Target object allocation method and device and computer equipment

Abstract

The invention discloses a target object allocation method and device and computer equipment. The method comprises the steps that demand information is acquired, and the demand information at least comprises conditions needing to be met by target objects distributed to a target set; analyzing the demand information by using a pre-trained allocation model to obtain an initial allocation result, the allocation result being used for representing a plurality of target objects allocated to the target set; feedback information of the multiple rounds of initial distribution results is received, the initial distribution results are adjusted based on the multiple rounds of feedback information, a target distribution result is obtained, and the number of rounds of receiving the feedback information is determined at least according to the sentiment analysis evaluation indexes corresponding to the feedback information. According to the method and the device, the technical problem that the adaptability of the established team is relatively low because the establishment of the business team is completed in a manual mode in related technologies is solved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more specifically, to a target object allocation method, device and computer equipment. Background Art

[0002] Currently, operators are providing customers with various emerging services. Different business scenarios require different technical professionals, and more emerging business scenarios require the joint participation of multiple professional business personnel. Therefore, for large-scale operator companies with a large number of employees, a wide range of professional fields, and a fast speed of loading and launching business needs, the current technology used is to use department information, employee job levels, and employee project experience as reference conditions, and the team initiator manually collects information and designates members to form a team. There are three problems: 1. Since information such as employee project experience is difficult to synchronize in real time, the team initiator needs to spend more time on information collection and collation, and the efficiency of team formation is low; 2. When the team initiator forms a team, there is a lack of effective cases for reference, and the historical experience of forming teams of the same type cannot provide a systematic reference for the current team formation needs. Summary of the invention

[0003] The embodiments of the present application provide a target object allocation method, apparatus and computer device to at least solve the technical problem in the related art that the business team is formed manually, resulting in low adaptability of the formed team.

[0004] According to one aspect of an embodiment of the present application, a target object allocation method is provided, comprising: obtaining demand information, the demand information comprising at least: conditions that the target objects allocated to the target set need to meet; analyzing the demand information using a pre-trained allocation model to obtain an initial allocation result, the allocation result being used to represent multiple target objects allocated to the target set; receiving multiple rounds of feedback information on the initial allocation result, adjusting the initial allocation result based on the multiple rounds of feedback information, and obtaining a target allocation result, wherein the number of rounds of receiving the feedback information is determined at least according to a sentiment analysis evaluation index corresponding to the feedback information.

[0005] Optionally, the pre-trained allocation model is trained in the following manner, including: obtaining a training data set for the allocation model, the training data set comprising at least: sample data and label data, wherein the sample data comprises: multiple objects, multi-dimensional information corresponding to each of the multiple objects and multiple demand information, and the label data comprises: allocation results matching the multiple demand information; and using the training data set to train the allocation model.

[0006] Optionally, the allocation model is trained using the training data set, including: constructing an objective function of the allocation model based on the cosine similarity between the first vector and the second vector and a variable parameter, and determining the objective function of the allocation model as a first objective function, wherein the first objective function is used to represent the degree of match between the object and the demand information, the first vector is used to represent the multi-dimensional information corresponding to each object of the multiple objects, the second vector is used to represent the demand information, and the variable parameter is used to represent the weight of the object being selected; determining the target value of the variable parameter based on the training data set.

[0007] Optionally, determining the variable parameter according to the training data set includes: obtaining an initial value of the variable parameter, and obtaining a state space, an action space and a reward, wherein the state space represents all demand information in the training data set, the action space represents a multi-dimensional information set of multiple objects corresponding to an allocation result obtained by the allocation model according to the demand information, and the reward is used to represent an evaluation index of the degree of matching between the allocation result and the label data; constructing an evaluation network and a target network according to the initial value of the variable parameter, the state space, the action space and the reward, respectively; and determining a target value of the variable parameter according to the evaluation network and the target network.

[0008] Optionally, determining the variable parameter according to the evaluation network and the target network includes: using the evaluation network to determine multiple samples, wherein each of the multiple samples includes: current demand information, a current allocation result corresponding to the current demand information, a reward corresponding to the current allocation result, and next demand information; storing the multiple samples in a preset experience pool, and randomly sampling a preset number of samples from the preset experience pool to determine a target sample set; obtaining a target parameter of each sample in the target sample set in the target network, the target parameter being used to represent the sum of the current reward and the maximum reward that can be obtained for the next demand under the current value of the preset parameter after obtaining the allocation result corresponding to the preset demand information; determining a second objective function of the evaluation network based on the target parameter; and when the function value of the second objective function is minimum, determining the current value of the variable parameter as the target value of the variable parameter.

[0009] Optionally, multiple rounds of feedback information are received, and the initial allocation result is adjusted based on the multiple rounds of feedback information, including: converting the initial allocation result into natural language and adding prompt words that evaluate the initial allocation result and give adjustment suggestions to obtain a first text; outputting the first text, and receiving feedback information corresponding to the first text; converting the feedback information corresponding to the first text into the initial allocation result adjustment information; adjusting the initial allocation result based on the adjustment information to obtain an adjusted allocation result; and cyclically obtaining the multiple rounds of feedback information to adjust the adjusted allocation result until the score corresponding to the adjusted allocation result reaches the target score to obtain the target allocation result.

