Psychological sandbox description method and device based on large language model
The method leverages a large language model to objectively analyze sand tray configurations, addressing subjectivity in analyst interpretations and improving the accuracy and efficiency of psychological assessments.
Patent Information
- Application Number
- CN202411019874.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-29
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-07-29
AI Technical Summary
Existing sand tray analysis in psychological counseling is highly subjective due to analyst experience, leading to inaccurate psychological assessments.
A psychological sandbox description method using a large language model to extract and fuse spatial and semantic information from sand table images, performing multimodal fusion and reasoning to provide objective and accurate descriptions.
Improves the accuracy and efficiency of psychological sand table descriptions, enhancing the reliability of psychological assessments.
Smart Images

Figure CN118553382B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of medical care informatics, and in particular to a psychological sandbox description method and device based on a large language model. Background Art
[0002] Sandplay is widely used in the field of psychological counseling. It provides individuals with a non-verbal expression channel, reduces the sense of being tested, and helps psychoanalysts to deeply understand the psychological characteristics and behavioral patterns of clients and conduct effective psychological intervention on clients. In the field of psychological assessment, sandplay has been studied as a diagnostic tool, realizing its psychological assessment function by comparing the differences between different individuals.
[0003] In the related technology, the existing sand tray theory mainly identifies those with positive psychological symptoms based on the sand tray description. Specifically, the sand tray analyst combines the sand tray theory with experience to describe the client's psychology as shown in the sand tray, thereby achieving a preliminary psychological assessment of the client and assisting in the clinical diagnosis of mental illness. However, using the analyst's experience to describe the sand tray works is highly subjective, and there are deviations in the interpretations given by different analysts, resulting in inaccurate descriptions of the sand tray works, which in turn affects the user's psychological health assessment effect. Summary of the invention
[0004] The present invention provides a psychological sand table description method and device based on a large language model, so as to solve the defect that the prior art uses the analyst's experience to describe the sand table works, which is very subjective and leads to inaccurate description of the sand table works, thereby improving the accuracy and efficiency of the psychological sand table description.
[0005] The present invention provides a psychological sandbox description method based on a large language model, comprising:
[0006] Extracting sand tool spatial information and semantic information from a sand table visual image; the sand table visual image includes at least one sand tool;
[0007] Performing multimodal fusion on the sand tool spatial information and the semantic information to obtain fusion features;
[0008] Based on the large language model, psychological sand table reasoning is performed according to the fusion features to obtain a psychological sand table description result, wherein the psychological sand table description result is used to represent at least one of the spatial configuration of the sand table, the expression of the at least one sand tool, and the analysis of the key sand tool, and the key sand tool belongs to the at least one sand tool.
[0009] According to a psychological sand table description method based on a large language model provided by the present invention, the extraction of sand table spatial information and semantic information from the sand table visual image includes:
[0010] Acquire background information corresponding to at least one sand tool in the sand table visual image from the electronic sand table system, and convert the background information into the sand tool space information and the semantic information; wherein the background information includes the absolute coordinates and types corresponding to the at least one sand tool and the sandbox, respectively.
[0011] According to a psychological sand table description method based on a large language model provided by the present invention, the background information includes the absolute coordinates and type of the at least one sand tool;
[0012] The converting the background information into the sand tool space information and the semantic information comprises:
[0013] The absolute coordinates and type of the at least one sand tool are converted respectively according to the natural language processing model to obtain an intuitive description of the coordinates and an intuitive description of the type.
[0014] According to a psychological sand table description method based on a large language model provided by the present invention, the semantic information includes a plurality of different sand tool semantics, the sand tool space information includes a plurality of different sand tool space descriptions, and different sand tool semantics correspond to different sand tool space descriptions;
[0015] The multimodal fusion of the sand tool spatial information and the semantic information to obtain the fusion features includes:
[0016] Combining the multiple different sand tool semantics with the corresponding sand tool space descriptions to obtain multiple sand tool phrases;
[0017] The plurality of sand tool phrases are spliced together according to a preset prompt word template to obtain the fusion feature; the preset prompt word template includes the splicing relationship and position of the plurality of sand tool phrases.
