Material information generation method and device, storage medium and electronic equipment

Abstract

The invention provides a material information generation method and device, a storage medium and electronic equipment, and the method comprises the steps: carrying out the retrieval in a pre-constructed material knowledge graph according to a keyword contained in a material information query request, and obtaining a target material node; and integrating the attribute information of the target material node and the association relationship into a node context. And inputting the material information query request and the node context into the large language model subjected to domain knowledge fine tuning, so that the large language model generates response information for the material information query request based on the node context. By combining the structured domain knowledge of the material knowledge graph with the fine-tuned large language model, the problem of illusion, namely imaginary information, existing in professional field application of a traditional large language model can be effectively solved, and the accuracy and reliability of professional field material information generation are greatly improved.

Description

Technical Field

[0001] The technical solution disclosed herein relates to the field of material science and technology, and in particular to a material information generation method and device, a storage medium, and an electronic device. Background Art

[0002] In recent years, large language models have made significant progress in the field of natural language processing, demonstrating excellent text understanding and generation capabilities.

[0003] However, when applied to professional fields such as materials science, large language models have obvious limitations: because their training data often lacks coverage of deep professional knowledge, the model is prone to the so-called "hallucination" phenomenon when responding to professional queries, that is, generating content that seems reasonable but actually lacks scientific basis or is inconsistent with the facts. Summary of the Invention

[0004] In view of this, embodiments of the present disclosure provide a material information generating method and device, a storage medium, and an electronic device.

[0005] According to a first aspect of the present disclosure, a method for generating material information is proposed, the method comprising:

[0006] According to the keywords contained in the material information query request, the pre-built material knowledge graph is searched to obtain the target material node;

[0007] Integrating the attribute information and association relationship of the target material node into a node context;

[0008] The material information query request and the node context are input into a large language model fine-tuned by domain knowledge, so that the large language model generates response information for the material information query request based on the node context.

[0009] In combination with any embodiment provided by the present disclosure, the keyword contained in the material information query request is searched in a pre-built material knowledge graph to obtain the target material node, including:

[0010] Mapping each keyword included in the material information query request into a corresponding first feature vector;

[0011] Mapping the name and attribute information of each material node in the material knowledge graph into a corresponding second feature vector;

[0012] Calculating the similarity between each first eigenvector and each second eigenvector respectively;

[0013] The material node corresponding to the second eigenvector that meets the preset similarity condition is determined as the target material node.

[0014] In combination with any embodiment provided by the present disclosure,

[0015] Fine-tune the large language model using at least one of the following question-answer pairs:

[0016] Material properties, material application, material generation.

[0017] In combination with any embodiment provided by the present disclosure, the material knowledge graph stores material information of at least one material science sub-field, and the material knowledge graph is constructed based on the following method:

[0018] For each target field in the at least one materials science sub-field, generate a search term, and obtain field documents related to the target field from a selected literature database based on the search term;

[0019] Extracting material information of at least one material from the literature in the field;

[0020] A material knowledge subgraph corresponding to the target field is constructed based on the material information of the at least one material; the nodes in the material knowledge subgraph represent materials, and the edges between the nodes represent the association relationship between the materials;

[0021] The material knowledge sub-graphs corresponding to each target field in the at least one materials science sub-field are integrated to obtain the material knowledge graph.

[0022] In conjunction with any embodiment provided in the present disclosure, the material information of at least one material extracted from the field literature includes:

[0023] Inputting a text-based ontology corresponding to the target domain into a large language model, so that the large language model generates information extraction prompt information adapted to the target domain based on the ontology;

[0024] The information extraction prompt information and at least a portion of the text content of the field document are input into a large language model, so that the large language model extracts material information of at least one material from at least a portion of the text content of the field document based on the information extraction prompt information.

[0025] According to a second aspect of the present disclosure, a material information generating device is provided, the device comprising:

[0026] The node retrieval module is used to search the pre-built material knowledge graph based on the keywords contained in the material information query request to obtain the target material node;

[0027] A context integration module, configured to integrate the attribute information and association relationships of the target material node into a node context;

[0028] An information generation module is configured to input the material information query request and the node context into a large language model fine-tuned by domain knowledge, so that the large language model generates response information for the material information query request based on the node context.

[0029] According to a third aspect of the present disclosure, a computer-readable storage medium is provided, wherein the machine-readable storage medium stores machine-readable instructions, which, when called and executed by a processor, prompt the processor to implement the material information generation method of any embodiment of the present disclosure.

