CN113345590B - User mental health monitoring method and system based on heterogeneous graph - Google Patents
User mental health monitoring method and system based on heterogeneous graph Download PDFInfo
- Publication number
- CN113345590B CN113345590B CN202110730536.9A CN202110730536A CN113345590B CN 113345590 B CN113345590 B CN 113345590B CN 202110730536 A CN202110730536 A CN 202110730536A CN 113345590 B CN113345590 B CN 113345590B
- Authority
- CN
- China
- Prior art keywords
- user
- emotion
- personality
- behavior
- bert
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
Abstract
A user mental health monitoring method and system based on heterogeneous graphs belong to the technical field of data processing and analysis, and solve the problems that how to establish a personality portrait of a user by using the heterogeneous graphs and detect and early warn the mental health state of the user through social network text information of the user and physiological information of wearable equipment; the method comprises the steps of constructing a heterogeneous graph through text information, analyzing emotion expressed by a text and user behaviors extracted from the text to analyze the personality of a user, establishing a personality portrait of the user, analyzing text contents issued by the user in real time in combination with the personality portrait, verifying in combination with physiological information provided by user wearing equipment if the personality portrait is different from a normal state, and giving an early warning to the mental health state of the user, so that the accuracy rate of mental health intervention of the user is improved, and the forward effect of mental health monitoring of the user is strengthened.
Description
Technical Field
The invention belongs to the technical field of data processing and analysis, and relates to a user mental health monitoring method and system based on a heterogeneous graph.
Background
Mental health refers to the state of mental aspects and activities in a good or normal state. With the development of economy and the aggravation of competition, the rhythm of life and work of people is faster and faster. This clearly requires greater psychological stress on Chinese in the economic transformation phase. The number of people suffering from psychological problems in China is increased continuously, and particularly, the psychological pressure of people is increased by the problems of population loss, urbanization, nervous working environment, solitary child families and the like caused by social changes. However, the current society generally has insufficient acceptance of psychologists and understanding of psychological diseases, and the psychological diseases can not be prevented in early stage. With the continuous development of social networks, young people tend to express themselves on a network platform, and many scholars also begin to analyze user psychological states in combination with user behaviors on the network platform, but the users are difficult to help in time.
In the prior art, a chinese patent application, which is published on 6/30/2010, of "method, device and terminal for monitoring a health state of a user" monitors a physical health state of the user through a mobile phone, obtains motion information and air dust information of the user by using a sensor, and judges whether the motion information and the air dust information are different from normal values, thereby performing early warning on the physical health of the user; the Chinese patent application 'a vision problem common sense inference model and method based on multi-domain heterogeneous graph guidance', published as 12 and 20 in 2019, adopts a heterogeneous graph inference model and a multi-domain feature inference model constructed by a pre-training model, visual information and context information, and is used for breaking barriers in the language field and the visual field and completely fusing and aligning multi-modal information involved in tasks.
However, none of the above prior documents solves the problem of how to use a heterogeneous map to create a personality portrait of a user, and detect and warn the mental health status of the user by combining social network text information of the user and physiological information of a wearable device.
Disclosure of Invention
The technical problem to be solved by the invention is how to establish a personality portrait of a user by adopting a heterogeneous graph through social network text information of the user and physiological information of wearable equipment, and detect and early warn the mental health state of the user.
The invention solves the technical problems through the following technical scheme:
a user mental health monitoring method based on heterogeneous graphs comprises the following steps:
s1, collecting text data from each platform, establishing an emotion training set by means of a crowdsourcing technology and crowdsourcing platforms, and learning and optimizing an emotion classifier BERT model and a behavior word recognition BERT + CRF model;
s2, describing a personality network mode of action relations among the five personality traits, emotion and behaviors according to a psychology five personality theory system, and establishing a time sequence personality element structure;
s3, constructing a heterogeneous map of the user through the historical social network text data of the user, discovering the emotion and the behavior of the user by adopting an emotion classifier BERT model and a behavior word recognition BERT + CRF model, establishing a user personality portrait by combining a time sequence personality element structure,
s4, acquiring social network text data of the user at the current moment and physiological information fed back by the wearable device, learning the emotion and behavior of the current user respectively by adopting an emotion classifier BERT model and a behavior word recognition BERT + CRF model, finding abnormal emotion and behavior of the user,
and S5, analyzing the emotional tendency of the user through a time sequence personality element structure according to the five personality portrait of the user, detecting the psychological state of the user, and realizing accurate intervention on the psychological health of the user.
The input of the method is text data published by a user social platform and physiological data input by wearing equipment; and outputting the five personality portraits and the real-time emotional states of the user, and recommending a method for eliminating negative emotions.
The method comprises the steps of constructing a heterogeneous graph through text information, carrying out personality analysis on a user through analyzing emotion expressed by a text and user behaviors extracted from the text, so as to establish a personality portrait of the user, analyzing text contents issued by the user in real time in combination with the personality portrait, and verifying in combination with physiological information provided by user wearing equipment if the personality portrait is different from a normal state, so as to early warn the mental health state of the user.
On the basis of historical social network data of a user, a user personality portrait is established by relying on a heterogeneous graph technology system, the social network data of the user at the current moment and physiological information fed back by wearing equipment are combined, a deep learning technology is adopted, a user psychological fluctuation mode and a user behavior mode are learned, abnormal psychological fluctuation and abnormal behaviors of the user are found, emotional tendency of the user is analyzed, and the psychological state of the user is detected, so that the accuracy of psychological health intervention of the user is improved, and the forward effect of psychological health monitoring of the user is strengthened.
As a further improvement of the technical scheme of the present invention, the method for collecting text data from each platform, establishing an emotion training set, learning and optimizing an emotion classifier BERT model and a behavior word recognition BERT + CRF model by means of a crowdsourcing technique and crowdsourcing platform in step S1 specifically comprises:
crawling text data from a platform expressing emotion;
segmenting the crawled text into phrase levels, labeling emotion categories of phrases in a crowdsourcing mode, performing truth value reasoning, and dividing the obtained data into training data and test data in proportion for training an emotion classifier BERT model;
inputting data into a BERT model, and learning and optimizing the BERT model by adjusting model parameters;
behavior word labeling is carried out on the crawled text data in a crowdsourcing mode, true value reasoning is carried out, and the obtained data are divided into training data and testing data according to the proportion;
and inputting data into a behavior word recognition model BERT + CRF model, and learning and optimizing the behavior word recognition BERT + CRF model by adjusting model parameters.
