JP2020021226A - QOL analysis device and program - Google Patents

Abstract

To analyze association between events about user's utterances in the viewpoint of a QOL evaluation.SOLUTION: An event acquisition unit 12 acquires an event about target user's utterances, and acquires context information representing a context that causes the event, a QOL estimation unit 14 estimates an evaluation item to which the event corresponds among a plurality of respective evaluation items related to a QOL (Quality of Life) representing the quality of life of the target user, and the evaluation value of each evaluation item in each event acquired by the event acquisition unit 12, an association analysis unit 16 analyzes association between events by using an association analysis model 26 for outputting an estimation result by the QOL estimation unit 14, contents shown by the events, the context information, and an association degree between the events, and an output unit 18 outputs an analysis result by the association analysis unit 16.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a QOL analysis device and a QOL analysis program, and more particularly, to a QOL analysis device and a QOL analysis program for analyzing a relation between events related to a QOL.

  2. Description of the Related Art Conventionally, a technology for measuring a quality of life (QOL) representing quality of life has been known.

  For example, a QOL measurement device capable of naturally measuring a QOL by estimating a QOL from an utterance spontaneously uttered by an elderly person who estimates a QOL from a listening dialogue between the elderly and a system has been proposed. . In this apparatus, an input of a user utterance is received, and an event included in the user utterance and a QOL of an event included in the user utterance are determined for each of a plurality of predetermined QOL evaluation items related to physical items, psychological items, social items, and environmental items. The relevance is calculated by comparing the event with the event included in the evaluation item. Then, based on the calculated relevance, the QOL is measured based on the QOL scores for each of the physical item, the psychological item, the social item, and the environmental item (see Patent Document 1).

  In addition, there has been proposed a home health condition remote diagnosis method in which the health status of an elderly living alone is remotely diagnosed at a management center and signs of deterioration of the health status are grasped at a mild stage. In this method, in a house where the elderly person lives alone, a plurality of different living data on the elderly person are measured and transmitted to the management center. The living data includes the standard deviation of the number of actuations of the mat on the treading mat, which measures the behavior pattern (behavior data) of the elderly provided at the entrances of bedrooms, toilets, entrances, living rooms, kitchens, bathrooms, washrooms, etc. Information value and correction value, mat operation frequency, movement frequency, movement speed, number of toilets, and the like (see Patent Document 2).

  Conventionally, a technique for analyzing the association between events is known.

  For example, an event analysis device capable of grasping the causal relationship between events has been proposed. In this device, an event log including an occurrence date and time, a device identifier, and an event type identifier is collected, and the event log is displayed in chronological order as to whether or not an occurrence has occurred for each device event obtained by combining the device identifier and the event type identifier. Converted to a matrix and saved, the event matrix is divided into blocks of a predetermined reference time width, and by determining whether or not each device event has occurred for each block, the conditional probability between device events is calculated. A Bayesian network is constructed, and by using the constructed Bayesian network, a device event considered to be a cause of a device event to be analyzed or a device event considered to occur later is output (see Patent Document 3).

  In addition, a context analysis device for analyzing a relationship between events has been proposed. In this device, using a combination of a predicate having a shared term, word meaning identification information for specifying the meaning of the predicate, and case type information indicating the type of the case of the shared term, the probability of occurrence of an event sequence is calculated, A context analysis is performed on the document to be analyzed (see Patent Document 4).

JP-A-2017-102774 JP 2003-275181 A JP 2014-127093 A International Publication No. 14/002172

  However, the technique described in Patent Literature 1 cannot analyze the association between events from the viewpoint of QOL evaluation.

  In addition, in the technology described in Patent Document 2, it is possible to measure the life activity of the elderly as one of the evaluation items of QOL, but to estimate the evaluation value of the evaluation item of QOL other than the life activity. It is not possible. In addition, even if only the evaluation item of the living activity is used, it is not possible to analyze the association between the events related to the living activity. For example, even if an event that the user is not sufficiently asleep can be grasped based on the evaluation value estimated for each event, it is not possible to specify what other events are not sufficiently asleep related to the event.

  Further, according to the technology described in Patent Document 3, it is possible to analyze the association between events in an environment in a limited environment such as a factory, using the event occurrence time. However, since events that occur in the life of a person such as an elderly person occur in various environments, the relationship between the events cannot be specified only by time.

  Further, with the technology described in Patent Document 4, it is possible to analyze that there is a relationship between events having the same subject or a closely related verb (event), but analysis according to the QOL evaluation value or category (evaluation item) is performed. Can not do.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a QOL analysis device and a program capable of analyzing the association between events related to a user's utterance from the viewpoint of QOL evaluation.

  In order to achieve the above object, a QOL analysis device according to the present invention includes: an event acquisition unit that acquires an event related to a user's utterance and acquires context information indicating a context that led to the occurrence of the event; For each event acquired by the acquisition unit, of a plurality of items relating to quality of life (QOL) representing the quality of life of the user, a QOL estimation for estimating an item to which the event corresponds and an evaluation value for each item And an association analysis unit that analyzes the association between the events using an analysis model that outputs an estimation result by the QOL estimation unit, the content indicated by the event, the context information, and the degree of association between the events, And an output unit that outputs an analysis result by the related analysis unit.

  In the QOL analysis device according to the present invention, the event acquisition unit acquires an event related to the utterance of the user, acquires context information indicating a context that led to the occurrence of the event, and the QOL estimation unit acquires the context information by the event acquisition unit. Of each of the plurality of QOL-related items representing the quality of life of the user for each of the events, the item to which the event applies and the evaluation value of each item are estimated, and the relation analysis unit estimates the result by the QOL estimation unit. Using an analysis model that outputs the contents indicated by the event, the context information, and the degree of association between the events, the association between the events is analyzed, and the output unit outputs the analysis result by the association analysis unit.

