Term processing method and device, storage medium and electronic equipment

Abstract

The invention discloses a term processing method and device, a storage medium and electronic equipment. The method comprises the following steps: performing term splitting processing on at least two original terms to obtain a first number of first term components; combining at least two first term components with the same expression meaning in the first number of first term components to obtain a second number of second term components; under the condition that the first term operation network and the second term operation network are obtained, at least one output node meeting the recall condition corresponding to the node of the first term operation network is recalled from the nodes of the second term operation network; the term component corresponding to at least one output node is determined as at least one output standard term, and the method can be applied to artificial intelligence scenes and relates to technologies such as artificial intelligence medical technologies. The technical problem that the recall efficiency of the standard terms is low is solved.

Description

Technical Field

[0001] The present application relates to the field of computers, and in particular, to a method and apparatus for term processing, a storage medium, and an electronic device. Background Art

[0002] In the recall scenario of standard terms, it is usually necessary to face batch processing of hundreds of thousands / millions of original words (original terms). If normalized one by one, it will be extremely time-consuming and laborious, which will lead to the problem of low recall efficiency of standard terms. Therefore, there is a problem of low recall efficiency of standard terms.

[0003] In response to the above problems, no effective solution has been proposed yet. Summary of the Invention

[0004] Embodiments of the present application provide a method and apparatus for term processing, a storage medium, and an electronic device, so as to at least solve the technical problem of low recall efficiency of standard terms.

[0005] According to one aspect of the embodiments of the present application, a method for term processing is provided, including: in response to a standard term recall request triggered by at least two original terms in a specific field, performing term splitting processing on the at least two original terms to obtain a first number of first term components, where the first term components are term components in the original terms, and the standard term recall request is used to request recall of at least one output standard term that matches the at least two original terms; merging at least two first term components with the same expressed meaning among the first number of first term components to obtain a second number of second term components, where the second number is less than the first number, and the second term components are the first term components or term components obtained after merging the at least two first term components; constructing a first term operation network for the at least two original terms, and constructing a second term operation network for the standard terms in the specific field; recalling at least one output node that satisfies the recall condition corresponding to the nodes of the first term operation network from the nodes of the second term operation network, where the first term operation network is a term operation network constructed for the at least two original terms, the second term operation network is a term operation network constructed for the standard terms in the specific field, the nodes of the first term operation network correspond to the term components in the second number of second term components, and the nodes of the second term operation network correspond to the term components in the standard terms; determining the term components corresponding to the at least one output node as the at least one output standard term.

[0006] According to another aspect of the embodiments of the present application, there is also provided a term processing device, including: a splitting unit, configured to perform term splitting processing on the at least two original terms in response to a standard term recall request triggered for at least two original terms in a specific field, to obtain a first quantity of first term components, where the first term components are term components in the original terms, and the standard term recall request is used to request recalling at least one output standard term that matches the at least two original terms; a merging unit, configured to merge at least two first term components with the same expressed meaning among the first quantity of first term components, to obtain a second quantity of second term components, where the second quantity is less than the first quantity, and the second term components are the first term components or term components obtained after merging the at least two first term components; a constructing unit, configured to construct a first term operation network for the at least two original terms, and construct a second term operation network for the standard terms in the specific field; a recalling unit, configured to recall at least one output node that meets the recall condition corresponding to the nodes of the first term operation network from the nodes of the second term operation network, where the first term operation network is a term operation network constructed for the at least two original terms, the second term operation network is a term operation network constructed for the standard terms in the specific field, the nodes of the first term operation network correspond to the term components in the second quantity of second term components, and the nodes of the second term operation network correspond to the term components in the standard terms; a determining unit, configured to determine the term components corresponding to the at least one output node as the at least one output standard term.

[0007] As an alternative, the above device further includes at least one of the following: a first acquisition unit, configured to convert the first quantity of first term components into semantic vectors; measure the vector similarity between the semantic vectors, and determine, among the first quantity of first term components, multiple term components whose vector similarity is greater than or equal to a first similarity threshold as the at least two first term components, where the vector similarity has a positive correlation with the semantic similarity between the first quantity of first term components; a second acquisition unit, configured to deeply analyze the context associated with the first quantity of first term components, and extract context information having a close relationship with the first quantity of first term components from the context; deeply fuse the context information with the first quantity of first term components, and capture the information expression after the deep fusion; determine the information expression as the expression meaning corresponding to each of the first term components in the first quantity of first term components; a third acquisition unit, configured to measure the information similarity between the first quantity of first term components and each corpus in a preset corpus through the semantic vectors; and determine the at least two first term components from the first quantity of first term components by using the information similarity.

[0008] As an alternative solution, the above device further includes: a fourth acquisition unit, configured to, before measuring the information similarity between the first term components of the first quantity and each corpus in the preset corpus through the semantic vector, acquire a first component and a second component belonging to a specific business scenario from the first term components of the first quantity; a fifth acquisition unit, configured to, before measuring the information similarity between the first term components of the first quantity and each corpus in the preset corpus through the semantic vector, acquire the corpus matching the specific business scenario, where the preset corpus includes the corpus matching the specific business scenario; the third acquisition unit includes: a first acquisition module, configured to acquire a first similarity between the first component and each corpus in the corpus matching the specific business scenario; a first determination module, configured to, when the first similarity between the first component and each corpus in the corpus matching the specific business scenario is acquired and any of the first similarities is greater than or equal to a second preset threshold, use the first corpus with the highest first similarity in the corpus matching the specific business scenario as the expression meaning corresponding to the first component, where the information similarity includes the first similarity; a second acquisition module, configured to acquire a second similarity between the second component and each corpus in the corpus matching the specific business scenario; a second determination module, configured to, when the second similarity between the second component and each corpus in the corpus matching the specific business scenario is acquired and any of the second similarities is greater than or equal to the second preset threshold, use the second corpus with the highest second similarity in the corpus matching the specific business scenario as the expression meaning corresponding to the second component, where the information similarity includes the second similarity; a third acquisition module, configured to acquire a corpus similarity between the first corpus and the second corpus; a third determination module, configured to, when the corpus similarity is greater than or equal to a third similarity threshold, determine that the first component and the second component have the same expression meaning.

[0009] As an alternative, the above-mentioned recall unit includes: a fourth acquisition module, configured to acquire a sub-structure belonging to the first level between the first term operation network and the second term operation network, wherein the sub-structure belonging to the first level includes a first sub-structure of the first term operation network and a second sub-structure of the second term operation network, nodes of the first sub-structure correspond to term components belonging to the first level among the second quantity of second term components, and nodes of the second sub-structure correspond to term components belonging to the first level among the standard terms; a fourth determination module, configured to determine at least one first node among the nodes of the second sub-structure that satisfies a first condition for recalling the nodes of the first sub-structure, wherein the recall condition includes the first condition, and the similarity between the term component corresponding to the first node and the term component corresponding to the node of the first sub-structure is greater than or equal to a threshold value matched by the first condition; a fifth acquisition module, configured to acquire a sub-structure belonging to the second level between the first term operation network and the second term operation network, wherein the sub-structure belonging to the second level includes a third sub-structure of the first term operation network and a fourth sub-structure of the second term operation network, nodes of the third sub-structure correspond to term components belonging to the second level among the second quantity of second term components, nodes of the second sub-structure correspond to term components belonging to the second level among the standard terms, and the second level is a higher-level corresponding to the first level; a fifth determination module, configured to determine at least one second node among the nodes of the fourth sub-structure that has an indexing relationship with the at least one first node and satisfies a second condition for recalling the nodes of the third sub-structure, wherein the recall condition includes the second condition, and having the indexing relationship indicates an upper-lower relationship between nodes belonging to different levels, and the similarity between the term component corresponding to the second node and the term component corresponding to the node of the third sub-structure is greater than or equal to a threshold value matched by the second condition; a sixth determination module, configured to, when the number of sub-structures at the higher level corresponding to the second level is less than or equal to a preset threshold value, determine the at least one second node as the at least one output node.

[0010] As an alternative solution, the above merging unit includes: a first merging module, configured to merge the term components that have the same meaning and belong to the above first level among the above first quantity of first term components, to obtain a plurality of first-level components, and use the plurality of first-level components as the term components corresponding to the nodes of the above first sub-structure, wherein the above second quantity of second term components includes the plurality of first-level components; a second merging module, configured to merge the term components that have the same meaning and belong to the above second level among the above first quantity of first term components, to obtain a plurality of second-level components, and use the plurality of second-level components as the term components corresponding to the nodes of the above second sub-structure, wherein the above second quantity of second term components includes the plurality of second-level components.

[0011] As an alternative solution, the above device further includes: a classification module, configured to, before obtaining the sub-structures belonging to the first level between the above first term operation network and the above second term operation network, classify the above second quantity of second term components according to the component type, wherein the top-level sub-structure in the above first term operation network corresponds to the above at least two original terms, and the nodes of the sub-structures at other different levels in the above first term operation network correspond to the second term components of different above component types.

[0012] As an alternative solution, the above merging unit includes: a third merging module, configured to, when the above first quantity is greater than or equal to the excess threshold, merge the above at least two first term components to obtain the above second quantity of second term components.

[0013] As an alternative solution, the above third merging module includes: a first merging sub-module, configured to, when the above first quantity is greater than or equal to the excess threshold and less than the massive threshold, merge the above at least two first term components to obtain a second term component quantity in the first interval, wherein the above massive threshold is greater than the above excess threshold; or, a second merging sub-module, configured to, when the above first quantity is greater than or equal to the excess threshold and greater than or equal to the above massive threshold, merge the above at least two first term components to obtain a second term component quantity in the second interval, wherein the above second interval quantity is greater than the above first interval quantity.

[0014] According to another aspect of the embodiments of the present application, a computer program product or a computer program is provided. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the electronic device executes the term processing method as described above.

[0015] According to another aspect of the embodiments of the present application, an electronic device is further provided, including a memory, a processor, and a computer program stored on the memory and executable on the processor. Wherein, the above-mentioned processor executes the above-mentioned term processing method through the computer program.

[0016] In the embodiments of the present application, in response to a standard term recall request triggered by at least two original terms in a specific field, the above-mentioned at least two original terms are subjected to term splitting processing to obtain a first number of first term components, wherein the above-mentioned first term components are term components in the above-mentioned original terms, and the above-mentioned standard term recall request is used to request to recall at least one output standard term that matches the above-mentioned at least two original terms; at least two first term components with the same meaning in the above-mentioned first number of first term components are merged to obtain a second number of second term components, wherein the above-mentioned second number is less than the above-mentioned first number, and the above-mentioned second term components are the above-mentioned first term components, or term components obtained after merging the above-mentioned at least two first term components; a first term operation network is constructed for the above-mentioned at least two original terms, and a second term operation network is constructed for the standard terms in the above-mentioned specific field; at least one output node that meets the recall conditions corresponding to the nodes of the above-mentioned first term operation network is recalled from the nodes of the above-mentioned second term operation network, wherein the above-mentioned first term operation network is a term operation network constructed for the above-mentioned at least two original terms, the above-mentioned second term operation network is a term operation network constructed for the standard terms in the above-mentioned specific field, the nodes of the above-mentioned first term operation network correspond to the term components in the above-mentioned second number of second term components, and the nodes of the above-mentioned second term operation network correspond to the term components in the above-mentioned standard terms; the term components corresponding to the above-mentioned at least one output node are determined as the above-mentioned at least one output standard term. By splitting the original terms into smaller term components and further merging the term components with the same meaning, the number of terms is reduced, which facilitates subsequent recall of standard terms related to the input terms by using the term operation network, thereby achieving the purpose of reducing the computational amount of recall using the term operation network, and thus realizing the technical effect of improving the recall efficiency of standard terms, and further solving the technical problem of low recall efficiency of standard terms. Description of the Drawings

[0017] The accompanying drawings described herein are used to provide a further understanding of the present application, and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application, and do not constitute an improper limitation of the present application. In the drawings:

[0018] Figure 1 is a schematic diagram of an application environment of an optional term processing method according to an embodiment of the present application;

[0019] Figure 2 is a schematic diagram of a process of an optional term processing method according to an embodiment of the present application;

[0020] Figure 3 is a schematic diagram of an optional term processing method according to an embodiment of the present application;

[0021] Figure 4 is a schematic diagram of another optional term processing method according to an embodiment of the present application;

[0022] Figure 5 is a schematic diagram of another optional term processing method according to an embodiment of the present application;

[0023] Figure 6 is a schematic diagram of another optional term processing method according to an embodiment of the present application;

[0024] Figure 7 is a schematic diagram of another optional term processing method according to an embodiment of the present application;

[0025] Figure 8 is a schematic diagram of another optional term processing method according to an embodiment of the present application;

[0026] Figure 9 is a schematic diagram of another optional term processing method according to an embodiment of the present application;

[0027] Figure 10 is a schematic diagram of an optional term processing device according to an embodiment of the present application;

[0028] Figure 11 is a schematic diagram of the structure of an optional electronic device according to an embodiment of the present application. Detailed implementation manners

[0029] To enable those skilled in the art to better understand the solution of this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of this application.