As a further improvement of the technical scheme of the present invention, the personality network mode for describing the action relationship among the five personalities, the emotion and the behavior according to the psychology five-personalities theory system in step S2, and the method for establishing the time sequence personality cell structure specifically comprises the following steps:
expressing the characteristics expressed by each type of personality on the emotion and behavior, linking the personality with the emotion and behavior, and constructing a personality network mode, wherein the personality network mode comprises three object types of personality, emotion and behavior, and three relation types of infection, expression, intervention and embodiment; constructing a series of element structures for calculating the sub-dimension score of each personality type of the user so as to obtain the personality type tendency score of the user, wherein the calculation formula of the sub-dimension score of each personality type is as follows:
wherein n is 1 Number of emotion types, n 2 The number of the behavior types.
The relation matrix is a relation matrix of emotion types and behavior types, the relation matrix is 1 if the relation matrix has a link relation, and otherwise the relation matrix is 0; the characteristic sub-dimension of the personality type and the link of the personality type are fixed and are used as attributes of the personality type, so that a meta-structure with time sequence information from the emotion and behavior types to the personality type is directly constructed.
As a further improvement of the technical scheme of the invention, the method for establishing the personality portrait of the user by using the historical social network text data of the user to construct the heterogeneous map of the user, adopting a sentiment classifier BERT model and a behavior word recognition BERT + CRF model to discover the sentiment and the behavior of the user and combining a time sequence personality element structure comprises the following steps:
s31, crawling historical social network text data published by a user on a social network;
s32, segmenting the crawled text data into phrase levels;
s33, performing emotion recognition on the phrases by using the trained emotion classifier BERT model, and establishing a relation between the phrases and emotion categories, so as to construct a user five-personality portrait by a heterogeneous image technology;
s34, simultaneously, using the trained behavior word recognition BERT + CRF model to perform behavior word recognition on the phrase, and if the behavior word is recognized, complementing the behavior word and inputting the complemented behavior word into the five-personality portrait of the user; simultaneously, the behavior and the emotion are linked, so that the relationship between the emotion category and the behavior category is established;
and S35, when the established user five-personality portrait approaches to be relatively stable, the historical social network text data volume of the user meets the requirement, and accordingly the five-personality tendency of the user is reflected.
As a further improvement of the technical scheme of the present invention, in step S4, the social network text data of the user at the current time and the physiological information fed back by the wearable device are obtained, the emotion and behavior of the current user are learned respectively by using an emotion classifier BERT model and a behavior word recognition BERT + CRF model, and the method for finding the abnormal emotion and behavior of the user specifically comprises:
s41, obtaining social network text data of a user at the current moment and physiological information fed back by user wearing equipment, and dividing the text data into phrase levels;
s42, recognizing the emotion of the phrase by using an emotion classifier BERT model, searching the recognized emotion in a heterogeneous image of the five personality portrait of the user, and preliminarily considering that the psychological state of the user is in a stable state if the emotion can be searched and the frequency of the emotion exceeds a threshold value in historical data; if the psychological state of the user cannot be searched or the occurrence frequency in the historical data is less than a threshold value, preliminarily considering that the psychological state of the user is in a fluctuation state;
s43, further performing behavior word recognition on the phrase by using a behavior word recognition BERT + CRF model, if the behavior word is recognized, predicting the emotion type expressed by the phrase according to the relation between the emotion type and the behavior type established in the step S3, and returning to the step S42 for comparison and verification;
s44, if the relationship between the behavior category and the emotion category does not exist in the user five-personality portrait, establishing a new relationship to be complementally input into the user five-personality portrait;
and S45, if the user psychology is judged to be in a fluctuation state through the steps, analyzing whether the abnormal psychology state of the user belongs to positive abnormality or negative abnormality by combining the physiological information of the user on the same day.
As a further improvement of the technical scheme of the present invention, the method for analyzing the emotional tendency of the user through the time sequence personality cell structure according to the five personality portraits of the user and detecting the psychological state of the user to realize the accurate intervention of the psychological health of the user, which is described in step S5, specifically comprises the following steps:
s51, visually presenting the analysis result of each phrase to a user, enabling the user to evaluate the analysis result, feeding back the user evaluation to a BERT model and a BERT + CRF model for behavior word recognition, further improving the BERT model and the BERT + CRF model for behavior word recognition, and updating the quintuple portrait of the user;
s52, if the user is satisfied with the analysis result, evaluating the current overall psychological state of the user; if the user is in the negative abnormal psychological state on the same day, providing corresponding suggestions to help the user eliminate the negative psychological state;
and S53, analyzing whether the user state is improved or not according to the real-time physiological information provided by the wearable device, and giving a prompt.
A user mental health monitoring system based on heterogeneous graphs, comprising:
the model optimization module is used for collecting text data from each platform, establishing an emotion training set by relying on a crowdsourcing technology and crowdsourcing platforms, and learning and optimizing an emotion classifier BERT model and a behavior word recognition BERT + CRF model;
the time sequence personality cell structure establishing module is used for describing a personality network mode of action relations among the five personalities, the emotion and the behaviors according to a psychology five-personality theory system and establishing a time sequence personality cell structure;
the user personality portrait establishing module is used for establishing a heterogeneous map of the user through historical social network text data of the user, discovering the emotion and the behavior of the user by adopting an emotion classifier BERT model and a behavior word recognition BERT + CRF model, and establishing a user personality portrait by combining a time sequence personality cell structure;
the abnormal emotion and behavior discovery module is used for acquiring social network text data of the user at the current moment and physiological information fed back by the wearing equipment, learning the emotion and behavior of the current user respectively by adopting an emotion classifier BERT model and a behavior word recognition BERT + CRF model, and discovering the abnormal emotion and behavior of the user;
and the mental health intervention module is used for analyzing the emotional tendency of the user through a time sequence personality cell structure according to the five personality figures of the user, detecting the mental state of the user and realizing the accurate intervention of the mental health of the user.
As a further improvement of the technical scheme of the invention, the method for collecting text data from each platform, establishing an emotion training set, learning and optimizing an emotion classifier BERT model and a behavior word recognition BERT + CRF model by relying on a crowdsourcing technology and a crowdsourcing platform in the model optimization module specifically comprises the following steps:
crawling text data from a platform expressing emotion;
segmenting crawled texts into phrase levels, labeling emotion classes of phrases in a crowdsourcing mode, performing true value reasoning, and proportionally dividing obtained data into training data and test data for training an emotion classifier BERT model;
inputting data into a BERT model, and learning and optimizing the BERT model by adjusting model parameters;
behavior word labeling is carried out on the crawled text data in a crowdsourcing mode, true value reasoning is carried out, and the obtained data are divided into training data and testing data according to the proportion;
inputting data into a behavior word recognition model BERT + CRF model, and learning and optimizing the behavior word recognition BERT + CRF model by adjusting model parameters;
the time sequence personality element structure establishing module describes a personality network mode of action relations among the five personality, emotion and behaviors according to a psychology five personality theory system, and the method for establishing the time sequence personality element structure specifically comprises the following steps:
expressing the characteristics expressed by each type of personality to the emotion and behavior, linking the personality with the emotion and behavior, and constructing a personality network mode, wherein the personality network mode comprises three object types of personality, emotion and behavior and three relation types of infection, expression, intervention and embodiment; constructing a series of element structures for calculating the sub-dimension score of each personality type of the user so as to obtain the personality type tendency score of the user, wherein the calculation formula of the sub-dimension score of each personality type is as follows:
wherein n is 1 Number of emotion types, n 2 The number of the behavior types.