[0018] According to a psychological sandbox description method based on a large language model provided by the present invention, the psychological sandbox reasoning is performed based on the large language model according to the fusion features to obtain the psychological sandbox description result, which includes:
[0019] The large language model fine-tuned based on the psychological sand table description performs psychological sand table reasoning according to the fusion features to obtain the psychological sand table description result; wherein the large language model fine-tuned based on the psychological sand table description is obtained based on any of the following methods;
[0020] The sample psychological sandbox description and large language model LLM training are obtained;
[0021] Acquired through transfer learning.
[0022] The present invention also provides a psychological sandbox description device based on a large language model, comprising:
[0023] An information extraction module is used to extract sand tool spatial information and semantic information from a sand table visual image; the sand table visual image includes at least one sand tool;
[0024] An information fusion module, used for multimodally fusing the sand tool spatial information and the semantic information to obtain fusion features;
[0025] The reasoning module is used to perform psychological sand tray reasoning according to the fusion features based on the large language model to obtain a psychological sand tray description result, wherein the psychological sand tray description result is used to represent at least one of the spatial configuration of the sand tray, the expression of the at least one sand tool, and the analysis of the key sand tool, and the key sand tool belongs to the at least one sand tool.
[0026] According to a psychological sandbox description device based on a large language model provided by the present invention, the information extraction module is specifically used for:
[0027] Acquire background information corresponding to at least one sand tool in the sand table visual image from the electronic sand table system, and convert the background information into the sand tool space information and the semantic information; wherein the background information includes the absolute coordinates and types corresponding to the at least one sand tool and the sandbox, respectively.
[0028] According to a psychological sand table description device based on a large language model provided by the present invention, the semantic information includes a plurality of different sand tool semantics, the sand tool space information includes a plurality of different sand tool space descriptions, and different sand tool semantics correspond to different sand tool space descriptions;
[0029] The information fusion module is specifically used for:
[0030] Combining the multiple different sand tool semantics with the corresponding sand tool space descriptions to obtain multiple sand tool phrases;
[0031] The plurality of sand tool phrases are spliced together according to a preset prompt word template to obtain the fusion feature; the preset prompt word template includes the splicing relationship and position of the plurality of sand tool phrases.
[0032] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the psychological sandbox description method based on a large language model as described above is implemented.
[0033] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described psychological sandbox description methods based on a large language model.
[0034] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above-mentioned psychological sandbox description methods based on a large language model.
[0035] The psychological sand table description method and device based on the large language model provided by the present invention extract the sand tool spatial information and semantic information from the sand table visual image, and perform multimodal fusion of the sand tool spatial information and semantic information to obtain fusion features. Finally, the large language model is used to perform psychological sand table reasoning according to the fusion features to obtain the psychological sand table description result, thereby improving the accuracy and efficiency of the psychological sand table description, and thus improving the efficiency of psychological counseling. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0037] Figure 1 It is one of the flow charts of the psychological sandbox description method based on the large language model provided by the present invention.
[0038] Figure 2 This is the second flow chart of the psychological sandbox description method based on the large language model provided by the present invention.
[0039] Figure 3 This is the third flow chart of the psychological sandbox description method based on the large language model provided by the present invention.
[0040] Figure 4 It is a structural schematic diagram of a psychological sandbox description device based on a large language model provided by the present invention.
[0041] Figure 5 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION
[0042] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0043] Combine the following Figure 1-Figure 4 The present invention describes the psychological sandbox description method and device based on a large language model.
[0044] Figure 1 This is one of the flow charts of the psychological sandbox description method based on the large language model provided by the present invention, such as Figure 1 As shown, the method comprises the following steps:
[0045] Step 110: extracting the spatial information and semantic information of the sand tool from the sand table visual image; the sand table visual image includes at least one sand tool.
[0046] In this step, sand tools are miniature models, also known as small objects, used by visitors through sandplay. They are also the "language" used by visitors. Visitors use a variety of sand tools to express their life experiences, psychological feelings, archetypes and symbolic meanings in the sand tray.
[0047] In this example, there can be many types of sand tools, and the more they cover, the better, so that the "vocabulary" of sand tools used by visitors will be richer and their creativity will be stronger.
