing corpus set; and
generating the trained
classification model when the classification model satisfies a preset training
condition.
Date Regue/Date Received 2022-07-27
[0179] Embodiment 4
[0180] Corresponding to all the above embodiments, this embodiment of the
present application
provides an electronic equipment that comprises:
[0181] one or more processor(s); and a memory, associated with the one or more
processor(s)
and used for storing a program instruction. The program instruction performs
the following
operations when it is read and executed by the one or more processor(s):
[0182] obtaining, from a preset public sentiment text library, a target public
sentiment text to
which a target object corresponds and whose generating time is within a preset
time range,
according to a keyword library to which the target object corresponds;
[0183] predicting an event category to which each target public sentiment text
corresponds by
employing a trained classification model;
[0184] generating a public sentiment risk score to which the target public
sentiment text
corresponds according to the event category to which the target public
sentiment text
corresponds as obtained by prediction and according to preset negative hot
words included
in the target public sentiment text; and
[0185] determining whether to send out first risk early warning to which the
target object
corresponds according to the public sentiment risk score and a preset first
risk early warning
threshold.
[0186] Fig. 4 exemplarily illustrates the framework of the electronic
equipment that can
specifically include a processor 1510, a video display adapter 1511, a
magnetic disk driver
1512, an input/output interface 1513, a network interface 1514, and a memory
1520. The
processor 1510, the video display adapter 1511, the magnetic disk driver 1512,
the
input/output interface 1513, the network interface 1514, and the memory 1520
can be
31
Date Regue/Date Received 2022-07-27
communicably connected with one another via a bus 1530.
[0187] The processor 1510 can be embodied as a general CPU (Central Processing
Unit), a
microprocessor, an ASIC (Application Specific Integrated Circuit), or one or
more integrated
circuit(s) for executing relevant program(s) to realize the technical
solutions provided by the
present application.
[0188] The memory 1520 can be embodied in such a form as an ROM (Read Only
Memory),
an RAM (Random Access Memory), a static storage device, or a dynamic storage
device.
The memory 1520 can store an operating system 1521 for controlling the running
of the
electronic equipment 1500, and a basic input/output system (BIOS) 1522 for
controlling
lower-level operations of the electronic equipment 1500. In addition, the
memory 1520 can
also store a web browser 1523, a data storage management system 1524, and an
icon font
processing system 1525, etc. The icon font processing system 1525 can be an
application
program that specifically realizes the aforementioned various step operations
in the
embodiments of the present application. To sum it up, when the technical
solutions provided
by the present application are to be realized via software or firmware, the
relevant program
codes are stored in the memory 1520, and invoked and executed by the processor
1510.
[0189] The input/output interface 1513 is employed to connect with an
input/output module to
realize input and output of information. The input/output module can be
equipped in the
device as a component part (not shown in the drawings), and can also be
externally connected
with the device to provide corresponding functions. The input means can
include a keyboard,
a mouse, a touch screen, a microphone, and various sensors etc., and the
output means can
include a display screen, a loudspeaker, a vibrator, an indicator light etc.
[0190] The network interface 1514 is employed to connect to a communication
module (not
32
Date Regue/Date Received 2022-07-27
shown in the drawings) to realize intercommunication between the current
device and other
devices. The communication module can realize communication in a wired mode
(via USB,
network cable, for example) or in a wireless mode (via mobile network, WIFI,
Bluetooth,
etc.).
[0191] The bus 1530 includes a passageway transmitting information between
various
component parts of the equipment (such as the processor 1510, the video
display adapter
1511, the magnetic disk driver 1512, the input/output interface 1513, the
network interface
1514, and the memory 1520).
[0192] Additionally, the electronic equipment 1500 may further obtain
information of specific
collection conditions from a virtual resource object collection condition
information database
for judgment on conditions, and so on.
[0193] As should be noted, although merely the processor 1510, the video
display adapter 1511,
the magnetic disk driver 1512, the input/output interface 1513, the network
interface 1514,
the memory 1520, and the bus 1530 are illustrated for the aforementioned
equipment, the
equipment may further include other component parts prerequisite for realizing
normal
running during specific implementation. In addition, as can be understood by
persons skilled
in the art, the aforementioned device may as well only include component parts
necessary for
realizing the solutions of the present application, without including the
entire component
parts as illustrated.
[0194] As can be known through the description to the aforementioned
embodiments, it is
clearly learnt by person skilled in the art that the present application can
be realized through
software plus a general hardware platform. Based on such understanding, the
technical
solutions of the present application, or the contributions made thereby over
the state of the
33
Date Regue/Date Received 2022-07-27
art, can be essentially embodied in the form of a software product, and such a
computer
software product can be stored in a storage medium, such as an ROM/RAM, a
magnetic disk,
an optical disk etc., and includes plural instructions enabling a computer
equipment (such as
a personal computer, a cloud server, or a network device etc.) to execute the
methods
described in various embodiments or some sections of the embodiments of the
present
application.
[0195] The various embodiments are progressively described in the Description,
identical or
similar sections among the various embodiments can be inferred from one
another, and each
embodiment stresses what is different from other embodiments. Particularly,
with respect to
the system or system embodiment, since it is essentially similar to the method
embodiment,
its description is relatively simple, and the relevant sections thereof can be
inferred from the
corresponding sections of the method embodiment. The system or system
embodiment as
described above is merely exemplary in nature, units therein described as
separate parts can
be or may not be physically separate, parts displayed as units can be or may
not be physical
units, that is to say, they can be located in a single site, or distributed
over a plurality of
network units. It is possible to select partial modules or the entire modules
to realize the
objectives of the embodied solutions based on practical requirements. It is
understandable
and implementable by persons ordinarily skilled in the art without spending
creative effort in
the process.
[0196] What the above describes is merely directed to preferred embodiments of
the present
application, and is not meant to restrict the present application. Any
modification, equivalent
substitution, and improvement makeable within the spirit and scope of the
present application
shall all be covered by the protection scope of the present application.
34
Date Regue/Date Received 2022-07-27
