CN116978522A

CN116978522A - Object allocation method, device, electronic equipment, storage medium and program product

Info

Publication number: CN116978522A
Application number: CN202310469446.8A
Authority: CN
Inventors: 陈裕通
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2023-04-24
Filing date: 2023-04-24
Publication date: 2023-10-31

Abstract

The application provides an object allocation method, an object allocation device, electronic equipment, a computer readable storage medium and a computer program product; the method comprises the following steps: acquiring consultation information of a service receiving object, and acquiring a plurality of service providing objects matched with the consultation information; performing ascending sort processing based on the load capacity on the plurality of service providing objects, and taking at least one service providing object with the top sort as a candidate service providing object; for each candidate service providing object, acquiring an allocation strategy weight positively correlated with the evaluation score of the candidate service providing object; acquiring sampling probability positively correlated with the allocation policy weight of each candidate service providing object; and performing weighted random selection processing based on the sampling probability of each candidate service providing object to obtain a target service providing object, and distributing the target service providing object to the service receiving object. According to the application, the evaluation scores of the service providing objects are balanced, and the load of each service providing object is balanced.

Description

Object allocation method, device, electronic equipment, storage medium and program product

Technical Field

The present application relates to artificial intelligence technology, and in particular, to an object allocation method, apparatus, electronic device, computer readable storage medium, and computer program product.

Background

Artificial intelligence (AI, artificial Intelligence) is the theory, method and technique and application system that uses a digital computer or a machine controlled by a digital computer to simulate, extend and extend human intelligence, sense the environment, acquire knowledge and use the knowledge to obtain optimal results.

In the related art, a conventional offline service platform is moved to the internet, so that service efficiency can be improved, for example, pressure of a hospital is shared through a medical consultation platform, for example, legal consultation efficiency is improved through a legal consultation platform, a problem input by a consultant is received in the related art, a service providing object, for example, a doctor or a lawyer is allocated to the consultant, but in the related art, a service providing object with a front recommended ranking is always preferentially allocated, so that load of the service providing object is uneven.

Disclosure of Invention

The embodiment of the application provides an object distribution method, an object distribution device, electronic equipment, a computer readable storage medium and a computer program product, which can balance the load of each service providing object under the condition of considering the evaluation score of the service providing object.

The technical scheme of the embodiment of the application is realized as follows:

the embodiment of the application provides an object distribution method, which comprises the following steps:

acquiring consultation information of a service receiving object, and acquiring a plurality of service providing objects matched with the consultation information;

performing ascending sort processing based on the load capacity on the plurality of service providing objects, and taking at least one service providing object which is sorted to the front as a candidate service providing object;

for each candidate service providing object, acquiring an allocation strategy weight positively correlated with the evaluation score of the candidate service providing object;

acquiring sampling probability positively correlated with the allocation strategy weight of each candidate service providing object;

and executing weighted random selection processing based on the sampling probability of each candidate service providing object to obtain a target service providing object, and distributing the target service providing object to the service receiving object.

An embodiment of the present application provides an object allocation apparatus, including:

the acquisition module is used for acquiring the consultation information of the service receiving object and acquiring a plurality of service providing objects matched with the consultation information;

the load module is used for carrying out ascending sort processing based on the load capacity on the plurality of service providing objects, and taking at least one service providing object which is ranked at the front as a candidate service providing object;

The weight module is used for acquiring an allocation strategy weight positively correlated to the evaluation score of each candidate service providing object;

the sampling module is used for acquiring sampling probability positively correlated with the distribution strategy weight of each candidate service providing object;

and the distribution module is used for executing weighted random selection processing based on the sampling probability of each candidate service providing object to obtain a target service providing object, and distributing the target service providing object to the service receiving object.

In the above scheme, the obtaining module is further configured to obtain a plurality of candidate service providing organizations; performing text matching processing on the consultation information to obtain a first probability of matching the consultation information with each candidate service providing organization; taking the candidate service providing organization with the largest first probability as the service providing organization matched with the consultation information; all the service providing objects belonging to the service providing organization are taken as a plurality of service providing objects matched with the consultation information.

In the above scheme, the load module is further configured to perform an ascending sort process on the plurality of service providing objects based on a load amount of each service providing object, to obtain an ascending sort result; acquiring the service providing objects ranked at the first position from the ascending ranking result, and taking the load capacity of the service providing objects ranked at the first position as target load capacity; and taking the service providing object with the target load capacity in the plurality of service providing objects as the candidate service providing object.

In the above scheme, the load module is further configured to respond to a newly added first consultation order, and add a process to the load capacity of the service providing object corresponding to the first consultation order, so as to obtain a new load capacity of the service providing object corresponding to the first consultation order; in response to canceling the second consultation order, subtracting the loading capacity of the service providing object corresponding to the second consultation order to obtain a new loading capacity of the service providing object corresponding to the second consultation order; responding to the change of a service providing object corresponding to a third consultation order, adding one to the load capacity of a new service providing object corresponding to the third consultation order to obtain the new load capacity of the new service providing object, and subtracting one to the load capacity of an old service providing object corresponding to the third consultation order to obtain the new load capacity of the old service providing object; and in response to completion of the fourth consultation order, subtracting the loading capacity of the service providing object corresponding to the fourth consultation order to obtain the new loading capacity of the service providing object corresponding to the fourth consultation order.

In the above aspect, the weighting module is further configured to perform, for each of the candidate service providing objects, the following processing: acquiring the quality scores of the candidate service providing objects; acquiring the liveness score of the candidate service providing object; obtaining a relevance score of the candidate service providing object; and forming at least one of the quality score, the liveness score and the relevance score into an evaluation score of the candidate service providing object.

In the above scheme, the weight module is further configured to obtain a rank score of the candidate service providing object, and obtain an interaction score of the candidate service providing object; and carrying out weighted summation processing on the grade scores and the interaction scores to obtain the quality scores of the candidate service providing objects.

In the above scheme, the weight module is further configured to obtain an organization grade score of a service providing organization to which the candidate service providing object belongs; acquiring the professional grade scores of the candidate service providing objects; and carrying out weighted summation processing on the organization grade scores and the professional grade scores to obtain grade scores of the candidate service providing objects.

In the above-mentioned scheme, the weighting module is further configured to obtain, from the historical service data of the candidate service providing object, a service score of the candidate service providing object, a consultation number score of the candidate service providing object, a response rate score of the candidate service providing object, and a response speed score of the candidate service providing object; and carrying out weighted summation processing on the service scores of the candidate service providing objects, the consultation quantity scores of the candidate service providing objects, the response rate scores of the candidate service providing objects and the response speed scores of the candidate service providing objects to obtain the interaction scores of the candidate service providing objects.

In the above solution, the weighting module is further configured to obtain a consultation number of each time period from the historical service data of the candidate service providing object, and sum up the consultation numbers of the time periods to obtain a total number of consultations to obtain a score that is positively correlated with the consultation number of the current time period and negatively correlated with the total number of consultations of the current time period as the liveness score corresponding to the current time period; the current time period is a time period to which the time of acquiring the consultation information in a plurality of time periods belongs.

In the above scheme, the weight module is further configured to extract a consultation keyword from the consultation information to form a consultation keyword set; acquiring a tag set of the candidate service providing object, wherein the tag set comprises tags used for representing professional keywords of the candidate service providing object; acquiring a union set between the consultation keyword set and the label set; a relevance score is obtained that is positively correlated to the number of elements of the union and negatively correlated to the number of elements of the set of consultation keywords.

In the above solution, the weighting module is further configured to, when the evaluation score includes at least two of the quality score, the liveness score, and the relevance score, perform weighted summation processing on the at least two scores included, to obtain an allocation policy weight of the candidate service providing object. When the evaluation score includes any one of the quality score, the liveness score, and the correlation score, the any one score is taken as an allocation policy weight of the candidate service providing object.

In the above scheme, the sampling module is further configured to perform summation processing on the allocation policy weights of the plurality of candidate service providing objects, so as to obtain a first summation result; for each candidate service providing object, taking the ratio between the distribution strategy weight of the candidate service providing object and the first summation result as the sampling probability.

acquiring medical consultation information of a medical service receiving object, and acquiring a plurality of medical service providing objects matched with the medical consultation information;

acquiring at least one medical service providing object with the smallest load amount in the plurality of medical service providing objects as a candidate medical service providing object;

for each medical candidate service providing object, acquiring an allocation strategy weight positively correlated with the evaluation score of the medical candidate service providing object;

acquiring sampling probability positively correlated with the distribution strategy weight of each candidate medical service providing object;

and executing weighted random selection processing based on the sampling probability of each candidate medical service providing object to obtain a target medical service providing object, and distributing the target medical service providing object to the medical service receiving object.

The embodiment of the application provides an object distribution device, which comprises:

the medical treatment acquisition module is used for acquiring medical treatment consultation information of a medical treatment service receiving object and acquiring a plurality of medical treatment service providing objects matched with the medical treatment consultation information;

the medical load module is used for acquiring at least one medical service providing object with the smallest load amount in the plurality of medical service providing objects as a candidate medical service providing object;

a medical weight module, configured to obtain, for each of the medical candidate service providing objects, an allocation policy weight positively correlated with an evaluation score of the medical candidate service providing object;

the medical sampling module is used for acquiring sampling probability positively correlated with the distribution strategy weight of each candidate medical service providing object;

and the medical distribution module is used for executing weighted random selection processing based on the sampling probability of each candidate medical service providing object to obtain a target medical service providing object and distributing the target medical service providing object to the medical service receiving object.

An embodiment of the present application provides an electronic device, including:

a memory for storing computer executable instructions;

And the processor is used for realizing the object allocation method provided by the embodiment of the application when executing the computer executable instructions stored in the memory.

The embodiment of the application provides a computer readable storage medium which stores computer executable instructions for realizing the object allocation method provided by the embodiment of the application when being executed by a processor.

The embodiment of the application provides a computer program product, which comprises a computer program or a computer executable instruction, and the computer program or the computer executable instruction realize the object allocation method provided by the embodiment of the application when being executed by a processor.

The embodiment of the application has the following beneficial effects:

according to the embodiment of the application, the consultation information of the service receiving object is obtained, and a plurality of service providing objects matched with the consultation information are obtained; taking at least one service providing object with the smallest load amount in the plurality of service providing objects as a candidate service providing object, so that the load of the service providing objects can be balanced, and determining the distribution strategy weight of each candidate service providing object based on the evaluation score; and acquiring sampling probability positively correlated with the distribution strategy weight of each candidate service providing object, so that the evaluation score of the service providing object is considered when the service providing object is selected, weighting random selection processing is performed on the basis of the sampling probability of each candidate service providing object to obtain a target service providing object, the target service providing object is distributed to the service receiving object, and finally the target service providing object is matched after the evaluation score is considered and belongs to the service providing object with the minimum load capacity, so that the load balance of service providing object distribution is realized.

Drawings

FIG. 1 is a schematic view of an application scenario of an object distribution system according to an embodiment of the present application;

FIG. 2 is a schematic structural diagram of an electronic device for object allocation according to an embodiment of the present application;

fig. 3A to fig. 3E are schematic flow diagrams of an object allocation method according to an embodiment of the present application;

FIG. 4 is a schematic interface diagram of a medical interrogation platform according to an embodiment of the present application;

FIG. 5 is a schematic illustration of doctor loading provided by an embodiment of the application;

FIG. 6 is a schematic diagram of an overall framework provided by an embodiment of the present application;

FIG. 7 is a training flow chart of a department identification model provided by an embodiment of the application;

FIG. 8 is a schematic diagram of a quality score configuration provided by an embodiment of the present application;

FIG. 9 is a diagram of load statistics provided by an embodiment of the present application;

fig. 10 is a schematic flow chart provided in an embodiment of the present application.

Detailed Description

The present application will be further described in detail with reference to the accompanying drawings, for the purpose of making the objects, technical solutions and advantages of the present application more apparent, and the described embodiments should not be construed as limiting the present application, and all other embodiments obtained by those skilled in the art without making any inventive effort are within the scope of the present application.

In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is to be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments and can be combined with one another without conflict.

In the following description, the terms "first", "second", "third" and the like are merely used to distinguish similar objects and do not represent a specific ordering of the objects, it being understood that the "first", "second", "third" may be interchanged with a specific order or sequence, as permitted, to enable embodiments of the application described herein to be practiced otherwise than as illustrated or described herein.

It will be appreciated that in the embodiments of the present application, related data such as user information is involved, and when the embodiments of the present application are applied to specific products or technologies, user permissions or agreements need to be obtained, and the collection, use and processing of related data need to comply with relevant laws and regulations and standards of relevant countries and regions.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of the application only and is not intended to be limiting of the application.

Before describing embodiments of the present application in further detail, the terms and terminology involved in the embodiments of the present application will be described, and the terms and terminology involved in the embodiments of the present application will be used in the following explanation.

1) The bidirectional encoder representation (Bidirectional Encoder Representations from Transformers, BERT) structure based on the converter is a pre-training language model, the pre-training is a concept of migration learning, the pre-training model is a model with strong generalization capability obtained by training by using huge data, when the model is required to be used in a specific scene, for example, medical named entity identification is carried out, only an output layer is required to be modified, incremental training is carried out by using data of a corresponding scene, and the weight of the BERT structure is slightly adjusted.

2) Weighted minimum connection policy: the weighted minimum connection algorithm is based on a minimum connection number scheduling algorithm, and different weights are distributed to each server according to different processing capacities of the servers, so that the server can accept service requests of corresponding weights.

3) Redis: the database is a memory-based database, is suitable for rapidly acquiring data on line, is an open-source log-type and Key-Value database which is written and supported by ANSI C language, can be based on memory and can be persistent, and provides APIs of multiple languages.

In the related art, a conventional offline service platform is moved to the internet, so that service efficiency can be improved, for example, pressure of a hospital is shared through a medical consultation platform, for example, legal consultation efficiency is improved through a legal consultation platform, a problem input by a consultant is received in the related art, a service providing object, for example, a doctor or a lawyer is allocated to the consultant, but in the related art, a service providing object with a front recommended ranking is always preferentially allocated, so that load of the service providing object is uneven. For example, in the related art, an allocation scheme is provided for medical service recommending work, and firstly, the load of TopN doctors in a recommended ranking list is obtained; acquiring a star doctor exceeding a preset work load threshold according to the load of each doctor, and generating a doctor queue to be replaced; the doctor to be replaced and the replaceable doctor are selected for replacement, and the replacement method in the related art is to select the doctor with the smallest service level difference between the candidate doctor and the doctor by providing the proper replaceable doctor for the user allocated with the doctor, so that the load of the doctor is reduced without influencing the service quality.

Applicants have found that the related art in practicing embodiments of the present application may result in TopN doctors having a ease of dispensing opportunities and non-TopN doctors having fewer dispensing opportunities. The more TopN doctors visit, the better the accumulated service data, the higher the probability of being distributed, resulting in the martai effect, which is unfavorable for the overall load balance of the inquiry platform. The related technology only considers the relativity of the doctor of inquiry and the illness state description information, but does not consider the historical service quality condition of the doctor, the doctor with high relativity can not necessarily provide good service quality, and the doctor with high service quality can obtain larger allocation probability. The related art does not consider the work and rest habit of doctors, and is easy to conduct inquiry and dispatch when the doctors rest or are busy, and the doctors are free to reply to the patients in time, so that the waiting time of the patients is long, and poor inquiry experience is brought.

The object allocation method provided by the embodiment of the application can be independently realized by a terminal/server; the terminal and the server may cooperate to perform, for example, the terminal alone bears the following object allocation method, or the terminal transmits the consultation information of the service reception object to the server, and the server executes the object allocation method based on the received consultation information of the service reception object.

The electronic device for object allocation provided by the embodiment of the application can be various types of terminals or servers, wherein the servers can be independent physical servers, can be a server cluster or a distributed system formed by a plurality of physical servers, and can be cloud servers for providing cloud computing services; the terminal may be, but is not limited to, a smart phone, a tablet computer, a notebook computer, a desktop computer, a smart speaker, a smart watch, a vehicle-mounted terminal, etc. The terminal and the server may be directly or indirectly connected through wired or wireless communication, and the present application is not limited herein.

Taking a server as an example, for example, a server cluster deployed in a cloud may be used, an artificial intelligence cloud Service (AI as a Service, AI aas) is opened to a user, a platform splits several types of common AI services and provides independent or packaged services in the cloud, and the Service mode is similar to an AI theme mall, and all users can access one or more artificial intelligence services provided by using the AI aas platform through an application programming interface.

As an example, one of the artificial intelligence cloud services may be a service for distributing objects, that is, a cloud server encapsulates a program for distributing objects provided by the embodiment of the present application. A user (for example, a medical service provider) invokes an object allocation service in a cloud service through a terminal (a client is operated, for example, a medical client and the like) so that a server deployed in a cloud end invokes a program of encapsulated object allocation, acquires consultation information of a service receiving object, and acquires a plurality of service providing objects matched with the consultation information; performing ascending sort processing based on the load capacity on the plurality of service providing objects, and taking at least one service providing object with the top sort as a candidate service providing object; for each candidate service providing object, acquiring an allocation strategy weight positively correlated with the evaluation score of the candidate service providing object; acquiring sampling probability positively correlated with the allocation policy weight of each candidate service providing object; and performing weighted random selection processing based on the sampling probability of each candidate service providing object to obtain a target service providing object, and distributing the target service providing object to the service receiving object.

Referring to fig. 1, fig. 1 is a schematic view of an application scenario of an object distribution system according to an embodiment of the present application, a terminal 400 is connected to a server 200 through a network 300, and the network 300 may be a wide area network or a local area network, or a combination of the two.

The terminal 400 (a terminal used by a service-receiving object, which may be a patient or a person seeking legal consultation) receives consultation information input by the service-receiving object, the terminal 400 transmits an object allocation request including the consultation information to the server 200, and the server 200 acquires a plurality of service-providing objects (symptomatic doctors or lawyers who are in a professional direction) matched with the consultation information according to the object allocation request; performing ascending sort processing based on the load capacity on the plurality of service providing objects, and taking at least one service providing object with the top sort as a candidate service providing object; for each candidate service providing object, acquiring an allocation strategy weight positively correlated with the evaluation score of the candidate service providing object; acquiring sampling probability positively correlated with the allocation policy weight of each candidate service providing object; the weighted random selection process is performed based on the sampling probability of each candidate service providing object to obtain a target service providing object, and the target service providing object is allocated to the service receiving object, and the server 200 transmits the allocation result to the terminal 400 to display, for example, a consultation window of the service receiving object and the target service providing object.

In some embodiments, the terminal 400 (a terminal used by a service-receiving object, which may be a patient, a person seeking legal consultation) receives consultation information input by the service-receiving object, and the terminal 400 acquires a plurality of service-providing objects (symptomatic doctors or lawyers who are in a professional direction) matched with the consultation information according to an object allocation request; performing ascending sort processing based on the load capacity on the plurality of service providing objects, and taking at least one service providing object with the top sort as a candidate service providing object; for each candidate service providing object, acquiring an allocation strategy weight positively correlated with the evaluation score of the candidate service providing object; acquiring sampling probability positively correlated with the allocation policy weight of each candidate service providing object; and performing weighted random selection processing based on the sampling probability of each candidate service providing object to obtain a target service providing object, and distributing the target service providing object to the service receiving object, and displaying the distribution result, for example, displaying a consultation window of the service receiving object and the target service providing object, in a man-machine interaction interface of the terminal 400.

In some embodiments, the server 200 may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, or a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDNs, and basic cloud computing services such as big data and artificial intelligence platforms. The terminal 400 may be, but is not limited to, a smart phone, a tablet computer, a notebook computer, a desktop computer, a smart speaker, a smart watch, etc. The terminal and the server may be directly or indirectly connected through wired or wireless communication, which is not limited in the embodiment of the present invention.

In some embodiments, the terminal or the server may implement the object allocation method provided by the embodiments of the present application by running a computer program. For example, the computer program may be a native program or a software module in an operating system; a local (Native) Application program (APP), i.e. a program that needs to be installed in an operating system to run, such as a live APP or an instant messaging APP; the method can also be an applet, namely a program which can be run only by being downloaded into a browser environment; but also an applet that can be embedded in any APP. In general, the computer programs described above may be any form of application, module or plug-in.

Referring to fig. 2, fig. 2 is a schematic structural diagram of an electronic device for object allocation according to an embodiment of the present application, and the electronic device is a server 200, which is illustrated as an example, and the server 200 for object allocation shown in fig. 2 includes: at least one processor 210, a memory 250, at least one network interface 220. The various components in server 200 are coupled together by bus system 240. It is understood that the bus system 240 is used to enable connected communications between these components. The bus system 240 includes a power bus, a control bus, and a status signal bus in addition to the data bus. But for clarity of illustration the various buses are labeled as bus system 240 in fig. 2.

The processor 210 may be an integrated circuit chip with signal processing capabilities such as a general purpose processor, such as a microprocessor or any conventional processor, or the like, a digital signal processor (DSP, digital Signal Processor), or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or the like.

Memory 250 includes volatile memory or nonvolatile memory, and may also include both volatile and nonvolatile memory. The non-volatile memory may be a read only memory (ROM, read Onl y Memory) and the volatile memory may be a random access memory (RAM, random Access Memory). The memory 250 described in embodiments of the present application is intended to comprise any suitable type of memory. Memory 250 optionally includes one or more storage devices physically located remote from processor 210.

In some embodiments, memory 250 is capable of storing data to support various operations, examples of which include programs, modules and data structures, or subsets or supersets thereof, as exemplified below.

An operating system 251 including system programs for handling various basic system services and performing hardware-related tasks, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and handling hardware-based tasks;

A network communication module 252 for reaching other electronic devices via one or more (wired or wireless) network interfaces 220, the exemplary network interfaces 220 include: bluetooth, wireless compatibility authentication (WiFi), and universal serial bus (USB, universal Serial Bus), etc.;

in some embodiments, the object allocation apparatus provided in the embodiments of the present application may be implemented in a software manner, for example, may be an object allocation plug-in the above terminal, and may be an object allocation service in the above server. Of course, the object allocation apparatus provided in the embodiments of the present application may be provided in various forms including application programs, software modules, scripts or codes, as various software embodiments. Fig. 2 shows an object distribution means 255 stored in a memory 250, which may be software in the form of programs and plug-ins, e.g. image processing plug-ins, and comprises a series of modules including an acquisition module 2551, a load module 2552, a weight module 2553, a sampling module 2554 and a distribution module 2555.

As described above, the object allocation method provided by the embodiment of the present application may be implemented by various types of electronic devices. Referring to fig. 3A, fig. 3A is a flowchart of an object allocation method according to an embodiment of the present application, and is described with reference to steps 101 to 105 shown in fig. 3A.

In step 101, advisory information of a service acceptance object is acquired, and a plurality of service provision objects matching the advisory information are acquired.

As an example, the service recipient may be a patient, a user who needs legal assistance, or the like who needs to perform counseling service, the counseling information may be illness state description information when the service recipient is a patient, and the counseling information may be description information of an event to be counseled when the service recipient is a user who needs legal assistance. A service providing object matching with the counseling information, for example, a doctor matching with the illness description information, may be acquired as a service providing object based on the counseling information, for example, the counseling information is counseling information for intestinal diseases, and the matching doctor may be a doctor of the enterology.

In some embodiments, referring to fig. 3B, the acquiring of the plurality of service providing objects matched with the counseling information in step 101 may be implemented through steps 1011 to 1014 shown in fig. 3B.

In step 1011, a plurality of candidate service providing organizations are acquired.

By way of example, a doctor-patient consultation scenario is taken as an illustration, and candidate service providing organizations include breast surgery, obstetrics and gynecology, internal medicine, cardiac surgery, orthopedics, and the like, which all belong to the candidate service providing organizations. Taking legal consultation scenario as an example for illustration, candidate service providing organizations include criminal case teams, civil case teams, and the like.

In step 1012, text matching is performed on the advisory information to obtain a first probability that the advisory information matches each candidate service providing organization.

As an example, the doctor-patient consultation scenario is illustrated, and step 1012 may be implemented by a department identification model, which is used to map the patient's condition description (consultation information) to the symptomatic department (service providing organization) through a text classification algorithm of natural language processing, so that the patient can quickly find out the doctor of the corresponding department to make a consultation. Firstly, feature extraction processing is carried out on consultation information through a department identification model to obtain text features of the consultation information, encoding processing and decoding processing are carried out on the text features, and first probability of corresponding text features can be obtained, wherein the first probability represents the probability that the consultation information is matched with each candidate service providing organization.

As an example, the structure of the department identification model may be a pre-trained BERT model, see fig. 7, training a specific flow as follows: constructing a department identification marking data set; performing multi-classification task training by using a department identification marking data set; and testing the trained department identification model, and calculating the identification accuracy. When a department identification labeling data set is constructed, medical staff labels the name of a department for each illness state description (sample consultation information), the specific form of each corpus is (illness state description, department name labeling), and each department maintains at least 10 corpus data. For each training sample data, inputting the illness state description in the training sample data into a pre-trained BERT model, carrying out text matching processing on the illness state description by the pre-trained BERT model to obtain the prediction probability of matching the illness state description with each candidate service providing organization, for each candidate service providing organization, when the candidate service providing organization is a marked department, marking the candidate service providing organization as 1, when the candidate service providing organization is not the marked department, marking the candidate service providing organization as 0, obtaining the error between the prediction probability of the candidate service providing organization and the marking for each candidate service providing organization, and carrying out parameter updating processing on the pre-trained BERT model based on the fusion result of the errors of a plurality of candidate service providing organizations to obtain the trained BERT model as a department identification model.

As an example, a legal consultation scenario is described as an example, in which legal questions (consultation information) of a user are mapped to legal team (service providing organization) of a professional pair, so that the user can quickly find lawyers of the corresponding legal team to consult. Firstly, feature extraction processing is carried out on consultation information through a legal team identification model to obtain text features of the consultation information, coding processing and decoding processing are carried out on the text features, and first probability of corresponding text features can be obtained, wherein the first probability represents the probability that the consultation information is matched with each candidate service providing organization.

As an example, the structure of the legal team recognition model may be a pre-trained BERT model, training a specific procedure as follows: constructing a legal team identification marking data set; performing multi-classification task training by using a legal team identification marking data set; and testing the trained legal team recognition model, and calculating the recognition accuracy. When a legal team identification marking data set is constructed, legal problems are used as sample consultation information, legal staff marks the names of legal teams on each legal problem, the specific form of each corpus is (legal problems, legal team names marks), and each legal team at least maintains 10 corpus data. For each training sample data, inputting legal questions in the training sample data into a pre-trained BERT model, carrying out text matching processing on the legal questions by the pre-trained BERT model to obtain the prediction probability of matching the legal questions with each candidate service providing organization, for each candidate service providing organization, when the candidate service providing organization is a marked legal team, marking the candidate service providing organization as 1, when the candidate service providing organization is not a marked legal team, marking the candidate service providing organization as 0, obtaining errors between the prediction probability of the candidate service providing organization and the marking for each candidate service providing organization, and carrying out parameter updating processing on the pre-trained BERT model based on the fusion result of the errors of a plurality of candidate service providing organizations to obtain the trained BERT model as a legal team identification model.

In step 1013, the candidate service providing organization having the largest first probability is regarded as the service providing organization matching the advisory information.

By way of example, taking doctor-patient consultation scenario as an illustration, a first probability of matching between gynaecology and obstetrics and the description of the illness state, for example, a first probability of matching between breast surgery and the description of the illness state, a first probability of matching between internal medicine and the description of the illness state, and a candidate department having the largest first probability among these first probabilities as a department (service providing organization) matching with the description of the illness state may be outputted through step 1012.

By way of example, taking a legal consultation scenario as an illustration, a first probability of each candidate legal team (candidate service providing organization) may be output through step 1012, for example, a first probability of matching between a criminal legal team and a legal problem (consultation information), a first probability of matching between a civil legal team and a legal problem (consultation information), a first probability of matching between an intellectual property team and a legal problem (consultation information), and a candidate legal team having the largest first probability among these first probabilities may be regarded as a legal team (service providing organization) matching with the legal problem.

In step 1014, all the service providing objects belonging to the service providing organization are treated as a plurality of service providing objects matching the advisory information.

As an example, taking doctor-patient consultation scenario as an example, after the department matching the condition description output in step 1013 is acquired, all doctors (service providing objects) in the department are regarded as doctors matching the condition description.

As an example, taking a legal consultation scenario as an example, after the legal team matching the legal problem outputted in step 1013 is obtained, all lawyers (service providing objects) in the legal team are regarded as lawyers matching the legal problem.

According to the embodiment of the application, the accurate service providing object can be matched in an artificial intelligence mode, so that the distribution accuracy is improved.

In step 102, an ascending sort process based on the load amount is performed on the plurality of service providing objects, and at least one service providing object ranked earlier is used as a candidate service providing object.

In some embodiments, referring to fig. 3C, the ascending sort process based on the load amount is performed on the plurality of service providing objects in step 102, and at least one service providing object that is ranked first is used as a candidate service providing object, which may be implemented through steps 1021 through 1023 shown in fig. 3C.

In step 1021, an ascending sort process is performed on the plurality of service providing objects based on the load amount of each service providing object, to obtain an ascending sort result.

In step 1021, the load capacity of each service providing object needs to be obtained, for example, the load capacity of all doctors is obtained, wherein the load capacity is the sum of the number of patients being processed by the doctor at the current moment and the number of patients waiting to be consulted with the doctor, the doctor can be subjected to ascending sort processing based on the load capacity of each doctor, the load capacity of the doctor with the front sorting is less, and the doctor with the rear sorting has larger load capacity. For example, the load of all lawyers is obtained, the load is the sum of the number of consultants being processed by the lawyers at the current moment and the number of consultants waiting to be consulted with the lawyers, ascending ranking processing can be performed on the lawyers based on the load of each lawyer, the load of the lawyers with the front ranking is less, and the load of the lawyers with the rear ranking is greater.

In some embodiments, in response to the newly added first consultation order, adding a process to the load capacity of the service providing object corresponding to the first consultation order to obtain a new load capacity of the service providing object corresponding to the first consultation order. And in response to canceling the second consultation order, subtracting the loading capacity of the service providing object corresponding to the second consultation order to obtain the new loading capacity of the service providing object corresponding to the second consultation order. And in response to the change of the service providing object corresponding to the third consultation order, adding one to the load capacity of the new service providing object corresponding to the third consultation order to obtain the new load capacity of the new service providing object, and subtracting one to the load capacity of the old service providing object corresponding to the third consultation order to obtain the new load capacity of the old service providing object. And in response to the completion of the fourth consultation order, subtracting the loading capacity of the service providing object corresponding to the fourth consultation order to obtain the new loading capacity of the service providing object corresponding to the fourth consultation order. By the embodiment of the application, the load change is updated in real time, so that the load real-time accuracy of each service providing object can be ensured.

For example, when an online consultation doctor is allocated, the current consultation quantity (loading quantity) of each doctor needs to be acquired, the real-time loading condition of the doctor is stored in a database in a form of (doctor ID, loading quantity), and the database can be a memory database such as Redis, memcache. Referring to fig. 9, the doctor load real-time statistics module detects doctor's inquiry event through the message queue kafka to update doctor's load, and the following are inquiry event and corresponding update operations: the user submits inquiry order time, and the loading amount of the order corresponding to the doctor ID is increased by one; the user cancels the order, and the loading capacity of the doctor ID corresponding to the order is reduced by one; the user replaces the doctor with the original doctor ID load reduced by one, and the doctor ID load after replacement is increased by one; the inquiry order is completed, and the loading quantity of the order corresponding to the doctor ID is reduced by one.

In step 1022, the service providing objects ranked first are obtained from the ascending ranking result, and the load of the service providing objects ranked first is taken as the target load.

As an example, the service providing object ranked first in the ascending order result is a doctor or lawyer with the smallest load, for example, the load of the doctor ranked first is 1, and thus 1 is taken as the target load.

In step 1023, a service providing object having a target load amount among the plurality of service providing objects is taken as a candidate service providing object.

As an example, when there are only 1 service providing objects having a target load amount in the ascending sort result, that is, only the load amount of the service providing object ordered first is 1, the service providing object ordered first is taken as a candidate service providing object; when there are a plurality of service providing objects with target load in the ascending sort result, all service providing objects with load of 1 in the ascending sort result are used as candidate service providing objects, for example, service providing objects in the first 3 sorting positions in the ascending sort result.

According to the embodiment of the application, the service providing objects can be screened out through the load quantity, so that the calculation of the strategy weight aiming at each service providing object is avoided, and the load of the candidate service providing objects can be ensured to be smaller based on the load screening of the service providing objects, so that the waiting time of the service receiving objects is reduced.

In step 103, for each candidate service providing object, an allocation policy weight that is positively correlated with the evaluation score of the candidate service providing object is acquired.

As an example, there are two ways to obtain the evaluation score before obtaining the policy weight, the first way to obtain the evaluation score is to read the evaluation score from the database after determining the candidate service providing object, that is, maintain the database M in advance, and maintain the evaluation scores of all the service providing objects (whether matching with the consultation information or not) in the database M, where the evaluation scores are calculated based on the historical service data of the service providing objects, and the calculation way may be referred to in the following description; and secondly, after the candidate service providing object is determined, the evaluation score of the candidate service providing object is directly calculated on line based on the historical service data of the candidate service providing object, and the calculation principles of the two modes are the same.

In some embodiments, referring to fig. 3D, the obtaining of the allocation policy weight positively correlated with the evaluation score of the candidate service providing object in step 103 may be implemented by step 1031 or step 1032 shown in fig. 3D.

In step 1031, when the evaluation score includes at least two of a quality score, an liveness score, and a relevance score, weighting and summing the at least two scores included to obtain an allocation policy weight of the candidate service providing object.

As an example, a quality of service score S for doctor i is obtained _service Liveness S of the current time period _active Correlation S _rel The allocation policy weight is calculated according to the following formula (1):

S _fianl ＝W _service *S _service +W _active *S _active +W _rel *S _rel (1)；

wherein the quality of service score S _service Namely the mass fraction and the liveness S _active Namely, the liveness score and the relativity S _rel Namely, the correlation score, W _service Is the weight of the mass fraction, W _active Is the weight of the liveness score, W _rel Is the weight of the relevance score.

In step 1032, when the evaluation score includes any one of the quality score, the liveness score, and the correlation score, the any one of the scores is taken as an assigned policy weight of the candidate service providing object.

As an example, when only the relevance score is included in the evaluation score, the relevance score is an assigned policy weight of the candidate service providing object. For example, for a doctor without any experience of service, he will not have a quality score as well as an liveness score, so the relevance score may be directly taken as an assigned policy weight.

According to the embodiment of the application, the indexes of each dimension can be comprehensively considered, so that the distribution strategy weight can represent the comprehensive capability of the candidate service providing object.

In some embodiments, the following processing is performed for each candidate service providing object: obtaining quality scores of candidate service providing objects; acquiring liveness scores of candidate service providing objects; obtaining a relevance score of a candidate service providing object; and forming at least one of the quality score, the liveness score and the relevance score into an evaluation score of the candidate service providing object. According to the embodiment of the application, the indexes of each dimension can be comprehensively considered, so that the distribution strategy weight can represent the comprehensive capability of the candidate service providing object.

As an example, if after determining candidate service providing objects, evaluation scores are read from the database, i.e., the database M needs to be maintained in advance, the evaluation scores of all the service providing objects (whether or not matched with the advisory information) are maintained in the database M, and these evaluation scores are calculated based on the historical service data of the service providing objects. Querying the quality scores of the candidate service providing objects in the database based on the identity of the candidate service providing objects; querying the liveness score of the candidate service providing object based on the identity of the candidate service providing object in the database; querying a relevance score of the candidate service providing object based on the identity of the candidate service providing object in a database; wherein the quality score, liveness score and relevance score of the full-scale service providing object are stored in the database.

As an example, the quality of service score (quality score) is one of the basis for assigning weights, in principle the better the quality of service doctor, the greater the probability of getting an assignment. And the activity degree (activity score) of the doctor in each time period is also acquired, when doctor allocation is carried out, the doctor weight with high activity degree (activity score) in the time period can be improved in different time periods, and the doctor weight with low activity degree (activity score) is reduced, so that the response speed of the doctor is improved, and the waiting time of the user is reduced. The relevance score is used for representing the relevance degree of a doctor and consultation information, and even though candidate service providing objects matched with the consultation information are obtained through an organization matching mode, diseases which are good for different candidate service providing objects still differ, so that the relevance degree of the doctor and the consultation information needs to be considered, and at least one of the quality score, the liveness score and the relevance score is finally formed into an evaluation score of the candidate service providing objects.

In some embodiments, the obtaining the quality score of the candidate service providing object may be achieved by the following technical solutions: and obtaining the grade scores of the candidate service providing objects and obtaining the interaction scores of the candidate service providing objects. And carrying out weighted summation processing on the grade scores and the interaction scores to obtain quality scores of the candidate service providing objects. According to the embodiment of the application, subjective scores and objective scores can be comprehensively considered, so that the characterization capability of quality scores is improved.

As an example, referring to fig. 8, the rating score is an objective factor score in a doctor quality scoring system, the rating score is an interaction score determined according to a doctor's rating and a hospital's rating is a subjective factor score in a doctor quality scoring system, for example, a return rate, a consultation amount, and the like.

In some embodiments, the obtaining the grade score of the candidate service providing object may be achieved by the following technical solutions: obtaining an organization grade score of a service providing organization to which the candidate service providing object belongs; acquiring professional grade scores of candidate service providing objects; and carrying out weighted summation processing on the organization grade scores and the professional grade scores to obtain grade scores of candidate service providing objects. According to the embodiment of the application, the grade score of the object can be accurately calculated, and the calculation accuracy of the quality score is improved.

As an example, for the class score, when the service providing organization is a department, the class score of the organization in which the doctor is located or the department may be obtained as the class score of the organization of the service providing organization, for example, different classes of the organization correspond to different class scores of the organization, and the class score of the doctor may be obtained as the professional class score of the candidate service providing object, for example, different roles of the doctor correspond to different scores.

In some embodiments, the obtaining the interaction score of the candidate service providing object may be achieved by the following technical scheme: acquiring service scores of candidate service providing objects, consultation quantity scores of the candidate service providing objects, response rate scores of the candidate service providing objects and response speed scores of the candidate service providing objects from historical service data of the candidate service providing objects; and carrying out weighted summation processing on the service scores of the candidate service providing objects, the consultation quantity scores of the candidate service providing objects, the response rate scores of the candidate service providing objects and the response speed scores of the candidate service providing objects to obtain interaction scores of the candidate service providing objects. The embodiment of the application can improve the calculation accuracy of the interaction score.

As an example, different rating star grades correspond to different service scores, different advisory numbers correspond to different advisory numbers scores, different return rates correspond to different return rates scores, response speed scores, different response times correspond to different response speeds scores.

In some embodiments, the obtaining the liveness score of the candidate service providing object may be achieved by the following technical solutions: acquiring the consultation quantity of each time period from the historical service data of the candidate service providing object, and summing the consultation quantity of a plurality of time periods to obtain the total consultation quantity; acquiring the score positively correlated with the consultation quantity of the current time period and negatively correlated with the total consultation quantity of the current time period as the liveness score of the corresponding current time period; the current time period is a time period to which a time point at which the advisory information is acquired from among the plurality of time periods. According to the embodiment of the application, the liveness score can be accurately calculated, so that the characterization capability of the distribution strategy weight is improved.

As an example, counting the total number of consultations of the doctor in each time period, calculating the liveness score of each time period by using the laplace smoothing method, and storing the liveness score in the form of (doctor ID, (time period, liveness)) into the database, see formula (2):

Wherein, c _i The order quantity of the doctor in the ith time period is represented by alpha, which is a smoothing parameter and can be set to 0.5, and the time period can be divided into early morning (00:00-5:00), early morning (5:00-6:00), early morning (6:00-8:00), morning (8:00-11:00), noon (11:00-13:00), afternoon (13:00-17:00), evening (17:00-18:00) and evening (18:00-24:00), and k is the time period quantity, wherein k=8.

In some embodiments, the obtaining the relevance score of the candidate service providing object may be achieved by the following technical solutions: and extracting consultation keywords from the consultation information to form a consultation keyword set. And acquiring a tag set of the candidate service providing object, wherein the tag included in the tag set is used for representing the professional keyword of the candidate service providing object. And acquiring a union set between the consultation keyword set and the label set. A relevance score is obtained that is positively correlated to the number of elements of the union and negatively correlated to the number of elements of the consultation keyword set.

As an example, an assignment policy weight calculation is performed for each doctor, and first a correlation calculation is performed for doctor i according to the condition description, see formula (3):

wherein D is _desc For the set of terms of illness and symptoms mentioned in the description of illness (consultation information), D _doctor For doctor i to be good at treating the disease corresponding to the disease name and symptom word set (namely label set), epsilon is super parameter, S _rel Is a relevance score.

In step 104, a sampling probability is obtained that is positively correlated with the assigned policy weight for each candidate service providing object.

In some embodiments, referring to fig. 3E, the obtaining of the sampling probability in step 104 that is positively correlated with the assigned policy weight of each candidate service providing object may be implemented by steps 1041 through 1042 shown in fig. 3E.

In step 1041, a summation is performed on the assigned policy weights of the plurality of candidate service providing objects, to obtain a first summation result.

In step 1042, for each candidate service providing object, the ratio between the assigned policy weight of the candidate service providing object and the first summation result is taken as the sampling probability.

For example, see formula (4):

wherein,,is the allocation strategy for doctor iWeight(s)>Is the first summation result, p _i Is the sampling probability.

According to the embodiment of the application, the sampling probability can be obtained based on the strategy distribution weight, so that the subsequent sampling result is influenced by the strategy distribution weight, and the accuracy of randomly selecting the service providing object is improved.

In step 105, weighted random selection processing is performed based on the sampling probability of each candidate service providing object, resulting in a target service providing object, and the target service providing object is assigned to the service accepting object.

As an example, assuming that there are 3 candidate service providing objects, respectively numbered 1 to 3, the sampling probability in the number 3 being decimated is 70%, the sampling probability in the number 2 being decimated is 25%, the sampling probability in the number 1 being decimated is 5%, a random number of uniform distribution between 0 and 100 is generated, the selected target service providing object is the candidate service providing object of the number 3 when the random number is 0 to 70, the selected target service providing object is the candidate service providing object of the number 2 when the random number is 70 to 95, and the selected target service providing object is the candidate service providing object of the number 1 when the random number is 95 to 100.

In some embodiments, the object allocation method provided by the embodiment of the present application may be implemented by the following technical solutions: acquiring medical consultation information of a medical service receiving object, and acquiring a plurality of medical service providing objects matched with the medical consultation information; acquiring at least one medical service providing object with the smallest load amount in the plurality of medical service providing objects as a candidate medical service providing object; for each medical candidate service providing object, acquiring an allocation strategy weight positively correlated with the evaluation score of the medical candidate service providing object; acquiring sampling probability positively correlated with the distribution strategy weight of each candidate medical service providing object; and executing weighted random selection processing based on the sampling probability of each candidate medical service providing object to obtain a target medical service providing object, and distributing the target medical service providing object to the medical service receiving object.

The medical consultation information in the embodiment of the present application is an example (illness description) of the consultation information in the step 101 in a doctor-patient consultation scene, the medical service providing object is an example (doctor) of the service providing object in the step 101 in the doctor-patient consultation scene, and the service providing organization is an example (department of a hospital) of the service providing organization in the step 1011 in the doctor-patient consultation scene. Reference may be made to the specific implementation of steps 101 to 105 for the specific implementation of the embodiment of the present application.

The embodiment of the application solves the problems of long time for waiting for the inquiry of the patient and high workload of the doctor, and ensures the accuracy of the doctor allocation, thereby improving the inquiry efficiency of the inquiry platform.

In the following, an exemplary application of the embodiment of the present application in a practical application scenario will be described.

Referring to fig. 4, fig. 4 shows an interface schematic diagram of a medical inquiry platform provided by the embodiment of the present application, taking the medical inquiry platform as an example, the medical inquiry platform provides a rapid online inquiry service, a patient inputs a disease description, a server distributes doctors to make inquiry, a terminal (a terminal used by the patient) accepts the disease description input by a service receiving object, the terminal sends a doctor distribution request including the disease description to the server, and the server acquires a plurality of doctors matched with consultation information according to the doctor distribution request; carrying out ascending sort processing based on the load quantity on a plurality of doctors, and taking at least one doctor with the front sort as a candidate doctor; acquiring an evaluation score of each candidate doctor, and determining an allocation strategy weight of each candidate doctor based on the evaluation score; acquiring sampling probability positively correlated with the assigned strategy weight of each candidate doctor; and performing weighted random selection processing based on the sampling probability of each candidate doctor to obtain a target doctor, distributing the target doctor to the patient, and sending the distribution result to a terminal by a server for display, for example, displaying a consultation window of the patient and the target doctor.

The embodiment of the application aims to reasonably allocate the on-line consultation doctor so as to reduce the workload of the head doctor and improve the workload of the tail doctor, thereby realizing the balanced utilization of medical resources, and reducing the waiting time of patients so as to improve the user consultation experience. Referring to fig. 5, fig. 5 shows a schematic diagram of doctor load provided by an embodiment of the present application, in which three patients are already in line in doctor 1, if a reassigned patient is given to him, he enters a high load state, and the patient needs a long waiting time, at this time, a stop order receiving process should be performed for doctor 1, and a consultation should be allocated to an idle doctor.

Referring to fig. 6 and 10, fig. 6 shows a schematic overall framework provided by the embodiment of the present application, and fig. 10 shows a schematic flow chart provided by the embodiment of the present application, where the overall framework of the on-line doctor questioning distribution system is divided into 4 modules, respectively: the department identification model training module; a quality of service calculation module; a doctor load real-time statistics module; and the doctor allocation module is a main system module, and the other 3 modules serve the doctor allocation module.

The department identification model training module is used for executing department identification model training, and the department identification model is used for mapping the illness state description to a symptomatic department through a text classification algorithm processed by natural language so as to facilitate a patient to quickly find out doctors of the corresponding departments to conduct inquiry. The training method and the training device for the department identification model based on the pre-training model BERT are used for training the department identification model, and see fig. 7, and the training specific flow is as follows: constructing a department identification marking data set; based on an open source Chinese pre-training model BERT (the BERT model can use a Chinese_L-12_H-768_A-12 or can be replaced by a pre-training language model such as RoBerta and the like); performing multi-classification task training by using a department identification marking data set; and testing the trained department identification model, and calculating the identification accuracy. When a department identification marking data set is constructed, medical staff marks the name of each department on each illness state description, the specific form of each corpus is (illness state description, department name), and each department maintains at least 10 corpus data.

The quality of service calculation module is described below. The service quality calculation module is used for scoring the doctor's inquiry service quality according to the doctor's history inquiry data. The quality of service score is one of the basis for assigning weights, in principle the better the quality of service the greater the probability of getting an assignment. In the embodiment of the application, doctor service quality scoring rules are formulated, referring to fig. 8, fine adjustment of the rules can be performed according to service conditions, the service quality computing module also counts the activity degree of a doctor in each time period, and when the doctor is distributed, the doctor weight with high activity degree in the time period can be improved in different time periods, and the doctor weight with low activity degree is reduced, so that the response speed of the doctor is improved, and the waiting time of a user is reduced. The specific flow of the service quality calculation is as follows: for each clinician d, the following is performed: acquiring hospital rank and doctor grade of a hospital where a doctor d is located, the name of the doctor d, and the last 3 month history data of the doctor d, including average evaluation star, average response time, consultation quantity, recovery rate and order starting time of inquiry, and calculating a doctor service quality score S according to a scoring rule shown in FIG. 8 _service ∈[0,100]And stored in the form of (doctor ID, quality of service score) in a database; counting the number of orders of a doctor in each time period, and smoothing by using LaplacianThe method calculates the liveness of each time period and stores the liveness in a database in the form of (doctor ID, (time period, liveness)) see formula (5):

wherein c _i The order quantity of the doctor in the ith time period is represented by alpha, which is a smoothing parameter and can be set to 0.5, and the time period can be divided into early morning (00:00-5:00), early morning (5:00-6:00), early morning (6:00-8:00), morning (8:00-11:00), noon (11:00-13:00), afternoon (13:00-17:00), evening (17:00-18:00) and evening (18:00-24:00), and k is the time period quantity, wherein k=8.

The doctor load real-time statistics module is described below, and when an online consultation doctor is allocated, the current consultation quantity of each doctor needs to be acquired, so that the doctor load real-time statistics module is needed. The real-time loading condition of the doctor is stored in a database in the form of (doctor ID, loading quantity), and the database can be a memory database such as Redis, memcache. Referring to fig. 9, the doctor load real-time statistics module detects doctor's inquiry event through the message queue kaf ka to update doctor's load, and the following are inquiry event and corresponding update operations: the user submits inquiry order time, and the loading amount of the order corresponding to the doctor ID is increased by one; the user cancels the order, and the loading capacity of the doctor ID corresponding to the order is reduced by one; the user replaces the doctor with the original doctor ID load reduced by one, and the doctor ID load after replacement is increased by one; the inquiry order is completed, and the loading quantity of the order corresponding to the doctor ID is reduced by one.

Finally introducing a doctor allocation module, wherein the doctor allocation module is responsible for allocating doctors with good service quality and quick response to the users according to the illness state descriptions of the users so as to meet the requirements of the users on inquiry. Different from a random distribution mode, the doctor distribution module utilizes the processing results of the modules to aim at providing doctors with high response speed and good service quality related to illness states for users, and gives consideration to the real-time load of each doctor. The proper department can be accurately found according to the disease description of the user through the department identification model; the service quality calculation module is used for evaluating the service quality and the activity of the doctor, and the higher the service quality and the activity of the doctor, the higher the probability of obtaining allocation; the doctor load real-time statistics module is responsible for recording the current load of the doctor, and reduces the allocation probability for the doctor with high load so as to reduce the waiting time of the user.

Firstly, acquiring the disease condition description of a patient, performing department identification by using a trained department identification model, acquiring the department of the patient, and continuously acquiring all doctors in the department of the patient. Doctor allocation is carried out by adopting a load balancing strategy of weighted minimum connection, and the load capacity c of each doctor is obtained from a database _i Taking doctor set D= { D with minimum load _i |c _i ＝min _j c _i }。

The assignment strategy weight calculation is carried out on each doctor, and firstly, the correlation calculation is carried out on the doctor i according to the disease state description, see the formula (6):

wherein D is _desc For the set of terms of disease and symptoms mentioned in the description of the condition, D _doctor For doctor i to be adept at describing the disease and the symptom word set, epsilon is a super parameter, the prevention denominator is 0, and 0.000001 is preferable.

Obtaining a quality of service score S for a patient i in a collection D from a database _service And the activity S of the current time period _active The allocation policy weight is calculated according to the following formula (7):

S _final ＝ w _rel *S _rel + w _service *S _service +w _active *S _active (7)；

s.t.w _rel +w _service +w _active ＝1

wherein w is _rel For the relevance score weight, w _service Scoring weights for doctor quality of service, w _active For liveness weight, the sum of the three is 1, and w is taken to be _rel Is 0.4, w _service Is 0.3, w _service 0.2.

Calculating sampling probability p of doctors in the set D, see formula (8):

wherein,,is to assign policy weights, D is the least loaded doctor set, p _i For probability, doctors are selected by a weighted random manner and assigned to users.

Compared with the traditional doctor inquiry distribution scheme, the doctor inquiry distribution method and device not only consider the correlation between doctors and disease descriptions, but also design a doctor distribution strategy according to the busy condition of the loads of the doctors, the service quality of the doctors and the liveness of the doctors, solve the problems of long time for waiting for inquiry of patients and high workload of the doctors, and simultaneously ensure the accuracy of doctor distribution, thereby improving the inquiry efficiency of an inquiry platform.

In the embodiment of the application, the load balancing strategy of the weighted minimum connection is utilized, the greedy strategy with the minimum load number is not adopted as in the prior scheme, but weighted random sampling is carried out according to the weight, so that newly-resident doctors can obtain the opportunity of distribution, doctors with common service quality also have the opportunity of improving the service quality of the doctors, the Martai effect is reduced, and the ecological balance of the consultation platform is maintained.

The embodiment of the application considers the activity of the doctor in the current time period and influences the allocation probability of the doctor as a factor. In the prior art, the work and rest habits of doctors are not considered, the consultation order is easy to be carried out when the doctors rest or are busy, the activity of the doctors can reflect the order receiving speed in the current time period, after the activity of the doctors in each time period is added, the allocation probability of the active doctors can be increased, and the waiting time of users is reduced.

Continuing with the description below of an exemplary architecture of the object distribution device 255 implemented as a software module provided by embodiments of the present application, in some embodiments, as shown in fig. 2, the software modules stored in the object distribution device 255 of the memory 250 may include: an acquisition module 2551, configured to acquire advisory information of a service receiving object, and acquire a plurality of service providing objects matched with the advisory information; a load module 2552, configured to perform an ascending sort process based on a load amount on the plurality of service providing objects, and use at least one service providing object that is ranked first as a candidate service providing object; a weight module 2553, configured to obtain, for each of the candidate service providing objects, an allocation policy weight that is positively correlated with an evaluation score of the candidate service providing object; a sampling module 2554, configured to obtain a sampling probability that is positively correlated with an allocation policy weight of each of the candidate service providing objects; an allocation module 2555, configured to perform weighted random selection processing based on the sampling probability of each candidate service providing object, to obtain a target service providing object, and allocate the target service providing object to the service receiving object.

In some embodiments, the obtaining module 2551 is further configured to obtain a plurality of candidate service providing organizations; performing text matching processing on the consultation information to obtain a first probability of matching the consultation information with each candidate service providing organization; taking the candidate service providing organization with the largest first probability as the service providing organization matched with the consultation information; all the service providing objects belonging to the service providing organization are taken as a plurality of service providing objects matched with the consultation information.

In some embodiments, the load module 2552 is further configured to perform an ascending sort process on a plurality of service providing objects based on a load amount of each service providing object, to obtain an ascending sort result; acquiring the service providing objects ranked at the first position from the ascending ranking result, and taking the load capacity of the service providing objects ranked at the first position as target load capacity; and taking the service providing object with the target load capacity in the plurality of service providing objects as the candidate service providing object.

In some embodiments, the load module 2552 is further configured to respond to a newly added first consultation order, and add a process to the load capacity of the service providing object corresponding to the first consultation order, so as to obtain a new load capacity of the service providing object corresponding to the first consultation order; in response to canceling the second consultation order, subtracting the loading capacity of the service providing object corresponding to the second consultation order to obtain a new loading capacity of the service providing object corresponding to the second consultation order; responding to the change of a service providing object corresponding to a third consultation order, adding one to the load capacity of a new service providing object corresponding to the third consultation order to obtain the new load capacity of the new service providing object, and subtracting one to the load capacity of an old service providing object corresponding to the third consultation order to obtain the new load capacity of the old service providing object; and in response to completion of the fourth consultation order, subtracting the loading capacity of the service providing object corresponding to the fourth consultation order to obtain the new loading capacity of the service providing object corresponding to the fourth consultation order.

In some embodiments, the weighting module 2553 is further configured to perform the following processing for each of the candidate service providing objects: acquiring the quality scores of the candidate service providing objects; acquiring the liveness score of the candidate service providing object; obtaining a relevance score of the candidate service providing object; and forming at least one of the quality score, the liveness score and the relevance score into an evaluation score of the candidate service providing object.

In some embodiments, the weighting module 2553 is further configured to obtain a ranking score of the candidate service providing object, and obtain an interaction score of the candidate service providing object; and carrying out weighted summation processing on the grade scores and the interaction scores to obtain the quality scores of the candidate service providing objects.

In some embodiments, the weighting module 2553 is further configured to obtain an organization level score of a service providing organization to which the candidate service providing object belongs; acquiring the professional grade scores of the candidate service providing objects; and carrying out weighted summation processing on the organization grade scores and the professional grade scores to obtain grade scores of the candidate service providing objects.

In some embodiments, the weighting module 2553 is further configured to obtain a service score of the candidate service providing object, a consultation quantity score of the candidate service providing object, a response rate score of the candidate service providing object, and a response speed score of the candidate service providing object from the historical service data of the candidate service providing object; and carrying out weighted summation processing on the service scores of the candidate service providing objects, the consultation quantity scores of the candidate service providing objects, the response rate scores of the candidate service providing objects and the response speed scores of the candidate service providing objects to obtain the interaction scores of the candidate service providing objects.

In some embodiments, the weighting module 2553 is further configured to obtain a consultation number of each time period from the historical service data of the candidate service providing object, and sum the consultation numbers of a plurality of time periods to obtain a total number of consultations to obtain a score that is positively correlated with the consultation number of the current time period and negatively correlated with the total number of consultations of the current time period as an activity score corresponding to the current time period; the current time period is a time period to which the time of acquiring the consultation information in a plurality of time periods belongs.

In some embodiments, the weighting module 2553 is further configured to extract consultation keywords from the consultation information to form a consultation keyword set; acquiring a tag set of the candidate service providing object, wherein the tag set comprises tags used for representing professional keywords of the candidate service providing object; acquiring a union set between the consultation keyword set and the label set; a relevance score is obtained that is positively correlated to the number of elements of the union and negatively correlated to the number of elements of the set of consultation keywords.

In some embodiments, the weighting module 2553 is further configured to, when the evaluation score includes at least two of the quality score, the liveness score, and the relevance score, perform weighted summation processing on the at least two scores included to obtain an allocation policy weight of the candidate service providing object. When the evaluation score includes any one of the quality score, the liveness score, and the correlation score, the any one score is taken as an allocation policy weight of the candidate service providing object.

In some embodiments, the sampling module 2554 is further configured to sum the assigned policy weights of the plurality of candidate service providing objects to obtain a first summation result; for each candidate service providing object, taking the ratio between the distribution strategy weight of the candidate service providing object and the first summation result as the sampling probability.

Continuing with the description below of exemplary structures implemented as software modules of an object distribution device provided by embodiments of the present application, in some embodiments, the software modules stored in the object distribution device of the memory may include: the medical treatment acquisition module is used for acquiring medical treatment consultation information of a medical treatment service receiving object and acquiring a plurality of medical treatment service providing objects matched with the medical treatment consultation information; the medical load module is used for acquiring at least one medical service providing object with the smallest load amount in the plurality of medical service providing objects as a candidate medical service providing object; a medical weight module, configured to obtain, for each of the medical candidate service providing objects, an allocation policy weight positively correlated with an evaluation score of the medical candidate service providing object; the medical sampling module is used for acquiring sampling probability positively correlated with the distribution strategy weight of each candidate medical service providing object; and the medical distribution module is used for executing weighted random selection processing based on the sampling probability of each candidate medical service providing object to obtain a target medical service providing object and distributing the target medical service providing object to the medical service receiving object.

Embodiments of the present application provide a computer program product comprising a computer program or computer-executable instructions stored in a computer-readable storage medium. The processor of the electronic device reads the computer-executable instructions from the computer-readable storage medium, and the processor executes the computer-executable instructions, so that the electronic device performs the object allocation method according to the embodiment of the present application.

Embodiments of the present application provide a computer-readable storage medium having stored therein computer-executable instructions which, when executed by a processor, cause the processor to perform an object allocation method provided by embodiments of the present application, for example, as shown in fig. 3A-3E.

In some embodiments, the computer readable storage medium may be FRAM, ROM, PROM, EP ROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; but may be a variety of devices including one or any combination of the above memories.

In some embodiments, computer-executable instructions may be written in any form of programming language, including compiled or interpreted languages, or declarative or procedural languages, in the form of programs, software modules, scripts, or code, and they may be deployed in any form, including as stand-alone programs or as modules, components, subroutines, or other units suitable for use in a computing environment.

As an example, computer-executable instructions may, but need not, correspond to files in a file system, may be stored as part of a file that holds other programs or data, such as in one or more scripts in a hypertext markup language (HTML, hyper Text Markup Language) document, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files that store one or more modules, sub-programs, or portions of code).

As an example, computer-executable instructions may be deployed to be executed on one electronic device or on multiple electronic devices located at one site or, alternatively, on multiple electronic devices distributed across multiple sites and interconnected by a communication network.

In summary, according to the embodiment of the application, the consultation information of the service receiving object is obtained, and a plurality of service providing objects matched with the consultation information are obtained; taking at least one service providing object with the smallest load amount in the plurality of service providing objects as a candidate service providing object, so that the load of the service providing objects can be balanced, and determining the distribution strategy weight of each candidate service providing object based on the evaluation score; and acquiring sampling probability positively correlated with the distribution strategy weight of each candidate service providing object, so that the evaluation score of the service providing object is considered when the service providing object is selected, weighting random selection processing is performed on the basis of the sampling probability of each candidate service providing object to obtain a target service providing object, the target service providing object is distributed to the service receiving object, and finally the target service providing object is matched after the evaluation score is considered and belongs to the service providing object with the minimum load capacity, so that the load balance of service providing object distribution is realized.

The foregoing is merely exemplary embodiments of the present application and is not intended to limit the scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and scope of the present application are included in the protection scope of the present application.

Claims

1. A method of object allocation, the method comprising:

2. The method of claim 1, wherein the acquiring a plurality of service providing objects matched with the counseling information comprises:

Acquiring a plurality of candidate service providing organizations;

performing text matching processing on the consultation information to obtain a first probability of matching the consultation information with each candidate service providing organization;

taking the candidate service providing organization with the largest first probability as the service providing organization matched with the consultation information;

all the service providing objects belonging to the service providing organization are taken as a plurality of service providing objects matched with the consultation information.

3. The method of claim 1, wherein said performing an ascending order based on a load amount on said plurality of service providing objects, wherein said ranking at least one of said service providing objects that is top as a candidate service providing object comprises:

performing ascending sort processing on a plurality of service providing objects based on the load capacity of each service providing object to obtain ascending sort results;

acquiring the service providing objects ranked at the first position from the ascending ranking result, and taking the load capacity of the service providing objects ranked at the first position as target load capacity;

and taking the service providing object with the target load capacity in the plurality of service providing objects as the candidate service providing object.

4. A method according to claim 3, characterized in that the method further comprises:

responding to a newly added first consultation order, and adding one to the load capacity of a service providing object corresponding to the first consultation order to obtain the new load capacity of the service providing object corresponding to the first consultation order;

in response to canceling the second consultation order, subtracting the loading capacity of the service providing object corresponding to the second consultation order to obtain a new loading capacity of the service providing object corresponding to the second consultation order;

responding to the change of a service providing object corresponding to a third consultation order, adding one to the load capacity of a new service providing object corresponding to the third consultation order to obtain the new load capacity of the new service providing object, and subtracting one to the load capacity of an old service providing object corresponding to the third consultation order to obtain the new load capacity of the old service providing object;

and in response to completion of the fourth consultation order, subtracting the loading capacity of the service providing object corresponding to the fourth consultation order to obtain the new loading capacity of the service providing object corresponding to the fourth consultation order.

5. The method according to claim 1, wherein the method further comprises:

the following processing is performed for each of the candidate service providing objects:

acquiring the quality scores of the candidate service providing objects;

acquiring the liveness score of the candidate service providing object;

obtaining a relevance score of the candidate service providing object;

and forming at least one of the quality score, the liveness score and the relevance score into an evaluation score of the candidate service providing object.

6. The method of claim 5, wherein the obtaining the quality score of the candidate service providing object comprises:

obtaining the grade scores of the candidate service providing objects and obtaining the interaction scores of the candidate service providing objects;

and carrying out weighted summation processing on the grade scores and the interaction scores to obtain the quality scores of the candidate service providing objects.

7. The method of claim 6, wherein the obtaining a ranking score for the candidate service providing object comprises:

obtaining an organization grade score of a service providing organization to which the candidate service providing object belongs;

Acquiring the professional grade scores of the candidate service providing objects;

and carrying out weighted summation processing on the organization grade scores and the professional grade scores to obtain grade scores of the candidate service providing objects.

8. The method of claim 6, wherein the obtaining the interaction score for the candidate service providing object comprises:

acquiring service scores of the candidate service providing objects, consultation quantity scores of the candidate service providing objects, response rate scores of the candidate service providing objects and response speed scores of the candidate service providing objects from historical service data of the candidate service providing objects;

and carrying out weighted summation processing on the service scores of the candidate service providing objects, the consultation quantity scores of the candidate service providing objects, the response rate scores of the candidate service providing objects and the response speed scores of the candidate service providing objects to obtain the interaction scores of the candidate service providing objects.

9. The method of claim 5, wherein the obtaining the liveness score of the candidate service providing object comprises:

acquiring the consultation quantity of each time period from the historical service data of the candidate service providing object, and summing the consultation quantity of a plurality of time periods to obtain the total consultation quantity

Obtaining a score which is positively correlated with the consultation quantity of the current time period and negatively correlated with the total consultation quantity of the current time period as an liveness score corresponding to the current time period;

the current time period is a time period to which the time of acquiring the consultation information in a plurality of time periods belongs.

10. The method of claim 5, wherein the obtaining the relevance score for the candidate service providing object comprises:

extracting consultation keywords from the consultation information to form a consultation keyword set;

acquiring a tag set of the candidate service providing object, wherein the tag set comprises tags used for representing professional keywords of the candidate service providing object;

acquiring a union set between the consultation keyword set and the label set;

a relevance score is obtained that is positively correlated to the number of elements of the union and negatively correlated to the number of elements of the set of consultation keywords.

11. The method of claim 5, wherein the step of determining the position of the probe is performed,

the obtaining the quality score of the candidate service providing object includes:

querying the quality scores of the candidate service providing objects in a database based on the identity of the candidate service providing objects;

The obtaining the liveness score of the candidate service providing object includes:

querying the liveness score of the candidate service providing object based on the identity of the candidate service providing object in the database;

the obtaining the relevance score of the candidate service providing object includes:

querying a relevance score of the candidate service providing object based on the identity of the candidate service providing object in the database;

wherein the database stores quality scores, liveness scores and relevance scores of the full-scale service providing objects.

12. The method of claim 1, wherein the obtaining an assigned policy weight that is positively correlated with the evaluation score of the candidate service providing object comprises:

when the evaluation score comprises at least two of a quality score, an liveness score and a correlation score, carrying out weighted summation processing on the at least two scores to obtain an allocation strategy weight of the candidate service providing object;

when the evaluation score includes any one of a quality score, an liveness score, and a correlation score, the any one score is taken as an allocation policy weight of the candidate service providing object.

13. The method of claim 1, wherein the obtaining a sampling probability positively correlated with an assigned policy weight for each of the candidate service providing objects comprises:

summing the distributed strategy weights of the candidate service providing objects to obtain a first summation result;

for each candidate service providing object, taking the ratio between the distribution strategy weight of the candidate service providing object and the first summation result as the sampling probability.

14. An artificial intelligence based object distribution method, the method comprising:

15. An object dispensing device, the device comprising:

16. An object dispensing device, the device comprising:

17. An electronic device, the electronic device comprising:

a memory for storing executable instructions;

A processor for implementing the object allocation method of any one of claims 1 to 13 or claim 14 when executing computer executable instructions stored in said memory.

18. A computer readable storage medium storing computer executable instructions which when executed by a processor implement the object allocation method of any one of claims 1 to 13 or claim 14.

19. A computer program product comprising computer executable instructions which when executed by a processor implement the object allocation method of any one of claims 1 to 13 or claim 14.