Projection test model training method and device, computer equipment and storage medium

Abstract

The invention relates to the field of psychological testing, and provides a training method, device and equipment for a projection test model and a computer storage medium, and the method comprises the steps: obtaining stimulation information of a projection test, reaction information generated by a user based on the stimulation information, and label information corresponding to the reaction information; the stimulation information and the reaction information are input into a preset deep learning model, label prediction information output by the deep learning model is obtained, and at least one middle layer of the deep learning model is adjusted through a parameter matrix; determining a loss function of the parameter matrix according to the label prediction information and the label information; and adjusting the parameter matrix according to the loss function until a target matrix is obtained.

Description

Technical Field

[0001] The present application relates to the field of psychological testing, and in particular to a training method, device, computer equipment and storage medium for a projective testing model. Background Art

[0002] Projective testing is a common psychometric tool that uses ambiguous or ambiguous stimuli to allow subjects to respond without restriction. When faced with ambiguity, individuals tend to project their inner feelings, thoughts, and conflicts onto these stimuli. Since the results of projective tests often reflect the subconscious content of individuals, they are often used to explore psychological characteristics that may not be easily obtained through direct questioning. In the field of psychology, commonly used projective tests include the Rorschach inkblot test, sentence completion test, drawing test, thematic apperception test (TAT), etc. Since projective tests do not have standardized answers and scoring systems like objective tests, projective tests rely more on the subjective judgment and experience of the analyst. Therefore, the analysis efficiency of projective tests is relatively low, and the objectivity and accuracy are poor. There is an urgent need for a projective test result analysis method that can improve the analysis efficiency. Summary of the invention

[0003] The main purpose of this application is to provide a training method, device, equipment and computer storage medium for a projective test model, aiming to improve the analysis efficiency of the projective test.

[0004] In a first aspect, the present application provides a method for training a projective test model, the method comprising the following steps:

[0005] Acquire stimulation information of the projective test, response information generated by the user based on the stimulation information, and label information corresponding to the response information;

[0006] Inputting the stimulus information and the response information into a preset deep learning model, and obtaining label prediction information output by the deep learning model, wherein at least one intermediate layer of the deep learning model is adjusted by a parameter matrix;

[0007] Determine a loss function of the parameter matrix according to the label prediction information and the label information;

[0008] The parameter matrix is ​​adjusted according to the loss function until a target matrix is ​​obtained.

[0009] In a second aspect, the present application further provides a training device for a projective test model, the training device for a projective test model comprising:

[0010] An information input module, used to obtain the stimulus information of the projective test, the response information generated by the user based on the stimulus information, and the label information corresponding to the response information;

[0011] An information prediction module, used to input the stimulus information and the response information into a preset deep learning model, and obtain label prediction information output by the deep learning model, wherein at least one intermediate layer of the deep learning model is adjusted by a parameter matrix;

[0012] A loss calculation module, used to determine the loss function of the parameter matrix according to the label prediction information and the label information;

[0013] The matrix adjustment module is used to adjust the parameter matrix according to the loss function until a target matrix is ​​obtained.

[0014] In a third aspect, the present application also provides a computer device, comprising a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein when the computer program is executed by the processor, the training method of the projective test model as described above is implemented.

[0015] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the training method of the projective test model as described above is implemented.

[0016] The present application provides a training method, device, equipment and computer storage medium for a projective test model, wherein the present application obtains the stimulus information of the projective test, the reaction information generated by the user based on the stimulus information and the label information corresponding to the reaction information; inputs the stimulus information and the reaction information into a preset deep learning model, and obtains the label prediction information output by the deep learning model, wherein at least one intermediate layer of the deep learning model is adjusted by a parameter matrix; the loss function of the parameter matrix is ​​determined according to the label prediction information and the label information; and the parameter matrix is ​​adjusted according to the loss function until a target matrix is ​​obtained. Since the target matrix obtained through training is used as a parameter matrix to adjust the intermediate layer of the deep learning model, the deep learning model can be used for projective testing, and the user's reaction information is automatically analyzed according to the stimulus information, thereby improving the analysis efficiency of the projective test. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 A flowchart of a method for training a projective test model provided in one embodiment of the present application;