[0048] In this embodiment, the spatial information of the sand tools includes the spatial coordinates of the sand tools, the relative position information and orientation information of the sand tools with respect to other objects in the sandbox, etc.; the semantic information of the sand tools includes the type, appearance, color and quantity of the sand tools, etc.
[0049] In this embodiment, the spatial information and semantic information of the sand tools can be extracted from the sand table visual image through an image feature extraction algorithm; for example, the image feature extraction algorithm can be a scale-invariant feature transform (SIFT) algorithm, a histogram of oriented gradients (HOG) algorithm, etc., or the spatial information of the image can be extracted through a neural network, for example, the corresponding spatial information and semantic information of the sand tools can be extracted through a convolutional neural network (CNN).
[0050] Step 120: Multimodally fuse the spatial information and semantic information of the sand tool to obtain fused features.
[0051] In this step, the multimodal fusion algorithm can be used to combine the spatial information and semantic information of the sand tools one by one, strengthen the association between the spatial position and semantic description of each sand tool, and thus obtain more accurate sand tool description information.
[0052] In some embodiments, the semantic information includes multiple different sand tool semantics, the sand tool space information includes multiple different sand tool space descriptions, and different sand tool semantics correspond to different sand tool space descriptions; the sand tool space information and the semantic information are multimodally fused to obtain fusion features, including: combining multiple different sand tool semantics with corresponding sand tool space descriptions to obtain multiple sand tool phrases; splicing multiple sand tool phrases according to a preset prompt word template to obtain a fusion feature; the preset prompt word template includes the splicing relationship and position of multiple sand tool phrases.
[0053] In this embodiment, for sand tool 1 and sand tool 2 in the sand table visual image, the corresponding sand tool semantics include the type and quantity of sand tools, for example, "Sand tool 1 is a car, and sand tool 1 includes 1", "Sand tool 2 is a bush", and sand tool 2 includes 2; the sand tool spatial description can be determined according to the coordinates of each sand tool. For example, according to the sand tool coordinates corresponding to sand tool 1 and sand tool 2, the sand tool spatial description of sand tool 1 is "Sand tool 1 is in the center of the picture", and the sand tool spatial description of sand tool 2 is "Sand tool 2 is on the left side of the picture"; the sand tool semantics and sand tool spatial description corresponding to each sand tool are combined to obtain the corresponding sand tool phrase, which can be "There are 6 sand tools in the center of the sand table, the car is in the center, and the bush is on the left side..."; then, according to the preset prompt word template, multiple sand tool phrases are connected into a complete prompt word, and the fusion feature can be expressed as "The car is in the center of the sand table, the bush is on the left side of the sand table..."
[0054] In this embodiment, by multimodally fusing the spatial information and semantic information of the sand tool, a fusion feature with stronger representation ability can be obtained, thereby obtaining a more accurate mental health reasoning result.
[0055] Step 130: Perform psychological sand tray reasoning based on the fusion features based on the large language model to obtain a psychological sand tray description result, which is used to represent at least one of the spatial configuration of the sand tray, the expression of at least one sand tool, and the analysis of the key sand tool, and the key sand tool belongs to at least one sand tool.
[0056] In this embodiment, the expression of the sand tools may be intuitive description information of the sand tools, such as the position, quantity, volume, shape, and color of the sand tray.
[0057] In this embodiment, the key sand tools can be set according to user needs. The key sand tools can be one or more. The descriptions corresponding to the key sand tools include key information for psychological assessment analysis.
[0058] In this embodiment, the large language model may be a prediction model trained for the scenario of predicting the visitor's mental health through psychological sandbox reasoning.
[0059] For example, the large language model can be a large language model fine-tuned by the psychological sandbox description.
[0060] Specifically, performing psychological sand table reasoning based on the fusion features based on the large language model to obtain the psychological sand table description result includes: performing psychological sand table reasoning based on the fusion features based on the large language model fine-tuned based on the psychological sand table description to obtain the psychological sand table description result; wherein the large language model fine-tuned based on the psychological sand table description is obtained based on any of the following methods; obtained by sample psychological sand table description and large language model LLM training; obtained through transfer learning.
[0061] In this embodiment, the large language model fine-tuned by the psychological sandbox description can be obtained through real-time training, or the network parameters of the available large model can be shared through transfer learning to obtain the corresponding large language model.