  As described above, since the estimation result of the QOL evaluation is used for the analysis between the events, the association between the events related to the utterance of the user can be analyzed from the viewpoint of the QOL evaluation. As a result, it is possible to look back on the relationship between events related to utterances that the user has not noticed, which may lead to an improvement in QOL.

  Further, the analysis model receives as input the evaluation value of the QOL and information including at least one of the items to which the event applies, and outputs a degree of association between the events. The relevance may be set higher as the difference between the QOL evaluation values between events is smaller. Further, the relevance may be set higher as the degree of coincidence between items corresponding to the event is higher.

  Further, the analysis model further receives information including at least one of an occurrence time of the event, a subject of an operation or a state indicated by the event, and an operation or a state indicated by the event, and a degree of association between the events. Is output. The relevance may be set higher as the difference between the occurrence times of the events between the events is smaller. Further, the degree of association may be set higher as the similarity of the subject of the operation or the state between events is higher. Further, the degree of association may be set higher as the degree of similarity of the operation or state between events is higher.

  In addition, the output unit can output the analysis result to at least one of an information processing terminal used by the user and an information processing terminal used by a person associated with the user. As a result, the user can look back on his / her own QOL, which may lead to an improvement in QOL. Further, the related person can look back on the user's QOL, and can provide information that triggers the user to improve the user's QOL. When the analysis result is output to both, the analysis result can be shared between the user and the related parties.

  Further, the QOL analysis device according to the present invention is configured such that, when the relevance between events indicated by the analysis result output by the output unit is corrected by the user or the related person, the analysis is performed based on the corrected content. It can be configured to further include a re-learning unit that re-learns the model. Thereby, an analysis model suitable for the user can be constructed.

  In addition, the event acquisition unit, based on the operation or state indicated by the event, and the context information, for each of the acquired events, an event indicating the current or past situation of the user, or an event indicating a request The output unit assigns attribute information of any of the events acquired by the event acquisition unit as the analysis result and the degree of association between the event and the one of the events by the association analysis unit is a predetermined value or more. It is possible to generate and output screen data for displaying another event that has been analyzed as being distinguished according to the attribute information given to the other event. As a result, useful information for improving the QOL of the user can be output in a more easily graspable state.

  A QOL analysis program according to the present invention is a program for causing a computer to function as each unit of the above-described QOL analysis device.

  ADVANTAGE OF THE INVENTION According to the QOL analysis apparatus and program of this invention, the relationship between the events regarding a user's utterance can be analyzed from a viewpoint of QOL evaluation by using the estimation result about QOL evaluation for analysis between events.

It is a functional block diagram of a QOL analysis device concerning a 1st and 3rd embodiment. FIG. 3 is a diagram illustrating an example of an event database according to the first embodiment. It is a figure for demonstrating the definition of QOL. It is a figure for explaining processing of a QOL estimation part. It is a figure showing an example of an estimation result by a QOL estimation part. It is a figure showing an example of an estimation result by a QOL estimation part. It is a figure which shows an example of the analysis result by an association analysis part notionally. FIG. 7 is a diagram illustrating an example of an analysis result screen according to the first embodiment. It is a flowchart which shows an example of an event acquisition process. 9 is a flowchart illustrating an example of an association analysis process according to the first and second embodiments. It is a functional block diagram of a QOL analysis device concerning a 2nd embodiment. FIG. 9 is a diagram for explaining correction of an analysis result by a user. It is a flowchart which shows an example of a relearning process. FIG. 14 is a diagram illustrating an example of an event database according to the third embodiment. It is a figure showing an example of an analysis result screen in a 3rd embodiment. 15 is a flowchart illustrating an example of a relation analysis process according to the third embodiment.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

<First embodiment>
In the first embodiment, as shown in FIG. 1, a QOL (Quality of Life) measurement target representing a quality of life (hereinafter referred to as “target user”) uses a QOL analysis on an information processing terminal 40 used by the user. The case where the device 10 is mounted will be described as an example. That is, the QOL analysis device 10 functions as an application that operates on the information processing terminal 40.

  The information processing terminal 40 includes the QOL analysis device 10, the input unit 42, and the display unit 44, and can be realized by, for example, a smartphone, a tablet terminal, a personal computer, or the like.

  The input unit 42 includes a microphone, a keyboard, a mouse, and the like for inputting information such as voice and text data. Further, the input unit 42 includes a communication interface with an external device such as an external microphone, a conversation robot, and other applications operating on the information processing terminal 40. The external microphone is installed in the home of the target user, a clock or glasses worn by the target user, a car used by the target user, a wheelchair, an interactive robot, or the like.

  For example, the input unit 42 inputs to the QOL analysis device 10 voice data detected by the microphone, such as the utterance of the target user, the utterance of another person other than the target user such as a cohabitant, and the utterance of the interactive robot. In addition, for example, the input unit 42 acquires text data or the like input to another application such as an email or an SNS (Social Networking Service) operating on the information processing terminal 40 from an email or an SNS log, and performs QOL analysis. Input to the device 10. In addition, for example, the input unit 42 acquires a conversation log of the conversation robot and inputs the acquired conversation log to the QOL analysis device 10.

  The display unit 44 is realized by a display or the like on which various information including the analysis result of the QOL analysis device 10 is displayed.

  Note that the input unit 42 and the display unit 44 may be integrally configured by a touch panel display or the like.

  Further, the information processing terminal 40 includes a CPU, a RAM, and a ROM storing a QOL analysis processing program for executing a QOL analysis processing routine including an event acquisition processing and a related analysis processing described later. By executing the processing program, it functions as the QOL analyzer 10.

  The QOL analysis device 10 functionally includes an event acquisition unit 12, a QOL estimation unit 14, a relation analysis unit 16, and an output unit 18, as shown in FIG. Further, an event database (DB) 22, a QOL estimation model 24, and a related analysis model 26 are stored in a predetermined storage area of the QOL analysis device 10.

  The event acquisition unit 12 acquires an event related to the utterance of the target user from information input via the input unit 42, and acquires context information indicating a context in which the event occurred.