[0030] According to a fourth aspect of the present disclosure, there is provided an electronic device comprising

[0031] processor;

[0032] a memory for storing processor-executable instructions;

[0033] The processor is configured to execute the material information generating method of any embodiment of the present disclosure.

[0034] The technical solutions provided by the embodiments of the present disclosure may have the following beneficial effects:

[0035] In the material information generation method and device, storage medium and electronic device provided by the embodiments of the present disclosure, first, according to the keywords contained in the material information query request, a search is performed in the pre-constructed material knowledge graph to obtain the target material node. Then, the attribute information and association relationship of the target material node are integrated into the node context, and the material information query request and the node context are input into the large language model fine-tuned by the domain knowledge, so that the large language model generates response information for the material information query request based on the node context. By combining the structured domain knowledge of the material knowledge graph with the fine-tuned large language model, the "hallucination" (i.e., fictitious information) problem existing in the application of traditional large language models in professional fields can be effectively overcome, and the accuracy and reliability of material information generation in professional fields can be greatly improved.

[0036] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0038] Figure 1 is a flow chart of a method for generating material information according to an exemplary embodiment of the present disclosure;

[0039] Figure 2 is a schematic flow chart of a material synthesis path according to an exemplary embodiment of the present disclosure;

[0040] Figure 3 is a flow chart of another material information generating method according to an exemplary embodiment of the present disclosure;

[0041] Figure 4 is a flow chart of another material information generating method according to an exemplary embodiment of the present disclosure;

[0042] Figure 5a This is a schematic diagram of a material information acquisition process in the field of metal catalysts according to an exemplary embodiment of the present disclosure;

[0043] Figure 5b is a structural diagram of a material knowledge graph according to an exemplary embodiment of the present disclosure;

[0044] Figure 6 is a structural diagram of a material information generating device according to an exemplary embodiment of the present disclosure;

[0045] Figure 7 It is a structural diagram of an electronic device according to an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION

[0046] Here, exemplary embodiments will be described in detail, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numerals in different drawings represent the same or similar elements.

[0047] In recent years, large language models have made significant progress in the field of natural language processing, demonstrating excellent text understanding and generation capabilities.

[0048] However, when applied to professional fields such as materials science, large language models have obvious limitations: because their training data often lacks coverage of deep professional knowledge, the model is prone to the so-called "hallucination" phenomenon when responding to professional queries, that is, generating content that seems reasonable but actually lacks scientific basis or is inconsistent with the facts.

[0049] In light of this, embodiments of the present disclosure provide a method for generating material information. This method effectively overcomes the "hallucination" (i.e., fabricated information) problem that exists in traditional large language models when applied in specialized fields by combining the structured domain knowledge of a material knowledge graph with a fine-tuned large language model.

[0050] The material information generating method of the embodiment of the present disclosure is described in detail below with reference to the accompanying drawings.

[0051] Figure 1 This is a flow chart of a material information generation method according to an exemplary embodiment of the present disclosure. The method can be executed by various computing devices, including but not limited to computer devices. Figure 1 As shown, the exemplary embodiment method may include the following steps:

[0052] In step 101, a search is performed in a pre-built material knowledge graph based on the keywords contained in the material information query request to obtain a target material node.

[0053] After receiving a material information query request from a user, keyword extraction can be performed on the material information query request to obtain at least one keyword contained in the material information query request. Then, for each extracted keyword, a multi-dimensional search and matching can be performed based on the keyword in a pre-built material knowledge graph, including but not limited to exact matching of material node names and fuzzy matching of material node attribute information, to obtain at least one material node that matches the keyword. This at least one material node can be referred to as a target material node.

[0054] Taking the pre-constructed material knowledge graph containing at least material node A and material node B as an example, the attribute information of the material node A includes at least: oxidant, and the attribute information of the material node B includes at least: catalyst, oxidant.

[0055] When a user enters a material information query request such as "What materials can be used as oxidants?", keyword extraction can be performed on the material information query request to obtain the keyword "oxidant." Then, based on the keyword "oxidant," a multi-dimensional search and matching can be performed in the pre-built material knowledge graph, including matching the names of each material node in the material knowledge graph and the attribute information of each material node. This results in obtaining material nodes A and B that match the keyword "oxidant." Material nodes A and B can be referred to as target material nodes.