[0062] In this embodiment, for the creative content of the visitor in the sand tray, the corresponding creative content image, that is, the sand tray visual image, is first obtained through the electronic sand tray system; the corresponding sand tool image can also be obtained through a third-party camera device, such as the visitor's mobile phone, SLR camera, etc.; the system obtains the sand tool spatial information and semantic information from the acquired image, and then performs multimodal fusion of the sand tool spatial information and semantic information to obtain fusion features; finally, the large language model is used to perform psychological sand tray reasoning based on the fusion features to obtain the psychological sand tray description result.
[0063] The psychological sand table description method based on a large language model provided by an embodiment of the present invention extracts sand tool spatial information and semantic information from the sand table visual image, and performs multimodal fusion of the sand tool spatial information and semantic information to obtain fusion features. Finally, the large language model is used to perform psychological sand table reasoning according to the fusion features to obtain a psychological sand table description result, thereby improving the accuracy and efficiency of the psychological sand table description, and thus improving the efficiency of psychological counseling.
[0064] In some embodiments, extracting sand tool spatial information and semantic information from a sand table visual image includes: obtaining background information corresponding to at least one sand tool in the sand table visual image from an electronic sand table system, and converting the background information into sand tool spatial information and semantic information; wherein the background information includes the absolute coordinates and type corresponding to at least one sand tool.
[0065] Specifically, the background information includes the absolute coordinates and type of at least one sand tool; converting the background information into sand tool spatial information and semantic information includes: converting the absolute coordinates and type of at least one sand tool according to a natural language processing model to obtain a coordinate intuitive description and a type intuitive description.
[0066] In this embodiment, the coordinate information of each sand tool can be converted into a text description using a natural language processing model, and the natural language processing model can be obtained through pre-training.
[0067] Figure 2This is the second flow chart of the psychological sandbox description method based on the large language model provided by the present invention. Figure 2 In the embodiment shown, for the sand tool 1 and the sand tool 2 in the sand table visual image (corresponding to the sand table image), the sand tool coordinates corresponding to each sand tool in the sand table visual image are first obtained through the electronic sand table system. For example, the coordinates of a sand tool are ( x,y ), and then use the natural language processing model to convert the coordinates of each sand tool into an intuitive description. For example, the intuitive description corresponding to sand tool 1 is "Sand tool 1 is in the center of the picture", and the intuitive description corresponding to sand tool 2 is "Sand tool 2 is on the left side of the picture"; then combine the sand tool semantics and sand tool space description corresponding to each sand tool, and the corresponding sand tool phrase can be "There are 6 sand tools in the center of the sand table, the car is in the center, and the grass is on the left..."; then connect multiple sand tool phrases into complete prompt words according to the preset prompt word template, and the fused features can be expressed as "The car is in the center of the sand table, the grass is on the left side of the sand table..."; finally, input the fused features (corresponding prompt words) into the large language model for reasoning to obtain the corresponding psychological sand table description (including sand table space configuration, sand tool expression and key sand tool analysis).
[0068] The psychological sand table description method based on a large language model provided by an embodiment of the present invention improves the efficiency of obtaining the sand table spatial information and semantic information by acquiring the background information corresponding to at least one sand tool in the sand table visual image from an electronic sand table system and converting the background information into sand tool spatial information and semantic information, thereby improving the algorithm calculation efficiency.
[0069] Figure 3 This is the third flow chart of the psychological sandbox description method based on the large language model provided by the present invention. Figure 3 In the illustrated embodiment, the psychological sand table description method based on the large language model is specifically implemented through the following steps: S100, obtaining a sand table visual image; S200, based on the sand table visual image, obtaining the sand tool spatial information and semantic information in the visual image through a sand table feature extraction algorithm; S300, fusing the sand tool spatial information and semantic information through a multimodal fusion algorithm; S400, performing information reasoning based on a large model fine-tuned based on the psychological sand table description; S500, obtaining a psychological sand table description, including the sand table space configuration, sand tool expression and specific sand tool analysis.
[0070] The following is a description of the psychological sandbox description device based on a large language model provided by the present invention. The psychological sandbox description device based on a large language model described below and the psychological sandbox description method based on a large language model described above can refer to each other.
[0071] Figure 4 is a structural diagram of a psychological sandbox description device based on a large language model provided by the present invention, such as Figure 4As shown, the psychological sandbox description device based on the large language model includes: an information extraction module 410, an information fusion module 420 and a reasoning module 430.
[0072] The information extraction module 410 is used to extract the spatial information and semantic information of the sand tool from the sand table visual image; the sand table visual image includes at least one sand tool;
[0073] The information fusion module 420 is used to perform multimodal fusion of the sand tool spatial information and semantic information to obtain fusion features;
[0074] The reasoning module 430 is used to perform psychological sand tray reasoning based on the large language model according to the fusion features to obtain a psychological sand tray description result, which is used to represent at least one of the spatial configuration of the sand tray, the expression of at least one sand tool, and the analysis of key sand tools, and the key sand tools belong to at least one sand tool.
[0075] The psychological sand table description device based on the large language model provided by the embodiment of the present invention extracts the sand tool spatial information and semantic information from the sand table visual image, and performs multimodal fusion of the sand tool spatial information and semantic information to obtain fusion features. Finally, the large language model is used to perform psychological sand table reasoning according to the fusion features to obtain the psychological sand table description result, thereby improving the accuracy and efficiency of the psychological sand table description, and thus improving the efficiency of psychological counseling.
[0076] In some embodiments, the information extraction module is specifically used to: obtain background information corresponding to at least one sand tool in the sand table visual image from the electronic sand table system, and convert the background information into sand tool spatial information and semantic information; wherein the background information includes the absolute coordinates and type corresponding to at least one sand tool.
[0077] The psychological sand table description device based on the large language model provided by the embodiment of the present invention obtains background information corresponding to at least one sand tool in the sand table visual image from the electronic sand table system, and converts the background information into sand tool space information and semantic information.
[0078] In some embodiments, the semantic information includes multiple different sand tool semantics, the sand tool space information includes multiple different sand tool space descriptions, and different sand tool semantics correspond to different sand tool space descriptions; the information fusion module is specifically used to: combine multiple different sand tool semantics with corresponding sand tool space descriptions to obtain multiple sand tool phrases; splice the multiple sand tool phrases according to a preset prompt word template to obtain a fusion feature; the preset prompt word template includes the splicing relationship and position of multiple sand tool phrases.
[0079] The psychological sand table description device based on a large language model provided by an embodiment of the present invention obtains background information corresponding to at least one sand tool in a sand table visual image from an electronic sand table system, and converts the background information into sand tool spatial information and semantic information, thereby improving the efficiency of obtaining sand tool spatial information and semantic information, thereby improving the algorithm calculation efficiency.
[0080] Figure 5 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 5 As shown, the electronic device may include: a processor 510, a communication interface 520, a memory 530 and a communication bus 540, wherein the processor 510, the communication interface 520 and the memory 530 communicate with each other through the communication bus 540. The processor 510 may call the logic instructions in the memory 530 to execute the psychological sand table description method based on the large language model, the method comprising: extracting the sand tool spatial information and semantic information from the sand table visual image; the sand table visual image includes at least one sand tool; multimodally fusing the sand tool spatial information and semantic information to obtain a fusion feature; performing psychological sand table reasoning based on the fusion feature based on the large language model to obtain a psychological sand table description result, the psychological sand table description result is used to represent at least one of the spatial configuration of the sand table, the expression of at least one sand tool and the analysis of the key sand tool, and the key sand tool belongs to at least one sand tool.
[0081] In addition, the logic instructions in the above-mentioned memory 530 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the 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 invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.
[0082] On the other hand, the present invention also provides a computer program product, which includes a computer program, and the computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the psychological sand tray description method based on a large language model provided by the above methods, the method including: extracting sand tool spatial information and semantic information from a sand tray visual image; the sand tray visual image includes at least one sand tool; multimodally fusing the sand tool spatial information and semantic information to obtain fusion features; performing psychological sand tray reasoning based on the fusion features based on the large language model to obtain a psychological sand tray description result, the psychological sand tray description result is used to represent at least one of the spatial configuration of the sand tray, the expression of at least one sand tool, and the analysis of key sand tools, and the key sand tools belong to at least one sand tool.
[0083] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the psychological sand tray description method based on a large language model provided by the above-mentioned methods, the method comprising: extracting sand tool spatial information and semantic information from a sand tray visual image; the sand tray visual image comprises at least one sand tool; multimodally fusing the sand tool spatial information and semantic information to obtain fusion features; performing psychological sand tray reasoning based on the fusion features based on a large language model to obtain a psychological sand tray description result, the psychological sand tray description result being used to represent at least one of the spatial configuration of the sand tray, the expression of at least one sand tool, and the analysis of key sand tools, wherein the key sand tools belong to at least one sand tool.
[0084] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.
[0085] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0086] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A psychological sandbox description method based on a large language model, characterized in that: include: Extracting the spatial information and semantic information of the sand tool from the sand table visual image; the sand table visual image includes at least one sand tool; the spatial information of the sand tool includes the spatial coordinates of the sand tool, the relative position information and the orientation information with respect to other objects in the sand table; the semantic information of the sand tool includes the type, appearance, color and quantity of the sand tool; Performing multimodal fusion on the sand tool spatial information and the semantic information to obtain fusion features; Based on the large language model, psychological sand table reasoning is performed according to the fusion features to obtain a psychological sand table description result, wherein the psychological sand table description result is used to represent at least one of the spatial configuration of the sand table, the expression of the at least one sand tool, and the analysis of the key sand tool, wherein the key sand tool belongs to the at least one sand tool; extracting the sand tool spatial information and semantic information from the sand table visual image includes: Acquire background information corresponding to at least one sand tool in the sand table visual image from the electronic sand table system, and convert the background information into the sand tool spatial information and the semantic information; wherein the background information includes the absolute coordinates and type corresponding to the at least one sand tool and the sandbox respectively; the background information includes the absolute coordinates and type of the at least one sand tool; The converting the background information into the sand tool space information and the semantic information comprises: The absolute coordinates and type of the at least one sand tool are respectively converted according to a natural language processing model to obtain a coordinate intuitive description and a type intuitive description; the semantic information includes a plurality of different sand tool semantics, and the sand tool space information includes a plurality of different sand tool space descriptions, and different sand tool semantics correspond to different sand tool space descriptions; The multimodal fusion of the sand tool spatial information and the semantic information to obtain the fusion features includes: Combining the multiple different sand tool semantics with the corresponding sand tool space descriptions to obtain multiple sand tool phrases; The plurality of sand tool phrases are spliced according to a preset prompt word template to obtain the fusion feature; the preset prompt word template includes the splicing relationship and position of the plurality of sand tool phrases; the psychological sand table reasoning is performed based on the large language model according to the fusion feature to obtain the psychological sand table description result including: The large language model fine-tuned based on the psychological sand table description performs psychological sand table reasoning according to the fusion features to obtain the psychological sand table description result; wherein the large language model fine-tuned based on the psychological sand table description is obtained based on any of the following methods; The sample psychological sandbox description and large language model LLM training are obtained; Acquired through transfer learning.
2. A psychological sandbox description device based on a large language model corresponding to the psychological sandbox description method based on a large language model according to claim 1, characterized in that: include: An information extraction module is used to extract the spatial information and semantic information of the sand tool from the sand table visual image; the sand table visual image includes at least one sand tool; the sand tool spatial information includes the spatial coordinates of the sand tool, the relative position information and orientation information with respect to other objects in the sand table; the sand tool semantic information includes the type, appearance, color and quantity of the sand tool; An information fusion module, used for multimodally fusing the sand tool spatial information and the semantic information to obtain fusion features; The reasoning module is used to perform psychological sand tray reasoning according to the fusion features based on the large language model to obtain a psychological sand tray description result, wherein the psychological sand tray description result is used to represent at least one of the spatial configuration of the sand tray, the expression of the at least one sand tool, and the analysis of the key sand tool, and the key sand tool belongs to the at least one sand tool.
3. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the psychological sandbox description method based on the large language model as claimed in claim 1 is implemented.
4. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the psychological sandbox description method based on a large language model as claimed in claim 1 is implemented.