[0019] Figure 2 A usage scenario diagram of a method for training a projective test model provided in one embodiment of the present application;

[0020] Figure 3 A schematic block diagram of a training device for a projective test model provided in one embodiment of the present application;

[0021] Figure 4 The present invention is a block diagram showing the structure of a computer device according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are 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 are within the scope of protection of this application.

[0023] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0024] Embodiments of the present application provide a method, apparatus, computer device, and computer-readable storage medium for training a projective test model.

[0025] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0026] Please refer to Figure 1 , Figure 1A flow chart of a method for training a projection test model provided in an embodiment of the present application. The training method of the projection test model can be used in a terminal or a server to automatically analyze the test results of the projection test and improve the analysis efficiency of the projection test. Among them, the terminal can be an electronic device such as a mobile phone, a tablet computer, a laptop computer, a desktop computer, a personal digital assistant, and a wearable device; the server can be an independent server, or a server cluster, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (Content Delivery Network, CDN), and big data and artificial intelligence platforms.

[0027] Please refer to Figure 2 , Figure 2 This is a usage scenario diagram provided by an embodiment of the present application. Figure 2 As shown, the deep learning model is adjusted through the parameter matrix, and based on the adjusted deep learning model, label prediction information is output according to the stimulus information and the response information. The parameter matrix is ​​adjusted according to the loss function between the label prediction information and the label information, and a target matrix is ​​trained that enables the deep learning model to accurately perform projection test analysis.

[0028] like Figure 1 As shown, the training method of the projection test model includes steps S101 to S104.

[0029] Step S101: Acquire stimulation information of the projective test, response information generated by the user based on the stimulation information, and label information corresponding to the response information.

[0030] For example, the projection experiment is a psychological research method that reveals the user's inner feelings, motivations and attitudes through the reaction information obtained by the user's interpretation of vague or ambiguous stimulus information. In the projection experiment, the stimulus information can be in the form of an image, and the user's description of the image is obtained as the reaction information. The content of the image can be a scene or ink. Of course, it is not limited to this. The stimulus information can also be an incomplete sentence, and the user's supplement to the sentence is obtained as the reaction information. This is not limited here.

[0031] In some implementations, the stimulus information includes image information, the reaction information includes text information, and the step of obtaining the stimulus information of the projective test, the reaction information generated by the user based on the stimulus information, and the label information corresponding to the reaction information includes:

[0032] An image identifier of the image information, text information describing the image information by the user, and tag information corresponding to the text information are obtained, wherein the tag information includes a corresponding preset tag of the user in at least one dimension.

[0033] For example, in an embodiment using image information as stimulus information, a series of image information may be stored in a designated folder, and the files corresponding to the image information may be named by image identifiers, for example:<des_img_001> ,<des_img_002> When training the model, the image identifier is obtained as stimulus information, and the model automatically reads the image in the folder according to the image identifier, thereby reducing the amount of data that needs to be transmitted for model training.

[0034] For example, the text information may be a user's description of the image information, and the label information may be a psychology expert's evaluation of the user in various dimensions based on the response information, such as the psychological expert giving the user an emotional tone, mental illness risk level, and personality label based on the response information.

[0035] Step S102: input the stimulus information and the response information into a preset deep learning model, and obtain label prediction information output by the deep learning model, wherein at least one intermediate layer of the deep learning model is adjusted by a parameter matrix.