[0030] It should be noted that the terms "first", "second", etc. in the description and claims of this application and the above-mentioned accompanying drawings are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of this application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or are inherent to these processes, methods, products or devices.

[0031] For ease of understanding, the following terms are explained:

[0032] Medical entity: refers to a noun phrase in the text, such as a disease, a drug, a treatment method, etc.

[0033] Term standardization (diagnostic normalization): is an essential task in medical statistics. Clinically, there are often hundreds or thousands of different ways to write about the same diagnosis. The problem that standardization (normalization) aims to solve is to find the corresponding standard medical term expression for various different expressions in clinical practice.

[0034] Artificial Intelligence (AI for short) is to use a digital computer or a machine controlled by a digital computer to simulate, extend, and expand human intelligence, a theory, method, technology, and application system that can perceive the environment, acquire knowledge, and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science. It attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a way similar to human intelligence. Artificial intelligence is also to study the design principles and implementation methods of various intelligent machines to enable the machines to have the functions of perception, reasoning, and decision-making.

[0035] Artificial intelligence technology is an interdisciplinary subject that involves a wide range of fields, including both hardware-level and software-level technologies. The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, pre-trained model technology, operation / interaction systems, mechatronics, etc. Among them, the pre-trained model, also known as the large model or the basic model, can be widely applied to downstream tasks in various directions of artificial intelligence after fine-tuning. The software technologies of artificial intelligence mainly include several major directions such as computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0036] With the research and progress of artificial intelligence technology, artificial intelligence technology has been studied and applied in multiple fields. For example, common applications include smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, driverless, autonomous driving, drones, digital twins, virtual humans, robots, artificial intelligence-generated content (AIGC), conversational interaction, intelligent healthcare, intelligent customer service, game AI, etc. It is believed that with the development of technology, artificial intelligence technology will be applied in more fields and play an increasingly important role.

[0037] The solution provided by the embodiments of this application relates to technologies such as artificial intelligence in healthcare, and is specifically described through the following embodiments:

[0038] According to one aspect of the embodiments of this application, a term processing method is provided. Optionally, as an alternative implementation, the above term processing method can be but is not limited to being applied to an environment such as Figure 1 shown. Among them, it can but is not limited to include a user device 102 and a server 112. The user device 102 can but is not limited to include a display 104, a processor 106, and a memory 108. The server 112 includes a database 114 and a processing engine 116.

[0039] The specific process can be as follows:

[0040] Step S102, the user device 102 obtains a standard term recall request triggered by at least two original terms in a specific field;

[0041] Step S104, the standard term recall request is sent to the server 112 through the network 110;

[0042] Steps S106 - S110, the server 112 responds to the standard term recall request, and through the processing engine 116, performs term splitting on at least two original terms to obtain a first quantity of first term components; and further combines at least two first term components with the same expressed meaning among the first quantity of first term components to obtain a second quantity of second term components; in the case of obtaining the first term operation network and the second term operation network, recall at least one output node that meets the recall condition corresponding to the nodes of the first term operation network from the nodes of the second term operation network; and determine the term components corresponding to the at least one output node as at least one output standard term.

[0043] Step S112, send the output standard term to the user device 102 through the network 110. The user device 102 displays the output standard term on the display 104 through the processor 106, and stores the above output standard term in the memory 108.

[0044] Among them, before step S108, it is also possible to obtain the vector similarity between the first quantity of first term components, and determine multiple term components with a vector similarity greater than or equal to the first similarity threshold among the first quantity of first term components as at least two first term components; or, obtain the context information associated with the first quantity of first term components, and use the context information to determine the expressed meaning corresponding to each first term component among the first quantity of first term components; or, obtain the information similarity between each first term component and each corpus in the preset corpus, and use the information similarity to determine at least two first term components from the first quantity of first term components.

[0045] Optionally, step S110 may further include: obtaining the sub - structures belonging to the first level between the first term operation network and the second term operation network. Among them, the sub - structures belonging to the first level include the first sub - structure of the first term operation network and the second sub - structure of the second term operation network. The nodes of the first sub - structure correspond to the term components belonging to the first level among the second quantity of second term components, and the nodes of the second sub - structure correspond to the term components belonging to the first level among the standard terms.

[0046] Determine at least one first node among the nodes of the second sub - structure that meets the first condition for recalling the nodes of the first sub - structure. Among them, the recall condition includes the first condition, and the similarity between the term component corresponding to the first node and the term component corresponding to the node of the first sub - structure is greater than or equal to the threshold matched by the first condition.

[0047] Obtain the sub-structures at the second level between the first term operation network and the second term operation network. Among them, the sub-structures at the second level include the third sub-structure of the first term operation network and the fourth sub-structure of the second term operation network. The nodes of the third sub-structure correspond to the term components at the second level among the second quantity of second term components, and the nodes of the second sub-structure correspond to the term components at the second level in the standard terms. The second level is the upper level corresponding to the first level;

[0048] Determine at least one second node among the nodes of the fourth sub-structure that has an indexing relationship with at least one first node and satisfies the second condition for recalling the nodes of the third sub-structure. Among them, the recall condition includes the second condition, and having an indexing relationship means that there is a hierarchical relationship between nodes belonging to different levels. The similarity between the term components corresponding to the second node and the term components corresponding to the nodes of the third sub-structure is greater than or equal to the threshold matched by the second condition;

[0049] In the case where the number of sub-structures at the upper level corresponding to the second level is less than or equal to the preset threshold, determine at least one second node as at least one output node.

[0050] Optionally, step S110 may further include: in the case where the first quantity is greater than or equal to the excess threshold and less than the massive threshold, merge at least two first term components to obtain the second term components with the first interval quantity; or,

[0051] In the case where the first quantity is greater than or equal to the excess threshold and greater than or equal to the massive threshold, merge at least two first term components to obtain the second term components with the second interval quantity, where the second interval quantity is greater than the first interval quantity.

[0052] Except Figure 1 In addition to the examples shown, the above terminal device may be a terminal device configured with a target client, and may include but are not limited to at least one of the following: mobile phone (such as Android mobile phone, iOS mobile phone, etc.), laptop computer, tablet computer, handheld computer, MID (Mobile Internet Devices), PAD, desktop computer, smart TV, etc. The target client may be a video client, instant messaging client, browser client, education client, etc. The above network may include but are not limited to: wired network, wireless network, where the wired network includes: local area network, metropolitan area network and wide area network, and the wireless network includes: Bluetooth, WIFI and other networks that implement wireless communication. The above server may be a single server, or a server cluster composed of multiple servers, or a cloud server. The above is only an example, and this embodiment does not make any limitation thereto.

[0053] Optionally, as an alternative implementation, as Figure 2 shown, the term processing method may be executed by an electronic device, which may be, for example, a Figure 1 user device or a server as shown, and the specific steps include:

[0054] S202, in response to a standard term recall request triggered by at least two original terms in a specific field, perform term splitting processing on the at least two original terms to obtain a first quantity of first term components, where the first term components are term components in the original terms, and the standard term recall request is used to request the recall of at least one output standard term that matches the at least two original terms;

[0055] S204, merge at least two first term components with the same expressed meaning among the first quantity of first term components to obtain a second quantity of second term components, where the second quantity is less than the first quantity, the second term components are first term components, or term components obtained after merging at least two first term components;

[0056] S206, construct a first term operation network for at least two original terms to obtain a first term operation network, and construct a second term operation network for standard terms in a specific field to obtain a second term operation network;

[0057] S208, recall at least one output node that meets the recall conditions corresponding to the nodes of the first term operation network from the nodes of the second term operation network, where the first term operation network is a term operation network constructed for at least two original terms, the second term operation network is a term operation network constructed for standard terms in a specific field, the nodes of the first term operation network correspond to the term components in the second quantity of second term components, and the nodes of the second term operation network correspond to the term components in the standard terms;

[0058] S210, determine the term components corresponding to the at least one output node as at least one output standard term.

[0059] Optionally, in this embodiment, the above term processing method may be but is not limited to being applied in the medical field (specific field). In the medical field, different doctors or research institutions may use different terms to describe the same medical concept. By merging term components with the same expressed meaning, this embodiment can unify these different descriptions and reduce the redundancy and complexity of terms. For example, merge "myocardial infarction" and "heart infarction" into a unified standard term.

[0060] In this embodiment, considering that the original terms in the medical field may be very complex and lengthy, such as the full names of diseases, the chemical names of drugs, etc. Through term splitting processing, this embodiment can split these complex medical terms into smaller and more manageable term components. For example, for a certain complex disease name, this embodiment can split it into term components such as disease type, cause, symptoms, etc., and deduplicate a relatively large number of original terms through more refined term components to obtain a controllable number of original terms for subsequent standard term recall.

[0061] In addition, there are usually a large number of standard medical term libraries in the medical field, such as UMLS (Unified Medical Language System), etc. By constructing a second term operation network for standard medical terms, this embodiment can efficiently match and recall standard medical terms related to the original terms. For example, for a specific disease name, standard medical terms that match these term components can be recalled in the second term operation network, such as the ICD-10 code of the disease, related anatomical structures, pathophysiological processes, etc.

[0062] Through the matching and recall with the standard medical term library, this embodiment can determine the standard medical terms that match the original medical terms. These standard medical terms can be the official names of diseases, the generic names of drugs, the standard terms of anatomical structures, etc. This can help doctors more accurately understand and describe medical concepts, promote communication and cooperation among doctors, and improve the quality and efficiency of medical services.

[0063] Optionally, in this embodiment, when a user or system in a specific field (such as the medical field) issues a standard term recall request and provides at least two original terms as input, this embodiment will respond to this request. The response method is to perform term splitting processing on these original terms to obtain smaller and more specific term components. These term components are then used for further recall processing to find the standard terms that match the input original terms.

[0064] For further illustration, optionally, for example, in the field of computer science, assume that a user issues a standard term recall request and provides "distributed system" and "cloud computing architecture" as original terms. The system first performs term splitting processing on these original terms, and the first term components that may be obtained include "distributed", "system", "cloud computing", "architecture", "computing resources", etc.

[0065] Optionally, in this embodiment, the original term refers to a term or vocabulary that is unprocessed or unstandardized in a specific field. These terms are usually used by experts or practitioners in that field and may have certain characteristics such as complexity, verbosity, ambiguity, or non-uniformity. In different applications or contexts, the original terms may vary, but the common feature is that they have not undergone systematic processing or standardization.

[0066] Optionally, in this embodiment, term splitting processing can be, but is not limited to, understood as a technique or method for processing original terms, the purpose of which is to split complex, verbose, or ambiguous original terms into smaller, more specific, and more easily processable term components. This processing is usually based on specific rules, algorithms, or natural language processing techniques, aiming to improve the clarity, accuracy, and operability of terms.

[0067] Among them, the specific steps and methods of term splitting processing may vary depending on the application field. In some cases, the splitting may be based on the boundaries of vocabulary, such as splitting a complex disease name in the medical field into parts such as "etiology", "symptoms", and "treatment methods". In other cases, the splitting may be based on semantics or concepts, such as splitting "cloud computing architecture" in the field of computer science into "cloud computing", "architecture", "computing resources", etc.

[0068] Optionally, during the process of term processing, after term splitting processing, this embodiment obtains a first number of first term components. In order to further simplify and refine these term components, this embodiment can merge the term components that express the same or similar meanings. Through the merging process, this embodiment can obtain a smaller number of second term components.

[0069] Taking a further example, optionally, for example, in the financial field, assume that after splitting processing, this embodiment obtains term components such as "stock split", "share spin-off", "capital restructuring", etc. In certain situations, "stock split" and "share spin-off" may have similar meanings. By merging these two term components, this embodiment can obtain a new and more refined term component, such as "share split / spin-off". And term components like "capital restructuring" remain unchanged.

[0070] Optionally, in this embodiment, the expressed meaning can be, but is not limited to, referring to the specific meaning or concept conveyed by a term or keyword. In term processing, each term component is usually associated with one or more specific meanings or concepts. These meanings may involve the professional knowledge, concepts, operations, objects, or other related aspects in that field.