  The event related to the utterance of the target user is an utterance by the target user and an input to an application such as a mail or SNS by the target user. The target user's utterances include the target user's solitude, the target user's utterance in a conversation with another person other than the target user such as a roommate, the target user's utterance in a telephone conversation, and the target user's utterance in a conversation with a dialog robot. Utterances and the like are included.

  Also, the context information includes the utterance of another person such as a cohabitant or a telephone partner who has led to the utterance of the target user, or the other user when the input to the mail or SNS of the target user is a reply to the other person. Of the mail or SNS input. In addition, the context information includes the classification of the other person (wife, husband, child, grandchild, friend, doctor, etc.), and information indicating who the utterance or input of the target user is directed to (self-talk, Family, friends, doctors, SNS (unspecified majority), etc.). Further, the context information may include the date and time, the current location (address information, location information such as home and park), and the attributes of the target user (age, chronic illness, hobby, family structure, etc.).

  Specifically, the event acquisition unit 12 acquires information input from the input unit 42, and temporarily stores the information as history information in a predetermined storage area that stores information for the latest predetermined time. The “predetermined time” for storing the history information preliminarily defines a time required for extracting the context information about the event.

  Further, each time the information input from the input unit 42 is obtained, the event obtaining unit 12 determines whether an event related to the utterance of the target user has occurred. For example, if the acquired information is voice data indicating the utterance of the target user, the event acquisition unit 12 determines that an event related to the utterance of the target user has occurred. The target user's utterance can be determined based on the distance between the speaker and the microphone, the direction of the sound source (speaker) with respect to the microphone, and collation with the characteristics of the voice data of the target user's utterance registered in advance. it can. Further, for example, when a mail or SNS log is acquired, the event acquisition unit 12 determines that an event related to the utterance of the target user has occurred if the sender or the contributor is the target user.

  When determining that an event related to the utterance of the target user has occurred, the event acquisition unit 12 acquires information acquired from the input unit 42 as the content of the event. If the information obtained from the input unit 42 is voice data, voice recognition processing is performed to obtain text data indicating the content of the event.

  In addition, the event acquisition unit 12 acquires history information for a predetermined time from a predetermined storage area, performs voice recognition processing when the history information is voice data, and then performs morphological analysis on the history information. , Syntax analysis, semantic analysis, etc., to obtain context information. The classification of the other person (wife, husband, child, grandchild, friend, doctor, etc.) included in the context information is, for example, the name included in the utterance or e-mail, the notification of the caller ID of the telephone, the reply by e-mail or SNS. It can be specified from information such as the destination.

  In addition, the event acquisition unit 12 performs morphological analysis, syntax analysis, semantic analysis, and the like on the text data indicating the content of the event, and specifies the operation or state indicated by the event and its subject using context information. .

  The event acquisition unit 12 uses the time of occurrence of the event, the content of the event (text data), the acquired context information, the operation or state indicated by the specified event, and information on the subject as event information, and identifies the event. Is stored, for example, in the event DB 22 as shown in FIG. Note that the event occurrence time may be the time when the event acquisition unit 12 acquires the information from the input unit 42 or the time when the input unit 42 acquires the information.

  The QOL estimating unit 14 calculates, for each event acquired by the event acquiring unit 12, an evaluation item to which the event applies and an evaluation value for each evaluation item among a plurality of evaluation items related to QOL indicating the quality of life of the target user. presume.

  In the present embodiment, six items shown in FIG. 3 are used as QOL evaluation items. In the example of FIG. 3, “item number” is assigned to each evaluation item, and hereinafter, the evaluation item of item number i is also described as “evaluation item i”. In addition, each evaluation item is classified into two subclasses: a case where the classification of the QOL evaluation value is “positive” and a case where the classification is “negative”. The QOL evaluation items were determined with reference to Reference Document 1.

Reference 1: Hisashi Ota, Hiroshi Haga, Hisao Nagata, Kiyoji Tanaka, Kiyoshi Maeda, Toshiro Takesaki, Nao Seki, Yasuo Oyama, Yoshiko Nakanishi, Kazuko Ishikawa, "Developing a QOL Questionnaire for the Elderly in the Community Evaluation ", Nihon Koei Magazine, Vol. 48, No. 4, 2001.

  Specifically, the QOL estimating unit 14 estimates an evaluation value of each event for each QOL evaluation item using each event information stored in the event DB 22 and the QOL estimation model 24. The QOL estimation model 24 is a model learned using, as learning data, an event relating to the utterance of the target user to which the evaluation value of the correct answer for each evaluation item has been given. This is a model that outputs values.

  For example, as shown in FIG. 4, the QOL estimating unit 14 includes, as the QOL estimation model 24, an evaluation item estimator 24A that estimates which evaluation item the input event information corresponds to, and an evaluation for each evaluation item. A model including the evaluation value estimator 24B for estimating the value can be used. In this case, for example, the QOL estimating unit 14 extracts an independent word and an auxiliary word from “event contents” included in the event information, and also extracts an independent word from the context information.

  As shown in FIG. 4, the QOL estimating unit 14 inputs the independent word included in the event content and the independent word included in the context information to the evaluation item estimator 24A, and obtains the estimation result of the evaluation item.

  The evaluation item estimator 24A learns learning data in which “correct output data (evaluation item of the corresponding QOL)” is added to the independent word included in the event content and the independent word included in the context information. Can be built. The learning data to which the correct answer output data is added is manually created, and the evaluation item estimator 24A uses a neural network or the like as a relationship learning model for learning.

  The evaluation item estimation result output by the evaluation item estimator 24A can be, for example, the probability that each event corresponds to each item. For example, when the event "I am sorry for one day every day" is input, the following evaluation item estimation result is output.