The relation matrix is a relation matrix of emotion types and behavior types, the relation matrix is 1 if the relation matrix has a link relation, and otherwise the relation matrix is 0; the characteristic sub-dimension of the personality type and the link of the personality type are fixed and are used as attributes of the personality type, so that a meta-structure with time sequence information from the emotion and behavior types to the personality type is directly constructed.
As a further improvement of the technical scheme of the invention, the method for establishing the user personality portrait by adopting the emotion classifier BERT model and the behavior word recognition BERT + CRF model to discover the emotion and behavior of the user and combining the time sequence personality cell structure comprises the following steps:
crawling historical social network text data published by a user on a social network;
segmenting the crawled text data into phrase levels;
performing emotion recognition on the phrases by using a trained emotion classifier BERT model, and establishing a relation between the phrases and emotion categories, so as to construct a user five-personality portrait by using a heterogeneous graph technology;
meanwhile, the trained behavior word recognition BERT + CRF model is used for performing behavior word recognition on the phrase, and if the behavior word is recognized, the behavior word is complemented and input into the five-personality portrait of the user; simultaneously, the behavior and the emotion are linked, so that the relationship between the emotion category and the behavior category is established;
when the established five-personality portrait of the user approaches to be relatively stable, the historical social network text data volume of the user meets the requirement, and therefore the five-personality tendency of the user is reflected.
As a further improvement of the technical scheme of the invention, the abnormal emotion and behavior discovery module acquires social network text data of the user at the current moment and physiological information fed back by the wearable device, and learns the emotion and behavior of the current user respectively by adopting an emotion classifier BERT model and a behavior word recognition BERT + CRF model, and the method for discovering the abnormal emotion and behavior of the user specifically comprises the following steps:
acquiring social network text data of a user at the current moment and physiological information fed back by user wearing equipment, and segmenting the text data into phrase levels;
performing emotion recognition on the phrases by using an emotion classifier BERT model, searching recognized emotions in a heterogeneous graph of a user's five-personality portrait, and preliminarily considering that the psychological state of the user is in a stable state if the recognized emotions can be searched and the frequency of the recognized emotions in historical data exceeds a threshold value; if the psychological state of the user cannot be searched or the occurrence frequency in the historical data is less than a threshold value, preliminarily considering that the psychological state of the user is in a fluctuation state;
further utilizing a behavior word recognition BERT + CRF model to perform behavior word recognition on the phrase, if the behavior word is recognized, predicting the emotion type expressed by the phrase through the established relationship between the emotion type and the behavior type, and returning to perform comparison verification;
if the relationship between the behavior category and the emotion category does not exist in the user five-personality portrait, establishing a new relationship to be complementally input to the user five-personality portrait;
if the user psychology is judged to be in a fluctuation state through the steps, analyzing whether the abnormal psychology state of the user belongs to positive abnormality or negative abnormality by combining the physiological information of the user on the same day;
the method for accurately intervening the mental health of the user by analyzing the emotional tendency of the user through the time sequence personality cell structure according to the five personality figures of the user and detecting the mental state of the user in the mental health intervening module specifically comprises the following steps:
visually presenting the analysis result of each phrase to a user, enabling the user to evaluate the analysis result, feeding back the user evaluation to an emotion classifier BERT model and a behavior word recognition BERT + CRF model, further improving the emotion classifier BERT model and the behavior word recognition BERT + CRF model, and updating the quintuple portrait of the user;
if the user is satisfied with the analysis result, evaluating the current overall psychological state of the user; if the user is in the negative abnormal psychological state on the same day, providing corresponding suggestions to help the user eliminate the negative psychological state;
and analyzing whether the user state is improved or not according to the real-time physiological information provided by the wearable device, and giving a prompt.
The invention has the advantages that:
according to the technical scheme, the social information of the user at the current moment is analyzed through the personality portrait constructed by the historical social information of the user, and then is further verified through the physiological information provided by the user wearing equipment, so that the psychological state of the user can be accurately judged, if a negative tendency occurs, reasonable suggestions are provided for the user to help the user to adjust the user, the accuracy rate of psychological health intervention of the user is improved, and the forward effect of psychological health monitoring of the user is strengthened.
Drawings
FIG. 1 is a flowchart of a user mental health monitoring method based on a heterogeneous graph according to an embodiment of the present invention;
fig. 2 is a flowchart of a specific implementation of a user mental health monitoring method based on a heterogeneous graph according to a first embodiment of the present invention;
FIG. 3 is an example chart of the five personality-emotion-behavior of a user mental health monitoring method based on a heterogeneous graph according to a first embodiment of the present invention;
FIG. 4 is a schematic diagram of a personality network of a method for monitoring mental health of a user based on a heterogeneous graph according to an embodiment of the present invention;
fig. 5 is a time sequence personality configuration diagram of a user mental health monitoring method based on a heterogeneous graph according to a first embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the embodiments of the present invention, and it is obvious that the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The technical scheme of the invention is further described by combining the drawings and the specific embodiments in the specification:
example one
As shown in fig. 1 and 2, a user mental health monitoring method based on heterogeneous graph includes the following steps:
1. text data are collected from all platforms, an emotion training set is established by means of crowdsourcing technology and crowdsourcing platforms, an emotion classifier BERT model and a behavior word recognition BERT + CRF model are learned and optimized, and the emotion classifier BERT + CRF model comprises the following concrete steps:
1-1, crawling mass text data from platforms expressing emotions with high probability such as microblogs, broad bean film reviews, internet surfing music reviews and the like;
1-2, segmenting the crawled text into phrase levels, labeling emotion categories of phrases in a crowdsourcing mode, performing truth value reasoning, and dividing the obtained data into training data and test data according to the proportion of 7:3 for training an emotion classifier BERT model;
1-3, inputting data into a BERT model, and learning and optimizing the BERT model by adjusting model parameters;
1-4, performing behavior word labeling on the crawled text data in a crowdsourcing mode, performing true value reasoning, and dividing the obtained data into training data and test data according to the proportion of 7:3;
and 1-5, inputting data into a behavior word recognition BERT + CRF model, and learning and optimizing the behavior word recognition BERT + CRF model by adjusting model parameters.