[0036] Exemplarily, the deep learning model used in the training method of the projective test model provided in the embodiment of the present application can be a large language model (such as BERT, GPT, etc.). In a large language model, the number of parameters may be very large, reaching hundreds of millions or even more, and directly fine-tuning these parameters requires a large amount of computing resources and storage space. In order to reduce the computing resources required for model training, the present application adopts Low-Rank Adaptation of Large Language Models (LORA for short) to adjust the weight matrix of a large language model through a low-rank parameter matrix. Specifically, one or more parameter matrices for adjusting at least one intermediate layer of a deep learning model can be pre-set, and the one or more parameter matrices can be adjusted during the training process so that the deep learning model adjusted by the parameter matrix can be suitable for projective testing.

[0037] Exemplarily, the number of intermediate layers adjusted by the parameter matrix can be one or more, such as adjusting a specific attention mechanism layer, convolution layer or word embedding layer. Of course, it is not limited to this. The object adjusted by the parameter matrix can also be the output layer, so that the prediction result finally output by the model is more in line with the needs of the projection test.

[0038] In some implementations, inputting the stimulus information and the response information into a preset deep learning model, and obtaining label prediction information output by the deep learning model, includes:

[0039] Determining an actual matrix for outputting the label prediction information according to the original matrix of the deep learning model and the parameter matrix;

[0040] The stimulus information and the response information are transformed by using the actual matrix to obtain the label prediction information.

[0041] Exemplarily, the actual matrix of the deep learning model is determined based on the original matrix and the parameter matrix. Assuming that the original matrix of the deep learning model can be W0 and the parameter matrix is ​​ΔW, the actual matrix is ​​represented by W=W0+ΔW, and the actual matrix W is used as a weight matrix for transforming the stimulus information and response information of the input model. In this way, the original weight matrix W0 is retained, and the parameter matrix ΔW is adjusted through training to adapt to specific downstream tasks.

[0042] In some implementations, determining the actual matrix for outputting the label prediction information according to the original matrix of the deep learning model and the parameter matrix includes:

[0043] A parameter matrix for adjusting at least one intermediate layer of the deep learning model is determined according to the transpose of the first parameter matrix and the second parameter matrix.

[0044] For example, the parameter matrix ΔW can be decomposed into two smaller matrices A and B, and ΔW = □·B T , the sizes of A and B are usually much smaller than the size of the original matrix, which helps to reduce the number of model parameters and the complexity of training and calculation. For example, if the size of the original matrix W0 is M×N, then the size of A may be M×k, and the size of B may be N×k, where k is much smaller than M and N, so that the product of A and B forms a low-rank matrix of size M×N.

[0045] Step S103: determining a loss function of the parameter matrix according to the label prediction information and the label information.

[0046] Exemplarily, the loss function can quantify the difference between the model's label prediction information and label information. After the deep learning model outputs label prediction information based on the input stimulus information and response information, the loss information of the deep learning model predicting the label information is determined based on the label prediction information and label information corresponding to each group of stimulus information and response information, so as to adjust the parameter matrix of the deep learning model according to the loss information.

[0047] In some implementations, determining the loss function of the parameter matrix according to the label prediction information and the label information includes:

[0048] Determine an autoregressive loss function according to a result of the deep learning model performing an autoregressive task on the stimulus information and the response information;

[0049] Determine a classification loss function based on the label prediction information and the label information determined by the deep learning model performing a classification task on the stimulus information and the response information;

[0050] The loss function is determined according to the autoregressive loss function and the classification loss function.

[0051] Exemplarily, in the training method of the projection test model provided in the embodiment of the present application, the deep learning model needs to understand the input response information by performing an autoregressive task, and also needs to predict the corresponding label information by performing a classification task. Therefore, the loss function in the deep learning model training process includes the loss function of the autoregressive task and the loss function of the classification task.

[0052] In some embodiments, determining the autoregressive loss function according to the result of the deep learning model performing the autoregressive task on the stimulus information and the response information includes:

[0053] Acquire information to be predicted obtained by performing mask processing on the reaction information, where the information to be predicted includes at least a portion of the reaction information and mask information;

[0054] Performing an autoregressive task through the deep learning model to predict the mask information in the information to be predicted, and obtaining reaction prediction information;

[0055] The autoregressive loss function is determined according to the reaction information and the reaction prediction information.