It is boring alone every day ⇒
Daily life activity: 0.20
Health satisfaction: 0.40
Human support satisfaction: 0.95
Economic comfort level: 0.35
Mental health: 0.98
Mental vitality: 0.80

  The QOL estimating unit 14 estimates that the target event is an event corresponding to an evaluation item whose probability of each evaluation item indicated by the evaluation item estimation result is equal to or greater than a predetermined threshold. For example, in the above example, if the threshold is set to “0.7”, it is estimated that the event “I am not bored every day alone” corresponds to the evaluation items of human support satisfaction, mental health, and mental vitality. Is done.

  In addition, as shown in FIG. 4, the QOL estimating unit 14 inputs the independent word and the adjunct word included in the event content and the context information to the evaluation value estimator 24B, and obtains an estimation result of the evaluation value.

  The evaluation value estimator 24B learns the learning data in which “correct output data (evaluation items of the corresponding QOL and its evaluation value)” are added to the independent word and the adjunct included in the event content and the context information. Can be constructed. As the evaluation value, an evaluation value of each category of “positive” and “negative” is given according to the definition of QOL shown in FIG. For example, the evaluation value of "positive" is set to five levels of 1, 2, 3, 4, and 5, and the evaluation value of "negative" is set to five levels of -1, -2, -3, -4, and -5. can do. Note that the way of giving the evaluation value is not limited to this example, and may be three values of “positive”, “negative”, and “neutral”, or may be a value in which steps are more finely divided. The learning data provided with the correct answer output data is manually created, and the evaluation value estimator 24B uses a neural network or the like as a relationship learning model for learning.

  The evaluation value estimation result output from the evaluation value estimator 24B may be, for example, any one of the above-described five-level value of “positive” and the value of five-level value of “negative”. Note that the evaluation value output by the evaluation value estimator 24B does not need to be a discrete value as in the case of learning, and may be a continuous value.

  The QOL estimating unit 14 sets the evaluation value of the evaluation item determined to be an evaluation item to which the event does not correspond to “0” based on the evaluation item estimation result among the evaluation value estimation results of the evaluation value estimator 24B. . 5 and 6 show QOL evaluation items to which each event corresponds by the QOL estimating unit 14 and estimation results of the evaluation values. In FIG. 5 and FIG. 6, notations of evaluation items other than life activity, health satisfaction, and human support satisfaction are omitted.

  The association analysis unit 16 analyzes the association between events using the estimation result of the QOL estimation unit 14, the event information of each event, and the association analysis model 26 that outputs the degree of association between events.

  The association analysis model 26 is a model that receives as input a QOL evaluation value of each event and information including at least one of the evaluation items to which each event applies, and outputs a degree of association between the events. Further, at least one of the occurrence time of the event, the subject of the operation or state indicated by the event, and the operation or state indicated by the event may be included in the input information.

  The degree of association between the events output by the association analysis model 26 increases as the difference between the QOL evaluation values between the events decreases, and increases as the degree of matching of the corresponding evaluation item between the events increases. Is smaller, the higher the similarity of the subject of the operation or state between the events, the higher the similarity, and the higher the similarity of the operation or the state between the events, the higher the similarity.

  In the following, the QOL evaluation value of each event, the evaluation item to which each event applies, the time of occurrence of the event, the subject of the operation or state indicated by the event, and the relationship between the events when all of the operation or state indicated by the event are used A calculation example of the degree will be described. In this case, the degree of association between event A and event B is expressed by the following equation (1).

Relevance (A, B) = f (difference in evaluation value, applicable evaluation item, occurrence time,
Subject similarity, action or subject similarity) (1)

  f () is a function using the parameters in (). When calculating the degree of association between event A and event B by comprehensively evaluating the evaluation values of the respective evaluation items of the QOL, the degree of association (A, B, overall) is, for example, the following equation (2) Can be.

Relevance (A, B, total) = a ₁ × 1 / difference in evaluation value
+ A ₂ × proximity of applicable evaluation items
+ A ₃ × 1 / time difference
+ A ₄ × similarity of subject
+ A ₅ × Similarity of operation or state (2)

  In the expression (2), the difference between the evaluation values is the sum of the difference between the evaluation values for each of the evaluation items for the event A and the event B for all the evaluation items. The closeness of the corresponding evaluation item is the number of evaluation items to which both event A and event B correspond. Further, the similarity of the evaluation item estimation result (probability for each evaluation item) output by the evaluation item estimator 24A may be used as the closeness of the corresponding evaluation item. In addition, the similarity of the subject or the similarity of the action or the state is obtained, for example, by converting the “subject” or “action or state” included in each event information into a word vector using word2vec or the like, and calculating the cosine similarity between the word vectors. Can be used.

  When the relevance between event A and event B is calculated for each QOL evaluation item, the relevance (A, B, evaluation item i) can be expressed by the following equation (3), for example.

Relevance (A, B, evaluation item i) = closeness of applicable evaluation item ×
(A ₁ × 1 / difference in evaluation value
+ A ₃ × 1 / time difference
+ A ₄ × similarity of subject
+ A ₅ × operation or state similarity) (3)

  In the formula (3), the proximity of the corresponding evaluation item is 1 when both the event A and the event B correspond to the evaluation item i, and is 0 when any one of them does not correspond to the evaluation item i. The difference between the evaluation values is a difference between the evaluation values of the evaluation items i of the event A and the event B. Others are the same as the equation (2).

In the above equations (2) and (3), a _{1 to} a ₅ are coefficients. Related analytical model 26, the two events information can be constructed by deriving a ₁ ~a ₅ by learning using learning data assigned relevance of answers between that event.

  FIG. 7 conceptually shows an example of an analysis result by the association analysis unit 16. In the example of FIG. 7, the association between events is represented for each evaluation item in a graph format in which nodes corresponding to the events are connected by an edge having a thickness corresponding to the degree of association between the events.

  The output unit 18 outputs an analysis result by the association analysis unit 16. For example, the output unit 18 generates screen data for displaying the analysis result as a graph as shown in FIG. 7 and outputs the screen data to the display unit 44. The output unit 18 may also display the estimation result by the QOL estimation unit 14 together.