2. According to a psychological five-personality theory system, a personality network mode for describing the action relationship among the five personalities, the emotion and the behaviors is condensed, and a time sequence personality element structure is established, which specifically comprises the following steps:
2-1, classifying the personality into five types according to the five personality theory: openness (O), accountability (C), camber (E), amenity (A) and nervousness (N), and the personality type of the user is quantized into a vector by adopting the principle of a five-personality scale:
P＝<O score ,C score ,E score ,A score ,N score >
2-2, each personality more specifically shows some characteristics, such as the camber is specifically shown in six sub-dimensions of enthusiasm, happy group, independence, vitality, demand stimulation and positive emotion, and the score condition of each sub-dimension of the user on each personality is further calculated, taking the camber as an example:
wherein e is i Sub-dimension, k, representing camber i For the weight of the influence of each dimension on the personality type, initializing to
Which is then adjusted based on user feedback. The remaining character characteristics are shown in table 1.
TABLE 1 characteristics of various sub-dimensions of the five-personality
2-3, the characteristics expressed by each type of personality are specific to the emotion and behavior, so as to link the personality with the emotion and behavior, for example, as shown in fig. 3: extroversion represents the number and density of interpersonal interactions, the need for stimulation, and the ability to achieve pleasure. This dimension contrasts social, active, personally directed individuals with silent, serious, \ 33148, caucasian, quiet people. This aspect can be measured by two qualities: the level of human involvement and the level of vitality. The former assesses the extent to which an individual likes others to accompany, while the latter reflects the individual's individual pace and vitality level. For example: the distraction is an important emotional expression of the camber, the demand partner is an important behavioral expression of the camber, the distraction and the seeking partner both represent the user's camber personality tendencies, and the behaviors and emotions generally influence each other, but the influence is influenced by personality types, the emotions and behavioral associations expressed by people with different personality tendencies are different, the camber points to distraction in fig. 3 for the intervention of the expression of the seeking partner behavior, because the camber generally likes to be in contact with people and is enthusiastic, and when they are in a group, a positive emotion is generally expressed, such as distraction; vice versa, the same reasoning applies to the behavior from emotion.
2-4, as shown in fig. 4, the emotion can affect the behavior to a certain extent, the behavior is often expressed in real-time emotion, and different personality can affect the interaction between the emotion and the behavior, so that a personality network mode can be constructed, wherein the personality network mode comprises three object types of personality, emotion and behavior, and three relation types of infection, expression, intervention and embodiment;
and 2-5, constructing a series of element structures for calculating the sub-dimension score of each personality type of the user so as to obtain the personality type tendency score of the user. The characteristic sub-dimension of the personality type and the link of the personality type are fixed, and the characteristic sub-dimension can be used as the attribute of the personality type, so that a meta-structure with time sequence information from the emotion and behavior types to the personality type can be directly constructed.
The meta structure can be regarded as a directed acyclic graph formed by combining a plurality of original paths with common nodes, and the original paths are special cases of the meta structure. By analyzing the personality network mode of the user, a meta-structure capable of analyzing the personality type is extracted, and the five personality type nodes have time information, so that the time sequence personality meta-structure is called.
As shown in fig. 5, the meta-structure is summarized as that the initial value is assigned according to the probability of occurrence of each behavior node b and emotion node m in the historical text information of the user, and it should be noted here that there are not only emotions and behaviors that promote personality tendencies in the forward direction but also emotions and behaviors that suppress personality tendencies in the reverse direction, so that positive values are assigned to the forward influence factors and negative values are assigned to the reverse influence factors. The user's historical five-personality portrait serves as an important adjustment factor. Behaviors can express emotions, emotions can infect behaviors, and the emotions can reflect the character of personality dimension more deeply when appearing together. Therefore, a characteristic sub-dimension score model of the personality type can be established as follows:
wherein n is 1 Number of emotion types, n 2 The number of behavior types.
The relation matrix is a relation matrix of emotion types and behavior types, the relation matrix is 1 if the relation matrix has a link relation, and otherwise the relation matrix is 0.
3. The method comprises the steps of constructing a heterogeneous graph of a user through historical social network text data of the user, discovering the emotion and the behavior of the user by adopting an emotion classifier BERT model and a behavior word recognition BERT + CRF model, and establishing a user personality portrait by combining a time sequence personality element structure, wherein the method specifically comprises the following steps:
3-1, crawling historical social network text data published by a user on a social network;
3-2, segmenting the crawled text data into phrase levels;
3-3, performing emotion recognition on the phrases by using the trained emotion classifier BERT model, and establishing a relation between the phrases and emotion categories, so as to construct a user's five-personality portrait by a heterogeneous graph technology;
3-4, simultaneously carrying out behavior word recognition on the phrase by using the trained behavior word recognition BERT + CRF model, and if the behavior word is recognized, complementing the behavior word and inputting the complemented behavior word into the five-personality portrait of the user; simultaneously, the behavior and the emotion are linked, so that the relationship between the emotion category and the behavior category is established;
and 3-5, when the established five-personality portrait of the user approaches to be relatively stable, the text data volume of the historical social network of the user meets the requirement, and therefore the five-personality tendency of the user is reflected.
4. Acquiring social network text data of a user at the current moment and physiological information fed back by wearing equipment, learning the emotion and behavior of the current user respectively by adopting an emotion classifier BERT model and a behavior word recognition BERT + CRF model, and finding abnormal emotion and behavior of the user as follows:
4-1, acquiring social network text data of a user at the current moment and physiological information fed back by user wearing equipment, and dividing the text data into phrase levels;
4-2, performing emotion recognition on the phrases by using an emotion classifier BERT model, searching recognized emotions in heterogeneous images of the five personality portraits of the user, and preliminarily considering that the psychological state of the user is in a stable state if the emotion recognition can be searched and the frequency of the emotion recognition exceeds a threshold value in historical data; if the psychological state of the user cannot be searched or the occurrence frequency in the historical data is less than a threshold value, preliminarily considering that the psychological state of the user is in a fluctuation state;
4-3, further utilizing a behavior word recognition BERT + CRF model to perform behavior word recognition on the phrase, if the behavior word is recognized, predicting the emotion type expressed by the phrase according to the relation between the emotion type and the behavior type established in the step 3, and returning to the step 4-2 for comparison and verification;
4-4, if the relation between the behavior type and the emotion type does not exist in the user's five-personality portrait, establishing a new relation to complement and input the relation into the user's five-personality portrait;
and 4-5, if the psychology of the user is judged to be in a fluctuating state through the steps, analyzing whether the abnormal psychology of the user belongs to positive abnormality or negative abnormality by combining the physiological information of the user in the current day.