[0010] Optionally, the score corresponding to the adjusted allocation result is determined in the following manner, including: obtaining a sentiment analysis evaluation index corresponding to the feedback information and a modification degree evaluation index for the demand information from the multiple rounds of feedback information; obtaining a first weight corresponding to the sentiment analysis evaluation index and a second weight corresponding to the modification degree evaluation index; and determining the score corresponding to the adjusted allocation result based on the sentiment analysis evaluation index, the modification degree evaluation index, the first weight and the second weight.

[0011] Optionally, the method also includes: determining all demand information, all allocation results and the scores corresponding to all allocation results in the process of determining the target allocation result according to the demand information as incremental data of one cycle; using the incremental data to re-train the allocation model to obtain the target allocation model, wherein, in the process of determining the variable parameters of the target allocation model, the reward is determined as the score corresponding to the allocation result.

[0012] According to another aspect of the embodiment of the present application, a computer device is provided, including: a memory and a processor, wherein the memory is used to store program instructions; and the processor is connected to the memory and is used to execute the above-mentioned target object allocation method.

[0013] According to another aspect of the embodiments of the present application, a non-volatile storage medium is provided, which includes a stored computer program, wherein the device where the non-volatile storage medium is located executes the above-mentioned target object allocation method by running the computer program.

[0014] According to another aspect of the embodiments of the present application, a computer program product is provided, including computer instructions, which implement the above-mentioned target object allocation method when executed by a processor.

[0015] In an embodiment of the present application, demand information is obtained, and the demand information includes at least: conditions that the target objects assigned to the target set need to meet; a pre-trained allocation model is used to analyze the demand information to obtain an initial allocation result, and the allocation result is used to represent multiple target objects assigned to the target set; multiple rounds of feedback information on the initial allocation result are received, and the initial allocation result is adjusted based on the multiple rounds of feedback information to obtain a target allocation result, wherein the number of rounds of receiving feedback information is determined at least according to the sentiment analysis evaluation index corresponding to the feedback information, thereby achieving the purpose of adjusting the allocation result through multiple rounds of feedback information, thereby achieving the technical effect of improving the adaptability of the team, and further solving the technical problem of the low adaptability of the formed team caused by the manual method of forming a business team in the related technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0017] Figure 1 It is a hardware structure block diagram of a computer terminal for implementing a target object allocation method according to an embodiment of the present application;

[0018] Figure 2 is a flow chart of a target object allocation method according to an embodiment of the present application;

[0019] Figure 3 is a schematic diagram of a human-computer interaction interface according to an embodiment of the present application;

[0020] Figure 4 is a flow chart of another target object allocation method according to an embodiment of the present application;

[0021] Figure 5 is a flow chart of another target object allocation method according to an embodiment of the present application;

[0022] Figure 6 It is a structural diagram of a target object allocation device according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0024] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0025] The information collected in the embodiments of the present application is information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data comply with the relevant laws, regulations and standards of the relevant regions, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or reject automated decision-making results; if the user chooses to reject, the expert decision-making process will be entered.

[0026] In order to solve the problems existing in the related art, the present application embodiment provides a target object allocation method, which can be run on Figure 1 In the computer terminal shown, the computer terminal is explained below.

[0027] The target object allocation method embodiment provided in the embodiment of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 FIG. 1 shows a hardware structure block diagram of a computer terminal for implementing a target object allocation method. Figure 1 As shown, the computer terminal 10 may include one or more (102a, 102b, ..., 102n are used to illustrate) processors (the processor may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission module 106 for communication functions connected via a wired and / or wireless network. In addition, it may also include: a display, a keyboard, a cursor control device, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, and a BUS bus. Those skilled in the art can understand that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1More or fewer components as shown, or with Figure 1 Different configurations shown.

[0028] It should be noted that the one or more processors and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computer terminal 10. As described in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0029] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the target object allocation method in the embodiment of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, realizing the above-mentioned target object allocation method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0030] The transmission module 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computer terminal 10. In one example, the transmission module 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission module 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0031] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 .

[0032] It should be noted that, in some optional embodiments, the above Figure 1 The computer terminal shown may include hardware elements (including circuits), software elements (including computer code stored on a computer-readable medium), or a combination of hardware elements and software elements. It should be noted that Figure 1This is merely one example of a particular embodiment and is intended to illustrate the types of components that may be present in the computer terminal described above.