[0071] Regarding the judgment of whether the expressed meanings are the same, considering that the meaning of a term is often influenced by its context. Therefore, analyzing the context, relevant texts, or other information sources where the term appears is the key to determining its meaning. Moreover, understanding the basic knowledge and professional terms in a specific field is crucial for determining the accurate meaning of a term. This usually requires the participation of domain experts or experienced practitioners. In addition, different terms may be semantically similar or related, even if they are literally different. Using natural language processing techniques, such as word embeddings or semantic networks, can help quantify the semantic similarity between terms. Further, through feedback and interaction with actual users, the understanding of the meaning of terms can be verified and adjusted. This can include questionnaires, expert reviews, or other forms of user testing.

[0072] Optionally, for how to merge the term components with the same expressed meaning, in this embodiment, the core concept or theme of each term component can be determined first, which can be accomplished by analyzing their definitions, usages, and contexts. Further compare these core concepts to determine which term components have semantic overlap or similarity. Natural language processing techniques can be used to assist, such as calculating the semantic similarity scores between term components. For term components with semantic similarity, decide on the merging method, which may include selecting a more widely accepted or more representative term, or creating a new term to cover multiple similar concepts. Further, after the merger, through testing and feedback with domain experts or users, verify whether the new merged term is accurate, clear, and widely accepted. Once the merger result is verified, update the relevant term library and documents, and ensure that the new merged term is widely disseminated and accepted in the relevant community. Thus, this embodiment can not only ensure the accuracy and consistency of terms but also promote the effective dissemination and sharing of knowledge in a specific field.

[0073] Optionally, in this embodiment, after the first term operation network for the original terms and the second term operation network for the standard terms in a specific field have been constructed, this embodiment can recall the output nodes in the second term operation network that meet the recall conditions corresponding to the nodes in the first term operation network. Simply put, it is to find the nodes in the standard term network that match or meet certain conditions with the nodes in the original term network.

[0074] Optionally, in this embodiment, the first term operation network is a term operation network constructed for at least two original terms. This network is mainly used to represent and process concepts, relationships, and operations related to the original terms. The first term operation network can, but is not limited to, consist of a series of nodes and edges connecting these nodes. Each node represents a specific term component, and the edges represent the relationships and operation rules between these term components. The construction of the first term operation network takes into account the semantics, context, and possible relationships between terms, aiming to provide a structured representation for subsequent processing and analysis.

[0075] Optionally, in this embodiment, the second term operation network is a term operation network constructed for standard terms in a specific field. Similar to the first term operation network, the second term operation network is also composed of nodes and edges, but the nodes therein represent standard terms in a specific field. These standard terms are usually widely accepted and used in the field and have clear and unified definitions. The construction of the second term operation network relies on domain knowledge, expert opinions, and authoritative resources to ensure accuracy and consistency.

[0076] Optionally, in this embodiment, the output node refers to the node in the second term operation network that meets the corresponding recall condition of the node in the first term operation network. In other words, the output nodes are those nodes that match or meet specific conditions with some nodes in the first term operation network. These nodes usually represent standard terms related to or similar to the original terms and can therefore be used as candidate standard terms for subsequent confirmation or use.

[0077] Optionally, in this embodiment, the recall condition can, but is not limited to, refer to the criteria or rules based on which output nodes are recalled from the second term operation network, used to screen and determine which nodes meet specific requirements to ensure the accuracy and relevance of the recall.

[0078] For further illustration, optionally, the recall condition can include the semantic similarity between terms, that is, it is required that the output nodes are similar or close in meaning to the nodes in the first term operation network. This can be measured by calculating the cosine similarity between the vector representations of the terms, using pre-trained word embedding models, etc.;

[0079] Or, the recall condition can also consider the matching degree of the terms in the context, that is, it is required that the output nodes match the nodes in the first term operation network in the context or context in which they are located. This can be achieved by analyzing the sentences, paragraphs, or documents in which the terms are located, extracting relevant context information, and comparing the similarity between the contexts;

[0080] Alternatively, the recall condition can also be set according to the frequency of occurrence of terms in a specific domain, that is, it is required that the standard terms corresponding to the output nodes have a relatively high usage frequency in the domain. This can be achieved by statistically analyzing the literature, corpus, or other resources in the domain to determine the frequency distribution of terms and setting corresponding thresholds as recall conditions;

[0081] Alternatively, the recall condition can also consider the relevance of terms to a specific domain, that is, it is required that the standard terms corresponding to the output nodes are associated with the core concepts, themes, or entities of the domain. This can be determined by analyzing the knowledge graph, ontology of the domain, or the opinions of domain experts to identify relevant domain terms and using them as recall conditions.

[0082] It should be noted that in this embodiment, by using term splitting and merging processing, the number of terms to be processed is reduced, thereby reducing the computational workload of using the term operation network for recall. This processing method improves the recall efficiency of standard terms, enabling a faster response to user requests and providing relevant standard terms. This is of great significance for the accurate understanding and sharing of domain knowledge, and also helps to improve the recall performance of standard terms and the user experience.

[0083] For further illustration, optionally, for example Figure 3 As shown, in response to a standard term recall request triggered by at least two original terms (such as original term 302 and original term 304) in a specific domain, the original terms 302 and 304 are subjected to term splitting processing to obtain a first number of first term components, where the first term components are the term components in the original terms 302 and 304, and the standard term recall request is used to request the recall of output standard terms that match the original terms 302 and 304; among the first number of first term components, at least two first term components with the same expressed meaning are merged to obtain a second number of second term components, where the second number is less than the first number, and the second term components are the first term components or the term components obtained after merging at least two first term components;

[0084] Further, in the case of obtaining the first term operation network 306 and the second term operation network 308, at least one output node that meets the recall condition corresponding to the nodes of the first term operation network 306 is recalled from the nodes of the second term operation network 308. The first term operation network 306 is a term operation network constructed for the original terms 302 and 304, and the second term operation network 308 is a term operation network constructed for the standard terms in a specific domain. The nodes of the first term operation network 306 correspond to the term components in the second quantity of second term components, and the nodes of the second term operation network 308 correspond to the term components in the standard terms. The term components corresponding to the at least one output node are determined as at least one output standard term, such as output standard term 310 and output standard term 312.

[0085] Through the embodiments provided in this application, in response to a standard term recall request triggered by at least two original terms in a specific domain, the at least two original terms are subjected to term splitting processing to obtain a first quantity of first term components, where the first term components are the term components in the original terms, and the standard term recall request is used to request the recall of at least one output standard term that matches the at least two original terms. At least two first term components with the same expressed meaning in the first quantity of first term components are merged to obtain a second quantity of second term components, where the second quantity is less than the first quantity, and the second term components are either the first term components or the term components obtained after merging at least two first term components. A first term operation network is constructed for the at least two original terms, and a second term operation network is constructed for the standard terms in a specific domain. At least one output node that meets the recall condition corresponding to the nodes of the first term operation network is recalled from the nodes of the second term operation network. The first term operation network is a term operation network constructed for the at least two original terms, and the second term operation network is a term operation network constructed for the standard terms in a specific domain. The nodes of the first term operation network correspond to the term components in the second quantity of second term components, and the nodes of the second term operation network correspond to the term components in the standard terms. The term components corresponding to the at least one output node are determined as at least one output standard term. By splitting the original terms into smaller term components and further reducing the number of terms by merging term components with the same meaning, it is convenient to subsequently recall the standard terms related to the input terms by using the term operation network, thereby achieving the purpose of reducing the computational amount of recall using the term operation network and thus realizing the technical effect of improving the recall efficiency of the standard terms.

[0086] As an alternative, before combining at least two first term components with the same expressed meaning among the first quantity of first term components to obtain the second quantity of second term components, the method further includes at least one of the following:

[0087] S1-1, converting the first quantity of first term components into semantic vectors; measuring the vector similarity between the semantic vectors, and determining multiple term components with a vector similarity greater than or equal to the first similarity threshold among the first quantity of first term components as at least two first term components, where the vector similarity has a positive correlation with the semantic similarity between the first quantity of first term components;

[0088] S1-2, deeply analyzing the context associated with the first quantity of first term components and extracting context information in the context that has a close relationship with the first quantity of first term components; deeply fusing the context information with the first quantity of first term components and capturing the information expression meaning after the deep fusion; determining the information expression meaning as the expressed meaning corresponding to each first term component among the first quantity of first term components;

[0089] S1-3, measuring the information similarity between the first quantity of first term components and each corpus in the preset corpus through semantic vectors; using the information similarity to determine at least two first term components from the first quantity of first term components.

[0090] Optionally, in this embodiment, to ensure the accuracy of the combination, in addition to the judgment based on vector similarity, it is also possible to consider obtaining the context information of the term components and using this information to determine the specific meaning of the term components. For example, if "AI technology" appears in the context of describing machine learning applications, and "artificial intelligence" appears in the context of describing intelligent robots, then even if their vector similarity is high, they may not be combined because they may have subtle differences in specific applications. In addition, it is also possible to further verify and determine which term components should be combined by comparing the information similarity with the corpora in the preset corpus.

[0091] It should be noted that before combining term components with the same expressed meaning, this embodiment may also include some preprocessing steps to ensure the accuracy and effectiveness of the combination. Such as calculating the similarity between term components, obtaining the context information of term components, and calculating the information similarity between term components and the corpora in the preset corpus, etc.

[0092] For further illustration, there are, for example, three first term components: "AI technology", "artificial intelligence", and "machine intelligence". By calculating the vector similarity between them, it is found that the similarity between "AI technology" and "artificial intelligence" exceeds the first similarity threshold. Therefore, "AI technology" and "artificial intelligence" are regarded as term components with the same expressed meaning and can be merged.

[0093] Through the embodiments provided by the present application, convert the first quantity of first term components into semantic vectors; measure the vector similarity between the semantic vectors, and determine, among the first quantity of first term components, multiple term components whose vector similarity is greater than or equal to the first similarity threshold as at least two first term components, where the vector similarity has a positive correlation with the semantic similarity between the first quantity of first term components; deeply analyze the context associated with the first quantity of first term components, and extract context information in the context that has a close relationship with the first quantity of first term components; deeply fuse the context information with the first quantity of first term components, and capture the information expression after the deep fusion; determine the information expression as the expressed meaning corresponding to each first term component among the first quantity of first term components; measure the information similarity between the first quantity of first term components and each corpus in the preset corpus through the semantic vectors; use the information similarity to determine at least two first term components from the first quantity of first term components, thereby achieving the purpose of being able to more accurately identify and merge term components with the same expressed meaning, so as to obtain more concise and accurate second term components, and thus realizing the technical effect of improving the accuracy and effectiveness of term component merging.

[0094] As an alternative solution, before measuring the information similarity between the first quantity of first term components and each corpus in the preset corpus through the semantic vectors, the method further includes:

[0095] S2-1, obtain a first component and a second component belonging to a specific business scenario from the first quantity of first term components;

[0096] S2-2, obtain a corpus matching the specific business scenario, where the preset corpus includes the corpus matching the specific business scenario;

[0097] As an alternative solution, using the information similarity to determine at least two first term components from the first quantity of first term components includes:

[0098] S3-1, obtain the first similarity between the first component and each corpus in the corpus matching the specific business scenario;

[0099] S3-2. When the first similarity between each piece of corpus in the corpus matching the first component and the specific business scenario is obtained, and any first similarity is greater than or equal to the second preset threshold, the first piece of corpus with the highest first similarity in the corpus matching the specific business scenario is used as the expression meaning corresponding to the first component, where the information similarity includes the first similarity;

[0100] S3-3. Obtain the second similarity between the second component and each piece of corpus in the corpus matching the specific business scenario;

[0101] S3-4. When the second similarity between the second component and each piece of corpus in the corpus matching the specific business scenario is obtained, and any second similarity is greater than or equal to the second preset threshold, the second piece of corpus with the highest second similarity in the corpus matching the specific business scenario is used as the expression meaning corresponding to the second component, where the information similarity includes the second similarity;

[0102] S3-5. Obtain the corpus similarity between the first piece of corpus and the second piece of corpus;

[0103] S3-6. When the corpus similarity is greater than or equal to the third similarity threshold, determine that the first component and the second component have the same expression meaning.

[0104] It should be noted that before obtaining the information similarity between each first term component and each piece of corpus in the preset corpus, in this embodiment, first, for a specific business scenario, two components are selected from the first quantity of first term components, namely the first component and the second component. Then, a corpus matching this specific business scenario is obtained, and the information similarity between these two components and the corpus in the corpus is calculated. Based on these information similarities, the method determines whether these two components have the same expression meaning.

[0105] Optionally, in this embodiment, the specific business scenario may but is not limited to refer to a specific situation or scenario in a certain specific business or application field. These business scenarios usually have specific industry characteristics, professional terms, and context backgrounds, and need to be specifically processed and analyzed. In the method described in the text, the specific business scenario is an important factor for term meaning determination and standard term recall.

[0106] In a specific business scenario, the meaning of a term may change or have a specific interpretation. For example, in different industries, the term "shopping cart" may have different meanings. In the e-commerce field, it usually refers to a virtual tool for users to temporarily store goods to be purchased; while in the physical store retail field, it may refer to a physical shopping cart. Therefore, in order to accurately understand and use terms, it is necessary to consider the specific business scenario in which they are located.

[0107] By optimizing and processing for specific business scenarios, this embodiment can more accurately determine the expression meaning of terms in such scenarios and identify which term components have the same expression meaning. This helps to more precisely match and recall standard terms related to the input terms during the subsequent standard term recall process. Therefore, specific business scenarios are an important consideration in this method, which helps to improve the accuracy and practicality of term meaning determination.

[0108] For further illustration, optionally, for example, assume that this embodiment has a specific business scenario in the e-commerce field, where the first component is "shopping cart" and the second component is "shopping basket". By calculating the information similarity of the corpora in the corpus matching the e-commerce field, it is found that "shopping cart" and "shopping basket" have similar expression meanings in this scenario, that is, both "shopping cart" and "shopping basket" refer to tools used by users to temporarily store goods to be purchased.

[0109] Through the embodiment provided by this application, from the first quantity of first term components, obtain the first component and the second component belonging to a specific business scenario; obtain the corpus matching the specific business scenario, where the preset corpus includes the corpus matching the specific business scenario; obtain the first similarity between the first component and each corpus in the corpus matching the specific business scenario; in the case where any first similarity is greater than or equal to the second preset threshold, use the first corpus with the highest first similarity in the corpus matching the specific business scenario as the expression meaning corresponding to the first component, where the information similarity includes the first similarity; obtain the second similarity between the second component and each corpus in the corpus matching the specific business scenario; in the case where any second similarity is greater than or equal to the second preset threshold, use the second corpus with the highest second similarity in the corpus matching the specific business scenario as the expression meaning corresponding to the second component, where the information similarity includes the second similarity; obtain the corpus similarity between the first corpus and the second corpus; in the case where the corpus similarity is greater than or equal to the third similarity threshold, determine that the first component and the second component have the same expression meaning, thereby achieving the purpose of more accurately determining the expression meaning of term components in a specific business scenario and identifying which term components have the same expression meaning, and thus realizing the technical effect of improving the recall accuracy of standard terms.

[0110] As an optional solution, recalling at least one output node that meets the recall condition corresponding to the node of the first term operation network from the nodes of the second term operation network includes:

[0111] S4-1. Obtain the sub-structures at the first level between the first term operation network and the second term operation network. Among them, the sub-structures at the first level include the first sub-structure of the first term operation network and the second sub-structure of the second term operation network. The nodes of the first sub-structure correspond to the term components at the first level among the second quantity of second term components, and the nodes of the second sub-structure correspond to the term components at the first level in the standard terms.

[0112] S4-2. Determine at least one first node among the nodes of the second sub-structure that satisfies the first condition for recalling the nodes of the first sub-structure. Among them, the recall condition includes the first condition. The similarity between the term component corresponding to the first node and the term component corresponding to the node of the first sub-structure is greater than or equal to the threshold value matched by the first condition.

[0113] S4-3. Obtain the sub-structures at the second level between the first term operation network and the second term operation network. Among them, the sub-structures at the second level include the third sub-structure of the first term operation network and the fourth sub-structure of the second term operation network. The nodes of the third sub-structure correspond to the term components at the second level among the second quantity of second term components, and the nodes of the second sub-structure correspond to the term components at the second level in the standard terms. The second level is the upper level corresponding to the first level.

[0114] S4-4. Determine at least one second node among the nodes of the fourth sub-structure that has an indexing relationship with at least one first node and satisfies the second condition for recalling the nodes of the third sub-structure. Among them, the recall condition includes the second condition. Having an indexing relationship means that there is a superordinate-subordinate relationship between nodes belonging to different levels. The similarity between the term component corresponding to the second node and the term component corresponding to the node of the third sub-structure is greater than or equal to the threshold value matched by the second condition.

[0115] S4-5. In the case where the number of sub-structures at the upper level corresponding to the second level is less than or equal to the preset threshold, determine at least one second node as at least one output node.

[0116] Optionally, in this embodiment, the first and second levels in the term operation network are only for illustration and do not limit the quantity. For example, the term operation network may include more than two levels of sub-structures, among which at least include the sub-structures at the first level and the sub-structures at the topmost level.

[0117] Optionally, in this embodiment, the index relationship can be, but is not limited to, understood as a hierarchical relationship existing between nodes at different levels. This hierarchical relationship can be understood as the connection between a more general or abstract concept (upper node) and a more specific or specialized concept (lower node). In a knowledge representation and terminology operation network, the index relationship is usually used to organize and manage the hierarchical structure between terms. By establishing an index relationship between the upper node and the lower node, the semantic association and hierarchical structure between terms can be expressed.

[0118] For further illustration, optionally, if a node is the upper node of another node, then it usually represents a broader and more general concept, while the lower node represents a specific instance or subcategory of this concept. For example, in the biological classification system, "animal" is an upper node, while "dog", "cat", etc. are the lower nodes of "animal". In this example, there is an index relationship between "animal" and "dog", "cat", expressing their hierarchical relationship.

[0119] It should be noted that to improve the recall accuracy of standard terms, this embodiment designs a method for recall between two sub-structures at different levels. This process aims to ensure that the nodes recalled from the second terminology operation network can meet the specific recall conditions of the nodes in the first terminology operation network, so as to achieve a more precise match in terms of semantics and structure.

[0120] First of all, this embodiment focuses on the sub-structure of the first level. This level usually contains more specific and refined term components, which are crucial for accurately expressing concepts in a specific field. By performing recall in the first level, this embodiment can ensure that the recalled nodes are semantically closely related to the target node. To achieve this goal, this embodiment determines the nodes that meet the recall conditions according to specific first conditions. These conditions may involve factors such as the similarity between nodes and the relevance between term components, to ensure that the recalled nodes are highly consistent with the target node in terms of semantics.

[0121] However, performing recall only at the first level may not be sufficient to cover a wider range of concepts and contexts. Therefore, this embodiment further considers the upper level corresponding to the first level, such as the second level (this is only an example here and does not limit the number of levels). The second level usually contains more abstract and general term components, which can capture broader concepts and semantic relationships. By performing recall at the second level, this embodiment can further expand the term network related to the target node, improving the coverage and accuracy of recall. To achieve this goal, this embodiment determines the nodes that meet the recall conditions according to specific second conditions, and these conditions may involve factors such as the index relationship between nodes and the similarity threshold, to ensure that the recalled nodes are reasonably related to the target node in terms of structure.

[0122] For further illustration, optionally based on Figure 3 the scenario shown, continue for example Figure 4 as shown, the sub-structures belonging to level 1, level 2, and level 3 between the first term operation network 306 and the second term operation network 308, where the sub-structure belonging to level 1 is the lowest-level sub-structure, which can be understood as the sub-structure of the first level, and the sub-structure belonging to level 4 is the highest-level sub-structure, or rather, the number of sub-structures of the upper level corresponding to level 4 is less than or equal to 0 (preset threshold).

[0123] By performing recall between at least two different-level sub-structures and screening the nodes that meet the recall conditions according to specific conditions at each level, this embodiment can improve the recall accuracy of standard terms. And since this embodiment comprehensively considers both semantic and structural factors, it ensures that the recalled nodes are not only semantically related to the target node but also have a reasonable hyponymy relationship in terms of structure. This comprehensive recall method helps to improve the performance and effectiveness of the standard term system, enabling users to understand and use terms more accurately.

[0124] In addition, in this embodiment, since the number of original terms is large, and the increase in the number of original terms means that more term information needs to be processed and compared, which increases the computational complexity and time cost in the recall process. Each term needs to be compared and matched with a large number of other terms to determine whether it meets the recall conditions, which makes the recall process more complex and time-consuming.

[0125] Secondly, in order to improve the recall accuracy of standard terms, the embodiment adopts a multi-level term operation network structure. Although this structure can better organize and express the hierarchical relationship between terms, it also increases the complexity of recall. Recalling between different levels requires considering factors such as the hierarchical relationship and index relationship between nodes, which further increases the complexity and difficulty of calculation. As the recall complexity of standard terms increases, the recall efficiency will inevitably decline. The increase in complexity means that more computing resources and processing time are required to complete the recall task, which may lead to a decrease in real-time and response speed.

[0126] In this regard, this embodiment improves the recall efficiency by removing duplicate term components (merging at least two first term components with the same meaning in the first number of first term components) and introducing an efficient index structure, while maintaining a high recall accuracy. It can be seen that this embodiment is not simply to improve the recall efficiency of standard terms, or to improve the recall accuracy of standard terms, but to achieve a technical effect that takes into account both the recall efficiency and accuracy of standard terms.

[0127] Through the embodiments provided in this application, sub-structures belonging to the first level between the first term operation network and the second term operation network are obtained. Among them, the sub-structures belonging to the first level include the first sub-structure of the first term operation network and the second sub-structure of the second term operation network. The nodes of the first sub-structure correspond to the term components belonging to the first level among the second quantity of second term components, and the nodes of the second sub-structure correspond to the term components belonging to the first level in the standard terms; at least one first node that satisfies the first condition for recalling the nodes of the first sub-structure is determined among the nodes of the second sub-structure. Among them, the recall condition includes the first condition, and the similarity between the term components corresponding to the first node and the term components corresponding to the nodes of the first sub-structure is greater than or equal to the threshold value matched by the first condition; sub-structures belonging to the second level between the first term operation network and the second term operation network are obtained. Among them, the sub-structures belonging to the second level include the third sub-structure of the first term operation network and the fourth sub-structure of the second term operation network. The nodes of the third sub-structure correspond to the term components belonging to the second level among the second quantity of second term components, and the nodes of the second sub-structure correspond to the term components belonging to the second level in the standard terms. The second level is the upper level corresponding to the first level; at least one second node that has an index relationship with at least one first node and satisfies the second condition for recalling the nodes of the third sub-structure is determined. Among them, the recall condition includes the second condition, and having an index relationship means that there is a hierarchical relationship between nodes belonging to different levels. The similarity between the term components corresponding to the second node and the term components corresponding to the nodes of the third sub-structure is greater than or equal to the threshold value matched by the second condition; in the case where the number of sub-structures at the upper level corresponding to the second level is less than or equal to the preset threshold value, at least one second node is determined as at least one output node, thereby achieving the purpose of improving the recall efficiency by methods such as term component deduplication and introducing an efficient index structure, while maintaining a high recall accuracy, thus realizing the technical effect of taking into account both the recall efficiency and accuracy of standard terms.

[0128] As an alternative solution, merging at least two first term components with the same expressed meaning among the first quantity of first term components to obtain the second quantity of second term components includes:

[0129] S5-1, merging the first term components with the same expressed meaning and belonging to the same first level among the first quantity of first term components to obtain a plurality of first-level components, and using the plurality of first-level components as the term components corresponding to the nodes of the first sub-structure. Among them, the second quantity of second term components includes a plurality of first-level components;

[0130] S5-2, in the first quantity of first-term components, merge the term components that have the same meaning and belong to the same second level to obtain multiple second-level components, and use the multiple second-level components as the term components corresponding to the nodes of the second sub-structure, where the second quantity of second-term components includes multiple second-level components.

[0131] It should be noted that in this embodiment, the term components with the same meaning are merged to reduce redundancy and standardize the terms. This process is carried out separately at two different levels (the first level and the second level). The purpose of the above merging is to reduce redundancy and improve the consistency of the terms. When this embodiment processes a large number of terms, it is easy to have terms with the same meaning but different expressions, which will increase the complexity of understanding and processing. By merging these terms, this embodiment can standardize the terms, thereby improving the efficiency and accuracy of subsequent processing.

[0132] For further illustration by example, optionally assume that this embodiment has the following first-term components: "automobile", "vehicle", "sedan", and "jeep" have similar meanings expressed in the first level and can be regarded as terms describing similar entities; while in the second level, "transportation vehicle" is a more abstract concept that encompasses those specific terms in the first level. After merging, in the first level, this embodiment may only retain "automobile" as a representative; in the second level, this embodiment may only retain "transportation vehicle".

[0133] Through the embodiment provided by this application, in the first quantity of first-term components, the term components that have the same meaning and belong to the same first level are merged to obtain multiple first-level components, and the multiple first-level components are used as the term components corresponding to the nodes of the first sub-structure, where the second quantity of second-term components includes multiple first-level components; in the first quantity of first-term components, the term components that have the same meaning and belong to the same second level are merged to obtain multiple second-level components, and the multiple second-level components are used as the term components corresponding to the nodes of the second sub-structure, where the second quantity of second-term components includes multiple second-level components, thereby achieving the purpose of reducing redundancy and improving the consistency of the terms, and thus realizing the technical effect of improving the recall efficiency of standard terms.

[0134] As an optional solution, before obtaining the sub-structure belonging to the first level between the first-term operation network and the second-term operation network, the method further includes:

[0135] Classify the second quantity of second-term components according to the component type. Among them, the sub-structure at the highest level in the first-term operation network corresponds to at least two original terms. Except for the sub-structures at the highest level in the first-term operation network, the nodes of sub-structures at other different levels correspond to second-term components of different component types.

[0136] Optionally, in this embodiment, the component type can be, but is not limited to, the classification or attribute of terms or term components. When there are a large number of terms in this embodiment, in order to better organize, understand, and manage them, it is usually necessary to classify these terms. The basis for classification can be the meaning, usage, context, or other relevant characteristics of the terms. Each classification can be regarded as a "component type". For example, in biology, "animals", "plants", and "microorganisms" can be three different component types; in the food field, "fruits", "vegetables", and "meats" can also be regarded as different component types. Classifying term components and grouping them according to their component types can make a large number of terms more organized and easier to manage, and also helps to process and retrieve these terms more efficiently.

[0137] Optionally, in this embodiment, the highest level can be, but is not limited to, the highest or starting level in a structure or system, and usually represents the most abstract, general, or fundamental concept. In the first-term operation network, the sub-structure at the highest level represents the most generalized terms or concepts. These terms or concepts are usually relatively broad and can cover multiple specific sub-concepts or sub-terms. For example, in a classification system for animals, the concept of "animal" may be the highest level, while specific animal species such as "cats" and "dogs" belong to lower-level sub-concepts.

[0138] It should be noted that in order to make the term components more organized and facilitate subsequent processing and operations, this embodiment classifies the second quantity of second-term components according to the component type. By classifying the second-term components, this embodiment can gather the same or similar types of term components together, which helps to more effectively process and compare these components in subsequent operations. At the same time, by allocating different types of components to different levels and sub-structures, this embodiment can construct a clearer and more organized term operation network, which helps to improve the recall accuracy and efficiency of terms.

[0139] For further illustration, optionally assume that this embodiment has the following second term components: "apple", "banana", "orange" belong to the fruit category, and "car", "motorcycle", "bicycle" belong to the transportation category. After classification processing, this embodiment can classify "apple", "banana", "orange" as fruit category components, and classify "car", "motorcycle", "bicycle" as transportation category components. In the first term operation network, the sub-structures at the highest level may correspond to the two original terms "food" and "transportation", while the nodes of the sub-structures at other different levels correspond to the second term components of different component types, such as fruits, transportation, etc.

[0140] Through the embodiment provided by the present application, the second quantity of second term components are classified according to the component type, wherein the sub-structure at the highest level in the first term operation network corresponds to at least two original terms, and except for the sub-structure at the highest level in the first term operation network, the nodes of the sub-structures at other different levels correspond to the second term components of different component types, thereby achieving the purpose of constructing a clearer and more organized term operation network, and thus realizing the technical effect of improving the recall accuracy and efficiency of terms.

[0141] As an optional solution, at least two first term components with the same expressed meaning in the first quantity of first term components are merged to obtain the second quantity of second term components, including:

[0142] When the first quantity is greater than or equal to the excess threshold, at least two first term components are merged to obtain the second quantity of second term components.

[0143] It should be noted that when the first quantity of first term components exceeds or reaches a preset excess threshold, this embodiment will only merge at least two term components with the same expressed meaning among them to obtain the second quantity of second term components. This conditional merging strategy helps to ensure that the number of term components is not too large while retaining sufficient details and accuracy. Only when the number of term components reaches or exceeds a certain threshold will the merging mechanism be activated, which can avoid information loss caused by unnecessary merging operations.

[0144] For further illustration, optionally assume that this embodiment has 100 first term components, and the preset excess threshold is 80. Since 100 is greater than 80, this embodiment needs to perform a merging operation. Among them, the two term components "computer science" and "computer science" have the same expressed meaning, so they can be merged into one term component "computer science / computer science". After such a merging operation, this embodiment obtains the second quantity of second term components.

[0145] Through the embodiments provided in this application, when the first quantity is greater than or equal to the excess threshold, at least two first term components are merged to obtain a second quantity of second term components, thereby achieving the purpose of ensuring that the quantity of term components is not overly large while retaining sufficient details and accuracy, and thus realizing the technical effect of avoiding information loss that may be caused by unnecessary merging operations.

[0146] As an alternative solution, when the first quantity is greater than or equal to the excess threshold, merging at least two first term components to obtain a second quantity of second term components includes:

[0147] S6-1, when the first quantity is greater than or equal to the excess threshold and less than the massive threshold, merging at least two first term components to obtain a second quantity of second term components within the first interval, where the massive threshold is greater than the excess threshold; or,

[0148] S6-2, when the first quantity is greater than or equal to the excess threshold and greater than or equal to the massive threshold, merging at least two first term components to obtain a second quantity of second term components within the second interval, where the second interval quantity is greater than the first interval quantity.

[0149] It should be noted that in this embodiment, when the first quantity is between the excess threshold and the massive threshold, and when the first quantity exceeds or is equal to the massive threshold, the quantities of the second term components obtained after merging belong to different intervals. Compared with a relatively single merging strategy, this embodiment allows for more flexible management of the quantity of term components. According to different first quantity intervals, this embodiment can control the granularity of merging, thereby avoiding an overly large quantity of term components while retaining sufficient details.

[0150] For further illustration by example, it is optionally assumed that the excess threshold is 1000 and the massive threshold is 5000. When this embodiment has 2000 first term components (between the excess threshold and the massive threshold), this embodiment will merge to obtain the quantity of second term components belonging to the first interval; while when this embodiment has 10000 first term components (exceeding the massive threshold), this embodiment will merge to obtain the quantity of second term components belonging to the second interval, and this quantity will be larger than that of the first interval.

[0151] Through the embodiments provided in this application, when the first quantity is greater than or equal to the excess threshold and less than the massive threshold, at least two first term components are merged to obtain second term components with a first interval quantity, where the massive threshold is greater than the excess threshold; or, when the first quantity is greater than or equal to the excess threshold and greater than or equal to the massive threshold, at least two first term components are merged to obtain second term components with a second interval quantity, where the second interval quantity is greater than the first interval quantity. Thus, while retaining sufficient details, the purpose of avoiding an overly large number of term components is achieved, enabling more flexible management of the quantity of term components, thereby realizing the technical effect of avoiding information loss that may be caused by unnecessary merging operations.

[0152] As an alternative solution, for ease of understanding, the above term processing method is applied to the medical field. In actual clinical practice, doctors / nurses / pharmacists may have different interpretations of the same medical entity (such as a disease). The technology of unifying these different expressions into a standardized expression (such as the ICD-10 medical standard system) is called term standardization technology.

[0153] In actual business scenarios, term standardization technology often needs to handle batches of original words in the order of millions / tens of millions. If normalized one by one, even with a processing efficiency of 30 QPS, it will take a long time to process. To address this challenge, this embodiment adopts a multi-level recall framework based on a term operation tree for batch processing acceleration, enabling parallel processing on the original word side.

[0154] Specifically, the (single) term operation tree structure contains 4 levels: diagnostic words / equivalent trees / subtrees / leaf nodes. Each level will achieve a huge reduction in quantity (for information compression) after normalization and deduplication. Therefore, standard words can be recalled at different levels through a multi-level index method.

[0155] This embodiment can achieve a processing efficiency faster than parallel acceleration in a task of batch processing 300,000 original words, further compressing the processing time to 40%.

[0156] It should be noted that with the rapid development of medical informatization, hospitals and medical institutions have accumulated a large amount of electronic data. However, this data is mainly stored in free text form and has characteristics such as colloquialism and diversity due to different personal writing habits and expression forms of doctors. This results in multiple different expression forms for the same medical concept, posing obstacles to medical data retrieval, analysis, and utilization. Therefore, diagnostic standardization plays a very important role in promoting the construction of medical informatization.

[0157] Optionally, in this embodiment, it is shown as Figure 5The user interface of the shown diagnostic standardization engine. This interface shows the important practical significance of diagnostic standardization in reducing the workload of medical record coders, improving work efficiency, and helping to connect data from all parties and provide a unified labeled diagnostic data interface. Through diagnostic standardization, in this embodiment, data from hospitals of multiple different levels and regions can be standardized and connected, enabling data to flow between multiple medical institutions, thereby promoting the analysis and utilization of medical data and accelerating the pace of medical informatization construction.

[0158] However, in the related technical solutions, there is a problem of long processing time for single-item normalization processing. Parallel normalization processing can improve processing efficiency, but its performance ceiling is limited by the configuration of the processing server. In some business scenarios, if the machine configuration is limited, the theoretical performance ceiling of parallel normalization processing is also relatively low. When the number of parallel threads is set to exceed the number of CPU cores, the CPU needs to frequently switch processes and process memory, which instead reduces the processing efficiency. Therefore, in practical applications, it is necessary to balance the relationship between processing efficiency and server configuration.

[0159] Optionally, this embodiment provides a solution based on diagnostic standardization to address the problems of data diversity and inconsistent expression forms in medical informatization construction. Through the user interface of the diagnostic standardization engine, the standardized processing and management of medical data can be achieved, improving work efficiency and data quality. However, the optimization of the normalization process still needs to consider actual limiting factors such as server configuration.

[0160] Furthermore, to simultaneously solve the problems of running efficiency and dependence on machine configuration, this embodiment proposes a multi-level recall framework based on a term operation tree. Specifically, this embodiment constructs a batch tree structure for the original terms based on the term operation tree structure (a structure that details various main components in diagnostic terms and their dependencies), and uses different levels in the term operation tree structure and different compression ratios of information at each level to achieve a unified description of the batch of original terms.

[0161] Secondly, this embodiment also applies the tree nodes at different levels in batches through multi-level operation tree recall to achieve the recall of candidate standard terms for the batch of original terms simultaneously.

[0162] Optionally, in this embodiment, for the construction of the above-mentioned term operation tree structure, the construction of the term operation tree is to identify the main components of terms in a sentence and the combination relationships between the components. It is constructed by combining a large number of medical logic rules and term vocabularies, and has a certain medical rationality in the organizational hierarchical structure. At the same time, this embodiment designs equivalent tree nodes in the operation tree structure to achieve multiple possible construction results for a single diagnostic term (in line with different interpretations of different physicians in clinical practice). Among them, the construction of the above-mentioned term operation tree structure includes multiple steps such as splitting of term main components, component recombination, and construction of the term operation tree.

[0163] For further illustration, optionally for the above-mentioned term principal component splitting, when a diagnostic word (or other medical term) is given, the present embodiment defines the key components in the term operation tree: anatomical component, disease component, operation component, modification component, orientation component, and relationship component.

[0164] Then the principal component splitting is to identify and locate the relationship component of the input original word. Here, the present embodiment uses the dictionary lookup method to identify the three splitting logics of "accompanied by", "and", and "or". The obtained relationship component serves as the root node of the entire term operation tree (the "relationship node" in the above figure), and the relationship component is of great significance for the subsequent expansion of the operation tree.

[0165] Furthermore, other components in the input diagnostic word need to be identified. Here, the present embodiment can also use the dictionary look-up method or the sequence labeling model method. It should be noted that these three components need to be identified and extracted separately and in parallel because there may be nesting or partial overlap. The specific steps are as follows:

[0166] S7-1. For a given diagnostic word M = {m1, m2, m3,..., m n}, where n is the length of the diagnostic word M, perform the following sequential term principal component extraction:

[0167] S7-2. Extract the span of the disease component, {m1, m2, m3,..., m n} --> {m k , m k:1 , m k:2 ,..., m k:l}, where l in the obtained span is the length of the disease root and k is the position of the disease root in the original word;

[0168] S7-3. Extract the span of the anatomical component, {m1, m2, m3,..., m n} --> {m k , m k:1 , m k:2 ,..., m k:l}, where l in the obtained span is the length of the anatomical part and k is the position of the anatomical part in the original word;

[0169] S7-4. Extract the span of the operation component, {m1, m2, m3,..., m n} --> {m k , m k:1 , m k:2 ,..., m k:l}, where in the obtained span, l is the length of the disease modification and k is the position of the disease modification in the original word;

[0170] S7-5, extract the span of the modifying components, {m1, m2, m3,..., m n} --> {m k , m k:1 , m k:2 ,..., m k:l} where in the obtained span, l is the length of the disease modification and k is the position of the disease modification in the original word;

[0171] After the above steps, all the main components of a diagnostic word can be obtained. Note that the main components at this time may intersect with each other, and various different combinations need to be located.

[0172] Optionally, in this embodiment, for component recombination, the component recombination engine only depends on the target length of the medical entity word and the span corresponding to each identified and split main component. For example Figure 6 the main component spans shown in are: facial muscle @ anatomical component (1, 3), tendon @ anatomical component (2, 4), tendinitis @ disease component (2, 5), tendonitis @ disease component (3, 5), inflammation @ disease component (4, 5), and the span of the entire medical entity word is (1, 5). It should be noted that the span in this embodiment includes the left side and does not include the right side.

[0173] In this regard, this embodiment proposes an algorithm called span - combo, which aims to use all candidate spans to obtain multiple combinations through various combination methods, where the span of each combination can "piece together" the span of the target medical entity word. The concept of "piecing together" here allows for overlapping situations between some spans. For example Figure 6 in the first group of results on the right in, (1, 3) and (2, 5) overlap the span (1, 2), corresponding to the text "muscle". Considering that the focus of this embodiment is to split a single medical entity in multiple possible ways, such overlapping situations are allowed in the span - combo algorithm. Here, the overlapping situations can also be filtered according to actual scenario requirements, that is, the span - combo algorithm is not allowed to have such component - to - component overlaps.

[0174] This embodiment implements the span - combo algorithm based on the backtracking algorithm and Depth First Search (DFS). The schematic flowchart of the DFS core algorithm is as Figure 7As shown, given the current relevant data, such as the given candidate list (cands), the starting position marker (begin), the number of candidates in the list (size), the current traversal path (path), the results of all traversal paths (res), and the length of the target entity word (target), first determine whether the result of the length of the target entity word is empty. If so, it means that a valid combination has been successfully found. Then, add the result of the current traversal path to the results of all traversal paths and return the results of all traversal paths. If not, traverse the candidate list starting from the starting position marker. In this process, this embodiment uses the variable of the length of the target entity word to obtain the variable at a certain position in the candidate list and obtain the variable result.

[0175] To further illustrate with an example, in the process of this embodiment obtaining the variable result, the variable of the length of the target entity word can be used to subtract the variable at the index position in the candidate list to obtain the variable result. For example, variable (0, 6) minus variable (1, 5) gives variable (0, 1) and variable (5, 6), or variable (2, 4) minus variable (3, 5) gives variable (2, 3). In this way, through calculation, multiple component candidates of a certain target medical entity can be automatically combined to ensure that there are no duplicate components in each combination and each combination must be able to "piece together" the original medical entity.

[0176] Next, this embodiment will also determine whether the obtained variable result is empty. If the variable result is not empty, add it to the current traversal path. If the variable result is empty, continue to traverse the next candidate.

[0177] Then, continue to traverse other candidates and repeat the above steps until all candidates have been traversed. In this process, this embodiment uses the backtracking algorithm to ensure that there are no duplicate components in each combination and each combination must be able to "piece together" the original medical entity.

[0178] Finally, when this embodiment obtains all valid combinations, these combinations can be used to achieve an efficient and reasonable reorganization of the split component results. In this way, not only can the requirements of medical rationality be met, but also the algorithm (backtracking / DFS) can be fully utilized to achieve an efficient and accurate processing of the split component results. This DFS-based span-combo algorithm implementation is not only efficient but also accurate and reliable, and is applicable to the automatic combination and recognition of components of various medical entities.

[0179] Optionally, this embodiment elaborates in detail on the construction of the (single) term operation tree. The construction of the operation tree is carried out by reasonably constructing the tree structure through the structures of the two steps of principal component splitting and component recombination. As Figure 8As shown, the operation tree structure includes four levels: the diagnosis term level, the equivalent tree level, the relationship / subtree level, and the leaf node level, which are specifically as follows:

[0180] First, the diagnosis term level is the original input diagnosis term. Each operation tree corresponds to only one diagnosis term and has only one diagnosis term node. This is the starting point of the entire operation tree.

[0181] Second, the equivalent tree nodes are different combinations of equivalent components obtained through component recombination in the previous step. This means that the tree structures below different equivalent tree nodes are all equivalent, even if the structures or components therein are inconsistent. This level is mainly used to represent different combinations of equivalent components.

[0182] Next is the relationship node, which represents components such as "accompanied by" / "and" / "or" that formulate diagnostic logic links and can be empty. The relationship node is mainly used to define the relationships between different subtree nodes and is not helpful for recalling standard terms and their core components.

[0183] The subtree node is located between the relationship node and the component leaf node and is a virtual node without actual meaning. Its role is to make the operation tree structure self-consistent and ensure the integrity and correctness of the entire tree structure.

[0184] Finally, the component leaf nodes represent specific different components, including diseases, anatomy, modifiers, etc. These leaf nodes are the basic building blocks of the operation tree and are used to describe specific medical concepts or terms.

[0185] It should be noted that this embodiment emphasizes the improvement of batch processing efficiency for large-scale data. Therefore, only the subtree nodes are considered in the third layer and the relationship nodes are not considered. This is because the relationship nodes mainly serve to define the relationships between different subtree nodes and are not helpful for recalling standard terms and their core components. By optimizing the structure of this layer, the construction efficiency and processing speed of the operation tree can be improved.

[0186] Optionally, the construction of the operation tree in this embodiment includes four hierarchical structures: the diagnosis term level, the equivalent tree level, the relationship / subtree level, and the leaf node level. Each level has its specific meaning and role, and through a reasonable construction method, efficient processing and analysis of medical terms are achieved. This construction method based on the operation tree not only improves the processing efficiency but also provides strong support for subsequent medical natural language processing tasks.

[0187] Optionally, as shown in the following table, in this embodiment, the number of deduplication nodes in different levels is experimented on multiple sets of diagnostic words. Among them, the set of diagnostic words includes the ICD-10 standard system and two sets of business scenario data (with 70,000 diagnostic words and 5,000 diagnostic words respectively). Note that the deduplication equivalent tree / deduplication subtree here refers to the structure obtained by deduplicating the tree structure serialized in a certain order. This part is not the focus of this embodiment, so it will not be elaborated.

[0188]

[0189]

[0190] It can be seen that there are 47,000 diagnostic words in ICD-10, and a total of 104,000 equivalent trees are used to express them, indicating that on average each of these diagnostic words has at least 2 ways of interpretation / variations. In the data of business scenario #1, each diagnostic word also corresponds to about 2 equivalent trees. At the subtree level, the magnitude after deduplication has decreased significantly compared to the equivalent tree level. In ICD-10, the deduplicated subtrees have decreased to 70,000, while in the data of business scenario #1, it has decreased to less than 60,000, which is already less than the number of diagnostic words. This shows that the basic units of the subtrees of a large number of words in the set of diagnostic words are the same, and these redundant subtree equivalent calculations can be avoided to improve efficiency. At the leaf node level, the magnitude after deduplication further decreases. There are only 14,000 deduplicated leaf nodes in ICD-10, while for the two sets of business scenario data, there are only less than 7,000 and 3,000 respectively. This shows that although the expressions of diagnostic words are strange, the core main components are always within a limited range, which lays a foundation for improving calculation efficiency. For example, even though there are 170,000 at the diagnostic word level, the basic component units are only 7,000. If only 7,000 leaf nodes can be compared, up to 96% of the calculation amount can be saved.

[0191] Optionally, in this embodiment, the multi-level recall framework based on the term operation tree is as Figure 9 shown. In this embodiment, the diagnostic words are defined as A1 / 2, B1 / 2, C1 / 2 respectively; the equivalent trees are defined as 1_1 / 2, 2_1 / 2, 3_1 / 2 respectively; the subtrees are defined as A1 / 2, B1 / 2, C1 / 2 respectively; the leaf nodes are defined as a1 / 2, b1 / 2, c1 / 2 respectively, where 1 represents the original word and 2 represents the standard word. Therefore, this embodiment can obtain Figure 9 two multi-level structures (corresponding to the left and right sides of the dotted line in Figure 9 respectively), where the multi-level structure can be understood as a network structure (term operation network) composed of common components of multiple term operation trees. At the same time, this embodiment establishes an index table between different levels, specifically:

[0192] Indicates the leaf node to subtree index table on the original word side,

[0193] Indicates the subtree to equivalent tree index table on the original word side,

[0194] Indicates the equivalent tree to diagnostic word index table on the original word side, Indicates the leaf node to subtree index table on the standard word side,

[0195] Indicates the subtree to equivalent tree index table on the standard word side,

[0196] Indicates the equivalent tree to diagnostic word index table on the standard word side,

[0197] Optionally, in this embodiment, Figure 9 The arrow crossing the middle dotted line represents the actual recall process, which includes the judgment of whether the leaf node names are the same, the judgment of whether there is an intersection in the synonym set, and the judgment of whether there is a hyponymy relationship between the leaf nodes. If any of these three types of judgments is satisfied, these two leaf nodes meet the recall conditions.

[0198] Further, taking the recall of b2 by a1 and the recall of a2 by b1 as an example to illustrate the running logic of this recall framework:

[0199] S8-1, judge and Whether the pairwise recall conditions of all index keys in these two index tables are satisfied. For example, if a1 recalls b2 and b1 recalls a2 are satisfied, record it as L = {(a1, b2), (b1, a2)};

[0200] S8-2, use a1 / b1 as the index key to index to the corresponding subtree sets A1, B1, D1 in ; Similarly, use a2 / b2 as the index key to index to the corresponding subtree sets A2, B2, C2 in . At this time, establish two temporary index tables to store the recall subtree to leaf node index

[0201] After that, it is necessary to judge whether the index keys in and satisfy that the corresponding index values are all in L. If not, delete the corresponding index key. After that, the recall subtree sets A1, B1 and the recall subtree sets A2, B2 can be obtained;

[0202] S8-3, index the corresponding equivalent tree sets 1_1 and 2_1 in using A1 / B1 as the index key; similarly, index the corresponding equivalent tree sets 1_2 and 2_2 in using A2 / B2 as the index key;

[0203] S8-4, index the corresponding diagnostic word sets A1, B1 in using 1_1 / 2_1 as the index key; similarly, index the corresponding diagnostic word sets A2, B2 in using 1_2 / 2_2 as the index key.

[0204] It should be noted that considering that the meanings of all equivalent trees under the same diagnostic word in the term operation tree are completely equal, the recall condition for step S8-4 is that any pair of equivalent trees is recalled, that is, the corresponding standard words are recalled; there is a logical relationship between all subtrees under the same equivalent tree, which is not considered during the recall stage. Therefore, the recall condition for step S8-3 is that any pair of subtrees is recalled, that is, the corresponding equivalent trees are recalled; while all leaf nodes under the same subtree are interdependent, and all of them must exist and be recalled for this pair of subtrees to be recalled. Therefore, in step S8-2, a temporary index table needs to be established to determine whether all leaf nodes under a pair of subtrees are recalled before the corresponding subtrees can be recalled.

[0205] It can be seen from the index table size and the recall logic steps that the multi-level recall framework compresses the number of matching comparisons between the original words and the standard words from 104685 * 172810 to 14083 * 6874 + a small number of index comparisons from subtrees to leaf nodes, achieving a theoretical time-consuming compression of 90% or more.

[0206] Through the embodiments provided in this application, the working efficiency of the diagnostic batch processing scenario is improved. And during the experiment, the CPU occupancy of the multi-level recall processing always remains at a relatively low level (compared with parallel processing), which also reflects the high adaptability of this embodiment and can be implemented in different machine configuration sites.

[0207] It can be understood that in the specific implementation of this application, data related to user information, etc. is involved. When the above embodiments of this application are applied to specific products or technologies, user permission or consent needs to be obtained, and the collection, use, and processing of relevant data need to comply with the relevant laws, regulations, and standards of relevant countries and regions.

[0208] It should be noted that, for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that this application is not limited by the described action sequence, because according to this application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0209] According to another aspect of the embodiments of the present application, there is also provided a term processing device for implementing the above-mentioned term processing method. As Figure 10 shown, the device includes:

[0210] A splitting unit 1002, configured to perform term splitting processing on at least two original terms in response to a standard term recall request triggered for at least two original terms in a specific field, to obtain a first quantity of first term components, where the first term components are term components in the original terms, and the standard term recall request is used to request the recall of at least one output standard term that matches the at least two original terms;

[0211] A merging unit 1004, configured to merge at least two first term components with the same expressed meaning among the first quantity of first term components, to obtain a second quantity of second term components, where the second quantity is less than the first quantity, and the second term components are the first term components or term components obtained after merging at least two first term components;

[0212] A constructing unit 1006, configured to construct a first term operation network for at least two original terms, and construct a second term operation network for standard terms in a specific field;

[0213] A recall unit 1008, configured to recall at least one output node that meets the recall condition corresponding to the nodes of the first term operation network from the nodes of the second term operation network, where the first term operation network is a term operation network constructed for at least two original terms, the second term operation network is a term operation network constructed for standard terms in a specific field, the nodes of the first term operation network correspond to the term components in the second quantity of second term components, and the nodes of the second term operation network correspond to the term components in the standard terms;

[0214] A determining unit 1010, configured to determine the term components corresponding to the at least one output node as at least one output standard term.

[0215] For specific embodiments, reference can be made to the examples shown in the above-mentioned term processing method, and details are not described herein again.

[0216] As an alternative, the apparatus further includes at least one of the following:

[0217] A first acquisition unit, configured to convert the first quantity of first term components into semantic vectors before merging at least two first term components with the same meaning in the first quantity of first term components to obtain a second quantity of second term components; measure the vector similarity between the semantic vectors, and determine, as at least two first term components, multiple term components in the first quantity of first term components whose vector similarity is greater than or equal to a first similarity threshold, where the vector similarity has a positive correlation with the semantic similarity between the first quantity of first term components;

[0218] A second acquisition unit, configured to deeply analyze the context associated with the first quantity of first term components and extract context information having a close relationship with the first quantity of first term components in the context before merging at least two first term components with the same meaning in the first quantity of first term components to obtain a second quantity of second term components; deeply fuse the context information with the first quantity of first term components, and capture the information expression meaning after the deep fusion; determine the information expression meaning as the expression meaning corresponding to each first term component in the first quantity of first term components;

[0219] A third acquisition unit, configured to measure the information similarity between the first quantity of first term components and each corpus in a preset corpus through semantic vectors before merging at least two first term components with the same meaning in the first quantity of first term components to obtain a second quantity of second term components; and determine at least two first term components from the first quantity of first term components by using the information similarity.

[0220] For specific embodiments, reference may be made to the examples shown in the above term processing method, and details are not described herein again.

[0221] As an alternative, the apparatus further includes:

[0222] A fourth acquisition unit, configured to acquire a first component and a second component belonging to a specific business scenario from the first quantity of first term components before measuring the information similarity between the first quantity of first term components and each corpus in a preset corpus through semantic vectors;

[0223] A fifth acquisition unit, configured to acquire a corpus matching the specific business scenario before measuring the information similarity between the first quantity of first term components and each corpus in a preset corpus through semantic vectors, where the preset corpus includes the corpus matching the specific business scenario;

[0224] The third acquisition unit includes:

[0225] A first acquisition module, configured to acquire a first similarity between each corpus in a corpus that matches a first component and a specific business scenario;

[0226] A first determination module, configured to, when the first similarity between each corpus in the corpus that matches the first component and the specific business scenario is acquired and any first similarity is greater than or equal to a second preset threshold, use the first corpus with the highest first similarity in the corpus that matches the specific business scenario as the expression meaning corresponding to the first component, where the information similarity includes the first similarity;

[0227] A second acquisition module, configured to acquire a second similarity between each corpus in a corpus that matches a second component and a specific business scenario;

[0228] A second determination module, configured to, when the second similarity between each corpus in the corpus that matches the second component and the specific business scenario is acquired and any second similarity is greater than or equal to a second preset threshold, use the second corpus with the highest second similarity in the corpus that matches the specific business scenario as the expression meaning corresponding to the second component, where the information similarity includes the second similarity;

[0229] A third acquisition module, configured to acquire a corpus similarity between the first corpus and the second corpus;

[0230] A third determination module, configured to determine that the first component and the second component have the same expression meaning when the corpus similarity is greater than or equal to a third similarity threshold.

[0231] For specific embodiments, reference may be made to the examples shown in the above-mentioned term processing method, and details are not described herein again.

[0232] As an optional solution, the recall unit 1008 includes:

[0233] A fourth acquisition module, configured to acquire sub-structures belonging to a first level between a first term operation network and a second term operation network, where the sub-structures belonging to the first level include a first sub-structure of the first term operation network and a second sub-structure of the second term operation network, nodes of the first sub-structure correspond to term components belonging to the first level among second-term components with a second quantity, and nodes of the second sub-structure correspond to term components belonging to the first level in standard terms;

[0234] A fourth determination module, configured to determine at least one first node in the nodes of the second sub-structure that satisfies a first condition for recalling the nodes of the first sub-structure, where the recall condition includes the first condition, and the similarity between the term component corresponding to the first node and the term component corresponding to the node of the first sub-structure is greater than or equal to a threshold matched by the first condition;

[0235] A fifth acquisition module, configured to acquire sub-structures belonging to a second level between a first term operation network and a second term operation network, where the sub-structures belonging to the second level include a third sub-structure of the first term operation network and a fourth sub-structure of the second term operation network. The nodes of the third sub-structure correspond to the term components belonging to the second level among the second quantity of second term components, and the nodes of the second sub-structure correspond to the term components belonging to the second level in the standard terms. The second level is the upper level corresponding to the first level.

[0236] A fifth determination module, configured to determine at least one second node among the nodes of the fourth sub-structure that has an indexing relationship with at least one first node and satisfies a second condition for recalling the nodes of the third sub-structure. The recall condition includes the second condition. Having an indexing relationship means that there is a hierarchical relationship between nodes belonging to different levels. The similarity between the term components corresponding to the second node and the term components corresponding to the nodes of the third sub-structure is greater than or equal to the threshold value matched by the second condition.

[0237] A sixth determination module, configured to, when the number of sub-structures at the upper level corresponding to the second level is less than or equal to a preset threshold, determine at least one second node as at least one output node.

[0238] For specific embodiments, reference may be made to the examples shown in the above-mentioned term processing method, and details are not described herein again.

[0239] As an alternative solution, the merging unit 1004 includes:

[0240] A first merging module, configured to merge the term components with the same expressed meaning and belonging to the same first level among the first quantity of first term components to obtain a plurality of first-level components, and use the plurality of first-level components as the term components corresponding to the nodes of the first sub-structure, where the second quantity of second term components includes a plurality of first-level components.

[0241] A second merging module, configured to merge the term components with the same expressed meaning and belonging to the same second level among the first quantity of first term components to obtain a plurality of second-level components, and use the plurality of second-level components as the term components corresponding to the nodes of the second sub-structure, where the second quantity of second term components includes a plurality of second-level components.

[0242] For specific embodiments, reference may be made to the examples shown in the above-mentioned term processing method, and details are not described herein again.

[0243] As an alternative solution, the apparatus further includes:

[0244] A classification module, configured to classify the second terms of the second quantity according to the component type before obtaining the sub-structures at the first level between the first term operation network and the second term operation network. Among them, the sub-structure at the highest level in the first term operation network corresponds to at least two original terms, and the nodes of the sub-structures at other different levels in the first term operation network correspond to the second term components of different component types.

[0245] For specific embodiments, reference may be made to the examples shown in the above-mentioned term processing method, and details are not described herein again.

[0246] As an alternative solution, the merging unit 1004 includes:

[0247] A third merging module, configured to merge at least two first term components to obtain the second terms of the second quantity when the first quantity is greater than or equal to the excess threshold.

[0248] For specific embodiments, reference may be made to the examples shown in the above-mentioned term processing method, and details are not described herein again.

[0249] As an alternative solution, the third merging module includes:

[0250] A first merging sub-module, configured to merge at least two first term components to obtain the second terms of the first interval quantity when the first quantity is greater than or equal to the excess threshold and less than the massive threshold, where the massive threshold is greater than the excess threshold; or,

[0251] A second merging sub-module, configured to merge at least two first term components to obtain the second terms of the second interval quantity when the first quantity is greater than or equal to the excess threshold and greater than or equal to the massive threshold, where the second interval quantity is greater than the first interval quantity.

[0252] For specific embodiments, reference may be made to the examples shown in the above-mentioned term processing method, and details are not described herein again.

[0253] According to another aspect of the embodiments of the present application, an electronic device for implementing the above-mentioned term processing method is further provided. The electronic device may be, but is not limited to, Figure 1 the user equipment 102 or the server 112 shown in Figure 11 As shown, the electronic device includes a memory 1102 and a processor 1104. The memory 1102 stores a computer program, and the processor 1104 is configured to execute the steps in any one of the above method embodiments through the computer program.

[0254] Optionally, in this embodiment, the above-mentioned electronic device may be at least one network device among multiple network devices of a computer network.

[0255] Optionally, in this embodiment, the above-mentioned processor may be configured to perform the following steps through a computer program:

[0256] S1. In response to a standard term recall request triggered by at least two original terms in a specific domain, perform term splitting processing on the at least two original terms to obtain a first quantity of first term components, where the first term components are term components in the original terms, and the standard term recall request is used to request the recall of at least one output standard term that matches the at least two original terms;

[0257] S2. Merge at least two first term components with the same expressed meaning among the first quantity of first term components to obtain a second quantity of second term components, where the second quantity is less than the first quantity, and the second term components are either the first term components or the term components obtained after merging the at least two first term components;

[0258] S3. Construct a first term operation network for the at least two original terms and a second term operation network for the standard terms in a specific domain;

[0259] S4. In the case of obtaining the first term operation network and the second term operation network, recall at least one output node that meets the recall conditions corresponding to the nodes of the first term operation network from the nodes of the second term operation network, where the first term operation network is a term operation network constructed for the at least two original terms, the second term operation network is a term operation network constructed for the standard terms in a specific domain, the nodes of the first term operation network correspond to the term components in the second quantity of second term components, and the nodes of the second term operation network correspond to the term components in the standard terms;

[0260] S5. Determine the term components corresponding to the at least one output node as the at least one output standard term.

[0261] Optionally, those of ordinary skill in the art can understand that Figure 11 the structure shown is only schematic Figure 11 and does not limit the structure of the above-mentioned electronic device. For example, the electronic device may further include more or fewer components (such as a network interface, etc.) than those shown Figure 11 or have a different configuration from that shown. Figure 11

[0262] ​Among them, the memory 1102 can be used to store software programs and modules, such as the program instructions / modules corresponding to the term processing method and device in the embodiments of the present application. The processor 1104 executes various functional applications and data processing by running the software programs and modules stored in the memory 1102, that is, to implement the above-mentioned term processing method. The memory 1102 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory 1102 may further include a memory remotely disposed relative to the processor 1104, and these remote memories can be connected to the electronic device through a network. Examples of the above network include but are not limited to the Internet, enterprise intranet, local area network, mobile communication network, and combinations thereof. Among them, the memory 1102 can specifically but not limitedly be used to store information such as original terms and output standard terms. As an example, as Figure 11 shown, the above-mentioned memory 1102 may but not limitedly include the splitting unit 1002, merging unit 1004, building unit 1006, recall unit 1008, and determination unit 1010 in the above-mentioned term processing device. In addition, it may also include but not limited to other module units in the above-mentioned term processing device, which will not be elaborated in this example.

[0263] Optionally, the above-mentioned transmission device 1106 is used to receive or send data via a network. Specific examples of the above network may include a wired network and a wireless network. In one instance, the transmission device 1106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices and routers through a network cable, so as to communicate with the Internet or local area network. In one instance, the transmission device 1106 is a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0264] In addition, the above-mentioned electronic device further includes: a display 1108, which is used to display information such as the above-mentioned original terms and output standard terms; and a connection bus 1110, which is used to connect each module component in the above-mentioned electronic device.

[0265] In other embodiments, the above-mentioned user equipment or server may be a node in a distributed system. Among them, the distributed system may be a blockchain system, and the blockchain system may be a distributed system formed by connecting multiple nodes through network communication. Among them, the nodes can form a peer-to-peer network, and any form of computing device, such as electronic devices such as servers and user equipment, can become a node in the blockchain system by joining the peer-to-peer network.

[0266] According to one aspect of the present application, a computer program product is provided. The computer program product includes a computer program / instructions, and the computer program / instructions contain program code for performing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication part, and / or installed from a removable medium. When the computer program is executed by the central processing unit, various functions provided by the embodiments of the present application are performed.

[0267] The serial numbers of the embodiments of the present application above are only for description and do not represent the superiority or inferiority of the embodiments.

[0268] It should be noted that the computer system of the electronic device is only an example and should not impose any limitations on the functions and usage scope of the embodiments of the present application.

[0269] The computer system includes a central processing unit (CPU), which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) or the program loaded from the storage part into the random access memory (RAM). In the random access memory, various programs and data required for system operation are also stored. The central processing unit, the read-only memory, and the random access memory are connected to each other through a bus. The input / output interface (Input / Output interface, i.e., I / O interface) is also connected to the bus.

[0270] The following components are connected to the input / output interface: an input part including a keyboard, a mouse, etc.; an output part including a cathode ray tube (CRT), a liquid crystal display (LCD), etc. and a speaker, etc.; a storage part including a hard disk, etc.; and a communication part including a network interface card such as a local area network card, a modem, etc. The communication part performs communication processing via a network such as the Internet. A drive is also connected to the input / output interface as needed. A removable medium, such as a disk, an optical disc, a magneto-optical disc, a semiconductor memory, etc., is installed on the drive as needed so that the computer program read from it can be installed into the storage part as needed.

[0271] In particular, according to an embodiment of the present application, the processes described in each method flowchart can be implemented as computer software programs. For example, an embodiment of the present application includes a computer program product that includes a computer program carried on a computer-readable medium. The computer program contains program code for performing the methods shown in the flowcharts. In such an embodiment, the computer program can be downloaded and installed from a network through a communication part, and / or installed from a removable medium. When the computer program is executed by a central processing unit, various functions defined in the system of the present application are executed.

[0272] According to one aspect of the present application, there is provided a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the methods provided in the above various alternative implementation manners.

[0273] Optionally, in this embodiment, the above computer-readable storage medium may be configured to store a computer program for performing the following steps:

[0274] S1. In response to a standard term recall request triggered by at least two original terms in a specific field, perform term splitting processing on the at least two original terms to obtain a first quantity of first term components, where the first term components are term components in the original terms, and the standard term recall request is used to request the recall of at least one output standard term that matches the at least two original terms;

[0275] S2. Merge at least two first term components with the same expressed meaning among the first quantity of first term components to obtain a second quantity of second term components, where the second quantity is less than the first quantity, the second term components are first term components, or term components obtained after merging at least two first term components;

[0276] S3. Construct a first term operation network for at least two original terms to obtain a first term operation network, and construct a second term operation network for standard terms in a specific field to obtain a second term operation network;

[0277] S4. In the case of obtaining the first term operation network and the second term operation network, recall at least one output node that meets the recall condition corresponding to the nodes of the first term operation network from the nodes of the second term operation network, where the first term operation network is a term operation network constructed for at least two original terms, the second term operation network is a term operation network constructed for standard terms in a specific field, the nodes of the first term operation network correspond to the term components in the second quantity of second term components, and the nodes of the second term operation network correspond to the term components in the standard terms;

[0278] S5. Determine the term components corresponding to at least one output node as at least one output standard term.

[0279] Optionally, in the embodiments of the present application, the term "module" or "unit" refers to a computer program with a predetermined function or a part of a computer program, which works together with other related parts to achieve a predetermined goal, and can be implemented in whole or in part by using software, hardware (such as a processing circuit or a memory), or a combination thereof. Similarly, one processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be a part of an overall module or unit that includes the functions of the module or unit.

[0280] Optionally, in this embodiment, those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the relevant hardware of the electronic device through a program, and the program can be stored in a computer-readable storage medium. The storage medium can include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, an optical disc, etc.

[0281] The serial numbers of the embodiments of the present application above are only for description and do not represent the superiority or inferiority of the embodiments.

[0282] If the integrated unit in the above embodiments is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in the above computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing one or more computer devices (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application.

[0283] In the above embodiments of the present application, the descriptions of the various embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0284] In several embodiments provided in the present application, it should be understood that the disclosed user equipment can be implemented in other ways. Among them, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces. The indirect coupling or communication connection of units or modules can be in electrical or other forms.

[0285] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0286] In addition, each functional unit in various embodiments of the present application can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0287] The above are only the preferred embodiments of the present application. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present application, several improvements and refinements can still be made, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A method for term processing, characterized in that, Including: In response to a standard term recall request triggered by at least two original terms in a specific domain, performing term splitting processing on the at least two original terms to obtain a first quantity of first term components, wherein the standard term recall request is used to request the recall of at least one output standard term that matches the at least two original terms; Merging at least two first term components with the same expressed meaning among the first quantity of first term components to obtain a second quantity of second term components, wherein the second quantity is less than the first quantity; Constructing a first term operation network for the at least two original terms and constructing a second term operation network for the standard terms in the specific domain; Recalling at least one output node that meets the recall conditions corresponding to the nodes of the first term operation network from the nodes of the second term operation network, wherein the nodes of the first term operation network correspond to the term components in the second quantity of second term components, and the nodes of the second term operation network correspond to the term components in the standard terms; Determining the term components corresponding to the at least one output node as the at least one output standard term.

2. The method according to claim 1, wherein Before merging at least two first term components with the same expressed meaning among the first quantity of first term components to obtain a second quantity of second term components, the method further includes: Converting the first quantity of first term components into semantic vectors; Measuring the vector similarity between the semantic vectors, and determining, among the first quantity of first term components, multiple term components with a vector similarity greater than or equal to a first similarity threshold as the at least two first term components, wherein the vector similarity has a positive correlation with the semantic similarity between the first quantity of first term components; or, deeply analyzing the context associated with the first quantity of first term components and extracting context information in the context that has a close relationship with the first quantity of first term components; Deeply fusing the context information with the first quantity of first term components and capturing the information expression meaning obtained after the deep fusion; Determining the information expression meaning as the expressed meaning corresponding to each first term component in the first quantity of first term components; or, Measuring the information similarity between the first quantity of first term components and each corpus in a preset corpus through the semantic vectors; Using the information similarity to determine the at least two first term components from the first quantity of first term components.

3. The method according to claim 2, wherein: Before measuring the information similarity between the first quantity of first term components and each corpus in a preset corpus through the semantic vectors, the method further includes: Obtaining a first component and a second component belonging to a specific business scenario from the first quantity of first term components; Obtain the corpus matching the specific business scenario, where the preset corpus includes the corpus matching the specific business scenario; The determining the at least two first term components from the first quantity of first term components by using the information similarity includes: When the first similarity between the first component and each corpus in the corpus matching the specific business scenario is obtained, and any one of the first similarities is greater than or equal to the second preset threshold, the first corpus with the highest first similarity in the corpus matching the specific business scenario is used as the expression meaning corresponding to the first component, where the information similarity includes the first similarity; When the second similarity between the second component and each corpus in the corpus matching the specific business scenario is obtained, and any one of the second similarities is greater than or equal to the second preset threshold, the second corpus with the highest second similarity in the corpus matching the specific business scenario is used as the expression meaning corresponding to the second component, where the information similarity includes the second similarity; Obtain the corpus similarity between the first corpus and the second corpus; When the corpus similarity is greater than or equal to the third similarity threshold, it is determined that the first component and the second component have the same expression meaning.

4. The method according to claim 1, wherein The recalling of at least one output node that meets the recall condition corresponding to the node of the first term operation network from the nodes of the second term operation network includes: Obtain the sub-structures belonging to the first level between the first term operation network and the second term operation network, where the sub-structures belonging to the first level include the first sub-structure of the first term operation network and the second sub-structure of the second term operation network. The nodes of the first sub-structure correspond to the term components belonging to the first level among the second quantity of second term components, and the nodes of the second sub-structure correspond to the term components belonging to the first level in the standard terms; Determine at least one first node in the nodes of the second sub-structure that meets the first condition for recalling the nodes of the first sub-structure, where the recall condition includes the first condition, and the similarity between the term components corresponding to the first node and the term components corresponding to the nodes of the first sub-structure is greater than or equal to the threshold matching the first condition; Obtain the sub-structures belonging to the second level between the first term operation network and the second term operation network, where the sub-structures belonging to the second level include the third sub-structure of the first term operation network and the fourth sub-structure of the second term operation network. The nodes of the third sub-structure correspond to the term components belonging to the second level among the second quantity of second term components, and the nodes of the second sub-structure correspond to the term components belonging to the second level in the standard terms. The second level is the upper level corresponding to the first level; Determine at least one second node among the nodes of the fourth sub-structure that has an indexing relationship with the at least one first node and satisfies a second condition for recalling the nodes of the third sub-structure. Among them, the recall condition includes the second condition, and having the indexing relationship means that there is a superordinate-subordinate relationship between nodes belonging to different levels. The similarity between the term components corresponding to the second node and the term components corresponding to the nodes of the third sub-structure is greater than or equal to the threshold matched by the second condition; In the case where the number of sub-structures at the upper level corresponding to the second level is less than or equal to a preset threshold, determine the at least one second node as the at least one output node.

5. The method according to claim 4, wherein The merging of at least two first term components with the same expressed meaning among the first quantity of first term components to obtain a second quantity of second term components includes: Merge the first term components with the same expressed meaning and belonging to the first level among the first quantity of first term components to obtain multiple first-level components, and use the multiple first-level components as the term components corresponding to the nodes of the first sub-structure. Among them, the second quantity of second term components includes the multiple first-level components; Merge the first term components with the same expressed meaning and belonging to the second level among the first quantity of first term components to obtain multiple second-level components, and use the multiple second-level components as the term components corresponding to the nodes of the second sub-structure. Among them, the second quantity of second term components includes the multiple second-level components.

6. The method according to claim 4, characterized in that, Before obtaining the sub-structure at the first level between the first term operation network and the second term operation network, the method further includes: Classify the second quantity of second term components according to the component type. Among them, the top-level sub-structure in the first term operation network corresponds to the at least two original terms, and the nodes of other different-level sub-structures in the first term operation network correspond to the second term components of different component types.

7. The method according to any one of claims 1 to 6, characterized in that, The merging of at least two first term components with the same expressed meaning among the first quantity of first term components to obtain a second quantity of second term components includes: In the case where the first quantity is greater than or equal to an excess threshold, merge the at least two first term components to obtain the second quantity of second term components.

8. The method according to claim 7, characterized in that The merging of the at least two first term components to obtain the second quantity of second term components in the case where the first quantity is greater than or equal to an excess threshold includes: In the case where the first quantity is greater than or equal to the excess threshold and less than the massive threshold, the at least two first term components are combined to obtain second term components with a first interval quantity, where the massive threshold is greater than the excess threshold; or, in the case where the first quantity is greater than or equal to the excess threshold and the first quantity is greater than or equal to the massive threshold, the at least two first term components are combined to obtain second term components with a second interval quantity, where the second interval quantity is greater than the first interval quantity.

9. A term processing device, characterized in that, Comprising: A splitting unit, configured to perform a term splitting process on the at least two original terms in response to a standard term recall request triggered for the at least two original terms in a specific field, to obtain first term components with a first quantity, where the standard term recall request is used to request the recall of at least one output standard term that matches the at least two original terms; A combining unit, configured to combine at least two first term components with the same expressed meaning among the first term components with the first quantity, to obtain second term components with a second quantity, where the second quantity is less than the first quantity; A constructing unit, configured to construct a first term operation network for the at least two original terms, and construct a second term operation network for the standard terms in the specific field; A recall unit, configured to recall at least one output node that meets the recall conditions corresponding to the nodes of the first term operation network from the nodes of the second term operation network, where the nodes of the first term operation network correspond to the term components in the second term components with the second quantity, and the nodes of the second term operation network correspond to the term components in the standard terms; A determining unit, configured to determine the term components corresponding to the at least one output node as the at least one output standard term.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, where the program, when run by an electronic device, executes the method described in any one of claims 1 to 8.

11. A computer program product, comprising a computer program / instructions, characterized in that, The computer program / instructions, when executed by a processor, implement the steps of the method described in any one of claims 1 to 8.

12. An electronic device, comprising a memory and a processor, characterized in that, A computer program is stored in the memory, and the processor is configured to execute the method described in any one of claims 1 to 8 through the computer program.

Images