  FIG. 8 shows an example of the analysis result screen 50 displayed on the display unit 44. In the example of FIG. 8, the analysis result screen 50 includes an event-related area 52 in which the association between events is displayed, an evaluation item selection area 54 for selecting an evaluation item for displaying an evaluation value, and an evaluation item selection area 54. And an evaluation value area 56 in which the evaluation value of each event with respect to the evaluation item selected in is displayed.

  The event-related area 52 includes, for example, a pull-down menu from which any of the evaluation items or the comprehensive evaluation can be selected, and the relationship between the events analyzed for the selected evaluation item or the comprehensive evaluation is, for example, as shown in FIG. It is displayed as a graph. In the evaluation item selection area 54, for example, one or more evaluation items or a total evaluation value (sum of evaluation values for each evaluation item for each event) can be selected by a check box or the like.

  The evaluation value area 56 displays, for example, a graph in which the horizontal axis represents the event ID of an event that occurred within a predetermined period, and the vertical axis represents the QOL evaluation value estimated for the event indicated by the event ID. The predetermined period can be set as appropriate, such as a month including today, the latest month, the latest week, and the like.

  As shown by P in FIG. 8, when a point on the graph displayed in the evaluation value area 56 is selected, an event corresponding to the point and an event related to the event are displayed in the event-related area 52. May be displayed.

  Next, the operation of the QOL analysis device 10 according to the first embodiment will be described.

  When the information acquired by the input unit 42 is input to the QOL analysis device 10, the event acquisition process illustrated in FIG. In addition, in an application operating as the QOL analysis device 10 on the information processing terminal 40, when the display of the analysis result screen is instructed by the target user, the QOL analysis device 10 executes the related analysis process illustrated in FIG.

  First, the event acquisition processing shown in FIG. 9 will be described.

  In step S12, the event acquisition unit 12 acquires information input from the input unit 42, and temporarily stores the information as history information in a predetermined storage area that stores information for the latest predetermined time.

  Next, in step S14, the event acquisition unit 12 determines whether an event related to the utterance of the target user has occurred based on the information acquired in step S12. When an event has occurred, the process proceeds to step S16, and when no event has occurred, the process returns to step S12.

  In step S16, the event acquisition unit 12 acquires the information acquired from the input unit 42 as the content of the event. If the information obtained from the input unit 42 is voice data, voice recognition processing is performed to obtain text data indicating the content of the event.

  Next, in step S18, the event acquisition unit 12 acquires history information for a predetermined time from a predetermined storage area, performs voice recognition processing when the history information is voice data, On the other hand, morphological analysis, syntax analysis, semantic analysis, and the like are performed to obtain context information.

  Next, in step S20, the event acquisition unit 12 performs morphological analysis, syntax analysis, semantic analysis, and the like on the text data indicating the content of the event, and also uses the context information to determine the operation or state indicated by the event. Identify the subject.

  Next, in step S22, the event acquiring unit 12 specifies the time of occurrence of the event, the content (text data) of the event acquired in step S16, the context information acquired in step S18, and the context information acquired in step S20. The operation or state indicated by the event and the information on the subject thereof are used as event information, and an event ID, which is identification information of the event, is assigned to the event information and stored in, for example, the event DB 22 as shown in FIG. Then, the process returns to step S12.

  Next, the association analysis processing shown in FIG. 10 will be described.

  In step S32, the QOL estimating unit 14 acquires, from the event DB 22, event information in which the “occurrence time” is included in a predetermined period such as, for example, a month including today, the latest month, the latest week, and the like.

  Next, in step S34, the QOL estimating unit 14 estimates the evaluation value of each event in the QOL evaluation item using each of the event information acquired in step S32 and the QOL estimation model 24.

  Next, in step S36, the association analysis unit 16 selects an unselected event pair from the events for which the QOL evaluation value has been estimated in step S34.

  Next, in step S38, the association analysis unit 16 uses the comprehensive evaluation value of the QOL estimated in step S34, the event information of each event acquired in step S32, and the Calculate the relevance.

  Next, in step S40, the relation analysis unit 16 uses the evaluation value for each QOL evaluation item estimated in step S34, the event information of each event acquired in step S32, and the relation analysis model 26, Calculate the degree of association between events.

  Next, in step S42, the association analysis unit 16 determines whether or not an unselected event pair exists in the events for which the QOL evaluation value has been estimated in step S34. If an unselected event pair exists, the process returns to step S36. If all pairs have been selected, the process proceeds to step S44.

  In step S44, the output unit 18 displays an analysis result screen including the estimation result of the evaluation value for each QOL evaluation item estimated in step S34 and the degree of association between the events calculated in steps S38 and S40. The screen data is generated and output to the display unit 44, and the related analysis processing ends.

  Thereby, an analysis result screen 50 as shown in FIG. 8, for example, is displayed on the display unit 44.

  As described above, according to the QOL analysis device according to the first embodiment, the evaluation value for each evaluation item of the QOL estimated for the event related to the utterance of the user, the context representing the context in which the event occurred Since the association between the events is analyzed using the information and the association analysis model that outputs the degree of association between the events, the association between the events related to the utterance of the user can be analyzed from the viewpoint of QOL evaluation.

  As described above, by being able to analyze the association between events related to the utterance of the user from the viewpoint of QOL evaluation, for example, "Eating nothing tastes good" is related to "lonely" And the fact that "stiff shoulders" is related to "lonely loneliness" and so on. It is expected to be a trigger. In addition, by sharing the analysis result with another person (related user), for example, the related user performs an action such as “I'm going to play on the weekend”, so that the QOL of the target user is expected to be improved.

<Second embodiment>
Next, a second embodiment will be described. In the QOL analysis device and the information processing terminal according to the second embodiment, the same components as those of the QOL analysis device 10 and the information processing terminal 40 according to the first embodiment are denoted by the same reference numerals and detailed description. Is omitted.

  As shown in FIG. 11, a QOL analysis system 200 according to the second embodiment includes a QOL analysis device 210, an information processing terminal 240A used by a target user, and an information processing terminal 240B used by a related user. It consists of. The QOL analysis device 210 and each of the information processing terminals 240A and 240B are connected via the Internet or the like and can communicate with each other. Note that the related user is a related person associated with the target user, such as a family, a friend, a specialist (a counselor, a care manager, a doctor, a caregiver, etc.) of the target user living away.

  Each of the information processing terminals 240A and 240B includes an input unit 42 and a display unit 44, and can be realized by, for example, a smartphone, a tablet terminal, a personal computer, or the like. Although the information processing terminal 240A and the information processing terminal 240B have the same configuration, different reference numerals are given for convenience of explanation.

  An analysis result screen 50 as shown in FIG. 8, for example, is displayed on the display unit 44 of the information processing terminal 240A used by the target user, as in the first embodiment. Similarly, the display unit 44 of the information processing terminal 240B used by the related user also displays the analysis result screen 50 of the target user associated with the related user.

  Each of the target user and the related user can correct the association between events and the evaluation value of the QOL with respect to the displayed analysis result screen 50. For example, as shown in FIG. 12, a new event is added to the event-related area 52 of the analysis result screen 50, and the thickness is designated so as to associate the added event with another event already displayed. By drawing an edge, the degree of association between the added event and another event can be specified. Further, the relevance may be corrected by changing the thickness of the edge between the already displayed events. Also, for example, by dragging the graph displayed in the evaluation value area 56 of the analysis result screen 50, the QOL evaluation value of each event can be corrected.

  When at least one of the association between the events and the evaluation value of the QOL is corrected, the correction information is transmitted from the information processing terminals 240A and 240B to the QOL analysis device 210.

  The QOL analysis device 210 includes a CPU, a RAM, and a ROM that stores a QOL analysis processing program for executing a QOL analysis processing routine including an event acquisition process, an association analysis process, and a relearning process, which will be described later. The QOL analysis device 210 can be realized by, for example, a server device, a personal computer, or the like.

  The QOL analysis device 210 functionally includes an event acquisition unit 12, a QOL estimation unit 14, an association analysis unit 16, an output unit 18, and a relearning unit 20, as shown in FIG. In a predetermined storage area of the QOL analysis device 210, an event DB 22, a QOL estimation model 24, and a related analysis model 26 are stored.

  Upon receiving the correction information from the information processing terminal 240A or 240B, the re-learning unit 20 re-learns the QOL estimation model 24 and the related analysis model 26 based on the received correction information.

  Specifically, when the re-learning unit 20 receives the correction information indicating that the event has been added, the relearning unit 20 gives the added event an evaluation value for each QOL evaluation item. For example, the relearning unit 20 acquires the event information of the added event by the event acquiring unit 12, and sets the QOL estimation value estimated by the QOL estimating unit 14 based on the acquired event information to the added event. Can be granted. In addition, the relearning unit 20 may give the evaluation value of the QOL estimated for another event associated with the added event to the added event. If there are a plurality of other events, the average of the QOL evaluation values estimated for each of those events is given to the added event. When there are a plurality of added events, the relearning unit 20 assigns a QOL evaluation value to each of the plurality of events.

  Further, the re-learning unit 20 generates learning data for the QOL estimation model based on the event information of the added event and the evaluation value of each of the QOL evaluation items. Further, the relearning unit 20 performs the association based on the event information of the added event and the other event associated with the added event, and the degree of association specified between the added event and the other event. Generate learning data for the analysis model.

  Further, when the re-learning unit 20 receives the correction information of the relevance between the existing events, the re-learning unit 20 performs a relevance analysis model based on the event information of the event whose relevance is corrected and the corrected relevance. Generate learning data for

  When the re-learning unit 20 receives the correction information of the evaluation value for each evaluation item of the QOL for any of the events, the relearning unit 20 determines the event information of the event whose evaluation value has been corrected and the event information of each of the corrected evaluation items. , Learning data for the QOL estimation model is generated.

  The re-learning unit 20 re-learns the QOL estimation model 24 using the generated learning data for the QOL estimation model, and re-learns the association analysis model 26 using the generated learning data for the association analysis model.

  Next, the operation of the QOL analysis device 210 according to the second embodiment will be described.

  As in the first embodiment, the QOL analysis device 210 executes the event acquisition processing shown in FIG. 9 and the related analysis processing shown in FIG. When the target user or the related user makes a correction to the analysis result screen 50 displayed on the display unit 44 of the information processing terminal 240A or 240B, the correction information is transmitted from the information processing terminal 240A or 240B to the QOL analysis device 210. Then, the re-learning process shown in FIG.

  In step S62 of the relearning process illustrated in FIG. 13, the relearning unit 20 determines whether an event has been added based on the correction information received from the information processing terminal 240A or 240B. If an event has been added, the process proceeds to step S64. If no event has been added, the process proceeds to step S70.

  In step S64, the relearning unit 20 assigns an evaluation value for each evaluation item of the QOL to the added event.

  Next, in step S66, the re-learning unit 20 generates learning data for the QOL estimation model based on the event information of the added event and the evaluation value for each QOL evaluation item given in step S64. I do.

  Next, in step S68, the relearning unit 20 determines whether the event information of the added event and the other event associated with the added event, the degree of association specified between the added event and the other event, and the like. The learning data for the related analysis model is generated based on.

  Next, in step S70, the re-learning unit 20 determines whether or not the degree of association between the existing events has been corrected based on the correction information received from the information processing terminal 240A or 240B. If the relevance has been corrected, the process proceeds to step S72. If the relevance has not been corrected, the process proceeds to step S74.

  In step S72, the re-learning unit 20 generates learning data for an association analysis model based on the event information of the event whose association degree has been modified and the modified association degree.

  Next, in step S74, it is determined whether the evaluation value of each QOL evaluation item for any event has been corrected. If the evaluation value has been corrected, the process proceeds to step S76. If the evaluation value has not been corrected, the process proceeds to step S78.