[0033] In the above-mentioned operating environment, an embodiment of the present application provides an embodiment of a target object allocation method. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0034] Figure 2 is a flow chart of a target object allocation method according to an embodiment of the present application. Figure 2 As shown, the method comprises the following steps:

[0035] Step S202, obtaining demand information, wherein the demand information at least includes: conditions that the target objects assigned to the target set need to meet;

[0036] In step S202, the demand information may be in natural language. After the demand information is obtained, the demand may be converted into a triple structure, for example, an (entity, entity, entity) structure such as (job requirement, requirement type, requirement result). For example, the requirement information for team formation is: to form a six-person team named "AI Competition Team" to participate in the "Network AI Big Model" competition. Team members must obtain professional L3 or above certification, and the team must have at least one employee from A, B, and C companies. At this time, the triplet corresponding to the requirement information is: {"question":"Form a six-person team named 'AI Competition Team' to participate in the "Network AI Big Model" competition. Team members must obtain professional L3 or above certification, and the team must have at least one employee from A, B, and C companies.","answer":{"Team name":["AI Competition Team"],"Competition name":["Network AI Big Model"],"Employee qualification requirements":["Professional L3 and above"],"Employee range":["A","B","C"]}}.

[0037] Step S204, using a pre-trained allocation model to analyze the demand information to obtain an initial allocation result, where the allocation result is used to represent multiple target objects allocated to the target set;

[0038] Step S206, receiving multiple rounds of feedback information on the initial allocation results, adjusting the initial allocation results based on the multiple rounds of feedback information, and obtaining a target allocation result, wherein the number of rounds of receiving feedback information is determined at least according to the sentiment analysis evaluation index corresponding to the feedback information.

[0039] Through the above steps S202 to S206, by obtaining demand information, the demand information at least includes: the conditions that the target objects assigned to the target set need to meet; using a pre-trained allocation model to analyze the demand information to obtain an initial allocation result, the allocation result is used to represent multiple target objects assigned to the target set; receiving multiple rounds of feedback information on the initial allocation result, adjusting the initial allocation result based on the multiple rounds of feedback information, and obtaining a target allocation result, wherein the number of rounds of receiving feedback information is determined at least according to the sentiment analysis evaluation index corresponding to the feedback information, achieving the purpose of adjusting the allocation result through multiple rounds of feedback information, thereby achieving the technical effect of improving the adaptability of the team, and further solving the technical problem of the low adaptability of the formed team caused by the manual completion of the formation of the business team in the related technology. The following is a detailed description.

[0040] In step S202 of the above-mentioned target object allocation method, the pre-trained allocation model is trained in the following manner, including: obtaining a training data set of the allocation model, the training data set at least including: sample data and label data, wherein the sample data includes: multiple objects, multi-dimensional information corresponding to each of the multiple objects and multiple demand information, and the label data includes: allocation results matching the multiple demand information; and using the training data set to train the allocation model.

[0041] Specifically, the multi-dimensional information corresponding to each of the multiple objects and the multiple objects can be obtained from a pre-acquired information database, and the information database includes: the multi-dimensional information of each object, taking the object as an employee account as an example, the multi-dimensional information includes: the employee's name, gender, education, date of birth, date of employment, professional field certification, etc., as well as the employee's professional ability certification level and other business skills scores. Table 1 shows the multi-dimensional information of an employee account in an information database.

[0042] The training data set of the allocation model consists of: information database, triple structure of demand information after entity extraction based on demand information. These two types of data are used as sample data in the training data set. The employee ID in the allocation plan that meets the demand information is used as the labeling result for the sample data. The sample data and the labeling result together constitute the training data set of the allocation model.

[0043] Table 1

[0044]

[0045] Among them, the specific method of using the training data set to train the allocation model is as follows: construct the objective function of the allocation model based on the cosine similarity between the first vector and the second vector and the variable parameter, and determine the objective function of the allocation model as the first objective function, wherein the first objective function is used to represent the degree of matching between the object and the demand information, the first vector is used to represent the multi-dimensional information corresponding to each object of the multiple objects, the second vector is used to represent the demand information, and the variable parameter is used to represent the weight of the object being selected; determine the target value of the variable parameter according to the training data set.

[0046] Among them, the first objective function is as follows:

[0047]

[0048] In the formula, similarity represents the cosine similarity between the first vector and the second vector, O represents the first objective function, and P i represents the first vector, I represents the second vector, i represents the object number, N represents the number of objects, ω i Indicates the weight of the i-th object being selected.

[0049] It should be noted that the variable parameter is formally a column vector, that is, a matrix consisting of a column containing N elements, where the N elements respectively represent the weights of the N objects selected and assigned to the target set.

[0050] In some embodiments of the present application, the specific process of determining the target value of the variable parameter based on the training data set is as follows: obtain the initial value of the variable parameter, and obtain the state space, action space and reward, wherein the state space represents all demand information in the training data set, the action space represents a multi-dimensional information set of multiple objects corresponding to the allocation result obtained by the allocation model according to the demand information, and the reward is used to represent an evaluation index of the degree of matching between the allocation result and the label data; construct an evaluation network and a target network according to the initial value of the variable parameter, the state space, the action space and the reward, respectively; determine the target value of the variable parameter according to the evaluation network and the target network.