5. According to the five-personality portrait of the user, the emotional tendency of the user is analyzed through the time sequence personality cell structure, the psychological state of the user is detected, and the accurate intervention of the psychological health of the user is realized, specifically as follows:
5-1, visually presenting the analysis result of each phrase to a user, evaluating the analysis result by the user, feeding back the user evaluation to a BERT model of an emotion classifier and a BERT + CRF model of behavior word recognition, further improving the BERT model of the emotion classifier and the BERT + CRF model of the behavior word recognition, and updating the five-personality portrait of the user;
5-2, if the user is satisfied with the analysis result, evaluating the current overall psychological state of the user; if the user is in the negative abnormal psychological state on the same day, providing corresponding suggestions to help the user eliminate the negative psychological state;
and 5-3, analyzing whether the user state is improved or not according to the real-time physiological information provided by the wearable device, and giving a prompt.
Example two
A user mental health monitoring system based on heterogeneous graphs, comprising:
the model optimization module is used for collecting text data from each platform, establishing an emotion training set by relying on a crowdsourcing technology and crowdsourcing platforms, and learning and optimizing an emotion classifier BERT model and a behavior word recognition BERT + CRF model;
the time sequence personality cell structure establishing module is used for describing a personality network mode of action relations among the five personalities, the emotion and the behaviors according to a psychology five-personality theory system and establishing a time sequence personality cell structure;
the user personality portrait establishing module is used for establishing a heterogeneous map of the user through historical social network text data of the user, discovering the emotion and the behavior of the user by adopting an emotion classifier BERT model and a behavior word recognition BERT + CRF model, and establishing a user personality portrait by combining a time sequence personality cell structure;
the abnormal emotion and behavior discovery module is used for acquiring social network text data of the user at the current moment and physiological information fed back by the wearing equipment, learning the emotion and behavior of the current user respectively by adopting an emotion classifier BERT model and a behavior word recognition BERT + CRF model, and discovering the abnormal emotion and behavior of the user;
and the mental health intervention module is used for analyzing the emotional tendency of the user through the time sequence personality element structure according to the five personality portrait of the user, detecting the mental state of the user and realizing the accurate intervention of the mental health of the user.
The method for collecting text data from each platform, establishing an emotion training set, learning and optimizing the emotion classifier BERT model and the behavior word recognition BERT + CRF model by means of a crowdsourcing technology and a crowdsourcing platform in the model optimization module specifically comprises the following steps:
crawling text data from a platform expressing emotion;
segmenting the crawled text into phrase levels, labeling emotion categories of phrases in a crowdsourcing mode, performing truth value reasoning, and dividing the obtained data into training data and test data in proportion for training an emotion classifier BERT model;
inputting data into a BERT model, and learning and optimizing the BERT model by adjusting model parameters;
behavior word labeling is carried out on the crawled text data in a crowdsourcing mode, true value reasoning is carried out, and the obtained data are divided into training data and testing data according to the proportion;
inputting data into a behavior word recognition model BERT + CRF model, and learning and optimizing the behavior word recognition BERT + CRF model by adjusting model parameters;
the personality network mode for describing the action relationship among the five personalities, the emotion and the behaviors according to the psychology five-personalities theory system in the time sequence personality cell structure establishing module specifically comprises the following steps:
expressing the characteristics expressed by each type of personality on the emotion and behavior, linking the personality with the emotion and behavior, and constructing a personality network mode, wherein the personality network mode comprises three object types of personality, emotion and behavior, and three relation types of infection, expression, intervention and embodiment; constructing a series of element structures for calculating the sub-dimension score of each personality type of the user so as to obtain the personality type tendency score of the user, wherein the calculation formula of the sub-dimension score of each personality type is as follows:
wherein n is 1 Number of emotion types, n 2 The number of the behavior types.
The relation matrix is a relation matrix of emotion types and behavior types, the relation matrix is 1 if the relation matrix has a link relation, and otherwise the relation matrix is 0; the characteristic sub-dimension of the personality type and the link of the personality type are fixed and are used as attributes of the personality type, so that a meta-structure with time sequence information from the emotion and behavior types to the personality type is directly constructed.
The method for establishing the user personality portrait by establishing the historical social network text data of the user in the user personality portrait establishing module, adopting a sentiment classifier BERT model and a behavior word recognition BERT + CRF model to discover the sentiment and the behavior of the user, and combining a time sequence personality element structure comprises the following steps:
crawling historical social network text data published by a user on a social network;
segmenting the crawled text data into phrase levels;
performing emotion recognition on the phrases by using a trained emotion classifier BERT model, and establishing a relation between the phrases and emotion categories, so as to construct a user five-personality portrait by using a heterogeneous graph technology;
meanwhile, the trained behavior word recognition BERT + CRF model is used for performing behavior word recognition on the phrase, and if the behavior word is recognized, the behavior word is complemented and input into the five-personality portrait of the user; simultaneously, the behavior and the emotion are linked, so that the relationship between the emotion category and the behavior category is established;
when the established user five-personality portrait approaches to be relatively stable, the historical social network text data volume of the user meets the requirement, and therefore the five-personality tendency of the user is reflected.
The method for discovering the abnormal emotion and behavior of the user comprises the following steps of obtaining social network text data of the user at the current moment and physiological information fed back by the wearable device in the abnormal emotion and behavior discovering module, respectively learning the emotion and behavior of the current user by adopting an emotion classifier BERT model and a behavior word recognition BERT + CRF model, and specifically:
acquiring social network text data of a user at the current moment and physiological information fed back by user wearing equipment, and segmenting the text data into phrase levels;
performing emotion recognition on the phrases by using an emotion classifier BERT model, searching recognized emotions in a heterogeneous graph of a user's five-personality portrait, and preliminarily considering that the psychological state of the user is in a stable state if the recognized emotions can be searched and the frequency of the recognized emotions in historical data exceeds a threshold value; if the psychological state of the user cannot be searched or the occurrence frequency in the historical data is less than a threshold value, preliminarily considering that the psychological state of the user is in a fluctuation state;
further utilizing a behavior word recognition BERT + CRF model to perform behavior word recognition on the phrase, if the behavior word is recognized, predicting the emotion type expressed by the phrase through the established relationship between the emotion type and the behavior type, and returning to perform comparison verification;
if the relationship between the behavior category and the emotion category does not exist in the user five-personality portrait, establishing a new relationship to be complementally input to the user five-personality portrait;
if the user psychology is judged to be in a fluctuation state through the steps, analyzing whether the abnormal psychology state of the user belongs to positive abnormality or negative abnormality by combining the physiological information of the user on the same day;
the method for accurately intervening the mental health of the user by analyzing the emotional tendency of the user through the time sequence personality cell structure according to the five personality figures of the user and detecting the mental state of the user in the mental health intervening module specifically comprises the following steps:
visually presenting the analysis result of each phrase to a user, enabling the user to evaluate the analysis result, feeding back the user evaluation to an emotion classifier BERT model and a behavior word recognition BERT + CRF model, further improving the emotion classifier BERT model and the behavior word recognition BERT + CRF model, and updating the quintuple portrait of the user;
if the user is satisfied with the analysis result, evaluating the current overall psychological state of the user; if the user is in the negative abnormal psychological state on the same day, providing corresponding suggestions to help the user eliminate the negative psychological state;
and analyzing whether the user state is improved or not according to the real-time physiological information provided by the wearable device, and giving a prompt.