  In step S76, the re-learning unit 20 generates learning data for the QOL estimation model based on the event information of the event whose evaluation value has been corrected and the evaluation value for each corrected evaluation item.

  Next, in step S78, the re-learning unit 20 re-learns the QOL estimation model 24 using the learning data for the QOL estimation model generated in at least one of steps S66 and S76. Further, the relearning unit 20 re-learns the association analysis model 26 using the association analysis model learning data generated in at least one of steps S68 and S72, and the relearning process ends.

  After the completion of the re-learning, the association analysis processing shown in FIG. 10 may be executed again.

  As described above, according to the QOL analysis apparatus according to the second embodiment, when a correction to the output analysis result is received from the user, the QOL estimation model and the related analysis model are based on the correction information. Re-learn. Accordingly, the QOL estimation model and the association analysis model are improved so that the estimation result of the QOL suitable for the target user and the analysis result of the association between the events can be obtained.

<Third embodiment>
Next, a third embodiment will be described. In the QOL analysis device and the information processing terminal according to the third embodiment, the same components as those of the QOL analysis device 10 and the information processing terminal 40 according to the first embodiment are denoted by the same reference numerals and detailed description. Is omitted.

  As shown in FIG. 1, the information processing terminal 340 includes a QOL analysis device 310, an input unit 42, and a display unit 44.

  The QOL analysis device 310 functionally includes an event acquisition unit 312, a QOL estimation unit 14, the association analysis unit 16, and an output unit 318, as shown in FIG. In a predetermined storage area of the QOL analysis device 310, an event DB 322, the QOL estimation model 24, and the association analysis model 26 are stored.

  The event acquisition unit 312 acquires an event from information input from the input unit 42 and generates event information based on the input information and history information, similarly to the event acquisition unit 12 of the first embodiment. I do. At this time, the event acquiring unit 312 determines whether the event is an event indicating the current or past situation of the target user or a request of the target user based on the “operation or state” and the “context information” included in the event information. Is added to the event information.

  FIG. 14 shows an example of the event DB 322. The event DB 322 includes an item of “attribute” in addition to each item of the event DB 22 in the first embodiment. In FIG. 14, items other than the event ID, the event content, and the attribute are omitted. In the example of FIG. 14, the “attribute” indicates “situation” when the corresponding event is an event indicating the current or past situation of the target user, and the corresponding event is an event indicating a request of the target user. Stores a “request”.

  The output unit 318 converts any one of the events acquired by the event acquisition unit 312 and another event analyzed by the association analysis unit 16 that the degree of association with the event is equal to or more than a predetermined value into another event. It generates screen data for distinction and display according to the assigned attribute information, and outputs the screen data to the display unit 44.

  FIG. 15 shows an example of the analysis result screen 350 displayed on the display unit 44. In the example of FIG. 15, the analysis result screen 350 includes a latest event area 60 in which the latest event input most recently is displayed, an estimation of an evaluation value for each QOL evaluation item for the latest event, and a relation between the events. And an execution button 62 selected when executing the analysis. The analysis result screen 350 also includes a similar event area 64 in which an event whose relevance to the latest event is equal to or more than a predetermined value and whose “attribute” is “situation” is displayed as “similar search result”, A desired event area 66 in which an event whose relevance to the latest event is equal to or more than a predetermined value and whose “attribute” is “desired” is displayed as “desired search result” is included. The analysis result screen 350 includes an evaluation item selection area 54 and an evaluation value area 56.

  In the graph displayed in the evaluation value area 56, the event displayed in the latest event area 60, the event displayed in the similar event area 64, and the event displayed in the desired event area 66 are displayed with different marks. be able to. In the example of FIG. 15, the event displayed in the latest event area 60 is indicated by a double circle, the event displayed in the similar event area 64 is indicated by a circle, and the event displayed in the desired event area 66 is indicated by a triangle mark. . Further, as shown by Q in FIG. 15, for an event selected from the similar event area 64 or the desired event area 66, the display of the similar event area 64 or the desired event area 66 and the display of the graph of the evaluation value area 56 are performed. May be highlighted.

  Next, the operation of the QOL analysis device 310 according to the third embodiment will be described.

  As in the first embodiment, the event acquisition process shown in FIG. Then, when it is determined that an event related to the utterance of the user has occurred during the activation of the application operating as the QOL analysis device 310 on the information processing terminal 340, the QOL analysis device 310 executes the related analysis process illustrated in FIG. Is done. Note that, in the relation analysis processing according to the third embodiment, the same processing as the relation analysis processing (FIG. 10) according to the first embodiment is denoted by the same step number, and detailed description is omitted.

  In step S332 of the association analysis process shown in FIG. Get event information.

  Next, in step S34, the QOL estimating unit 14 estimates the evaluation value of each event in the QOL evaluation item using each of the event information acquired in step S332 and the QOL estimation model 24.

  Next, in step S336, the relation analysis unit 16 selects one event other than the latest event from among the events for which the QOL evaluation value has been estimated in step S34.

  Next, in step S38, the association analysis unit 16 calculates the degree of association between the latest event and the other event selected in step S336 by using the overall evaluation value of the QOL estimated in step S34, and in step S332. It is calculated using the acquired event information of each event and the related analysis model 26.

  Next, in step S40, the association analysis unit 16 determines the degree of association between the latest event and the other event selected in step S336 by using the evaluation value for each QOL evaluation item estimated in step S34. The calculation is performed using the event information of each event acquired in step S332 and the association analysis model 26.

  Next, in step S342, the association analysis unit 16 determines whether there is another unselected event among the events for which the QOL evaluation value was estimated in step S34. If there is another unselected event, the process returns to step S336. If all other events have been selected, the process proceeds to step S343.

  In step S343, the output unit 318 extracts another event whose relevance with the latest event is equal to or more than a predetermined value, and indicates the extracted event as an attribute of the event, that is, the current or past situation of the target user. The event is classified according to the event or the event indicating the request of the target user.