[0051] Based on the training data set, the state space is defined to represent the set of all demand information in the training data set; secondly, the action space is defined to represent the set of information of each object in the allocation result obtained by the allocation model according to the corresponding demand information; then, the reward is defined as the matching evaluation index between the employee allocation plan given by the allocation model and the target set actually corresponding to the demand information.

[0052] In actual application scenarios, a random column vector can be used as the initial value of the variable parameter, and then the evaluation network can be initialized with the variable parameter, and the same variable parameter can be copied to initialize the target network and the preset experience pool.

[0053] Among them, the evaluation network is used to determine which action to take in the current state, that is, under the input of current demand information, learn to approximate the optimal allocation result, and use the gradient descent method to update the value of the variable parameter by minimizing the value of the second objective loss function.

[0054] The target network is used to determine the best action to take in the next state, that is, to select the best allocation plan based on the next job requirement for the evaluation network to minimize the value of the second objective loss function.

[0055] The target network copies the values ​​of the variable parameters of the evaluation network to the target network every preset number of training steps (or every preset time in another optional manner) to ensure that the target network maintains a certain stability while gradually moving closer to the learning of the evaluation network. The evaluation network and the target network jointly promote the training effect and performance improvement of the employee allocation model.

[0056] In some embodiments of the present application, the preset experience replay pool is used to solve the impact of strong correlation of continuous experience in the allocation model. For example, if the same demand information appears continuously in the training data set, the allocation model will be affected by these training data during the training and learning process, limiting the generalization ability. The preset experience pool breaks this correlation by randomly extracting small batches of samples, reducing the dependence between time steps, and making the training more stable.

[0057] The specific method of determining the target value of the variable parameter according to the evaluation network and the target network is as follows: using the evaluation network to determine a plurality of samples, wherein each of the plurality of samples includes: current demand information, a current allocation result corresponding to the current demand information, a reward corresponding to the current allocation result, and next demand information; storing the plurality of samples in a preset experience pool, and randomly sampling a preset number of samples from the preset experience pool to determine as a target sample set; obtaining a target parameter of each sample in the target sample set in the target network, the target parameter being used to represent the summation of the current reward and the maximum reward that can be obtained for the next demand under the current value of the preset parameter after obtaining the allocation result corresponding to the preset demand information; determining a second objective function of the evaluation network based on the target parameter; and determining the current value of the variable parameter as the target value of the variable parameter when the function value of the second objective function is minimum.

[0058] In the actual application scenario, the selection action is executed in the current state S, and the evaluation network uses a greedy strategy to execute the selection action A. The selection action A refers to the action of selecting a number of bit objects from the information database to form a set of possible employee allocation solutions.

[0059] After executing the selection action A in the current state, calculate the reward R of the selection action A. The current reward R is the matching evaluation score between the allocation scheme given by the allocation model and the actual required object under the current demand information. The environment state is switched to the state S′ of the next demand information in the training data set; the samples (S, A, R, S′) are stored in the preset experience pool. When the number of samples in the preset experience pool reaches the quantity threshold, a small batch of preset number of samples is randomly sampled from the preset experience pool. Each sample contains the current demand information, its corresponding allocation result, the evaluation index of the corresponding allocation result and the next new demand information. For each sample, the target parameter is calculated in the target network, which means that after the demand information and the allocation result are given, under the current variable parameter value, the current reward plus the next demand information can obtain the maximum reward, which reflects the quality of the allocation result of the allocation model under the current variable parameter value.

[0060] The target parameters are calculated as follows:

[0061]

[0062] in, Y j represents the target parameter of the jth sample, R j represents the reward of the jth sample, γ represents the discount factor, which is usually set to a value close to 1 to indicate the importance of future rewards. represents the target network, ω - represents the current value of the variable parameter in the target network, A j+1 represents the next action (the allocation result corresponding to the next demand information), S j+1 Indicates the next requirement information.

[0063] The second objective function is as follows:

[0064]

[0065] Where L represents the value of the second objective function, Y j represents the target parameter value of the jth sample, Q ω (S j , A j ) represents the evaluation network, ω represents the current value of the variable parameter in the evaluation network, M represents the number of samples in the target sample set, S jIndicates current demand information, A j Indicates the current action (the allocation result corresponding to the current demand information).

[0066] In actual application scenarios, the second objective loss function is minimized, and the values ​​of the variable parameters of the evaluation network are updated using the gradient descent method. The values ​​of the variable parameters are iteratively adjusted within a certain interval to minimize the value of the second objective function, so that the values ​​of the variable parameters of the evaluation network can be as close as possible to the maximum evaluation index that can be obtained from the allocation result corresponding to the next demand information of the model. A fixed period is set to regularly copy the values ​​of the variable parameters from the evaluation network to update the parameters in the target network, ensuring that the target network maintains a certain stability and gradually moves closer to the evaluation network.

[0067] After the training is completed, the values ​​of the current variable parameters in the target network are copied and placed in the allocation model to obtain an allocation model whose performance reaches the preset performance.