The above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (6)
1. A user mental health monitoring method based on heterogeneous graphs is characterized by comprising the following steps:
s1, collecting text data from each platform, establishing an emotion training set by means of a crowdsourcing technology and crowdsourcing platforms, and learning and optimizing an emotion classifier BERT model and a behavior word recognition BERT + CRF model;
s2, describing a personality network mode of action relations among the five personalities, emotion and behaviors according to a psychology five-personality theory system, and establishing a time sequence personality element structure; the specific method comprises the following steps:
expressing the characteristics expressed by each type of personality on the emotion and behavior, linking the personality with the emotion and behavior, and constructing a personality network mode, wherein the personality network mode comprises three object types of personality, emotion and behavior, and three relation types of infection, expression, intervention and embodiment; constructing a series of element structures for calculating the sub-dimension score of each personality type of the user so as to obtain the personality type tendency score of the user, wherein the calculation formula of the sub-dimension score of each personality type is as follows:
wherein n is 1 Number of emotion types, n 2 The number of the behavior types;
the relation matrix is a relation matrix of emotion types and behavior types, the relation matrix is 1 if the relation matrix has a link relation, and otherwise the relation matrix is 0; the characteristic sub-dimension of the personality type and the link of the personality type are fixed, and the characteristic sub-dimension is taken as the attribute of the personality type, so that a meta-structure with time sequence information from the emotion type and the behavior type to the personality type is directly constructed;
s3, constructing a heterogeneous graph of the user through historical social network text data of the user, discovering the emotion and behavior of the user by adopting an emotion classifier BERT model and a behavior word recognition BERT + CRF model, and establishing a personality portrait of the user by combining a time sequence personality cell structure; the specific method comprises the following steps:
s31, crawling historical social network text data published by a user on a social network;
s32, segmenting the crawled text data into phrase levels;
s33, performing emotion recognition on the phrases by using the trained emotion classifier BERT model, and establishing a relation between the phrases and emotion categories, so as to construct a user five-personality portrait by a heterogeneous image technology;
s34, simultaneously, using the trained behavior word recognition BERT + CRF model to perform behavior word recognition on the phrase, and if the behavior word is recognized, complementing the behavior word and inputting the complemented behavior word into the five-personality portrait of the user; simultaneously, the behavior and the emotion are linked, so that the relationship between the emotion category and the behavior category is established;
s35, when the established user five-personality portrait approaches to be relatively stable, the historical social network text data volume of the user meets the requirement, and accordingly the five-personality tendency of the user is reflected;
s4, acquiring social network text data of the user at the current moment and physiological information fed back by the wearable device, learning the emotion and behavior of the current user respectively by adopting an emotion classifier BERT model and a behavior word recognition BERT + CRF model, and finding abnormal emotion and behavior of the user; the specific method comprises the following steps:
s41, obtaining social network text data of a user at the current moment and physiological information fed back by user wearing equipment, and dividing the text data into phrase levels;
s42, recognizing the emotion of the phrase by using an emotion classifier BERT model, searching the recognized emotion in a heterogeneous image of the five personality portrait of the user, and preliminarily considering that the psychological state of the user is in a stable state if the emotion can be searched and the frequency of the emotion exceeds a threshold value in historical data; if the psychological state of the user cannot be searched or the occurrence frequency in the historical data is less than a threshold value, preliminarily considering that the psychological state of the user is in a fluctuation state;
s43, further performing behavior word recognition on the phrase by using a behavior word recognition BERT + CRF model, if the behavior word is recognized, predicting the emotion type expressed by the phrase according to the relation between the emotion type and the behavior type established in the step S3, and returning to the step S42 for comparison and verification;
s44, if the relation between the behavior type and the emotion type does not exist in the user 'S five-personality portrait, establishing a new relation to complement and input the relation into the user' S five-personality portrait;
s45, if the user psychology is judged to be in a fluctuating state through the steps, analyzing whether the abnormal psychological state of the user belongs to positive abnormality or negative abnormality by combining the physiological information of the user on the same day;
and S5, analyzing the emotional tendency of the user through a time sequence personality cell structure according to the five personality portraits of the user, detecting the psychological state of the user, and realizing accurate intervention on the psychological health of the user.
2. The method for monitoring user mental health based on the heterogeneous map according to claim 1, wherein the method for collecting text data from each platform, establishing an emotion training set, learning and optimizing an emotion classifier BERT model and a behavior word recognition BERT + CRF model by means of a crowdsourcing technique and a crowdsourcing platform in step S1 specifically comprises the following steps:
crawling text data from a platform expressing emotion;
segmenting the crawled text into phrase levels, labeling emotion categories of phrases in a crowdsourcing mode, performing truth value reasoning, and dividing the obtained data into training data and test data in proportion for training an emotion classifier BERT model;
inputting data into a BERT model, and learning and optimizing the BERT model by adjusting model parameters;
behavior word labeling is carried out on the crawled text data in a crowdsourcing mode, true value reasoning is carried out, and the obtained data are divided into training data and testing data according to the proportion;
and inputting the data into a behavior word recognition model BERT + CRF model, and learning and optimizing the behavior word recognition BERT + CRF model by adjusting model parameters.
3. The method for monitoring the mental health of the user based on the heterogeneous map as claimed in claim 2, wherein the method for analyzing the emotional tendency of the user through the time sequence personality element structure according to the five personality portrait of the user and detecting the mental state of the user to realize the accurate intervention of the mental health of the user in the step S5 specifically comprises the following steps:
s51, visually presenting the analysis result of each phrase to a user, enabling the user to evaluate the analysis result, feeding back the user evaluation to a BERT model and a BERT + CRF model for behavior word recognition, further improving the BERT model and the BERT + CRF model for behavior word recognition, and updating the quintuple portrait of the user;
s52, if the user is satisfied with the analysis result, evaluating the current overall psychological state of the user; if the user is in the negative abnormal psychological state on the same day, providing corresponding suggestions to help the user eliminate the negative psychological state;
and S53, analyzing whether the user state is improved or not according to the real-time physiological information provided by the wearable device, and giving a prompt.