[0056] Exemplarily, the deep learning network learns how to understand the reaction information by predicting the mask information part in the reaction information through the autoregressive task. Specifically, a part of the characters in the reaction information is replaced with the mask information to obtain the information to be predicted, and the deep learning network is used to predict the mask information in the information to be predicted to obtain the reaction prediction information, and the autoregressive loss function is determined according to the difference between the reaction prediction information and the reaction information. Among them, the autoregressive loss function can be determined according to the similarity between the reaction prediction information and the reaction information. The greater the similarity, the smaller the loss; the smaller the similarity, the greater the loss.

[0057] In some embodiments, determining a classification loss function based on the label prediction information and the label information determined by the deep learning model performing a classification task on the stimulus information and the response information includes:

[0058] Performing a classification task by the deep learning model, predicting at least one of the user's emotional tone, mental illness risk level, and personality label according to the stimulus information and the response information, and obtaining the label prediction information;

[0059] The classification loss function is determined according to the label prediction information and the label information.

[0060] Exemplarily, the dimensions of label information may include emotional tone, mental illness risk level, and personality label. Among them, the preset labels corresponding to emotional tone may include, for example, very positive, positive, neutral, negative, very negative, etc.; the preset labels corresponding to mental illness risk may include, for example, high, medium, and low risk of depression, high, medium, and low risk of mania, etc.; the preset labels corresponding to personality labels may include, for example, bold and cowardly types representing behavior, thinking and emotional types representing thinking, challenging and submissive types representing communication, etc. Therefore, predicting label information as a result of a projective test is actually a classification task, and the classification loss function is determined according to the difference between the label prediction information and the corresponding label information.

[0061] Step S104: adjusting the parameter matrix according to the loss function until a target matrix is ​​obtained.

[0062] Exemplarily, the loss function of the deep learning model is minimized through optimization algorithms such as gradient descent, and the parameter matrix is ​​adjusted so that the deep learning model can predict the label information under the adjustment of the parameter matrix, thereby obtaining a projective test model that can determine the label information based on the stimulus information and the user's response information, thereby improving the efficiency, accuracy and objectivity of the analysis of the projective test results.

[0063] The training method of the projective test model provided in the above embodiment obtains the stimulus information of the projective test, the reaction information generated by the user based on the stimulus information, and the label information corresponding to the reaction information; inputs the stimulus information and the reaction information into a preset deep learning model, and obtains the label prediction information output by the deep learning model, wherein at least one intermediate layer of the deep learning model is adjusted by a parameter matrix; the loss function of the parameter matrix is ​​determined according to the label prediction information and the label information; the parameter matrix is ​​adjusted according to the loss function until a target matrix is ​​obtained. Since the target matrix obtained through training is used as a parameter matrix to adjust the intermediate layer of the deep learning model, the deep learning model can be used for projective testing, and the user's reaction information is automatically analyzed according to the stimulus information, thereby improving the analysis efficiency of the projective test.

[0064] See also Figure 3 , Figure 3 It is a schematic diagram of a training device for a projective test model provided in one embodiment of the present application. The training device for the projective test model can be configured in a server or a terminal to execute the aforementioned training method for the projective test model.

[0065] like Figure 3 As shown, the training device of the projection test model includes: an information input module 110, an information prediction module 120, a loss calculation module 130, and a matrix adjustment module 140.

[0066] An information input module 110, used to obtain the stimulus information of the projective test, the response information generated by the user based on the stimulus information, and the label information corresponding to the response information;

[0067] An information prediction module 120, configured to input the stimulus information and the response information into a preset deep learning model, and obtain label prediction information output by the deep learning model, wherein at least one intermediate layer of the deep learning model is adjusted by a parameter matrix;

[0068] A loss calculation module 130, configured to determine a loss function of the parameter matrix according to the label prediction information and the label information;

[0069] The matrix adjustment module 140 is used to adjust the parameter matrix according to the loss function until a target matrix is ​​obtained.