[0068] In some embodiments of the present application, multiple rounds of feedback information are received, and the initial allocation result is adjusted based on the multiple rounds of feedback information, including: converting the initial allocation result into natural language and adding prompt words that evaluate the initial allocation result and give adjustment suggestions to obtain a first text; outputting the first text, and receiving feedback information corresponding to the first text; converting the feedback information corresponding to the first text into the initial allocation result adjustment information; adjusting the initial allocation result based on the adjustment information to obtain an adjusted allocation result; and cyclically obtaining the multiple rounds of feedback information to adjust the adjusted allocation result until the score corresponding to the adjusted allocation result reaches the target score to obtain the target allocation result.

[0069] Specifically, in the human-job matching optimization algorithm based on the multi-round feedback mechanism, first, the initial job requirements of the team initiator are extracted using the Bert NER entity extraction technology of natural language processing (NLP) to obtain the initial requirement information triple I A , where I A Contains s job requirements, each job requirement i x Contains t x The requirements are as follows:

[0070]

[0071] Secondly, the initial demand information is input into the allocation model to obtain the initial allocation result, and then the initial allocation result is input into the human-computer dialogue system. The human-computer dialogue system converts the initial allocation result into natural language and adds prompt words that can guide the evaluation of the allocation result to obtain the first text, guide the evaluation of the initial allocation result, and give adjustment suggestions, or receive further refined demand information.

[0072] The new demand information in the feedback information is processed by NLP to obtain the latest job demand triple I B , as shown below:

[0073]

[0074] Based on the latest job requirements triple I B As the basis for updating, in the initial job requirement triple I A Based on the update, the updated demand information is obtained. C .

[0075] Figure 3 A schematic diagram of a human-computer dialogue window is shown, Figure 3 As shown, the initial demand information is continuously modified to obtain the target allocation result, and the initial allocation result can also be adjusted to obtain the target allocation result.

[0076] In some embodiments of the present application, the score corresponding to the adjusted allocation result is determined in the following manner, including: obtaining a sentiment analysis evaluation index corresponding to the feedback information and a modification degree evaluation index for the demand information from the multiple rounds of feedback information; obtaining a first weight corresponding to the sentiment analysis evaluation index and a second weight corresponding to the modification degree evaluation index; and determining the score corresponding to the adjusted allocation result based on the sentiment analysis evaluation index, the modification degree evaluation index, the first weight and the second weight.

[0077] Specifically, by calculating the user's evaluation score for the employee allocation plan, it is determined whether the user is satisfied with the current latest employee allocation plan. The evaluation score includes the degree of affirmation and negation of the user's sentiment analysis and the degree of modification of the original job requirements by the optimized feedback content. The two weighted scores are used as the user's evaluation score for the employee allocation plan in this round of employee allocation. Among them, the degree of affirmation and negation of sentiment analysis can be obtained by the sentiment analysis calculation method in NLP. The specific process of sentiment analysis calculation is: in the demand interaction process of team formation, firstly, the feedback evaluation content of the team initiator on the current employee allocation plan is vectorized in real time using the sentence-bert model; secondly, the sentiment words such as general sentiment words, degree adverbs, positive and negative words have been vectorized and stored in the sentiment analysis model; therefore, the vector of the current feedback content of the team initiator is matched with the preset vector library of the sentiment analysis model, and the similarity between the user feedback content and the preset sentiment word set is calculated using semantic similarity to obtain the degree of affirmation and negation of the user's sentiment analysis. In addition, the sentiment analysis model supports recording user feedback during real-time interaction. For example, by asking users whether they are dissatisfied with the current solution, when the user answers "yes", the user's feedback can be recorded as a negative sentiment data set. As the user continues to accumulate data sets, the sentiment analysis matching can not only match with the preset sentiment word set, but also match with the data set accumulated during use, so as to further improve the accuracy of the sentiment analysis method. In addition, the degree of modification of the original job requirements can be calculated by the proportion of text summaries in NLP.

[0078] The calculation standard of the evaluation score W is as follows: W is the user's evaluation score for the employee allocation plan of this round, α is the user's sentiment analysis affirmation or negation degree (sentiment analysis evaluation index), β is the modification degree of the optimization feedback content to the original job requirements (modification degree evaluation index), θ1 is the priority of the user's sentiment analysis affirmation and negation degree (first weight), θ2 is the priority of the modification degree of the optimization feedback content to the original job requirements (second weight), where the calculation method of the evaluation score is as follows:

[0079]

[0080] The calculation method of the evaluation score must meet the following constraints:

[0081]

[0082] In some embodiments of the present application, when the evaluation score meets the preset scoring condition, the current evaluation score is determined to be the target score, and the preset scoring condition is as shown in the following formula:

[0083]

[0084] In an optional embodiment, η can be set to 0.8.

[0085] In some embodiments of the present application, all demand information, all allocation results, and the scores corresponding to all allocation results in the process of determining the target allocation result based on the demand information are determined as incremental data of one cycle; the allocation model is retrained using the incremental data to obtain the target allocation model, wherein, in the process of determining the variable parameters of the target allocation model, the reward is determined as the score corresponding to the allocation result.