[0056] When a user enters a material information query request such as "What are the synthesis pathways for material A?", keyword extraction can be performed on the material information query request to obtain the keyword "material A." Then, based on the keyword "material A," a multi-dimensional search and matching can be performed within the pre-built material knowledge graph. This includes matching the names of each material node in the material knowledge graph and the attribute information of each material node. This results in obtaining a material node A that matches the keyword "material A." This material node A can be referred to as the target material node.

[0057] In step 102, the attribute information and association relationship of the target material node are integrated into a node context.

[0058] The association relationship includes but is not limited to reaction relationship, catalytic relationship, adsorption relationship, etc.

[0059] For example, association relationships include reaction relationships, which are used to characterize the transformation or generation relationship between different materials. Specifically, when material A can be transformed into material C through a specific physical or chemical process, a reaction relationship pointing from material A to material C can be established in the material knowledge graph.

[0060] In this example, the reaction relationships of the target material node specifically refer to a series of reaction relationships that use the target material as a product. Specifically, these reaction relationships record key information such as the various precursor materials that can synthesize the target material and their corresponding reaction conditions, process parameters, etc.

[0061] In the aforementioned example where the material information query request is "What materials can be used as oxidants", the attribute information and association relationship of material node A (i.e., at least one synthesis path using material A to generate the product) can be integrated into the first node context, and the attribute information and association relationship of material node B (i.e., at least one synthesis path using material B to generate the product) can be integrated into the second node context.

[0062] In the aforementioned example where the material information query request is "What are the synthesis paths of material A?", the attribute information and association relationship of material node A (i.e., at least one synthesis path with material A as the generated product) can be integrated into the third node context.

[0063] In step 103 , the material information query request and the node context are input into a large language model fine-tuned with domain knowledge, so that the large language model generates response information for the material information query request based on the node context.

[0064] In the example of the aforementioned material information query request being "What materials can be used as oxidants", the material information query request "What materials can be used as oxidants" and the aforementioned first node context and second node context can be input into a large language model fine-tuned by domain knowledge, so that the large language model generates response information based on the input first node context and second node context, such as "material A and material B".

[0065] In the example of the aforementioned material information query request being "What are the synthesis paths of material A?", the material information query request "What are the synthesis paths of material A?" and the aforementioned third node context can be input into the large language model fine-tuned by domain knowledge, so that the large language model generates response information based on the input third node context, for example, Figure 2 As shown, multiple synthetic pathways to generate material A (the substance on the far right) are illustrated.

[0066] Optionally, when the user inputs a material information query request such as "What is the optimal synthesis path of material A?", the large language model obtains the following based on the input third node context: Figure 2 After the multiple synthesis paths are shown, the optimal synthesis path can be screened out from the multiple synthesis paths and output.

[0067] Furthermore, after outputting the optimal synthesis path, when receiving user information regarding the optimal synthesis path, such as inquiry information about specific operation steps, the large language model can generate corresponding answers based on the information in the material knowledge graph.

[0068] In an optional embodiment, the Qwen2-7b model can be used as the aforementioned large language model. In this case, to enable the large language model to better understand the information in the pre-built material knowledge graph, question-answer pairs containing at least material properties, material applications, and material generation can be established based on the pre-built material knowledge graph. The LoRA fine-tuning method can then be used to fine-tune the large language model using the aforementioned question-answer pairs to enable it to understand the domain knowledge of multiple types of materials.

[0069] For example, the question-and-answer pairs for material properties may be:

[0070] Question: What are the conductive properties of graphene?

[0071] Answer: Graphene has excellent electrical conductivity, and its electron mobility can reach 200,000 cm 2 / (V·s), it is one of the materials with the best electrical conductivity at room temperature.

[0072] Examples of question-and-answer sessions for material applications include:

[0073] Question: What is the role of solvents in organic reactions?

[0074] Answer: 1) Provide an interaction medium for reactants; 2) Affect reaction rate; 3) Absorb or transfer reaction heat; 4) Stabilize intermediates or transition states; 5) Affect reaction equilibrium; 6) Provide a specific polar environment.

[0075] The question-answer pairs generated for the material can be:

[0076] Question: What are the common operating steps in organic synthesis?

[0077] Answer: The routine operating steps of organic synthesis include: adding materials, stirring, heating, cooling, refluxing, distillation, filtration, extraction, concentration, crystallization, drying, chromatographic separation and yield determination, etc.

[0078] It should be noted that the aforementioned description of establishing question-and-answer pairs encompassing material properties, material applications, and material generation is merely illustrative, intended to facilitate a better understanding of the technical solutions of the embodiments of the present disclosure by those skilled in the art. In practical applications, question-and-answer pairs encompassing molecular structure representations, reaction conditions, operational steps, and the like may also be established, and this disclosure does not limit this.

[0079] In the material information generation method provided by the embodiment of the present disclosure, by combining the structured domain knowledge of the material knowledge graph with the fine-tuned large language model, the "hallucination" (i.e., fictitious information) problem existing in the application of traditional large language models in professional fields can be effectively overcome, and the accuracy and reliability of material information generation in professional fields can be greatly improved. At the same time, when screening materials, the method can quickly and accurately locate suitable materials based on user functional requirements. When planning the synthesis path, the method can recommend the optimal synthesis path with the help of the reaction relationship between each material in the graph and the knowledge reserve of the large language model. For example, the synthesis path with the highest efficiency can be recommended, which can greatly reduce the time cost of material research and development and accelerate the development process of new materials.

[0080] In an optional embodiment, if Figure 3 As shown, the aforementioned step 101 may specifically include:

[0081] In step 301, each keyword included in the material information query request is mapped to a corresponding first feature vector.

[0082] As previously described, after receiving a material information query request input by a user, keyword extraction can be performed on the material information query request to obtain at least one keyword contained in the material information query request. Then, for each extracted keyword, the keyword can be mapped to a corresponding first feature vector.

[0083] In the aforementioned example where the material information query request is "What materials can be used as oxidants?", the keyword "oxidant" can be mapped to a corresponding first feature vector.

[0084] In the aforementioned example where the material information query request is "What are the synthesis paths of material A?", the keyword "material A" may be mapped to a corresponding first feature vector.

[0085] In step 302, the name and attribute information of each material node in the material knowledge graph are mapped to a corresponding second feature vector.

[0086] Still taking the example of a pre-built material knowledge graph containing at least material node A and material node B, the name and attribute information of material node A can be mapped into corresponding second eigenvectors, and the name and attribute information of material node B can be mapped into corresponding second eigenvectors respectively.

[0087] In step 303 , the similarity between each first eigenvector and each second eigenvector is calculated respectively.

[0088] In the example where the material information query request is "What materials can be used as oxidants?", after mapping the keyword "oxidant" to a corresponding first feature vector, the similarity between the first feature vector and the aforementioned second feature vectors can be calculated.

[0089] In the example where the material information query request is "What are the synthesis paths of material A?", after mapping the keyword "material A" to the corresponding first feature vector, the similarity between the first feature vector and the aforementioned second feature vectors can be calculated.

[0090] In step 304, the material node corresponding to the second eigenvector that meets the preset similarity condition is determined as the target material node.

[0091] The preset similarity condition may include any of the following: the similarity value is greater than or equal to a preset similarity threshold, after the calculated similarities are arranged in descending order, it is ranked in the top N positions, where N is a positive integer, etc.

[0092] For ease of understanding, the following embodiments are described by taking the preset similarity condition that the similarity value is greater than or equal to a preset similarity threshold as an example.

[0093] In combination with the above, in the example where the material information query request is "What materials can be used as oxidants", it can be determined that the similarity between the second eigenvectors corresponding to the attribute information of material node A and material node B and the first eigenvector corresponding to the keyword "oxidant" is greater than the preset similarity threshold. At this time, material node A and material node B can be determined as the target material nodes.

[0094] In the aforementioned example where the material information query request is "What are the synthesis paths of material A", it can be determined that the similarity between the second feature vector corresponding to the name of material node A and the first feature vector corresponding to the keyword "material A" is greater than the preset similarity threshold. At this time, material node A can be determined as the target material node.

[0095] In the material information generation method provided in the embodiment of the present disclosure, intelligent retrieval of material knowledge graphs is achieved through feature vector mapping and similarity calculation, which can significantly improve the efficiency and accuracy of retrieval.

[0096] In an optional embodiment, the aforementioned material knowledge graph stores material information of at least one material science sub-field, such as Figure 4 As shown, the material knowledge graph can be constructed based on the following methods:

[0097] In step 401, for each target field in the at least one materials science sub-field, a search term is generated, and field documents related to the target field are acquired from a selected literature database based on the search term.

[0098] The at least one materials science sub-field may include but is not limited to: metal catalyst materials, inorganic non-metallic materials, polymer materials, small molecule materials, etc.

[0099] Taking the target field of metal catalyst materials as an example, corresponding search terms can be generated, such as the search term "metal catalyst materials". Then, relevant field documents can be obtained from a pre-selected literature database based on "metal catalyst materials".

[0100] In step 402, material information of at least one material is extracted from the field literature.

[0101] After obtaining the field literature related to the field of metal catalyst materials, material information of at least one metal catalyst material can be extracted from these field literatures. The material information includes but is not limited to the property information and reaction relationship of the metal material.

[0102] Optionally, different text-based ontologies corresponding to different material science sub-fields can be obtained in advance. Then, for the field of metal catalyst materials, the text-based ontology corresponding to the field of metal catalyst materials can be first input into the large language model, so that the large language model generates information extraction prompt information that is compatible with the field of metal catalyst materials based on the ontology. Then, the information extraction prompt information and at least part of the text content in the field literature of the field of metal catalyst materials obtained above, such as the text content in the main body of the document, or the text content in the abstract of the document, can be input into the large language model, so that the large language model extracts material information of at least one metal catalyst material from at least part of the text content of the field literature based on the information extraction prompt information.

[0103] Figure 5a The material information acquisition process in the field of "metal catalysts" is illustrated. First, the search term "metal catalyst" is generated for the field of "metal catalysts". Then, relevant field documents are obtained from the open database based on "metal catalyst". At the same time, the text-based ontology corresponding to the field of "metal catalysts" can be input into the large language model to obtain information extraction prompt information (prompt) output by the large language model. Finally, the information extraction prompt information and at least part of the text content of the aforementioned field document can be input into the large language model, so that the large language model can extract material information of at least one metal catalyst material from at least part of the text content of the field document based on the information extraction prompt information.

[0104] In step 403, a material knowledge subgraph corresponding to the target field is constructed based on the material information of the at least one material.

[0105] A material knowledge subgraph corresponding to the field of metal materials can be constructed based on the material information of at least one metal material obtained above. The nodes in the material knowledge subgraph represent metal materials, and the edges between the nodes represent the association relationship between the metal materials.

[0106] In step 404, the material knowledge sub-graphs corresponding to each target field in the at least one material science sub-field are integrated to obtain the material knowledge graph.

[0107] Based on similar methods as mentioned above, material knowledge sub-graphs corresponding to the fields of inorganic non-metallic materials, polymer materials, small molecule materials, etc. can be constructed respectively.

[0108] Then, the material knowledge subgraphs corresponding to the fields of metal materials, inorganic non-metallic materials, polymer materials, small molecule materials, etc. can be integrated to obtain the material knowledge graph. For example, the structure of the integrated material knowledge graph can be as follows Figure 5b shown.

[0109] The material information generation method provided in the embodiments of the present disclosure achieves intelligent integration of knowledge in materials science sub-fields, such as metals, inorganic non-metals, polymers, and small molecules, by constructing a multi-domain integrated material knowledge graph. This material knowledge graph can provide comprehensive and reliable knowledge support for materials research and development, enabling researchers to quickly obtain accurate material information, discover potential material combinations and reaction pathways, and significantly improve the efficiency of new material research and development. Furthermore, this method uses ontology-guided information extraction technology to accurately obtain material information from domain literature, significantly improving the efficiency of knowledge graph construction.

[0110] For the sake of simplicity, the aforementioned method embodiments are all expressed as a series of action combinations. However, those skilled in the art should know that the present disclosure is not limited to the order of the actions described, because according to the present disclosure, certain steps can be performed in other orders or simultaneously.

[0111] Corresponding to the aforementioned method embodiments, the present disclosure also provides apparatus embodiments.

[0112] Figure 6 is a structural diagram of a material information generating device according to an exemplary embodiment of the present disclosure. Figure 6 The material information generating device may include:

[0113] The node retrieval module 61 is used to search the pre-built material knowledge graph according to the keywords contained in the material information query request to obtain the target material node.

[0114] The context integration module 62 is used to integrate the attribute information and association relationship of the target material node into a node context.

[0115] The information generation module 63 is configured to input the material information query request and the node context into a large language model fine-tuned by domain knowledge, so that the large language model generates response information for the material information query request based on the node context.

[0116] Optionally, the node retrieval module 61, when used to search the pre-built material knowledge graph based on the keywords included in the material information query request to obtain the target material node, includes:

[0117] Each keyword included in the material information query request is mapped to a corresponding first feature vector.

[0118] The name and attribute information of each material node in the material knowledge graph are respectively mapped to corresponding second feature vectors.

[0119] The similarity between each first eigenvector and each second eigenvector is calculated respectively.

[0120] The material node corresponding to the second eigenvector that meets the preset similarity condition is determined as the target material node.

[0121] Optionally, the large language model is fine-tuned using at least one of the following question-answer pairs:

[0122] Material properties, material application, material generation.

[0123] Optionally, the material knowledge graph stores material information of at least one material science sub-field, and the material knowledge graph is constructed based on the following modules:

[0124] The field literature acquisition module is used to generate search terms for each target field in the at least one materials science sub-field, and acquire field literature related to the target field from a selected literature database based on the search terms.

[0125] The material information extraction module is used to extract material information of at least one material from the field literature.

[0126] A subgraph construction module is used to construct a material knowledge subgraph corresponding to the target field based on the material information of the at least one material; the nodes in the material knowledge subgraph represent materials, and the edges between the nodes represent the association relationship between the materials.

[0127] The sub-graph integration module is used to integrate the material knowledge sub-graphs corresponding to each target field in the at least one material science sub-field to obtain the material knowledge graph.

[0128] Optionally, the material information extraction module, when used to extract material information of at least one material from the field literature, includes:

[0129] An ontology in text form corresponding to the target domain is input into a large language model, so that the large language model generates information extraction prompt information adapted to the target domain based on the ontology.

[0130] The information extraction prompt information and at least a portion of the text content of the field document are input into a large language model, so that the large language model extracts material information of at least one material from at least a portion of the text content of the field document based on the information extraction prompt information.

[0131] As for the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can be referred to the partial description of the method embodiment.

[0132] Figure 7 FIG2 is a schematic diagram showing the structure of an electronic device 700 according to an exemplary embodiment of the present disclosure. The electronic device may be any type of computing device, including but not limited to a computer device.

[0133] Reference Figure 7 , electronic device 700 may include one or more of the following components: a processing component 702 , a memory 704 , a power component 706 , a multimedia component 708 , an audio component 710 , an input / output (I / O) interface 712 , a sensor component 714 , and a communication component 716 .

[0134] The processing component 702 generally controls the overall operation of the electronic device 700, such as operations associated with display, phone calls, data communications, camera operation, and recording operations. The processing component 702 may include one or more processors 720 to execute instructions to perform all or part of the steps of the above-described method. In addition, the processing component 702 may include one or more modules to facilitate interaction between the processing component 702 and other components. For example, the processing component 702 may include a multimedia module to facilitate interaction between the multimedia component 708 and the processing component 702.

[0135] The memory 704 is configured to store various types of data to support operations on the device 700. Examples of such data include instructions for any application or method operating on the electronic device 700, contact data, phone book data, messages, pictures, videos, etc. The memory 704 can be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0136] The power supply component 706 provides power to the various components of the electronic device 700. The power supply component 706 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the electronic device 700.

[0137] The multimedia component 708 includes a screen that provides an output interface between the electronic device 700 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, slides, and gestures on the touch panel. The touch sensors may not only sense the boundaries of a touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 708 includes a front camera and / or a rear camera. When the electronic device 700 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera may receive external multimedia data. Each front camera and rear camera may be a fixed optical lens system or have focal length and optical zoom capabilities.

[0138] The audio component 710 is configured to output and / or input audio signals. For example, the audio component 710 includes a microphone (MIC), which is configured to receive external audio signals when the electronic device 700 is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signal can be further stored in the memory 704 or transmitted via the communication component 716. In some embodiments, the audio component 710 also includes a speaker for outputting audio signals.

[0139] I / O interface 712 provides an interface between processing component 702 and peripheral interface modules, such as a keyboard, click wheel, buttons, etc. These buttons may include but are not limited to: a home button, volume buttons, a start button, and a lock button.

[0140] The sensor assembly 714 includes one or more sensors for providing various aspects of status assessment for the electronic device 700. For example, the sensor assembly 714 can detect the open / closed state of the electronic device 700, the relative positioning of components, such as the display and keypad of the electronic device 700. The sensor assembly 714 can also detect changes in the position of the electronic device 700 or a component of the electronic device 700, the presence or absence of user contact with the electronic device 700, the orientation or acceleration / deceleration of the electronic device 700, and temperature changes of the electronic device 700. The sensor assembly 714 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 714 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 714 may also include an accelerometer, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0141] The communication component 716 is configured to facilitate wired or wireless communication between the electronic device 700 and other devices. The electronic device 700 can access a wireless network based on a communication standard, such as WiFi, 4G or 5G, 4G LTE, 5G NR or a combination thereof. In an exemplary embodiment, the communication component 716 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 716 further includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0142] In an exemplary embodiment, the electronic device 700 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above methods.

[0143] In an exemplary embodiment, a non-temporary computer-readable storage medium is also provided, such as a memory 704 including instructions. When the instructions in the storage medium are executed by the processor 720 of the electronic device 700, the electronic device 700 is enabled to execute the material information generation method of any embodiment of the present disclosure.

[0144] The non-transitory computer-readable storage medium may be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.

[0145] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A material information generation method, characterized in that: The method comprises: According to the keywords contained in the material information query request, the pre-built material knowledge graph is searched to obtain the target material node; Integrating the attribute information and association relationship of the target material node into a node context; The material information query request and the node context are input into a large language model fine-tuned by domain knowledge, so that the large language model generates response information for the material information query request based on the node context.

2. The method according to claim 1, characterized in that The method of searching the pre-built material knowledge graph based on the keywords contained in the material information query request to obtain the target material node includes: Mapping each keyword included in the material information query request into a corresponding first feature vector; Mapping the name and attribute information of each material node in the material knowledge graph into a corresponding second feature vector; Calculating the similarity between each first eigenvector and each second eigenvector respectively; The material node corresponding to the second eigenvector that meets the preset similarity condition is determined as the target material node.

3. The method according to claim 1, characterized in that Fine-tune the large language model using at least one of the following question-answer pairs: Material properties, material application, material generation.

4. The method according to claim 1, wherein The material knowledge graph stores material information of at least one material science sub-field, and is constructed based on the following method: For each target field in the at least one materials science sub-field, generate a search term, and obtain field documents related to the target field from a selected literature database based on the search term; Extracting material information of at least one material from the literature in the field; A material knowledge subgraph corresponding to the target field is constructed based on the material information of the at least one material; the nodes in the material knowledge subgraph represent materials, and the edges between the nodes represent the association relationship between the materials; The material knowledge sub-graphs corresponding to each target field in the at least one materials science sub-field are integrated to obtain the material knowledge graph.

5. The method according to claim 4, characterized in that The material information of at least one material is extracted from the field literature, including: Inputting a text-based ontology corresponding to the target domain into a large language model, so that the large language model generates information extraction prompt information adapted to the target domain based on the ontology; The information extraction prompt information and at least a portion of the text content of the field document are input into a large language model, so that the large language model extracts material information of at least one material from at least a portion of the text content of the field document based on the information extraction prompt information.

6. A material information generating device, characterized in that: The device comprises: The node retrieval module is used to search the pre-built material knowledge graph based on the keywords contained in the material information query request to obtain the target material node; A context integration module, configured to integrate the attribute information and association relationships of the target material node into a node context; An information generation module is configured to input the material information query request and the node context into a large language model fine-tuned by domain knowledge, so that the large language model generates response information for the material information query request based on the node context.

7. The device according to claim 6, characterized in that The node retrieval module, when used to search the pre-built material knowledge graph based on the keywords contained in the material information query request to obtain the target material node, includes: Mapping each keyword included in the material information query request into a corresponding first feature vector; Mapping the name and attribute information of each material node in the material knowledge graph into a corresponding second feature vector; Calculating the similarity between each first eigenvector and each second eigenvector respectively; The material node corresponding to the second eigenvector that meets the preset similarity condition is determined as the target material node.

8. The device according to claim 6, characterized in that Fine-tune the large language model using at least one of the following question-answer pairs: Material properties, material application, material generation.

9. A computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

10. An electronic device comprising: processor; a memory for storing processor-executable instructions; The processor is configured to execute the steps of any one of the methods of claims 1-5.

Images