[0070] In some implementations, the information input module 110, during the process of acquiring the stimulus information of the projective test, the response information generated by the user based on the stimulus information, and the label information corresponding to the response information, is used to perform:

[0071] An image identifier of the image information, text information describing the image information by the user, and tag information corresponding to the text information are obtained, wherein the tag information includes a corresponding preset tag of the user in at least one dimension.

[0072] In some implementations, the information prediction module 120, during the process of inputting the stimulus information and the response information into a preset deep learning model and obtaining the label prediction information output by the deep learning model, is configured to perform:

[0073] Determining an actual matrix for outputting the label prediction information according to the original matrix of the deep learning model and the parameter matrix;

[0074] The stimulus information and the response information are transformed by using the actual matrix to obtain the label prediction information.

[0075] In some implementations, the information prediction module 120, during the process of determining the actual matrix for outputting the label prediction information according to the original matrix of the deep learning model and the parameter matrix, is configured to perform:

[0076] A parameter matrix for adjusting at least one intermediate layer of the deep learning model is determined according to the transpose of the first parameter matrix and the second parameter matrix.

[0077] In some implementations, the loss calculation module 130, in the process of determining the loss function of the parameter matrix according to the label prediction information and the label information, is configured to perform:

[0078] Determine an autoregressive loss function according to a result of the deep learning model performing an autoregressive task on the stimulus information and the response information;

[0079] Determine a classification loss function based on the label prediction information and the label information determined by the deep learning model performing a classification task on the stimulus information and the response information;

[0080] The loss function is determined according to the autoregressive loss function and the classification loss function.

[0081] In some embodiments, the loss calculation module 130, in the process of determining the autoregressive loss function according to the result of the deep learning model performing the autoregressive task on the stimulus information and the response information, is used to perform:

[0082] Acquire information to be predicted obtained by performing mask processing on the reaction information, where the information to be predicted includes at least a portion of the reaction information and mask information;

[0083] Performing an autoregressive task through the deep learning model to predict the mask information in the information to be predicted, and obtaining reaction prediction information;

[0084] The autoregressive loss function is determined according to the reaction information and the reaction prediction information.

[0085] In some embodiments, the loss calculation module 130, in the process of determining the classification loss function based on the label prediction information and the label information determined by the deep learning model for performing the classification task for the stimulus information and the response information, is used to perform:

[0086] Performing a classification task by the deep learning model, predicting at least one of the user's emotional tone, mental illness risk level, and personality label according to the stimulus information and the response information, and obtaining the label prediction information;

[0087] The classification loss function is determined according to the label prediction information and the label information.

[0088] It should be noted that those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and modules and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0089] The method and apparatus of the present application can be used in many general or special computing system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0090] Exemplarily, the above method and apparatus may be implemented in the form of a computer program. The computer program may be implemented in Figure 4 Runs on the computer device shown.

[0091] See also Figure 4 , Figure 4A schematic block diagram of the structure of a computer device provided in an embodiment of the present application. The computer device may be a server or a terminal.

[0092] like Figure 4 As shown, the computer device includes a processor, a memory, and a network interface connected via a system bus, wherein the memory may include a storage medium and an internal memory.

[0093] The storage medium can store an operating system and a computer program. The computer program includes program instructions, and when the program instructions are executed, the processor can execute any training method of the projection test model.

[0094] The processor is used to provide computing and control capabilities and support the operation of the entire computer equipment.

[0095] The internal memory provides an environment for the operation of the computer program in the storage medium. When the computer program is executed by the processor, the processor can execute any training method of the projection test model.

[0096] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0097] It should be understood that the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0098] In one embodiment, the processor is used to run a computer program stored in the memory to implement the following steps:

[0099] Acquire stimulation information of the projective test, response information generated by the user based on the stimulation information, and label information corresponding to the response information;

[0100] Inputting the stimulus information and the response information into a preset deep learning model, and obtaining label prediction information output by the deep learning model, wherein at least one intermediate layer of the deep learning model is adjusted by a parameter matrix;

[0101] Determine a loss function of the parameter matrix according to the label prediction information and the label information;

[0102] The parameter matrix is ​​adjusted according to the loss function until a target matrix is ​​obtained.

[0103] In some implementations, the processor, in the process of acquiring the stimulus information of the projective test, the response information generated by the user based on the stimulus information, and the label information corresponding to the response information, is used to implement:

[0104] An image identifier of the image information, text information describing the image information by the user, and tag information corresponding to the text information are obtained, wherein the tag information includes a corresponding preset tag of the user in at least one dimension.

[0105] In some implementations, the processor, in the process of inputting the stimulus information and the response information into a preset deep learning model and obtaining the label prediction information output by the deep learning model, is used to implement:

[0106] Determining an actual matrix for outputting the label prediction information according to the original matrix of the deep learning model and the parameter matrix;

[0107] The stimulus information and the response information are transformed by using the actual matrix to obtain the label prediction information.

[0108] In some implementations, the processor, in the process of determining the actual matrix for outputting the label prediction information according to the original matrix of the deep learning model and the parameter matrix, is configured to implement:

[0109] A parameter matrix for adjusting at least one intermediate layer of the deep learning model is determined according to the transpose of the first parameter matrix and the second parameter matrix.

[0110] In some implementations, the processor, in the process of implementing the loss function of determining the parameter matrix according to the label prediction information and the label information, is configured to implement:

[0111] Determine an autoregressive loss function according to a result of the deep learning model performing an autoregressive task on the stimulus information and the response information;

[0112] Determine a classification loss function based on the label prediction information and the label information determined by the deep learning model performing a classification task on the stimulus information and the response information;

[0113] The loss function is determined according to the autoregressive loss function and the classification loss function.

[0114] In some embodiments, the processor, in the process of determining the autoregressive loss function according to the result of the deep learning model performing the autoregressive task on the stimulus information and the response information, is used to implement:

[0115] Acquire information to be predicted obtained by performing mask processing on the reaction information, where the information to be predicted includes at least a portion of the reaction information and mask information;

[0116] Performing an autoregressive task through the deep learning model to predict the mask information in the information to be predicted, and obtaining reaction prediction information;

[0117] The autoregressive loss function is determined according to the reaction information and the reaction prediction information.

[0118] In some embodiments, the processor, in the process of determining the classification loss function based on the label prediction information and the label information determined by performing the classification task on the stimulus information and the response information according to the deep learning model, is used to implement:

[0119] Performing a classification task by the deep learning model, predicting at least one of the user's emotional tone, mental illness risk level, and personality label according to the stimulus information and the response information, and obtaining the label prediction information;

[0120] The classification loss function is determined according to the label prediction information and the label information.

[0121] It should be noted that technical personnel in the relevant field can clearly understand that, for the convenience and conciseness of description, the specific working process of the training of the projection test model described above can refer to the corresponding process in the aforementioned embodiment of the training control method of the projection test model, and will not be repeated here.

[0122] An embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. The computer program includes program instructions. The method implemented when the program instructions are executed can refer to the various embodiments of the training method of the projection test model of the present application.

[0123] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiment, such as a hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc., equipped on the computer device.

[0124] It should be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0125] It should also be understood that the term "and / or" used in the specification of this application and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, including these combinations. It should be noted that, in this article, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "including a..." does not exclude the presence of other identical elements in the process, method, article or system including the element.