[0086] If the current allocation result does not meet the requirements, the next round of allocation will be carried out, and the training data corresponding to the demand information, allocation results and evaluation scores of this round of allocation results will be determined as incremental data, and a data buffer will be used to store the incremental data. The cycle is from proposing a demand information to completing this allocation task, and the newly collected incremental data is periodically used to form a data set to train the model.

[0087] By adopting the incremental data training method, only the newly added data is used for training, avoiding retraining the entire model, which greatly reduces the training time and computing power. At the same time, it can also make the model have better performance indicators for current users or recent assigned tasks.

[0088] In the process of training with incremental data, after selecting action A in the current state S, the reward R for selecting action A is calculated. The current reward R no longer needs to calculate the matching evaluation score between the allocation result given by the allocation model under the current demand information and the actual required target set. Instead, R=W, that is, the evaluation score of the current allocation result is used as the reward for the current allocation result given by the model.

[0089] By embedding multiple rounds of feedback mechanisms, the allocation model records the user's evaluation of the model while performing the task. Periodic training enables the allocation model to continuously learn and optimize its performance.

[0090] In order to avoid the problem that the employee allocation model still has no optimal solution for the current job requirements after multiple rounds of feedback optimization, a feasible improvement plan is further proposed, that is, setting the maximum number of rounds of multiple feedback optimization to be performed. When the optimal solution cannot be given after multiple rounds of feedback optimization between the human-computer dialogue model and the user, and the current number of feedback optimization rounds has reached the maximum number of rounds supported by the system, each requirement item in the current round of demand information triples is optimized to the maximum extent.

[0091] Under the premise of satisfying the requirements in the demand information triplet to the greatest extent, select the appropriate objects for allocation, that is, give the suboptimal solution x max . Suboptimal solution x maxIt should be the solution with the largest evaluation score W in the solution set:

[0092]

[0093] Or to meet the requirements of the job requirement triple to the greatest extent possible:

[0094]

[0095] Among them, the solution set X of the allocation result includes several object allocation solutions x, and the solution x contains the information of the employees selected in the employee allocation plan; (x) is the number of sub-items of the job requirements that solution x satisfies. For example, the latest demand information is:

[0096]

[0097] Current job requirements I c Contains sub-items of s job requirements, namely triple item i s , when the solution x in the employee allocation solution set X cannot fully satisfy the sub-items of the s job requirements, then find the solution x that can satisfy the sub-items of the most job requirements from the employee allocation solution set max .

[0098] Figure 4 Another target object assignment method is shown; Figure 5 Yet another target object allocation method is shown.

[0099] Figure 6 is a structural diagram of a target object allocation device according to an embodiment of the present application, such as Figure 6 As shown, the device comprises:

[0100] The acquisition module 60 is used to acquire demand information, wherein the demand information at least includes: conditions that the target objects assigned to the target set need to meet;

[0101] An analysis module 62, configured to analyze the demand information using a pre-trained allocation model to obtain an initial allocation result, wherein the allocation result is used to represent a plurality of target objects allocated to the target set;

[0102] The adjustment module 64 is used to receive multiple rounds of feedback information on the initial allocation results, and adjust the initial allocation results based on the multiple rounds of feedback information to obtain a target allocation result, wherein the number of rounds of receiving feedback information is determined at least according to the sentiment analysis evaluation index corresponding to the feedback information.

[0103] By obtaining demand information, the demand information at least includes: the conditions that the target objects assigned to the target set need to meet; using a pre-trained allocation model to analyze the demand information to obtain an initial allocation result, and the allocation result is used to represent multiple target objects assigned to the target set; receiving multiple rounds of feedback information on the initial allocation result, adjusting the initial allocation result based on the multiple rounds of feedback information, and obtaining a target allocation result, wherein the number of rounds of receiving feedback information is determined at least according to the sentiment analysis evaluation index corresponding to the feedback information, thereby achieving the purpose of adjusting the allocation result through multiple rounds of feedback information, thereby achieving the technical effect of improving the adaptability of the team, and further solving the technical problem of low adaptability of the formed team caused by the manual method of forming a business team in the related technology.

[0104] The analysis module 62 in the above-mentioned target object allocation device includes: a training submodule, which is used for training the pre-trained allocation model in the following manner, including: obtaining a training data set for the allocation model, the training data set at least including: sample data and label data, wherein the sample data includes: multiple objects, multi-dimensional information corresponding to each of the multiple objects and multiple demand information, and the label data includes: allocation results matching the multiple demand information; and using the training data set to train the allocation model.

[0105] A training submodule includes: a training unit, which is used to train the allocation model using the training data set, including: constructing an objective function of the allocation model based on the cosine similarity between the first vector and the second vector and a variable parameter, and determining the objective function of the allocation model as a first objective function, wherein the first objective function is used to represent the degree of matching between the object and the demand information, the first vector is used to represent the multi-dimensional information corresponding to each object of the multiple objects, the second vector is used to represent the demand information, and the variable parameter is used to represent the weight of the object being selected; determining the target value of the variable parameter based on the training data set.