4. A system for monitoring mental health of a user based on a heterogeneous graph, comprising:
the model optimization module is used for collecting text data from each platform, establishing an emotion training set by relying on a crowdsourcing technology and crowdsourcing platforms, and learning and optimizing an emotion classifier BERT model and a behavior word recognition BERT + CRF model;
the time sequence personality cell structure establishing module is used for describing a personality network mode of action relations among the five personalities, the emotion and the behaviors according to a psychology five-personality theory system and establishing a time sequence personality cell structure, and the specific method comprises the following steps:
expressing the characteristics expressed by each type of personality on the emotion and behavior, linking the personality with the emotion and behavior, and constructing a personality network mode, wherein the personality network mode comprises three object types of personality, emotion and behavior, and three relation types of infection, expression, intervention and embodiment; constructing a series of element structures for calculating the sub-dimension score of each personality type of the user so as to obtain the personality type tendency score of the user, wherein the calculation formula of the sub-dimension score of each personality type is as follows:
wherein n is 1 Number of emotion types, n 2 The number of the behavior types;
the relation matrix is a relation matrix of emotion types and behavior types, the relation matrix is 1 if the relation matrix has a link relation, and otherwise the relation matrix is 0; the characteristic sub-dimension of the personality type and the link of the personality type are fixed, and the characteristic sub-dimension is taken as the attribute of the personality type, so that a meta-structure with time sequence information from the emotion type and the behavior type to the personality type is directly constructed;
the user personality portrait establishing module is used for establishing a heterogeneous map of the user through historical social network text data of the user, discovering the emotion and the behavior of the user by adopting an emotion classifier BERT model and a behavior word recognition BERT + CRF model, and establishing a user personality portrait by combining a time sequence personality cell structure; the specific method comprises the following steps:
crawling historical social network text data published by a user on a social network;
segmenting the crawled text data into phrase levels;
performing emotion recognition on the phrases by using a trained emotion classifier BERT model, and establishing a relation between the phrases and emotion categories, so as to construct a user five-personality portrait by using a heterogeneous graph technology;
meanwhile, the trained behavior word recognition BERT + CRF model is used for performing behavior word recognition on the phrase, and if the behavior word is recognized, the behavior word is complemented and input to the five-character portrait of the user; simultaneously, the behavior and the emotion are linked, so that the relationship between the emotion category and the behavior category is established;
when the established user five-personality portrait approaches to be relatively stable, the historical social network text data volume of the user meets the requirement, and therefore the five-personality tendency of the user is reflected;
the abnormal emotion and behavior discovery module is used for acquiring social network text data of the user at the current moment and physiological information fed back by the wearing equipment, learning the emotion and behavior of the current user respectively by adopting an emotion classifier BERT model and a behavior word recognition BERT + CRF model, and discovering the abnormal emotion and behavior of the user; the specific method comprises the following steps:
acquiring social network text data of a user at the current moment and physiological information fed back by user wearing equipment, and segmenting the text data into phrase levels;
performing emotion recognition on the phrases by using an emotion classifier BERT model, searching recognized emotions in a heterogeneous graph of a user's five-personality portrait, and preliminarily considering that the psychological state of the user is in a stable state if the recognized emotions can be searched and the frequency of the recognized emotions in historical data exceeds a threshold value; if the psychological state of the user cannot be searched or the occurrence frequency in the historical data is less than a threshold value, preliminarily considering that the psychological state of the user is in a fluctuation state;
further utilizing a behavior word recognition BERT + CRF model to perform behavior word recognition on the phrase, if the behavior word is recognized, predicting the emotion type expressed by the phrase through the established relationship between the emotion type and the behavior type, and returning to perform comparison verification;
if the relationship between the behavior category and the emotion category does not exist in the user five-personality portrait, establishing a new relationship to be complementally input to the user five-personality portrait;
if the user psychology is judged to be in a fluctuation state through the steps, analyzing whether the abnormal psychology state of the user belongs to positive abnormality or negative abnormality by combining the physiological information of the user on the same day;
and the mental health intervention module is used for analyzing the emotional tendency of the user through the time sequence personality element structure according to the five personality portrait of the user, detecting the mental state of the user and realizing the accurate intervention of the mental health of the user.
5. The system according to claim 4, wherein the user mental health monitoring system based on heterogeneous graph,
the method for collecting text data from each platform, establishing an emotion training set, learning and optimizing the emotion classifier BERT model and the behavior word recognition BERT + CRF model by means of a crowdsourcing technology and a crowdsourcing platform in the model optimization module specifically comprises the following steps:
crawling text data from a platform expressing emotion;
segmenting the crawled text into phrase levels, labeling emotion categories of phrases in a crowdsourcing mode, performing truth value reasoning, and dividing the obtained data into training data and test data in proportion for training an emotion classifier BERT model;
inputting data into a BERT model, and learning and optimizing the BERT model by adjusting model parameters;
behavior word labeling is carried out on the crawled text data in a crowdsourcing mode, true value reasoning is carried out, and the obtained data are divided into training data and testing data according to the proportion;
and inputting the data into a behavior word recognition model BERT + CRF model, and learning and optimizing the behavior word recognition BERT + CRF model by adjusting model parameters.