[0126] The serial numbers of the embodiments of the present application are for description only and do not represent the advantages and disadvantages of the embodiments. The above description is only a specific implementation mode of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for training a projective test model, characterized in that: The method comprises: Acquire stimulation information of the projective test, response information generated by the user based on the stimulation information, and label information corresponding to the response information; Inputting the stimulus information and the response information into a preset deep learning model, and obtaining label prediction information output by the deep learning model, wherein at least one intermediate layer of the deep learning model is adjusted by a parameter matrix; Determine a loss function of the parameter matrix according to the label prediction information and the label information; The parameter matrix is ​​adjusted according to the loss function until a target matrix is ​​obtained.

2. The method for training a projective test model according to claim 1, characterized in that: The stimulus information includes image information, the reaction information includes text information, and the obtaining of the stimulus information of the projection test, the reaction information generated by the user based on the stimulus information, and the label information corresponding to the reaction information include: An image identifier of the image information, text information describing the image information by the user, and tag information corresponding to the text information are obtained, wherein the tag information includes a corresponding preset tag of the user in at least one dimension.

3. The method for training a projective test model according to claim 1, characterized in that: The step of inputting the stimulus information and the response information into a preset deep learning model and obtaining label prediction information output by the deep learning model includes: Determining an actual matrix for outputting the label prediction information according to the original matrix of the deep learning model and the parameter matrix; The stimulus information and the response information are transformed by using the actual matrix to obtain the label prediction information.

4. The method for training a projective test model according to claim 3, characterized in that: The determining, according to the original matrix of the deep learning model and the parameter matrix, an actual matrix for outputting the label prediction information comprises: A parameter matrix for adjusting at least one intermediate layer of the deep learning model is determined according to the transpose of the first parameter matrix and the second parameter matrix.

5. The method for training a projective test model according to claim 1, characterized in that: The determining the loss function of the parameter matrix according to the label prediction information and the label information includes: Determine an autoregressive loss function according to a result of the deep learning model performing an autoregressive task on the stimulus information and the response information; Determine a classification loss function based on the label prediction information and the label information determined by the deep learning model performing a classification task on the stimulus information and the response information; The loss function is determined according to the autoregressive loss function and the classification loss function.

6. The method for training a projective test model according to claim 5, characterized in that: The determining of the autoregressive loss function according to the result of the deep learning model performing the autoregressive task on the stimulus information and the response information comprises: Acquire information to be predicted obtained by performing mask processing on the reaction information, where the information to be predicted includes at least a portion of the reaction information and mask information; Performing an autoregressive task through the deep learning model to predict the mask information in the information to be predicted, and obtaining reaction prediction information; The autoregressive loss function is determined according to the reaction information and the reaction prediction information.

7. The method for training a projective test model according to claim 5, characterized in that: The determining of the classification loss function based on the label prediction information and the label information determined by performing the classification task on the stimulus information and the response information according to the deep learning model comprises: Performing a classification task by the deep learning model, predicting at least one of the user's emotional tone, mental illness risk level, and personality label according to the stimulus information and the response information, and obtaining the label prediction information; The classification loss function is determined according to the label prediction information and the label information.

8. A training device for a projective test model, characterized in that: The training device of the projective test model comprises: An information input module, used to obtain the stimulus information of the projective test, the response information generated by the user based on the stimulus information, and the label information corresponding to the response information; An information prediction module, used to input the stimulus information and the response information into a preset deep learning model, and obtain label prediction information output by the deep learning model, wherein at least one intermediate layer of the deep learning model is adjusted by a parameter matrix; A loss calculation module, used to determine the loss function of the parameter matrix according to the label prediction information and the label information; The matrix adjustment module is used to adjust the parameter matrix according to the loss function until a target matrix is ​​obtained.

9. A computer device, characterized in that: The computer device includes a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein when the computer program is executed by the processor, the steps of the training method of the projective test model as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the steps of the method for training a projective test model as claimed in any one of claims 1 to 7 are implemented.

Images