[0106] The training unit includes: a training subunit and a determination subunit. The training subunit is used to determine the variable parameter according to the training data set, including: obtaining the initial value of the variable parameter, and obtaining the state space, action space and reward, wherein the state space represents all demand information in the training data set, the action space represents the multi-dimensional information set of multiple objects corresponding to the allocation result obtained by the allocation model according to the demand information, and the reward is used to represent the evaluation index of the degree of matching between the allocation result and the label data; constructing an evaluation network and a target network according to the initial value of the variable parameter, the state space, the action space and the reward respectively; and determining the target value of the variable parameter according to the evaluation network and the target network.

[0107] A determination subunit, used to determine the variable parameter according to the evaluation network and the target network, comprising: using the evaluation network to determine a plurality of samples, wherein each of the plurality of samples comprises: current demand information, a current allocation result corresponding to the current demand information, a reward corresponding to the current allocation result, and next demand information; storing the plurality of samples in a preset experience pool, and randomly sampling a preset number of samples from the preset experience pool to determine as a target sample set; obtaining a target parameter of each sample in the target sample set in the target network, the target parameter being used to represent the summation of the current reward and the maximum reward that can be obtained for the next demand under the current value of the preset parameter after obtaining the allocation result corresponding to the preset demand information; determining a second objective function of the evaluation network based on the target parameter; and determining the current value of the variable parameter as the target value of the variable parameter when the function value of the second objective function is minimum.

[0108] The adjustment module 64 includes: an adjustment submodule, which is used to receive multiple rounds of feedback information and adjust the initial allocation result based on the multiple rounds of feedback information, including: converting the initial allocation result into natural language and adding prompt words that evaluate the initial allocation result and give adjustment suggestions to obtain a first text; outputting the first text and receiving feedback information corresponding to the first text; converting the feedback information corresponding to the first text into the initial allocation result adjustment information; adjusting the initial allocation result based on the adjustment information to obtain an adjusted allocation result; cyclically obtaining the multiple rounds of feedback information to adjust the adjusted allocation result until the score corresponding to the adjusted allocation result reaches the target score to obtain the target allocation result.

[0109] The adjustment submodule includes: a determination unit and an increment unit, wherein the determination unit is used to determine the score corresponding to the adjusted allocation result in the following manner, including: obtaining a sentiment analysis evaluation index corresponding to the feedback information and a modification degree evaluation index for the demand information from the multiple rounds of feedback information; obtaining a first weight corresponding to the sentiment analysis evaluation index and a second weight corresponding to the modification degree evaluation index; and determining the score corresponding to the adjusted allocation result according to the sentiment analysis evaluation index, the modification degree evaluation index, the first weight and the second weight.

[0110] An incremental unit is used to determine all the demand information, all the allocation results and the scores corresponding to all the allocation results in the process of determining the target allocation result according to the demand information as incremental data of one cycle; use the incremental data to re-train the allocation model to obtain the target allocation model, wherein, in the process of determining the variable parameters of the target allocation model, the reward is determined as the score corresponding to the allocation result.

[0111] It should be noted that Figure 6 The target object allocation device shown is used to execute Figure 2 The target object allocation method shown in the figure, therefore the relevant explanations in the above target object allocation method are also applicable to the target object allocation device, and will not be repeated here.

[0112] An embodiment of the present application also provides a computer device, including: a memory and a processor, wherein the memory is used to store program instructions; and the processor is connected to the memory and is used to execute the above-mentioned target object allocation method.

[0113] An embodiment of the present application further provides a non-volatile storage medium, which includes a stored computer program, wherein a device where the non-volatile storage medium is located executes the above-mentioned target object allocation method by running the computer program.

[0114] An embodiment of the present application also provides a computer program product, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the target object allocation method in the present application are implemented.

[0115] An embodiment of the present application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the target object allocation method in the present application.

[0116] An embodiment of the present application also provides a computer program, which, when executed by a processor, implements the steps of the target object allocation method in the present application.

[0117] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0118] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0119] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0120] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0121] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0122] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk and other media that can store program codes.

[0123] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A target object allocation method, characterized in that: include: Obtaining demand information, wherein the demand information at least includes: conditions that need to be met by the target objects assigned to the target set; Using a pre-trained allocation model to analyze the demand information to obtain an initial allocation result, where the allocation result is used to represent a plurality of target objects allocated to the target set; Receive multiple rounds of feedback information on the initial allocation results, adjust the initial allocation results based on the multiple rounds of feedback information, and obtain a target allocation result, wherein the number of rounds of receiving feedback information is determined at least according to the sentiment analysis evaluation index corresponding to the feedback information.

2. The method according to claim 1, characterized in that The pre-trained allocation model is trained in the following manner, including: Acquire a training data set of the allocation model, the training data set at least comprising: sample data and label data, wherein the sample data comprises: a plurality of objects, multi-dimensional information corresponding to each of the plurality of objects and a plurality of demand information, and the label data comprises: an allocation result matching the plurality of demand information; The allocation model is trained using the training data set.

3. The method according to claim 2, characterized in that The allocation model is trained using the training data set, including: constructing an objective function of the allocation model according to the cosine similarity between the first vector and the second vector and the variable parameter, and determining the objective function of the allocation model as a first objective function, wherein the first objective function is used to represent the degree of matching between the object and the demand information, the first vector is used to represent the multi-dimensional information corresponding to each object in the multiple objects, the second vector is used to represent the demand information, and the variable parameter is used to represent the weight of the object being selected; A target value of the variable parameter is determined based on the training data set.

4. The method according to claim 3, characterized in that Determining the variable parameter according to the training data set includes: Obtaining the initial value of the variable parameter, and obtaining the state space, action space and reward, wherein the state space represents all demand information in the training data set, the action space represents a multi-dimensional information set of multiple objects corresponding to the allocation result obtained by the allocation model according to the demand information, and the reward is used to represent an evaluation index of the degree of matching between the allocation result and the label data; Constructing an evaluation network and a target network respectively according to the initial value of the variable parameter, the state space, the action space and the reward; A target value of the variable parameter is determined based on the evaluation network and the target network.

5. The method according to claim 4, characterized in that Determining the variable parameter according to the evaluation network and the target network includes: Determine a plurality of samples using the evaluation network, wherein each of the plurality of samples includes: current demand information, a current allocation result corresponding to the current demand information, a reward corresponding to the current allocation result, and next demand information; The plurality of samples are stored in a preset experience pool, and a preset number of samples are randomly sampled from the preset experience pool to be determined as a target sample set; Obtaining a target parameter of each sample in the target sample set in the target network, wherein the target parameter is used to indicate that after obtaining the allocation result corresponding to the preset demand information, when the variable parameter is the current value, the current reward is summed with the maximum reward that can be obtained by the next demand; Determining a second objective function of the evaluation network based on the objective parameter; When the function value of the second objective function is minimum, the value of the current variable parameter is determined as the target value of the variable parameter.

6. The method according to claim 1, characterized in that Receiving multiple rounds of feedback information, and adjusting the initial allocation result based on the multiple rounds of feedback information, including: Converting the initial allocation result into natural language and adding prompt words for evaluating the initial allocation result and providing adjustment suggestions to obtain a first text; Outputting the first text, and receiving feedback information corresponding to the first text; Converting the feedback information corresponding to the first text into the initial allocation result adjustment information; Adjust the initial allocation result based on the adjustment information to obtain an adjusted allocation result; The multiple rounds of feedback information are cyclically acquired to adjust the adjusted allocation result until the score corresponding to the adjusted allocation result reaches the target score, thereby obtaining the target allocation result.

7. The method according to claim 6, characterized in that The score corresponding to the adjusted allocation result is determined by the following method, including: Obtaining sentiment analysis evaluation indicators corresponding to the feedback information and modification degree evaluation indicators of the demand information from the multiple rounds of feedback information; Obtaining a first weight corresponding to the sentiment analysis evaluation index and a second weight corresponding to the modification degree evaluation index; A score corresponding to the adjusted allocation result is determined according to the sentiment analysis evaluation index, the modification degree evaluation index, the first weight, and the second weight.

8. The method according to claim 6, characterized in that The method further comprises: Determine all demand information, all allocation results, and scores corresponding to all allocation results in the process of determining the target allocation result according to the demand information as incremental data for one cycle; The allocation model is retrained using the incremental data to obtain the target allocation model, wherein in the process of determining the variable parameters of the target allocation model, the reward is determined as the score corresponding to the allocation result.

9. A target object allocation device, characterized in that: include: An acquisition module, used for acquiring demand information, wherein the demand information at least includes: conditions that need to be met by the target objects assigned to the target set; An analysis module, used for analyzing the demand information by using a pre-trained allocation model to obtain an initial allocation result, wherein the allocation result is used to represent a plurality of target objects allocated to the target set; An adjustment module is used to receive multiple rounds of feedback information on the initial allocation results, and adjust the initial allocation results based on the multiple rounds of feedback information to obtain a target allocation result, wherein the number of rounds of receiving feedback information is determined at least according to the sentiment analysis evaluation index corresponding to the feedback information.

10. A computer device, characterized in that: include: A memory and a processor, wherein the memory is used to store program instructions; The processor is connected to the memory, and is used to execute the target object allocation method described in any one of claims 1 to 8.

11. A non-volatile storage medium, characterized in that: The non-volatile storage medium includes a stored computer program, wherein the device where the non-volatile storage medium is located executes the target object allocation method according to any one of claims 1 to 7 by running the computer program.

12. A computer program product comprising computer instructions, characterized in that: When the computer instructions are executed by a processor, the target object allocation method according to any one of claims 1 to 7 is implemented.

Images