6. The system according to claim 5, wherein the user mental health monitoring system based on heterogeneous graph,
the method for analyzing the emotional tendency of the user through the time sequence personality element structure according to the five personality portrait of the user and detecting the psychological state of the user to realize the accurate intervention of the psychological health of the user in the psychological health intervention module specifically comprises the following steps:
visually presenting the analysis result of each phrase to a user, enabling the user to evaluate the analysis result, feeding back the user evaluation to an emotion classifier BERT model and a behavior word recognition BERT + CRF model, further improving the emotion classifier BERT model and the behavior word recognition BERT + CRF model, and updating the quintuple portrait of the user;
if the user is satisfied with the analysis result, evaluating the current overall psychological state of the user; if the user is in a negative abnormal psychological state in the same day, providing corresponding suggestions to help the user eliminate the negative psychological state;
and analyzing whether the user state is improved or not according to the real-time physiological information provided by the wearable device, and giving a prompt.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110730536.9A CN113345590B (en) | 2021-06-29 | 2021-06-29 | User mental health monitoring method and system based on heterogeneous graph |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110730536.9A CN113345590B (en) | 2021-06-29 | 2021-06-29 | User mental health monitoring method and system based on heterogeneous graph |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113345590A CN113345590A (en) | 2021-09-03 |
| CN113345590B true CN113345590B (en) | 2022-12-16 |
Family
ID=77481566
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110730536.9A Active CN113345590B (en) | 2021-06-29 | 2021-06-29 | User mental health monitoring method and system based on heterogeneous graph |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113345590B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114445043B (en) * | 2022-01-26 | 2022-12-16 | 安徽大学 | Open ecological cloud ERP-based heterogeneous graph user demand accurate discovery method and system |
| CN115269845B (en) * | 2022-08-01 | 2023-06-23 | 安徽大学 | Network alignment method and system based on social network user personality |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8768869B1 (en) * | 2012-04-19 | 2014-07-01 | The United States Of America As Represented By The Secretary Of The Navy | BRIAN: a basic regimen for intelligent analysis using networks |
| CN106919673A (en) * | 2017-02-21 | 2017-07-04 | 浙江工商大学 | Text mood analysis system based on deep learning |
| CN109635207A (en) * | 2018-12-18 | 2019-04-16 | 上海海事大学 | A kind of social network user personality prediction technique based on Chinese text analysis |
| CN109903851A (en) * | 2019-01-24 | 2019-06-18 | 暨南大学 | A kind of automatic Observation technology of the psychological abnormality variation based on social networks |
| CN110009455A (en) * | 2019-04-02 | 2019-07-12 | 长安大学 | It is a kind of based on network representation study net about share out administrative staff's matching process |
| CN111966800A (en) * | 2020-07-27 | 2020-11-20 | 腾讯科技（深圳）有限公司 | Emotional dialogue generation method and device and emotional dialogue model training method and device |
| CN112185493A (en) * | 2020-08-26 | 2021-01-05 | 山东大学 | Personality preference diagnosis device and project recommendation system based on same |
| CN112948541A (en) * | 2021-02-01 | 2021-06-11 | 华南理工大学 | Financial news text emotional tendency analysis method based on graph convolution network |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108805089B (en) * | 2018-06-14 | 2021-06-29 | 南京云思创智信息科技有限公司 | Multi-modal-based emotion recognition method |
| KR102215584B1 (en) * | 2020-03-23 | 2021-02-10 | 최선희 | Method of analyzing user mentality by analyzing sentence and server performing the same |
| CN112417172B (en) * | 2020-11-23 | 2021-09-07 | 东北大学 | Construction and display method of multi-modal emotion knowledge graph |
| CN114445043B (en) * | 2022-01-26 | 2022-12-16 | 安徽大学 | Open ecological cloud ERP-based heterogeneous graph user demand accurate discovery method and system |
-
2021
- 2021-06-29 CN CN202110730536.9A patent/CN113345590B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8768869B1 (en) * | 2012-04-19 | 2014-07-01 | The United States Of America As Represented By The Secretary Of The Navy | BRIAN: a basic regimen for intelligent analysis using networks |
| CN106919673A (en) * | 2017-02-21 | 2017-07-04 | 浙江工商大学 | Text mood analysis system based on deep learning |
| CN109635207A (en) * | 2018-12-18 | 2019-04-16 | 上海海事大学 | A kind of social network user personality prediction technique based on Chinese text analysis |
| CN109903851A (en) * | 2019-01-24 | 2019-06-18 | 暨南大学 | A kind of automatic Observation technology of the psychological abnormality variation based on social networks |
| CN110009455A (en) * | 2019-04-02 | 2019-07-12 | 长安大学 | It is a kind of based on network representation study net about share out administrative staff's matching process |
| CN111966800A (en) * | 2020-07-27 | 2020-11-20 | 腾讯科技（深圳）有限公司 | Emotional dialogue generation method and device and emotional dialogue model training method and device |
| CN112185493A (en) * | 2020-08-26 | 2021-01-05 | 山东大学 | Personality preference diagnosis device and project recommendation system based on same |
| CN112948541A (en) * | 2021-02-01 | 2021-06-11 | 华南理工大学 | Financial news text emotional tendency analysis method based on graph convolution network |
Non-Patent Citations (3)
| Title |
|---|
| DSM系统中人格障碍诊断的新进展;孙青青等;《医学与哲学(B)》;20181123(第11期);第79-83页 * |
| Enhancing evidence-based medicine with natural language argumentative analysis of clinical trials;Tobias Mayer 等;《Artificial Intelligence in Medicine》;20210507;第1-15页 * |
| 基于用户社交网络数据的人格倾向性分析及预测模型的建立;娜迪热等;《电脑知识与技术》;20180305(第07期);第11-16页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113345590A (en) | 2021-09-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109165692B (en) | User character prediction device and method based on weak supervised learning | |
| CN113345590B (en) | User mental health monitoring method and system based on heterogeneous graph | |
| JP6835204B2 (en) | Learning material recommendation method, learning material recommendation device and learning material recommendation program | |
| CN111814454A (en) | Multi-modal network spoofing detection model on social network | |
| CN111898384A (en) | Text emotion recognition method and device, storage medium and electronic equipment | |
| CN112927782B (en) | Heart health state early warning system based on text emotion analysis | |
| CN114429817A (en) | Method and system for predicting health risk of old people and electronic equipment | |
| CN113571184A (en) | Dialogue interaction design method and system for mental health assessment | |
| CN115270807A (en) | Method, device and equipment for judging emotional tendency of network user and storage medium | |
| Qiu et al. | Predicting the quality of answers with less bias in online health question answering communities | |
| CN109979592A (en) | Mental health method for early warning, user terminal, server and system | |
| CN112735564A (en) | Mental health state prediction method, mental health state prediction apparatus, mental health state prediction medium, and computer program product | |
| CN106202052A (en) | Adolescent psychology pressure range and pressure source event perception method and device thereof | |
| Kalatzis et al. | Interactive dimensionality reduction for improving patient adherence in remote health monitoring | |
| Rezgui et al. | Theoretical formulas of semantic measure: a survey | |
| CN114022698A (en) | Multi-tag behavior identification method and device based on binary tree structure | |
| Yin et al. | Stroke risk prediction: Comparing different sampling algorithms | |
| Batalha et al. | Gender and personality: Beyond gender stereotypes to social identity and the dynamics of social change | |
| Kolla et al. | LLM-Mod: Can Large Language Models Assist Content Moderation? | |
| Singh et al. | Malaria parasite recognition in thin blood smear images using squeeze and excitation networks | |
| Zhou et al. | Predicting mental disorder from noisy questionnaires: an anomaly detection approach based on keyword extraction and machine learning techniques | |
| Shah et al. | Machine Learning-Based Approaches for Early Prediction of Depression | |
| CN115269845B (en) | Network alignment method and system based on social network user personality | |
| KR102205955B1 (en) | System and methods for evaluating and deducing interviewer's ability to use interview game | |
| CN112101033B (en) | Emotion analysis method and device for automobile public praise |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |