A physical examination information distribution method and device

Abstract

The present application provides a method, device, electronic device, computer-readable storage medium and computer program product for distributing physical examination information. First, the physical examination information and the information of multiple distribution objects are obtained; then the physical examination information and the information of the distribution objects are input into a pre-trained information matching model to obtain the matching degree between the physical examination information and the distribution objects; then, according to the matching degree between the physical examination information and each distribution object, the target distribution object is determined from the multiple distribution objects, and the physical examination information is distributed to the target distribution object. The matching degree between the physical examination information and the distribution object is calculated by the information matching model, and the matching degree between the distribution object and the physical examination information can be comprehensively considered from the two aspects of professional matching and workload matching, ensuring that the physical examination information is distributed to the professional matching distribution object while taking into account the workload of each distribution object, so that the workload between each distribution object is balanced.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method, device, electronic device, computer-readable storage medium, and computer program product for allocating physical examination information. Background Art

[0002] After the user completes a series of physical examination items (such as internal medicine examination, surgical examination, electrocardiogram, etc.), the physical examination information will include not only the examination results of each physical examination item (such as electrocardiogram: normal sinus rhythm), but also the general examination conclusion and health recommendations. The general examination conclusion and health recommendations are health assessments and recommendations made by the general examination doctor based on the various examination results in the physical examination information. In actual applications, the accuracy of health assessments and recommendations can only be ensured by assigning physical examination information to appropriate general examination doctors.

[0003] Therefore, how to match physical examination information and the chief examiner is a technical problem that needs to be solved by those working in this field. Summary of the invention

[0004] The present application provides a physical examination information allocation method, device, electronic device, computer-readable storage medium and computer program product to match physical examination information with allocation objects. The technical solution of the present application is as follows:

[0005] According to a first aspect of the present application, a method for distributing physical examination information is provided, the method comprising:

[0006] Acquire physical examination information and information of multiple assignment objects, wherein the physical examination information includes abnormal physical examination information and workload information of the physical examination information, and the assignment object information includes specialty information and workload information of the assignment object;

[0007] Inputting the physical examination information and the information of the assigned object into a pre-trained information matching model to obtain a matching degree between the physical examination information and the assigned object, wherein the information matching model is used to calculate the matching degree between the physical examination information and the assigned object based on the abnormal physical examination information, the workload information of the physical examination information, and the specialty information and workload information of the assigned object;

[0008] According to the matching degree between the physical examination information and each of the assignment objects, a target assignment object is determined from the multiple assignment objects, and the physical examination information is assigned to the target assignment object.

[0009] In an optional implementation, the information matching model includes a semantic vector model and a vector matching model, and the step of inputting the physical examination information and the information of the assigned object into the pre-trained information matching model to obtain the matching degree between the physical examination information and the assigned object includes:

[0010] Inputting the physical examination information and the information of the assigned object into the semantic vector model, and outputting a first vector;

[0011] The first vector is input into the vector matching model, and the matching degree between the physical examination information and the assigned object is output.

[0012] In an optional implementation, the step of inputting the physical examination information and the information of the assigned object into the semantic vector model and outputting a first vector includes:

[0013] The physical examination information and the information of the assigned object are spliced ​​to obtain spliced ​​information, wherein the header of the spliced ​​information uses a marking symbol [CLS], and the physical examination information and the information of the assigned object are connected using a separator symbol [SEP];

[0014] The splicing information is input into the semantic vector model, and the first vector is output.

[0015] In an optional implementation, the step of inputting the first vector into the vector matching model and outputting the matching degree between the physical examination information and the assigned object includes:

[0016] The vector matching model calculates the matching degree between the physical examination information and the assigned object according to the following formula:

[0017] The score represents the matching degree between the physical examination information and the assigned object, and the V [cls] represents the first vector, the W 1 represents the first mapping matrix of the vector matching model, the The second mapping matrix W representing the vector matching model 2 The transpose of b 1 represents the bias matrix of the vector matching model, and σ represents the activation function of the vector matching model.

[0018] In an optional implementation, the semantic vector model includes a BERT model.

[0019] In an optional implementation, the vector matching model includes a fully connected layer network model.

[0020] In an optional implementation, before the step of inputting the physical examination information and the information of the assigned object into a pre-trained information matching model to obtain the matching degree between the physical examination information and the assigned object, the method further includes:

[0021] Acquire multiple training sample pairs, each of the training samples is a first sample pair or a second sample pair, the first sample pair includes sample physical examination information and information of a first sample assignment object, and the second sample pair includes the sample physical examination information and information of a second sample assignment object, wherein the matching degree of the first sample pair is greater than the matching degree of the second sample pair, the sample physical examination information includes abnormal sample physical examination information, the information of the sample assignment object includes specialty information of the sample assignment object, and the sample assignment object is the first sample assignment object or the second sample assignment object;

[0022] Inputting the training sample pair into the model to be trained to obtain the matching degree of the training sample pair;

[0023] The model to be trained is trained according to the matching degree of the first sample pair, the matching degree of the second sample pair, the workload information of the sample physical examination information, and the workload information of each of the sample allocation objects, and the trained model to be trained is determined as the information matching model.

[0024] In an optional implementation, the step of training the model to be trained according to the matching degree of the first sample pair, the matching degree of the second sample pair, the workload information of the sample physical examination information, and the workload information of each of the sample allocation objects, and determining the trained model to be trained as the information matching model includes:

[0025] The model to be trained is trained with the goal that the difference between the matching degree of the first sample pair and the matching degree of the second sample pair is greater than a first preset threshold and the first parameter of each of the sample allocation objects is equal, wherein the first preset threshold is greater than zero, the first parameter is used to characterize the workload of each of the sample allocation objects, and the first parameter is determined by the workload information of the sample physical examination information.

[0026] In an optional implementation, the step of training the model to be trained with the goal that the difference between the matching degree of the first sample pair and the matching degree of the second sample pair is greater than a first preset threshold and the first parameters of the sample allocation objects are equal includes:

[0027] Calculating a difference between a matching degree of the first sample pair and a matching degree of the second sample pair, and determining a first loss portion according to the difference and the first preset threshold;

[0028] Determine a second loss part according to a second parameter of the plurality of sample physical examination information, a third parameter of the plurality of sample allocation objects, and a first parameter of each of the sample allocation objects, wherein the second parameter is used to characterize the workload of the plurality of sample physical examination information, and the third parameter is the number of the plurality of sample allocation objects;

[0029] Performing a weighted summation of the first loss part and the second loss part to obtain a loss function;

[0030] The model to be trained is trained with the goal of minimizing the loss function.

[0031] In an optional implementation, the first loss portion is determined according to the following formula:

[0032] Wherein, L1 represents the first loss part, and u j represents the sample physical examination information, the d ij Indicates information about the first sample allocation object, the d il Indicates information about the second sample allocation object, the (u j ,d ij ) represents the first sample pair, the (u j ,d il ) represents the second sample pair, the score(u j ,d il ) represents the matching degree of the second sample pair, and the score(u j ,d ij ) represents the matching degree of the first sample pair, and γ represents the first preset threshold;

[0033] The second loss part is determined according to the following formula:

[0034] Wherein, L2 represents the second loss part, N represents the second parameter, M represents the third parameter, and h i A first parameter representing the sample allocation object i;

[0035] The loss function is determined according to the following formula:

[0036] L=L1+βL2, wherein L represents the loss function and β represents the harmonic parameter.

[0037] In an optional implementation, when the information matching model includes a semantic vector model and a vector matching model, the model to be trained includes a first model to be trained and a second model to be trained, and the step of inputting the training sample pair into the model to be trained to obtain the matching degree of the training sample pair includes:

[0038] Inputting the training sample pair into the first to-be-trained model to obtain a vector of the training sample pair;

[0039] Inputting the vector of the training sample pair into the second to-be-trained model to obtain the matching degree of the training sample pair;

[0040] The step of training the model to be trained according to the matching degree of the first sample pair, the matching degree of the second sample pair, the workload information of the sample physical examination information, and the workload information of each sample allocation object, and determining the trained model to be trained as the information matching model includes:

[0041] According to the matching degree of the first sample pair, the matching degree of the second sample pair, the workload information of the sample physical examination information, and the workload information of each of the sample allocation objects, the first model to be trained and the second model to be trained are trained, and the trained first model to be trained is determined as the semantic vector model, and the trained second model to be trained is determined as the vector matching model.

[0042] In an optional implementation, the matching degree between the physical examination information and the target allocation object is greater than or equal to a second preset threshold; or,

[0043] The matching degree between the physical examination information and the target allocation object is greater than the matching degree between the physical examination information and other allocation objects, and the other allocation objects are allocation objects other than the target allocation object among the multiple allocation objects.

[0044] According to a second aspect of the present application, a physical examination information distribution device is provided, the device comprising:

[0045] An information acquisition module is configured to acquire physical examination information and information of a plurality of assigned objects, wherein the physical examination information includes abnormal physical examination information and workload information of the physical examination information, and the information of the assigned objects includes specialty information and workload information of the assigned objects;

[0046] an information processing module, configured to input the physical examination information and the information of the assigned object into a pre-trained information matching model to obtain a matching degree between the physical examination information and the assigned object, wherein the information matching model is used to calculate the matching degree between the physical examination information and the assigned object based on the abnormal physical examination information, the workload information of the physical examination information, and the specialty information and workload information of the assigned object;

[0047] The information allocation module is configured to determine a target allocation object from the multiple allocation objects according to the matching degree between the physical examination information and each of the allocation objects, and allocate the physical examination information to the target allocation object.

[0048] According to a third aspect of the present application, an electronic device is provided, the electronic device comprising:

[0049] processor;

[0050] a memory for storing instructions executable by the processor;

[0051] The processor is configured to execute the instructions to implement the physical examination information allocation method as described in the first aspect.

[0052] According to a fourth aspect of the present application, a storage medium is provided. When instructions in the storage medium are executed by a processor of an electronic device, the electronic device can execute the physical examination information distribution method as described in the first aspect.

[0053] According to a fifth aspect of the present application, a computer program product is provided. When instructions in the computer program product are executed by a processor of an electronic device, the electronic device is enabled to execute the physical examination information distribution method as described in the first aspect.

[0054] Compared with the prior art, this application has the following advantages:

[0055] The technical solution of the present application provides a method, device, electronic device, computer-readable storage medium and computer program product for distributing physical examination information, wherein the method for distributing physical examination information includes: first obtaining physical examination information and information of multiple distribution objects; then inputting the physical examination information and the information of the distribution objects into a pre-trained information matching model to obtain the matching degree between the physical examination information and the distribution objects, wherein the information matching model is used to calculate the matching degree between the physical examination information and the distribution objects according to the abnormal physical examination information, the workload information of the physical examination information and the specialty information and workload information of the distribution objects; then, according to the matching degree between the physical examination information and each distribution object, determining the target distribution object from the multiple distribution objects, and distributing the physical examination information to the target distribution object. The technical solution of the present application calculates the matching degree between the physical examination information and the distribution object through the information matching model, and can comprehensively consider the matching degree between the distribution object and the physical examination information from the two aspects of professional matching and workload matching, ensuring that the physical examination information is distributed to the distribution object with professional matching, while also taking into account the workload of each distribution object, so that the workload between each distribution object is balanced, thereby achieving the matching between the physical examination information and the distribution object.

[0056] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0058] Figure 1 The figure is a flow chart of a method for allocating physical examination information according to an exemplary embodiment.

[0059] Figure 2 The figure is a flowchart of a matching degree calculation according to an exemplary embodiment.

[0060] Figure 3 The figure is a schematic diagram showing how a semantic vector model processes information according to an exemplary embodiment.

[0061] Figure 4 The figure is a schematic diagram showing information processing by a vector matching model according to an exemplary embodiment.

[0062] Figure 5The figure is a flow chart showing a model obtaining method according to an exemplary embodiment.

[0063] Figure 6 The figure is a flowchart of a model training method according to an exemplary embodiment.

[0064] Figure 7 It is a structural block diagram of a physical examination information distribution device according to an exemplary embodiment.

[0065] Figure 8 It is a block diagram of an electronic device according to an exemplary embodiment.

[0066] Fig. 9 is a block diagram of another electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0067] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0068] Assume that the physical examination center has N physical examination information and M general inspectors. A physical examination information is distributed to each general inspector in turn. When the general inspector completes the health assessment and recommendations, the system will distribute a physical examination information to the doctor in sequence from the remaining physical examination information. Assume that there are two general inspectors A and B, A is good at internal medicine, and B is good at surgery. The internal medicine examination results in the physical examination information u are abnormal, and the surgical examination results are normal. Intuitively, the physical examination information u should be distributed to the general inspector A. If this physical examination information u is distributed to the general inspector B, there will be a problem that the physical examination information does not match the assigned general inspector.

[0069] In addition, after the chief inspector completes the work, the reports are distributed in order. If the total number of physical examination reports is constant, the workload of the chief inspector with high efficiency will be very large. For example, assuming that the efficiency of chief inspector A is twice that of chief inspector B, then chief inspector A will complete 2N / 3 reports, and chief inspector B will complete N / 3 reports. The workload of chief inspector A is twice that of chief inspector B.

[0070] In order to accurately match physical examination information and assigned objects, Figure 1 is a flow chart of a method for distributing physical examination information according to an exemplary embodiment. Figure 1 As shown, the method may include the following steps.

[0071] Step S11: obtaining physical examination information and information of multiple assigned objects, wherein the physical examination information includes abnormal physical examination information and workload information of the physical examination information, and the information of the assigned objects includes specialty information and workload information of the assigned objects.

[0072] The execution subject of this embodiment may be an electronic device, such as a server, etc., or a user terminal device, such as a mobile phone, a computer, etc.

[0073] The physical examination information may be, for example, a physical examination report that has completed a single examination result but needs to provide a general examination result. The physical examination information may include the examination results of a series of physical examination items (such as general examination, internal examination, electrocardiogram, head CT, etc.) (for example, the examination results of the electrocardiogram items include: sinus rhythm, etc.).

[0074] In this embodiment, the physical examination information may include abnormal physical examination information, and the abnormal physical examination information may include abnormal examination results or abnormal physical examination items, etc., and this embodiment does not limit the specific content of the abnormal physical examination information. The physical examination information may also include workload information of the physical examination information, and the workload information of the physical examination information may be determined by the number of words in the abnormal examination results, etc., and this embodiment does not limit this.

[0075] The assigned object may be, for example, a chief inspector. The information of the assigned object may include the specialty information of the assigned object, and the specialty information may include a description of the chief inspector's graduate school, major, years of work, and specialized inspection items, etc. The specific content of the specialty information is not limited in this embodiment. For example, the specialty information may be Doctor A, who graduated from **, ..., worked for ** years, ..., and is specialized in internal medicine... The information of the assigned object may also include the workload information of the assigned object, and the workload information may be determined by the number of physical examination reports that the chief inspector has processed, such as Doctor A has processed M physical examination reports, Doctor B has processed N physical examination reports, and so on.

[0076] The physical examination information can be obtained from each single examination device, for example, the examination results of each single examination device are given by each sub-examination doctor or automatically analyzed by the device. For example, the information of the assigned object can be pre-entered and stored, for example, and can be entered by each doctor through a user interaction interface.

[0077] Step S12: Input the physical examination information and the information of the assigned object into a pre-trained information matching model to obtain the matching degree between the physical examination information and the assigned object. The information matching model is used to calculate the matching degree between the physical examination information and the assigned object based on the abnormal physical examination information, the workload information of the physical examination information, and the specialty information and workload information of the assigned object.

[0078] Specifically, the information matching model can calculate the professional matching degree between the physical examination information and the assigned object based on the abnormal physical examination information of the physical examination information and the specialty information of the assigned object, and can also calculate the workload matching degree between the physical examination information and the assigned object based on the workload information of the physical examination information and the workload information of the assigned object. That is, the matching degree between the physical examination information and the assigned object calculated in step S12 includes the professional matching degree and the workload matching degree between the physical examination information and the assigned object.

[0079] When the physical examination information is a physical examination report and the assigned object is a chief physician, the professional match between the physical examination information and the assigned object refers to whether the examination results in the physical examination information are interpreted and processed by the chief physician who is good at the examination results. For example, if the internal medicine examination results in the physical examination information u are abnormal, then the internist A is more professionally suitable to handle the physical examination report than the surgeon B, that is, the professional match between the chief physician A and the physical examination report u is higher.

[0080] The workload matching between the physical examination information and the assigned objects means that the workload between multiple assigned objects can be balanced through the allocation of the physical examination information.

[0081] In this embodiment, the information matching model can be obtained by iterative training in advance based on the sample physical examination information, the first sample allocation object, and the second sample allocation object, wherein the professional matching degree between the first sample allocation object and the sample physical examination information is greater than the professional matching degree between the second sample allocation object and the sample physical examination information. Specifically, during the training process, the first matching degree between the first sample allocation object and the sample physical examination information, as well as the second matching degree between the second sample allocation object and the sample physical examination information can be calculated, and the information matching model is trained with the goal that the difference between the first matching degree and the second matching degree is greater than the first preset threshold and the workload between the first sample allocation object and the second sample allocation object is balanced. The subsequent embodiments will provide a detailed introduction to the specific training process of the information matching model.

[0082] The use of the information matching model can comprehensively consider the matching degree between the assigned objects and the physical examination information from the two aspects of professional matching and workload matching, ensuring that the physical examination information is assigned to the assigned objects with professional matching while also taking into account the workload of each assigned object, so that the workload among the assigned objects is balanced.

[0083] Step S13: determining a target assignment object from multiple assignment objects according to the matching degree between the physical examination information and each assignment object, and assigning the physical examination information to the target assignment object.

[0084] In a specific implementation, the matching degree between the physical examination information and each assigned object calculated in step S12 can be used. Assuming there are M assigned objects, the matching degree scores between the M assigned objects and the physical examination information u can be obtained, and then the target assigned object can be determined according to the matching degree between the physical examination information and each assigned object, and the physical examination information can be distributed to the target assigned object.

[0085] In some embodiments of the present disclosure, the execution order of each step is not limited and can be adjusted according to actual needs.

[0086] In this embodiment, there are multiple implementation methods for determining the target allocation object from multiple allocation objects.

[0087] In one implementation, the matching degree between the physical examination information and the target allocation object may be greater than or equal to a second preset threshold. The specific value of the second preset threshold may be determined according to actual needs and is not limited in this embodiment.

[0088] In another implementation, the matching degree between the physical examination information and the target allocation object may be greater than the matching degree between the physical examination information and other allocation objects. The other allocation objects are allocation objects other than the target allocation object among the multiple allocation objects. In this implementation, the allocation object with the greatest matching degree is determined as the target allocation object.

[0089] Given the physical examination information u, the information set of multiple assigned objects {d 1 ,d 2 ,…,d N}, function f(u,d i )∈R+ (R+ represents a positive real number) can be used to measure the physical examination information u and the information d of the assigned object i The larger the function value, the more the physical examination information u matches the assigned object. In practical applications, the physical examination report u is assigned to the target assigned object, where the information of the target assigned object can be expressed as That is, the target allocation object can be the allocation object with the highest matching degree with the physical examination report among multiple allocation objects. In this embodiment, the function f(u,d i ) represents the pre-trained information matching model.

[0090] The physical examination information allocation method provided in this embodiment first obtains the physical examination information and the information of multiple allocation objects; then the physical examination information and the information of the allocation objects are input into the pre-trained information matching model to obtain the matching degree between the physical examination information and the allocation objects; then, according to the matching degree between the physical examination information and each allocation object, the target allocation object is determined from the multiple allocation objects, and the physical examination information is allocated to the target allocation object. In this embodiment, the matching degree between the physical examination information and the allocation object is calculated by the information matching model, and the matching degree between the allocation object and the physical examination information can be comprehensively considered from the two aspects of professional matching and workload matching, ensuring that the physical examination information is allocated to the professionally matched allocation object while taking into account the workload of each allocation object, so that the workload between each allocation object is balanced, thereby achieving the matching between the physical examination information and the allocation object.

[0091] In an optional implementation, the information matching model includes a semantic vector model and a vector matching model. Figure 2 Step S12 may specifically include:

[0092] Step S21: input the physical examination information and the information of the assigned object into the semantic vector model, and output a first vector.

[0093] Step S22: input the first vector into a vector matching model, and output the matching degree between the physical examination information and the assigned object.

[0094] The semantic vector model is used to map the physical examination information and the information of the assigned object into the same vector space; and the vector matching model is used to calculate the matching degree between the physical examination information and the assigned object in the vector space.

[0095] The physical examination information and the information of the assigned object are very different in word form, but they are semantically related. Therefore, they can be mapped to the semantic vector space. In the semantic vector space, semantically related content is located close to each other. Therefore, the matching degree between the physical examination information and the information of the assigned object can be measured in the semantic vector space.

[0096] In this step, the physical examination information and the information of the assigned object are mapped into the same vector space through the pre-trained semantic vector model, and then the matching degree between the physical examination information and the assigned object is further calculated in the vector space.

[0097] The semantic vector model can be a language model such as a BERT model, an RNN model, or an LSTM model. Since the BERT model is a truly bidirectional language model, each word can use the context information of the word at the same time, so the semantic vector model using the BERT model can obtain a more accurate semantic vector and further improve the matching accuracy.

[0098] In this embodiment, the semantic vector model is trained using the BERT model. In an optional implementation, the physical examination information and the information of the assigned object may be first concatenated to obtain concatenated information, the header of the concatenated information uses a marker symbol [CLS], and the physical examination information and the information of the assigned object are connected using a separator symbol [SEP]; then the concatenated information is input into the semantic vector model to output a first vector.

[0099] In the specific implementation, we can first use technologies such as word embedding to convert the text-based physical examination information and the information of the assigned object into word vectors. Figure 3 "Character i" represents the i-th character in the corresponding text (physical examination information or information of the assigned object). The text "character i" corresponding to each character is converted into a vector Ei or Ei' through word embedding and other methods. Among them, Ei is the word vector of the physical examination information, and Ei' is the word vector of the information of the assigned object. Then, Ei and Ei' can be used as initialization vectors, and the physical examination information in the form of word vectors and the information of the assigned object can be spliced ​​using the marking symbol [CLS] and the separator symbol [SEP] to obtain the spliced ​​information. Specifically, the marking symbol [CLS] can be added to the header of the spliced ​​information, and the separator symbol [SEP] can be added between the physical examination information u and the information d of the assigned object. Afterwards, the spliced ​​information can be input as a whole into a semantic vector model such as the BERT model to obtain the first vector.

[0100] Among them, the marking symbol [CLS] is used as the head of the splicing information, indicating that the output vector V[cls] corresponding to the symbol can be used as the semantic representation or semantic vector of the splicing information. The marking symbol [CLS] can more "fairly" integrate the semantic information of each word or phrase in the splicing information. The separation symbol [SEP] is used to distinguish the physical examination information in the splicing information from the information of the assigned object.

[0101] like Figure 3 As shown, the first vector V[cls] is the output result of the semantic vector model, and the first vector V[cls] contains the physical examination information u and the information d of the assigned object. In some embodiments, the first vector V[cls] represents the vector corresponding to the input physical examination information and the assigned object information. When the semantic vector model adopts the BERT model, the dimension of V[cls] is 768 dimensions. In this embodiment, the default vector dimension of the adopted BERT model is 768 dimensions. Of course, the vector in the BERT model can also be of other dimensions, which is not limited in this embodiment. Figure 3 The Vi or Vi' in represents the vector output after "word i" is processed by the semantic vector model.

[0102] In a specific implementation, the information pair consisting of the physical examination information and the information of each assigned object can be input into a pre-trained semantic vector model to obtain a first vector corresponding to the input physical examination information and the information of the assigned object. In this way, the physical examination information and the information of each assigned object can be mapped to the same vector space through the semantic vector model.

[0103] In an optional implementation, the vector matching model may adopt a fully connected layer network model. It should be noted that the vector matching model is not limited to the fully connected layer network model, and any network model that can transform an input vector into a scalar may be used.

[0104] In an optional implementation, the vector matching model may calculate the matching degree between the physical examination information and the assigned object according to the following formula:

[0105] Among them, score represents the matching degree between the physical examination information and the assigned object, V [cls] represents the first vector, W 1 Represents the first mapping matrix of the vector matching model, The second mapping matrix W representing the vector matching model 2 The transpose of 1 represents the bias matrix of the vector matching model, and σ represents the activation function of the vector matching model.

[0106] like Figure 4 As shown, the first vector V[cls] can be input into the vector matching model to output the matching score between the physical examination information and the assigned object. Figure 4 x in i Represents the value of the i-th dimension in the first vector V[cls]. When the semantic vector model adopts the BERT model, W 1 Can be a 384*768 matrix, W 2 and b 1 It can be a 384*1 matrix.

[0107] The parameters of the semantic vector model and the parameters such as matrices W1, W2 and b1 in the vector matching model can be obtained through joint training. Among them, the parameters of the semantic vector model can use the parameters in the pre-trained model as initial values, and the matrices W1, W2 and b1 in the vector matching model can obtain initial values ​​through random initialization. For example, a large amount of data is used to pre-train the model to obtain the various parameters in the pre-trained model, and then the parameters are carefully adjusted on specific tasks. The specific training process of the semantic vector model and the vector matching model will be described in detail in the subsequent embodiments.

[0108] In this embodiment, the physical examination information and the information of the assigned object are mapped to the same vector space through a semantic vector model, the matching degree between the physical examination information and each assigned object is calculated in the vector space, and then the physical examination information is assigned to a suitable assigned object according to the matching degree.

[0109] In order to obtain the information matching model, in an optional implementation, refer to Figure 5 Before step S12, the physical examination information distribution method provided in the embodiment of the present application may further include the following steps.

[0110] Step S51: Acquire multiple training sample pairs, each training sample is a first sample pair or a second sample pair, the first sample pair includes sample physical examination information and information of a first sample assignment object, and the second sample pair includes sample physical examination information and information of a second sample assignment object, wherein the matching degree of the first sample pair is greater than the matching degree of the second sample pair, the sample physical examination information includes abnormal sample physical examination information, the information of the sample assignment object includes specialty information of the sample assignment object, and the sample assignment object is the first sample assignment object or the second sample assignment object.

[0111] The sample physical examination information may be, for example, a sample physical examination report that has completed a single examination result but needs to provide a general examination result. The sample physical examination information may include the examination results of a series of physical examination items (such as general examination, internal examination, electrocardiogram, head CT, etc.) (for example, the examination results of the electrocardiogram item include: sinus rhythm, etc.).

[0112] In this embodiment, the sample physical examination information may include abnormal sample physical examination information, and the abnormal sample physical examination information may include abnormal examination results or abnormal physical examination items, etc. This embodiment does not limit the specific content of the abnormal sample physical examination information.

[0113] The sample allocation object may be, for example, a chief physician. The information of the sample allocation object may include the specialty information of the sample allocation object. The specialty information may include a description of the chief physician's graduation school, major, years of work, and specialized examination items. This embodiment does not limit the specific content of the specialty information. For example, the specialty information may be Doctor A, graduated from **, ..., worked for ** years, ..., specialized in internal medicine...

[0114] In this embodiment, the first sample allocation object can be a sample allocation object that matches the professional of the sample physical examination information, that is, the specialty information of the first sample allocation object matches the abnormal sample physical examination information in the sample physical examination information. The second sample allocation object can be a sample allocation object that does not match the professional of the sample physical examination information, that is, the specialty information of the second sample allocation object does not match the abnormal sample physical examination information in the sample physical examination information. In the process of obtaining training sample pairs, for example, the first sample allocation object and the second sample allocation object can be manually specified based on the sample physical examination information and the information of the sample allocation object. For example, if the professional matching degree of the first sample allocation object is greater than the professional matching degree of the second sample allocation object, it can be considered that the first sample allocation object is a sample allocation object that matches the professional of the sample physical examination information, and the second sample allocation object is a sample allocation object that does not match the professional of the sample physical examination information.

[0115] The first sample pair (u j ,d ij ), including sample physical examination information u j and the information d of the first sample allocation object ij (where the subscript ij∈[1,M], M is the total number of sample allocation objects); the second sample pair (u j ,d il ), including sample physical examination information u j and the information d of the second sample allocation object il (where subscript il∈[1,M] and ij≠il). j ,d ij ) can be greater than the second sample pair (u j ,d il ), that is, the sample physical examination information is more matched with the major of the first sample assignment object.

[0116] Step S52: input the training sample pair into the model to be trained to obtain the matching degree of the training sample pair.

[0117] Step S53: Train the model to be trained according to the matching degree of the first sample pair, the matching degree of the second sample pair, the workload information of the sample physical examination information, and the workload information of each sample allocation object, and determine the trained model to be trained as the information matching model.

[0118] The workload information of the sample physical examination information may be determined, for example, by the number of sample physical examination reports included in the sample physical examination information, or by the number of words in abnormal examination results, etc., which is not limited in this embodiment.

[0119] The workload information of the sample allocation object can be determined by the workload of the sample physical examination information matching the profession of the sample allocation object, for example, it can be the number of sample physical examination information matching the profession of the sample allocation object.

[0120] In an optional implementation, when the information matching model includes a semantic vector model and a vector matching model, the model to be trained includes a first model to be trained and a second model to be trained, and step S52 may specifically include: first, inputting the training sample pair into the first model to be trained to obtain the vector of the training sample pair; then inputting the vector of the training sample pair into the second model to be trained to obtain the matching degree of the training sample pair, and step S53 may specifically include: training the first model to be trained and the second model to be trained according to the matching degree of the first sample pair, the matching degree of the second sample pair, the workload information of the sample physical examination information, and the workload information of each sample allocation object, and determining the trained first model to be trained as a semantic vector model, and determining the trained second model to be trained as a vector matching model.

[0121] The first model to be trained can be a language model such as a BERT model, an RNN model or an LSTM model. Since the BERT model is a truly bidirectional language model, each word can simultaneously utilize the context information of the word, so the first model to be trained can obtain a more accurate semantic vector by using the BERT model, further improving the matching accuracy.

[0122] The second model to be trained may adopt a fully connected layer network model. It should be noted that the second model to be trained is not limited to the fully connected layer network model, and any network model that can transform an input vector into a scalar may be used.

[0123] In a specific implementation, the training sample pair can be input into the first model to be trained to obtain a vector corresponding to the input training sample pair, and then the vector of the training sample pair can be input into the second model to be trained to obtain the matching degree corresponding to the input training sample pair. In this way, the matching degree of all training sample pairs can be obtained, including the first sample pair (u j ,d ij )'s matching degree score(u j ,d ij ) and the second sample pair (u j ,d il )'s matching degree score(u j ,d il ). Afterwards, the first model to be trained and the second model to be trained or the model to be trained can be trained according to the matching degree of the first sample pair, the matching degree of the second sample pair, the workload information of the sample physical examination information, and the workload information of each sample allocation object.

[0124] In an optional implementation, in step S53, the matching degree score (u j ,d ij ) and the matching degree score(u j ,d il ) is greater than the first preset threshold γ, and the first parameter of each sample allocation object is equal as the goal, the training model to be trained is trained, wherein the first preset threshold is greater than zero, the sample allocation object is the first sample allocation object or the second sample allocation object, the first parameter is used to characterize the workload of each sample allocation object, and the first parameter is determined by the workload information of the sample physical examination information. For example, the first parameter can be the workload of the sample physical examination information that matches the specialty of the sample allocation object.

[0125] In some embodiments, “equal” can mean that the two are completely equal, or the difference between the two can be less than a certain threshold, and the threshold can be set as needed.

[0126] Specifically, refer to Figure 6 , in step S53, it may include:

[0127] Step S61: Calculate the difference between the matching degree of the first sample pair and the matching degree of the second sample pair, and determine the first loss part according to the difference and a first preset threshold.

[0128] In practical applications, the matching score (u j ,d ij ) is better than the matching score(u j ,d il ) is at least higher than the first preset threshold γ as the first training target, that is, score(u j ,d ij )>score(u j ,d il )+γ, where il∈[1,M], ij∈[1,M] and ij≠il. The first preset threshold is greater than zero, for example, it can be set to 1, and the specific data can be set according to actual conditions, which is not limited in this embodiment.

[0129] In a specific implementation, the first loss part may be determined according to the following formula:

[0130] Among them, L1 represents the first loss part, u j Indicates sample physical examination information, d ij Indicates the information of the first sample allocation object, d il Indicates the information of the second sample allocation object, (u j ,d ij) represents the first sample pair, (u j ,d il ) represents the second sample pair, score(u j ,d il ) represents the matching degree of the second sample pair, score(u j ,d ij ) represents the matching degree of the first sample pair, and γ represents the first preset threshold. If a≤0, then [a]+=0; otherwise [a]+=a.

[0131] Step S62: Determine the second loss part according to the second parameter of the multiple sample physical examination information, the third parameter of the multiple sample allocation objects and the first parameter of each sample allocation object, wherein the second parameter is used to characterize the workload of the multiple sample physical examination information, and the third parameter is the number of the multiple sample allocation objects.

[0132] In this embodiment, the second training goal can be to make the first parameter of each sample allocation object equal, or to minimize the difference between the first parameter of each sample allocation object and the average workload, so that the trained model can balance the workload among the allocation objects.

[0133] In a specific implementation, the ratio of the second parameter to the third parameter may be calculated first; and then the second loss portion may be determined based on the difference between the first parameter and the ratio.

[0134] Assume that the workload of completing each sample physical examination information u is u_c i , where i∈[1,k], k represents the total number of sample physical examination information, then the total workload is the second parameter The average workload is M is the number of sample assignment objects as the third parameter. Assume that the sample physical examination information matching the sample assignment object j includes {u j1 ,…,u jh}, the workload imbalance of the sample allocation object can be recorded as in, The first parameter is the workload of sample physical examination information that matches the specialty of sample assignment object j.

[0135] In order to simplify the description, in this embodiment, it is assumed that the workload of each piece of physical examination information is the same, which is set to 1, that is, but Where N is the total number of sample physical examination information as the second parameter, and M is the number of sample allocation objects, i.e. the third parameter. So the first parameter of the sample allocation object j is but

[0136] In a specific implementation, the second loss part can be determined according to the following formula:

[0137] Where L2 represents the second loss part, N represents the second parameter, M represents the third parameter, and h i The first parameter representing the sample assignment object i.

[0138] According to the calculation formula of the second loss part, when the number of sample physical examination information processed by each sample allocation object is the average N / M, the workload of each sample allocation object is the same, which is the average workload. In this case, the minimum loss of workload balance is 0. When the number of sample physical examination information processed by each sample allocation object varies greatly, the loss will increase.

[0139] Step S63: Perform a weighted summation on the first loss part and the second loss part to obtain a loss function.

[0140] In the specific implementation, the loss function can be determined according to the following formula:

[0141] L=L1+βL2, where L represents the loss function and β represents the harmonic parameter.

[0142] Step S64: Train the model to be trained with the goal of minimizing the loss function.

[0143] In the specific implementation, batch gradient descent can be used to minimize the loss function to learn the model parameters.

[0144] In this embodiment, the first loss part is used to measure the professional matching loss between the sample physical examination information and each sample allocation object, and the second loss part is used to measure the balance loss between the workloads of each sample allocation object. j ,d ij ) is better than the matching score(u j ,d il ) is higher than γ and the workload (first parameter) processed by each assigned object is the same, the loss function L is minimized to 0.

[0145] From the formula of loss function L, we can know that if each piece of physical examination information is distributed to the assigned object with the best professional match (highest matching degree), some assigned objects will process a lot of physical examination information, while some assigned objects will process very little total inspection information. In this case, the first loss part is small, but the second loss part will be large. During model training, the total loss function L can be minimized by adjusting the parameters, so that part of the physical examination information can be distributed to the assigned objects with a higher professional match (such as the second or third highest professional match), so that the workload between the assigned objects is relatively balanced.

[0146] By using the above loss function to train the model to be trained, an information matching model can be obtained. By using the above loss function to train the first model to be trained and the second model to be trained, a semantic vector model and a vector matching model can be obtained. By using the trained information matching model or the semantic vector model and the vector matching model, the matching degree between the new physical examination information and each assigned object can be calculated, so that a target assigned object with a high professional matching degree and balanced workload can be selected from multiple assigned objects.

[0147] By constructing a semantic vector model and a vector matching model, the physical examination information and the information of the assigned object are mapped into the same vector space, and the target assigned object is determined in the vector space according to the professional matching degree and the workload balance, and then the physical examination information is distributed to the target assigned object. This scheme takes into account the professional matching degree between the physical examination information and the assigned object and the workload balance between the assigned objects. Compared with the traditional method, this scheme can assign the physical examination information to the assigned object with a higher matching degree, and at the same time can reasonably distribute the workload between the assigned objects, so as to achieve workload balance between the assigned objects.

[0148] In practical applications, the physical examination information distribution method provided in this embodiment can be executed by a server and sent to a terminal, or executed by a terminal. This embodiment does not limit the execution subject of the physical examination information distribution method.

[0149] Figure 7 FIG. 1 is a block diagram of a physical examination information distribution device according to an exemplary embodiment. Figure 7 , which may include:

[0150] The information acquisition module 71 is configured to acquire physical examination information and information of a plurality of assignment objects, wherein the physical examination information includes abnormal physical examination information and workload information of the physical examination information, and the assignment object information includes specialty information and workload information of the assignment object;

[0151] The information processing module 72 is configured to input the physical examination information and the information of the assigned object into a pre-trained information matching model to obtain a matching degree between the physical examination information and the assigned object, wherein the information matching model is used to calculate the matching degree between the physical examination information and the assigned object based on the abnormal physical examination information, the workload information of the physical examination information, and the specialty information and workload information of the assigned object;

[0152] The information allocation module 73 is configured to determine a target allocation object from the multiple allocation objects according to the matching degree between the physical examination information and each of the allocation objects, and allocate the physical examination information to the target allocation object.

[0153] The physical examination information distribution device provided in this embodiment can be applied to electronic devices, such as servers, etc., and can also be applied to user terminal devices, such as mobile phones, computers, etc.

[0154] The physical examination information may be, for example, a physical examination report that has completed a single examination result but needs to provide a general examination result. The physical examination information may include the examination results of a series of physical examination items (such as general examination, internal examination, electrocardiogram, head CT, etc.) (for example, the examination results of the electrocardiogram items include: sinus rhythm, etc.).

[0155] In this embodiment, the physical examination information may include abnormal physical examination information, and the abnormal physical examination information may include abnormal examination results or abnormal physical examination items, etc., and this embodiment does not limit the specific content of the abnormal physical examination information. The physical examination information may also include workload information of the physical examination information, and the workload information of the physical examination information may be determined, for example, by the number of copies of the physical examination report included in the physical examination information, or by the number of words in the abnormal examination results, etc., and this embodiment does not limit this.

[0156] The assigned object may be, for example, a chief inspector. The information of the assigned object may include the specialty information of the assigned object, and the specialty information may include a description of the chief inspector's graduate school, major, years of work, and specialized inspection items. The specific content of the specialty information is not limited in this embodiment. For example, the specialty information may be Doctor A, who graduated from **, ..., worked for ** years, ..., and is specialized in internal medicine... The information of the assigned object may also include the workload information of the assigned object, and the workload information may be determined by the number of physical examination reports that the chief inspector has processed, such as Doctor A has processed M physical examination reports, Doctor B has processed N physical examination reports, and so on.

[0157] The physical examination information can be obtained from each single examination device by the information acquisition module 71, for example, the examination results of each single examination device are given by each sub-examination doctor or automatically analyzed by the device. For example, the information of the assigned object can be pre-entered and stored, for example, and can be entered by each doctor through a user interaction interface.

[0158] Specifically, the information matching model can calculate the professional matching degree between the physical examination information and the assigned object based on the abnormal physical examination information of the physical examination information and the specialty information of the assigned object, and can also calculate the workload matching degree between the physical examination information and the assigned object based on the workload information of the physical examination information and the workload information of the assigned object. That is, the matching degree between the physical examination information and the assigned object calculated by the information processing module 72 includes the professional matching degree and the workload matching degree between the physical examination information and the assigned object.

[0159] When the physical examination information is a physical examination report and the assigned object is a chief physician, the professional match between the physical examination information and the assigned object refers to whether the examination results in the physical examination information are interpreted and processed by the chief physician who is good at the examination results. For example, if the internal medicine examination results in the physical examination information u are abnormal, then the internist A is more professionally suitable to handle the physical examination report than the surgeon B, that is, the professional match between the chief physician A and the physical examination report u is higher.

[0160] The workload matching between the physical examination information and the assigned objects means that the workload between multiple assigned objects can be balanced through the allocation of the physical examination information.

[0161] In this embodiment, the information matching model can be obtained by iterative training in advance based on the sample physical examination information, the first sample allocation object, and the second sample allocation object, wherein the professional matching degree between the first sample allocation object and the sample physical examination information is greater than the professional matching degree between the second sample allocation object and the sample physical examination information. Specifically, during the training process, the first matching degree between the first sample allocation object and the sample physical examination information, as well as the second matching degree between the second sample allocation object and the sample physical examination information can be calculated, and the information matching model is trained with the goal that the difference between the first matching degree and the second matching degree is greater than the first preset threshold and the workload between the first sample allocation object and the second sample allocation object is balanced. The subsequent embodiments will provide a detailed introduction to the specific training process of the information matching model.

[0162] The information processing module 72 uses an information matching model to comprehensively consider the matching degree between the assigned object and the physical examination information from the two aspects of professional matching and workload matching, ensuring that the physical examination information is assigned to the assigned object with professional matching while also taking into account the workload of each assigned object, so that the workload between the assigned objects is balanced.

[0163] In a specific implementation, the information distribution module 73 may use the matching degree between the physical examination information and each distribution object calculated by the information processing module 72. Assuming there are M distribution objects, the matching degree score between the M distribution objects and the physical examination information u can be obtained, and then the information distribution module 73 can determine the target distribution object according to the matching degree between the physical examination information and each distribution object, and distribute the physical examination information to the target distribution object.

[0164] In some embodiments of the present disclosure, the execution order of each module is not limited and can be adjusted according to actual needs.

[0165] In this embodiment, there are multiple implementation methods for the information distribution module 73 to determine the target distribution object from multiple distribution objects.

[0166] In one implementation, the matching degree between the physical examination information and the target allocation object may be greater than or equal to a second preset threshold. The specific value of the second preset threshold may be determined according to actual needs and is not limited in this embodiment.

[0167] In another implementation, the matching degree between the physical examination information and the target allocation object may be greater than the matching degree between the physical examination information and other allocation objects. The other allocation objects are allocation objects other than the target allocation object among the multiple allocation objects. In this implementation, the allocation object with the greatest matching degree is determined as the target allocation object.

[0168] Given the physical examination information u, the information set of multiple assigned objects {d 1 ,d 2 ,…,d N}, function f(u,d i )∈R+ (R+ represents a positive real number) can be used to measure the physical examination information u and the information d of the assigned object i The larger the function value, the more the physical examination information u matches the assigned object. In practical applications, the physical examination report u is assigned to the target assigned object, where the information of the target assigned object can be expressed as That is, the target allocation object can be the allocation object with the highest matching degree with the physical examination report among multiple allocation objects. In this embodiment, the function f(u,d i ) represents the pre-trained information matching model.

[0169] The physical examination information distribution device provided in this embodiment calculates the matching degree between the physical examination information and the distribution object through the information matching model, and can comprehensively consider the matching degree between the distribution object and the physical examination information from the two aspects of professional matching and workload matching, thereby ensuring that the physical examination information is distributed to the distribution object with professional matching while also taking into account the workload of each distribution object, so that the workload between each distribution object is balanced, thereby achieving a match between the physical examination information and the distribution object.

[0170] In an optional implementation, the information matching model includes a semantic vector model and a vector matching model, and the information processing module 72 is specifically configured as follows:

[0171] Inputting the physical examination information and the information of the assigned object into the semantic vector model, and outputting a first vector;

[0172] The first vector is input into the vector matching model, and the matching degree between the physical examination information and the assigned object is output.

[0173] In an optional implementation, the information processing module 72 is specifically configured as follows:

[0174] The physical examination information and the information of the assigned object are spliced ​​to obtain spliced ​​information, wherein the header of the spliced ​​information uses a marking symbol [CLS], and the physical examination information and the information of the assigned object are connected using a separator symbol [SEP];

[0175] The splicing information is input into the semantic vector model, and the first vector is output.

[0176] In an optional implementation, the information processing module 72 is specifically configured as follows:

[0177] The vector matching model calculates the matching degree between the physical examination information and the assigned object according to the following formula:

[0178] The score represents the matching degree between the physical examination information and the assigned object, and the V [cls] represents the first vector, the W 1 represents the first mapping matrix of the vector matching model, The second mapping matrix W representing the vector matching model 2 The transpose of b 1 represents the bias matrix of the vector matching model, and σ represents the activation function of the vector matching model.

[0179] In an optional implementation, the semantic vector model includes a BERT model.

[0180] In an optional implementation, the vector matching model includes a fully connected layer network model.

[0181] The semantic vector model is used to map the physical examination information and the information of the assigned object into the same vector space; and the vector matching model is used to calculate the matching degree between the physical examination information and the assigned object in the vector space.

[0182] The physical examination information and the information of the assigned object are very different in word form, but are semantically related. Therefore, the information processing module 72 can map them into a semantic vector space, where semantically related contents are located close to each other. Therefore, the matching degree between the physical examination information and the information of the assigned object can be measured in the semantic vector space.

[0183] The information processing module 72 maps the physical examination information and the information of the assigned object into the same vector space through the pre-trained semantic vector model, and further calculates the matching degree between the physical examination information and the assigned object in the vector space.

[0184] The semantic vector model can be a language model such as a BERT model, an RNN model, or an LSTM model. Since the BERT model is a truly bidirectional language model, each word can use the context information of the word at the same time, so the semantic vector model using the BERT model can obtain a more accurate semantic vector and further improve the matching accuracy.

[0185] In this embodiment, the semantic vector model is trained using the BERT model. In an optional implementation, the information processing module 72 may first concatenate the physical examination information with the information of the assigned object to obtain concatenated information, wherein the header of the concatenated information uses a marking symbol [CLS], and the physical examination information and the information of the assigned object are connected using a separator symbol [SEP]; and then the concatenated information is input into the semantic vector model to output a first vector.

[0186] In a specific implementation, the information processing module 72 may first convert the text-based physical examination information and the information of the assigned object into word vectors using a technology such as word embedding. Figure 3"Character i" represents the i-th character in the corresponding text (physical examination information or information of the assigned object). The text "character i" corresponding to each character is converted into a vector Ei or Ei' through word embedding and other methods. Among them, Ei is the word vector of the physical examination information, and Ei' is the word vector of the information of the assigned object. Then, Ei and Ei' can be used as initialization vectors, and the physical examination information in the form of word vectors and the information of the assigned object can be spliced ​​using the marking symbol [CLS] and the separator symbol [SEP] to obtain the spliced ​​information. Specifically, the marking symbol [CLS] can be added to the header of the spliced ​​information, and the separator symbol [SEP] can be added between the physical examination information u and the information d of the assigned object. Afterwards, the spliced ​​information can be input as a whole into a semantic vector model such as the BERT model to obtain the first vector.

[0187] Among them, the marking symbol [CLS] is used as the head of the splicing information, indicating that the output vector V[cls] corresponding to the symbol can be used as the semantic representation or semantic vector of the splicing information. The marking symbol [CLS] can more "fairly" integrate the semantic information of each word or phrase in the splicing information. The separation symbol [SEP] is used to distinguish the physical examination information in the splicing information from the information of the assigned object.

[0188] like Figure 3 As shown, the first vector V[cls] is the output result of the semantic vector model, and the first vector V[cls] contains the physical examination information u and the information d of the assigned object. In some embodiments, the first vector V[cls] represents the vector corresponding to the input physical examination information and the assigned object information. When the semantic vector model adopts the BERT model, the dimension of V[cls] is 768 dimensions. In this embodiment, the default vector dimension of the adopted BERT model is 768 dimensions. Of course, the vector in the BERT model can also be of other dimensions, which is not limited in this embodiment. Figure 3 The Vi or Vi' in represents the vector output after "word i" is processed by the semantic vector model.

[0189] In a specific implementation, the information processing module 72 can input the information pair consisting of the physical examination information and the information of each assigned object into a pre-trained semantic vector model to obtain a first vector corresponding to the input physical examination information and the information of the assigned object. In this way, the physical examination information and the information of each assigned object can be mapped to the same vector space through the semantic vector model.

[0190] In an optional implementation, the vector matching model may adopt a fully connected layer network model. It should be noted that the vector matching model is not limited to the fully connected layer network model, and any network model that can transform an input vector into a scalar may be used.

[0191] In an optional implementation, the vector matching model may calculate the matching degree between the physical examination information and the assigned object according to the following formula:

[0192] Among them, score represents the matching degree between the physical examination information and the assigned object, V [cls] represents the first vector, W 1 Represents the first mapping matrix of the vector matching model, The second mapping matrix W representing the vector matching model 2 The transpose of 1 represents the bias matrix of the vector matching model, and σ represents the activation function of the vector matching model.

[0193] like Figure 4 As shown, the information processing module 72 can input the first vector V[cls] into the vector matching model and output the matching degree score between the physical examination information and the assigned object. Figure 4 x in i Represents the value of the i-th dimension in the first vector V[cls]. When the semantic vector model adopts the BERT model, W 1 Can be a 384*768 matrix, W 2 and b 1 It can be a 384*1 matrix.

[0194] The parameters of the semantic vector model and the parameters such as matrices W1, W2 and b1 in the vector matching model can be obtained through joint training. Among them, the parameters of the semantic vector model can use the parameters in the pre-trained model as initial values, and the matrices W1, W2 and b1 in the vector matching model can obtain initial values ​​through random initialization. For example, a large amount of data is used to pre-train the model to obtain the various parameters in the pre-trained model, and then the parameters are carefully adjusted on specific tasks. The specific training process of the semantic vector model and the vector matching model will be described in detail in the subsequent embodiments.

[0195] In this embodiment, the information processing module 72 maps the physical examination information and the information of the assigned object into the same vector space through a semantic vector model, calculates the matching degree between the physical examination information and each assigned object in the vector space, and then assigns the physical examination information to the appropriate assigned object according to the matching degree.

[0196] In an optional implementation, the device further includes a model training module, and the model training module is specifically configured to:

[0197] Acquire multiple training sample pairs, each of the training samples is a first sample pair or a second sample pair, the first sample pair includes sample physical examination information and information of a first sample assignment object, and the second sample pair includes the sample physical examination information and information of a second sample assignment object, wherein the matching degree of the first sample pair is greater than the matching degree of the second sample pair, the sample physical examination information includes abnormal sample physical examination information, the information of the sample assignment object includes specialty information of the sample assignment object, and the sample assignment object is the first sample assignment object or the second sample assignment object;

[0198] Inputting the training sample pair into the model to be trained to obtain the matching degree of the training sample pair;

[0199] The model to be trained is trained according to the matching degree of the first sample pair, the matching degree of the second sample pair, the workload information of the sample physical examination information, and the workload information of each of the sample allocation objects, and the trained model to be trained is determined as the information matching model.

[0200] In an optional implementation, the model training module is specifically configured as follows:

[0201] The model to be trained is trained with the goal that the difference between the matching degree of the first sample pair and the matching degree of the second sample pair is greater than a first preset threshold and the first parameter of each of the sample allocation objects is equal, wherein the first preset threshold is greater than zero, the first parameter is used to characterize the workload of each of the sample allocation objects, and the first parameter is determined by the workload information of the sample physical examination information.

[0202] Calculating a difference between a matching degree of the first sample pair and a matching degree of the second sample pair, and determining a first loss portion according to the difference and the first preset threshold;

[0203] Determine a second loss part according to a second parameter of the plurality of sample physical examination information, a third parameter of the plurality of sample allocation objects, and a first parameter of each of the sample allocation objects, wherein the second parameter is used to characterize the workload of the plurality of sample physical examination information, and the third parameter is the number of the plurality of sample allocation objects;

[0204] Performing a weighted summation of the first loss part and the second loss part to obtain a loss function;

[0205] The model to be trained is trained with the goal of minimizing the loss function.

[0206] In an optional implementation, the model training module is specifically configured as follows:

[0207] The first loss part is determined according to the following formula:

[0208] Wherein, L1 represents the first loss part, and u j represents the sample physical examination information, the d ij Indicates information about the first sample allocation object, the d i x represents the information of the second sample allocation object, and the (u j ,d ij ) represents the first sample pair, the (u j ,d il ) represents the second sample pair, the score(u j ,d il ) represents the matching degree of the second sample pair, and the score(u j ,d ij ) represents the matching degree of the first sample pair, and γ represents the first preset threshold;

[0209] The second loss part is determined according to the following formula:

[0210] Wherein, L2 represents the second loss part, N represents the second parameter, M represents the third parameter, and h i A first parameter representing the sample allocation object i;

[0211] The loss function is determined according to the following formula:

[0212] L=L1+βL2, wherein L represents the loss function and β represents the harmonic parameter.

[0213] In an optional implementation, when the information matching model includes a semantic vector model and a vector matching model, the model to be trained includes a first model to be trained and a second model to be trained, and the model training module is specifically configured as follows:

[0214] Inputting the training sample pair into the first to-be-trained model to obtain a vector of the training sample pair;

[0215] Inputting the vector of the training sample pair into the second to-be-trained model to obtain the matching degree of the training sample pair;

[0216] According to the matching degree of the first sample pair, the matching degree of the second sample pair, the workload information of the sample physical examination information, and the workload information of each of the sample allocation objects, the first model to be trained and the second model to be trained are trained, and the trained first model to be trained is determined as the semantic vector model, and the trained second model to be trained is determined as the vector matching model.

[0217] The sample physical examination information may be, for example, a sample physical examination report that has completed a single examination result but needs to provide a general examination result. The sample physical examination information may include the examination results of a series of physical examination items (such as general examination, internal examination, electrocardiogram, head CT, etc.) (for example, the examination results of the electrocardiogram item include: sinus rhythm, etc.).

[0218] In this embodiment, the sample physical examination information may include abnormal sample physical examination information, and the abnormal sample physical examination information may include abnormal examination results or abnormal physical examination items, etc. This embodiment does not limit the specific content of the abnormal sample physical examination information.

[0219] The sample allocation object may be, for example, a chief physician. The information of the sample allocation object may include the specialty information of the sample allocation object. The specialty information may include a description of the chief physician's graduation school, major, years of work, and specialized examination items. This embodiment does not limit the specific content of the specialty information. For example, the specialty information may be Doctor A, graduated from **, ..., worked for ** years, ..., specialized in internal medicine...

[0220] In this embodiment, the first sample allocation object can be a sample allocation object that matches the professional of the sample physical examination information, that is, the specialty information of the first sample allocation object matches the abnormal sample physical examination information in the sample physical examination information. The second sample allocation object can be a sample allocation object that does not match the professional of the sample physical examination information, that is, the specialty information of the second sample allocation object does not match the abnormal sample physical examination information in the sample physical examination information. In the process of obtaining training sample pairs, for example, the first sample allocation object and the second sample allocation object can be manually specified based on the sample physical examination information and the information of the sample allocation object. For example, if the professional matching degree of the first sample allocation object is greater than the professional matching degree of the second sample allocation object, it can be considered that the first sample allocation object is a sample allocation object that matches the professional of the sample physical examination information, and the second sample allocation object is a sample allocation object that does not match the professional of the sample physical examination information.

[0221] The first sample pair (u j ,d ij ), including sample physical examination information u j and the information d of the first sample allocation object ij (where the subscript ij∈[1,M], M is the total number of sample allocation objects); the second sample pair (u j,d il ), including sample physical examination information u j and the information d of the second sample allocation object il (where subscript il∈[1,M] and ij≠il). j ,d ij ) can be greater than the second sample pair (u j ,d il ), that is, the sample physical examination information is more matched with the major of the first sample assignment object.

[0222] The workload information of the sample physical examination information may be determined, for example, by the number of sample physical examination reports included in the sample physical examination information, or by the number of words in abnormal examination results, etc., which is not limited in this embodiment.

[0223] The workload information of the sample allocation object can be determined by the workload of the sample physical examination information matching the profession of the sample allocation object, for example, it can be the number of sample physical examination information matching the profession of the sample allocation object.

[0224] The first model to be trained can be a language model such as a BERT model, an RNN model or an LSTM model. Since the BERT model is a truly bidirectional language model, each word can simultaneously utilize the context information of the word, so the first model to be trained can obtain a more accurate semantic vector by using the BERT model, further improving the matching accuracy.

[0225] The second model to be trained may adopt a fully connected layer network model. It should be noted that the second model to be trained is not limited to the fully connected layer network model, and any network model that can transform an input vector into a scalar may be used.

[0226] In a specific implementation, the model training module can input the training sample pair into the first model to be trained to obtain a vector corresponding to the input training sample pair, and then input the vector of the training sample pair into the second model to be trained to obtain the matching degree corresponding to the input training sample pair. In this way, the matching degree of all training sample pairs can be obtained, including the first sample pair (u j ,d ij )'s matching degree score(u j ,d ij ) and the second sample pair (u j ,d il )'s matching degree score(u j ,d il). Afterwards, the first model to be trained and the second model to be trained or the model to be trained can be trained according to the matching degree of the first sample pair, the matching degree of the second sample pair, the workload information of the sample physical examination information, and the workload information of each sample allocation object.

[0227] In an optional implementation, the model training module can use the matching score (u j ,d ij ) and the matching degree score(u j ,d il ) is greater than the first preset threshold γ, and the first parameter of each sample allocation object is equal as the goal, the training model to be trained is trained, wherein the first preset threshold is greater than zero, the sample allocation object is the first sample allocation object or the second sample allocation object, the first parameter is used to characterize the workload of each sample allocation object, and the first parameter is determined by the workload information of the sample physical examination information. For example, the first parameter can be the workload of the sample physical examination information that matches the specialty of the sample allocation object.

[0228] In some embodiments, “equal” can mean that the two are completely equal, or the difference between the two can be less than a certain threshold, and the threshold can be set as needed.

[0229] In practical applications, the model training module can use the matching score (u j ,d ij ) is better than the matching score(u j ,d il ) is at least higher than the first preset threshold γ as the first training target, that is, score(u j ,d ij )>score(u j ,d il )+γ, where il∈[1,M], ij∈[1,M] and ij≠il. The first preset threshold is greater than zero, for example, it can be set to 1, and the specific data can be set according to actual conditions, which is not limited in this embodiment.

[0230] In a specific implementation, the model training module can determine the first loss part according to the following formula:

[0231] Among them, L1 represents the first loss part, u j Indicates sample physical examination information, d ij Indicates the information of the first sample allocation object, d il Indicates the information of the second sample allocation object, (u j ,d il ) represents the first sample pair, (u j ,dil ) represents the second sample pair, score(u j ,d il ) represents the matching degree of the second sample pair, score(u j ,d ij ) represents the matching degree of the first sample pair, and γ represents the first preset threshold. If a≤0, then [a]+=0; otherwise [a]+=a.

[0232] In this embodiment, the model training module can use the equality of the first parameters of each sample allocation object as the second training goal, or use the minimization of the difference between the first parameter of each sample allocation object and the average workload as the second training goal, so that the trained model can balance the workload between each allocation object.

[0233] In a specific implementation, the model training module may first calculate the ratio of the second parameter to the third parameter; and then determine the second loss part based on the difference between the first parameter and the ratio.

[0234] Assume that the workload of completing each sample physical examination information u is u_c i , where i∈[1,k], k represents the total number of sample physical examination information, then the total workload is the second parameter The average workload is M is the number of sample assignment objects as the third parameter. Assume that the sample physical examination information matching the sample assignment object j includes {u j1 ,…,u jh}, the workload imbalance of the sample allocation object can be recorded as in, The first parameter is the workload of sample physical examination information that matches the specialty of sample assignment object j.

[0235] In order to simplify the description, in this embodiment, it is assumed that the workload of each piece of physical examination information is the same, which is set to 1, that is, but Where N is the total number of sample physical examination information as the second parameter, and M is the number of sample allocation objects, i.e. the third parameter. So the first parameter of the sample allocation object j is but

[0236] In a specific implementation, the model training module can determine the second loss part according to the following formula:

[0237] Where L2 represents the second loss part, N represents the second parameter, M represents the third parameter, and h i The first parameter representing the sample assignment object i.

[0238] According to the calculation formula of the second loss part, when the number of sample physical examination information processed by each sample allocation object is the average N / M, the workload of each sample allocation object is the same, which is the average workload. In this case, the minimum loss of workload balance is 0. When the number of sample physical examination information processed by each sample allocation object varies greatly, the loss will increase.

[0239] In the specific implementation, the model training module can determine the loss function according to the following formula:

[0240] L=L1+βL2, where L represents the loss function and β represents the harmonic parameter.

[0241] In a specific implementation, the model training module can use batch gradient descent to minimize the loss function to learn model parameters.

[0242] In this embodiment, the model training module uses the first loss part to measure the professional matching loss between the sample physical examination information and each sample allocation object, and uses the second loss part to measure the balance loss between the workloads of each sample allocation object. j ,d ij ) is better than the matching score(u j ,d il ) is higher than γ and the workload (first parameter) processed by each assigned object is the same, the loss function L is minimized to 0.

[0243] From the formula of loss function L, we can know that if each piece of physical examination information is distributed to the assigned object with the best professional match (highest matching degree), some assigned objects will process a lot of physical examination information, while some assigned objects will process very little total inspection information. In this case, the first loss part is small, but the second loss part will be large. During model training, the total loss function L can be minimized by adjusting the parameters, so that part of the physical examination information can be distributed to the assigned objects with a higher professional match (such as the second or third highest professional match), so that the workload between the assigned objects is relatively balanced.

[0244] The model training module uses the above loss function to train the model to be trained, and an information matching model can be obtained. The above loss function is used to train the first model to be trained and the second model to be trained, and a semantic vector model and a vector matching model can be obtained. Using the trained information matching model or semantic vector model and vector matching model, the matching degree between the new physical examination information and each assigned object can be calculated, so that a target assigned object with a high professional matching degree and balanced workload can be selected from multiple assigned objects.

[0245] In this embodiment, by constructing a semantic vector model and a vector matching model, the physical examination information and the information of the assigned object are mapped into the same vector space, and the target assigned object is determined in the vector space according to the professional matching degree and the workload balance, and then the physical examination information is distributed to the target assigned object. This scheme takes into account the professional matching degree between the physical examination information and the assigned object and the workload balance between the assigned objects. Compared with the traditional method, this scheme can assign the physical examination information to the assigned object with a higher matching degree, and at the same time can reasonably allocate the workload between the assigned objects, so as to achieve workload balance between the assigned objects.

[0246] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, for example, implemented using software, hardware, firmware, etc., and will not be elaborated here.

[0247] Figure 8 800 is a block diagram of an electronic device 800 shown in the present application. For example, the electronic device 800 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0248] Reference Figure 8 , the electronic device 800 may include one or more of the following components: a processing component 802 , a memory 804 , a power component 806 , a multimedia component 808 , an audio component 810 , an input / output (I / O) interface 812 , a sensor component 814 , and a communication component 816 .

[0249] The processing component 802 generally controls the overall operation of the electronic device 800, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the physical examination information distribution method described in any embodiment. In addition, the processing component 802 may include one or more modules to facilitate the interaction between the processing component 802 and other components. For example, the processing component 802 may include a multimedia module to facilitate the interaction between the multimedia component 808 and the processing component 802.

[0250] The memory 804 is configured to store various types of data to support operations on the device 800. Examples of such data include instructions for any application or method operating on the electronic device 800, contact data, phone book data, messages, pictures, videos, etc. The memory 804 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0251] The power supply component 806 provides power to the various components of the electronic device 800. The power supply component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the electronic device 800.

[0252] The multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 808 includes a front camera and / or a rear camera. When the device 800 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera may receive external multimedia data. Each front camera and the rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

[0253] The audio component 810 is configured to output and / or input audio signals. For example, the audio component 810 includes a microphone (MIC), and when the electronic device 800 is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in the memory 804 or sent via the communication component 816. In some embodiments, the audio component 810 also includes a speaker for outputting audio signals.

[0254] I / O interface 812 provides an interface between processing component 802 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include but are not limited to: home button, volume button, start button, and lock button.

[0255] The sensor assembly 814 includes one or more sensors for providing various aspects of status assessment for the electronic device 800. For example, the sensor assembly 814 can detect the open / closed state of the device 800, the relative positioning of components, such as the display and keypad of the electronic device 800, and the sensor assembly 814 can also detect the position change of the electronic device 800 or a component of the electronic device 800, the presence or absence of contact between the user and the electronic device 800, the orientation or acceleration / deceleration of the electronic device 800, and the temperature change of the electronic device 800. The sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 814 may also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 may also include an accelerometer, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0256] The communication component 816 is configured to facilitate wired or wireless communication between the electronic device 800 and other devices. The electronic device 800 can access a wireless network based on a communication standard, such as WiFi, a carrier network (such as 2G, 3G, 4G or 5G), or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0257] In an exemplary embodiment, the electronic device 800 can be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to execute the physical examination information distribution method described in any embodiment.

[0258] In an exemplary embodiment, the processing component 1922 executes the above-mentioned physical examination information allocation method, and the input interface 1958 receives the information input by the user, such as physical examination information and information of the allocation object, and the output interface 1958 outputs the allocation result, for example, allocating the physical examination information to the target allocation object.

[0259] In an exemplary embodiment, after determining the degree of match between the physical examination information and each assigned object, each assigned object is sorted according to the degree of match, and the sorting result is output through the output interface 1958. For example, the user can select each assigned object through the user interaction interface, or can edit the order, such as adjusting the arrangement sequence.

[0260] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 804 including instructions, and the instructions can be executed by the processor 820 of the electronic device 800 to complete the physical examination information distribution method described in any embodiment. For example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0261] In an exemplary embodiment, a computer program product is also provided, the computer program product includes a readable program code, the readable program code can be executed by the processor 820 of the device 800 to complete the physical examination information allocation method described in any embodiment. Optionally, the program code can be stored in a storage medium of the device 800, and the storage medium can be a non-transitory computer-readable storage medium, for example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0262] Fig. 9 1 is a block diagram of an electronic device 1900 shown in the present application. For example, the electronic device 1900 may be provided as a server.

[0263] Reference Fig. 9 The electronic device 1900 includes a processing component 1922, which further includes one or more processors, and a memory resource represented by a memory 1932 for storing instructions executable by the processing component 1922, such as an application. The application stored in the memory 1932 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 1922 is configured to execute instructions to perform the physical examination information distribution method described in any embodiment.

[0264] The electronic device 1900 may also include a power supply component 1926 configured to perform power management of the electronic device 1900, a wired or wireless network interface 1950 configured to connect the electronic device 1900 to a network, and an input / output (I / O) interface 1958. The electronic device 1900 may operate based on an operating system stored in the memory 1932, such as Windows Server TM, Mac OS X TM, Unix TM, Linux TM, FreeBSD TM or the like.

[0265] References herein to "one embodiment," "embodiment," or "one or more embodiments" mean that a particular feature, structure, or characteristic described in conjunction with the embodiment is included in at least one embodiment of the present invention. In addition, please note that examples of the term "in one embodiment" do not necessarily all refer to the same embodiment.

[0266] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0267] In the claims, any reference signs placed between brackets shall not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention may be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third etc. does not indicate any order. These words may be interpreted as names.

[0268] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for distributing physical examination information, characterized in that: The method comprises: Acquire physical examination information and information of multiple assigned objects, wherein the physical examination information includes abnormal physical examination information and workload information of the physical examination information, and the information of the assigned objects includes specialty information and workload information of the assigned objects; wherein the workload information of the assigned objects includes the number of physical examination reports processed by the assigned objects; Inputting the physical examination information and the information of the assigned object into a pre-trained information matching model to obtain a matching degree between the physical examination information and the assigned object, wherein the information matching model is used to calculate a matching degree between the physical examination information and the assigned object based on the abnormal physical examination information, the workload information of the physical examination information, and the specialty information and workload information of the assigned object, wherein the matching degree includes a professional matching degree and a workload matching degree; Wherein, before inputting the physical examination information and the information of the assigned object into the pre-trained information matching model, the method further includes: obtaining a plurality of training sample pairs, each of the training samples is a first sample pair or a second sample pair, the first sample pair includes the sample physical examination information and the information of the first sample assigned object, the second sample pair includes the sample physical examination information and the information of the second sample assigned object, wherein the matching degree of the first sample pair is greater than the matching degree of the second sample pair, the sample physical examination information includes abnormal sample physical examination information, the information of the sample assigned object includes the specialty information of the sample assigned object, and the sample assigned object is the first sample assigned object or the second sample assigned object; Inputting the training sample pair into the model to be trained to obtain the matching degree of the training sample pair; According to the difference between the matching degree of the first sample pair and the matching degree of the second sample pair, and the first parameter corresponding to each sample allocation object, the model to be trained is trained, and the trained model to be trained is determined as the information matching model; the first parameter is the workload of the sample physical examination information professionally matched with the sample allocation object; According to the matching degree between the physical examination information and each of the assignment objects, a target assignment object is determined from the multiple assignment objects, and the physical examination information is assigned to the target assignment object.

2. The method for distributing physical examination information according to claim 1, characterized in that: The information matching model includes a semantic vector model and a vector matching model. The step of inputting the physical examination information and the information of the assigned object into the pre-trained information matching model to obtain the matching degree between the physical examination information and the assigned object includes: Inputting the physical examination information and the information of the assigned object into the semantic vector model, and outputting a first vector; The first vector is input into the vector matching model, and the matching degree between the physical examination information and the assigned object is output.

3. The method for distributing physical examination information according to claim 2, characterized in that: The step of inputting the physical examination information and the information of the assigned object into the semantic vector model and outputting a first vector comprises: The physical examination information and the information of the assigned object are spliced ​​to obtain spliced ​​information, wherein the header of the spliced ​​information uses a marking symbol [CLS], and the physical examination information and the information of the assigned object are connected using a separator symbol [SEP]; The splicing information is input into the semantic vector model, and the first vector is output.

4. The method for distributing physical examination information according to claim 2, characterized in that: The step of inputting the first vector into the vector matching model and outputting the matching degree between the physical examination information and the assigned object comprises: The vector matching model calculates the matching degree between the physical examination information and the assigned object according to the following formula: The score represents the matching degree between the physical examination information and the assigned object, and the V [cls] represents the first vector, W1 represents the first mapping matrix of the vector matching model, represents the transpose of the second mapping matrix W2 of the vector matching model, b1 represents the bias matrix of the vector matching model, and σ represents the activation function of the vector matching model.

5. The method for distributing physical examination information according to any one of claims 2 to 4, characterized in that: The semantic vector model includes a BERT model.

6. The method for distributing physical examination information according to any one of claims 2 to 4, characterized in that: The vector matching model includes a fully connected layer network model.

7. The method for distributing physical examination information according to claim 1, characterized in that: The step of training a model to be trained based on the difference between the matching degree of the first sample pair and the matching degree of the second sample pair, and the first parameter corresponding to each sample allocation object, and determining the trained model to be trained as the information matching model includes: The model to be trained is trained with the goal that the difference between the matching degree of the first sample pair and the matching degree of the second sample pair is greater than a first preset threshold and the first parameter of each of the sample allocation objects is equal, wherein the first preset threshold is greater than zero, the first parameter is used to characterize the workload of each of the sample allocation objects, and the first parameter is determined by the workload information of the sample physical examination information.

8. The method for distributing physical examination information according to claim 7, characterized in that: The step of training the model to be trained with the goal that the difference between the matching degree of the first sample pair and the matching degree of the second sample pair is greater than a first preset threshold and the first parameters of the sample allocation objects are equal includes: Calculating a difference between a matching degree of the first sample pair and a matching degree of the second sample pair, and determining a first loss portion according to the difference and the first preset threshold; Determine a second loss part according to a second parameter of the plurality of sample physical examination information, a third parameter of the plurality of sample allocation objects, and a first parameter of each of the sample allocation objects, wherein the second parameter is used to characterize the workload of the plurality of sample physical examination information, and the third parameter is the number of the plurality of sample allocation objects; Performing a weighted summation of the first loss part and the second loss part to obtain a loss function; The model to be trained is trained with the goal of minimizing the loss function.

9. The method for distributing physical examination information according to claim 8, characterized in that: The first loss part is determined according to the following formula: Wherein, L1 represents the first loss part, and u j represents the sample physical examination information, the d ij Indicates information about the first sample allocation object, the d il Indicates information about the second sample allocation object, the (u j , d ij ) represents the first sample pair, the (u j , d il ) represents the second sample pair, the score(u j , d il ) represents the matching degree of the second sample pair, and the score(u j , d ij ) represents the matching degree of the first sample pair, and γ represents the first preset threshold; The second loss part is determined according to the following formula: Wherein, L2 represents the second loss part, N represents the second parameter, M represents the third parameter, and h i A first parameter representing the sample allocation object i; The loss function is determined according to the following formula: LL=L1+βL2, wherein L represents the loss function and β represents the harmonic parameter.

10. The method for distributing physical examination information according to any one of claims 8 to 9, characterized in that: When the information matching model includes a semantic vector model and a vector matching model, the model to be trained includes a first model to be trained and a second model to be trained, and the step of inputting the training sample pair into the model to be trained to obtain the matching degree of the training sample pair includes: Inputting the training sample pair into the first to-be-trained model to obtain a vector of the training sample pair; Inputting the vector of the training sample pair into the second to-be-trained model to obtain the matching degree of the training sample pair; The step of training the model to be trained according to the matching degree of the first sample pair, the matching degree of the second sample pair, the workload information of the sample physical examination information, and the workload information of each sample allocation object, and determining the trained model to be trained as the information matching model includes: According to the matching degree of the first sample pair, the matching degree of the second sample pair, the workload information of the sample physical examination information, and the workload information of each of the sample allocation objects, the first model to be trained and the second model to be trained are trained, and the trained first model to be trained is determined as the semantic vector model, and the trained second model to be trained is determined as the vector matching model.

11. The method for distributing physical examination information according to any one of claims 1 to 4 and 7 to 9, characterized in that: The matching degree between the physical examination information and the target allocation object is greater than or equal to a second preset threshold; or, The matching degree between the physical examination information and the target allocation object is greater than the matching degree between the physical examination information and other allocation objects, and the other allocation objects are allocation objects other than the target allocation object among the multiple allocation objects.

12. A physical examination information distribution device, characterized in that: The device comprises: An information acquisition module is configured to acquire physical examination information and information of a plurality of assigned objects, wherein the physical examination information includes abnormal physical examination information and workload information of the physical examination information, and the information of the assigned objects includes specialty information and workload information of the assigned objects; wherein the workload information of the assigned objects includes the number of physical examination reports processed by the assigned objects; an information processing module, configured to input the physical examination information and the information of the assigned object into a pre-trained information matching model to obtain a matching degree between the physical examination information and the assigned object, wherein the information matching model is used to calculate a matching degree between the physical examination information and the assigned object based on the abnormal physical examination information, the workload information of the physical examination information, and the specialty information and workload information of the assigned object, wherein the matching degree includes a professional matching degree and a workload matching degree; The device further includes a model training module, which is specifically configured as follows: Acquire multiple training sample pairs, each of the training samples is a first sample pair or a second sample pair, the first sample pair includes sample physical examination information and information of a first sample assignment object, and the second sample pair includes the sample physical examination information and information of a second sample assignment object, wherein the matching degree of the first sample pair is greater than the matching degree of the second sample pair, the sample physical examination information includes abnormal sample physical examination information, the information of the sample assignment object includes specialty information of the sample assignment object, and the sample assignment object is the first sample assignment object or the second sample assignment object; Inputting the training sample pair into the model to be trained to obtain the matching degree of the training sample pair; According to the difference between the matching degree of the first sample pair and the matching degree of the second sample pair, and the first parameter corresponding to each sample allocation object, the model to be trained is trained, and the trained model to be trained is determined as the information matching model; the first parameter is the workload of the sample physical examination information professionally matched with the sample allocation object; The information allocation module is configured to determine a target allocation object from the multiple allocation objects according to the matching degree between the physical examination information and each of the allocation objects, and allocate the physical examination information to the target allocation object.

13. An electronic device, characterized in that: The electronic device comprises: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the physical examination information allocation method according to any one of claims 1 to 11.

14. A computer-readable storage medium, characterized in that: When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the physical examination information distribution method according to any one of claims 1 to 11.

15. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the physical examination information distribution method according to any one of claims 1 to 11 is implemented.

Images