  In step S344, the output unit 318 generates screen data of an analysis result screen including the estimation result of the QOL evaluation value estimated in step S34 and the events extracted and classified in step S343, and displays the data. The result is output to the unit 44, and the association analysis process ends.

  Thus, an analysis result screen 350 is displayed on the display unit 44, for example, as shown in FIG.

  As described above, according to the QOL analysis device according to the third embodiment, an event related to a certain event is an event indicating the current or past situation of the target user or a request of the target user. Display differently according to the event. As a result, useful information for improving the QOL of the target user can be output in a more easily understandable state.

  The present invention is not limited to the above-described embodiment, and various modifications and applications can be made without departing from the gist of the present invention.

  For example, in each of the above embodiments, the evaluation value of the QOL, the evaluation item to which the event applies, the time at which the event occurs, the subject of the operation or state indicated by the event, and the operation or state indicated by the event are all used for the Although the case of calculating the degree of association has been described, the present invention is not limited to this. What is necessary is to use at least one of the evaluation value of the QOL and the evaluation item to which the event corresponds. In addition, when the event occurrence time, the entity of the operation or state indicated by the event, and the operation or state indicated by the event are used, one or more of them can be used in an appropriate combination. Further, the expression for calculating the degree of association is not limited to the example of the above embodiment.

  Further, in each of the above-described embodiments, the case where the relation analysis model is a model constructed by learning the coefficients of the calculation formula of the degree of relevance has been described, but the calculation formula in which the coefficients are set in advance may be used as a model. Good. Further, the relation analysis model may be a model obtained by learning a model such as a neural network that does not depend on a predetermined calculation formula.

  In each of the above embodiments, the case where the analysis result including both the estimation result of the QOL estimating unit and the analysis result of the related analyzing unit is output is described. However, in the present invention, at least the related analysis result is output. I just need.

  In each of the above embodiments, a case has been described in which, on the analysis result screen, the QOL evaluation value of the event that occurred during the predetermined period is displayed, and the analysis result related to the event that occurred during the predetermined period is displayed. The predetermined period need not be the same. For example, the graph of the evaluation value area of the analysis result screen displays the QOL evaluation value of the event that occurred in the last week, and the event-related area displays the relevance calculated for the event that occurred in the latest month. May be displayed.

  In the second embodiment, the case where both the QOL estimation model and the related analysis model are re-learned has been described. However, in the present invention, it is sufficient that at least the re-training of the related analysis model is performed. Further, the configuration for performing re-learning can be applied to a QOL analysis device mounted on an information processing terminal as in the first embodiment.

  Further, as in the third embodiment, the configuration for distinguishing and displaying related events according to the attribute of the event is different from that in the second embodiment in that the QOL analysis device and the information processing terminal are different from each other. The present invention is applicable even in the case of a separate configuration.

  Further, in the third embodiment, a case has been described in which the latest event input most recently is analyzed for its association with other events. However, any event stored in the event DB is selected. Then, the selected event may be targeted.

10, 210, 310 QOL analysis device 12, 312 Event acquisition unit 14 QOL estimation unit 16 Related analysis unit 18, 318 Output unit 20 Re-learning unit 22, 322 Event DB
24 QOL estimation model 26 Related analysis models 40, 240A, 240B, 340 Information processing terminal 42 Input unit 44 Display unit 50, 350 Analysis result screen 200 QOL analysis system

Claims (12)

An event acquisition unit that acquires an event related to the utterance of the user, and acquires context information indicating a context in which the event has occurred;
For each event acquired by the event acquisition unit, among a plurality of items relating to QOL (Quality of Life) representing the quality of life of the user, an item to which the event corresponds and an evaluation value for each item are estimated. A QOL estimator;
An association analysis unit that analyzes the association between the events using an analysis model that outputs an estimation result by the QOL estimation unit, the content indicated by the event, the context information, and the degree of association between the events;
An output unit that outputs an analysis result by the association analysis unit,
A QOL analysis device including:   The QOL analysis device according to claim 1, wherein the analysis model receives information including at least one of an evaluation value of the QOL and an item to which the event applies, and outputs a degree of association between the events.   The QOL analysis device according to claim 2, wherein the degree of association increases as the difference between the evaluation values of the QOL between events decreases.   4. The QOL analysis device according to claim 2, wherein the degree of association increases as the degree of coincidence between items corresponding to the event among the events increases. 5.   The analysis model further receives information including at least one of the occurrence time of the event, the subject of the operation or state indicated by the event, and the operation or state indicated by the event, and outputs the degree of association between the events. The QOL analyzer according to any one of claims 2 to 4.   The QOL analysis device according to claim 5, wherein the degree of association increases as the difference between the occurrence times of the events among the events decreases.   7. The QOL analysis device according to claim 5, wherein the degree of association increases as the similarity of the subject of the operation or state between events increases.   The QOL analysis device according to any one of claims 5 to 7, wherein the degree of association increases as the degree of similarity of the operation or state between events increases.   9. The output unit according to claim 1, wherein the output unit outputs the analysis result to at least one of an information processing terminal used by the user and an information processing terminal used by a related person associated with the user. 2. The QOL analyzer according to claim 1.   A re-learning unit that re-learns the analysis model based on the corrected content when the user or the related party modifies the degree of association between the events indicated by the analysis result output by the output unit. The QOL analysis device according to claim 9. The event acquisition unit, based on the operation or state indicated by the event, and the context information, for each of the acquired events, the attribute of the event indicating the current or past situation of the user, or the event indicating the request Give information,
The output unit may include, as the analysis result, any one of the events acquired by the event acquisition unit and another association analyzed by the association analysis unit to indicate that the degree of association with any one of the events is equal to or greater than a predetermined value. The QOL analysis device according to any one of claims 1 to 10, wherein screen data for displaying an event according to attribute information assigned to the other event is generated and output.   A QOL analysis program for causing a computer to function as each unit of the QOL analysis device according to any one of claims 1 to 11.