Candidate entity screening method, device, readable storage medium and electronic device

Abstract

The present disclosure provides a candidate entity screening method, which constructs a trigger word library, an entity internal word formation rule library and a standard semantic part-of-speech rule library based on a training set; for candidate entities, from the perspective of part-of-speech tagging, the candidate entities are double-checked by trigger words and internal word formation characteristics to determine the candidate entity with the highest similarity. The present disclosure also provides a candidate entity screening device, a readable storage medium and an electronic device.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and more particularly to a candidate entity screening method, a selected entity screening device, a readable storage medium, and an electronic device. Background Art

[0002] Smart medical construction refers to the entire process from patient admission, physician diagnosis and treatment, patient treatment to discharge after recovery, using intelligent electronic medical record archives as the information basis, using artificial intelligence technology to assist diagnosis and treatment and disease decision-making, and using intelligent equipment to achieve integrated scientific management of patient treatment.

[0003] Named entity recognition technology can mine potential information in electronic medical records, convert the text of unstructured data into structured information, and lay the foundation for the construction of smart medical care.

[0004] For medical named entities, it is crucial to improve the quality of new annotated data augmented with pre-models.

[0005] Labeling unlabeled data and adding it to the training set can effectively alleviate the problem of unlabeled data waste. However, the model trained with limited initial data is less effective in recognizing complex entity types, and adding such labeled data to the training set will introduce noise. Summary of the invention

[0006] In order to solve at least one of the above technical problems, the present disclosure proposes a candidate entity screening method, device, readable storage medium and electronic device.

[0007] According to one aspect of the present disclosure, a candidate entity screening method is provided, comprising the following steps:

[0008] S1: Build trigger word library, entity internal word formation rule library and standard semantic part-of-speech rule library based on the training set;

[0009] S2: Using the trigger words in the trigger word library to perform entity extraction on the medical text to obtain a candidate entity set;

[0010] S3: Compare the entities marked by the model with the entities in the candidate entity set to determine whether they are completely consistent: if they are completely consistent, proceed to step S4; otherwise, proceed to step S6;

[0011] S4: performing part-of-speech tagging on the entities that are completely consistent in step S3, and using the edit distance to calculate the semantic similarity set between the part-of-speech tagging combination of the entity and each part-of-speech combination in the entity internal word formation rule library;

[0012] S5: Based on a preset semantic similarity threshold, determine whether there is a value greater than the semantic similarity threshold in the semantic similarity set: if there is, the entity is correct, and the entity screening ends;

[0013] S6: Entities and trigger words in the candidate entity set are combined into key semantic segments, and part-of-speech tagging is performed on the key semantic segments respectively;

[0014] S7: using the edit distance to calculate the semantic similarity between each key semantic segment and the existing standard semantic segments in the standard semantic part-of-speech rule library, and taking the entities contained in the N key semantic segments with the highest similarity as the entities to be processed;

[0015] S8: Perform part-of-speech tagging on the entity to be processed, calculate the semantic similarity between the part-of-speech tag combination of the entity to be processed and the standard entity rule base using the edit distance, and determine the candidate entity with the highest similarity.

[0016] According to the candidate entity screening method of at least one embodiment of the present disclosure, the trigger vocabulary is constructed by segmenting sentences containing entities in a training set, and constructing the trigger vocabulary according to context information of the entities after segmentation.

[0017] According to the candidate entity screening method of at least one embodiment of the present disclosure, the entity internal word formation rule library is constructed for the entity by performing part-of-speech tagging on the entity and based on the part-of-speech tagging result.

[0018] According to the candidate entity screening method of at least one embodiment of the present disclosure, the standard semantic part-of-speech rule library defines the combination of trigger words and entities as standard semantic segments, performs lexical analysis on key semantic segments, and constructs a standard semantic part-of-speech rule library based on the lexical analysis results.

[0019] According to the candidate entity screening method of at least one embodiment of the present disclosure, in step S2, for a given sentence of a medical text, the positions of all trigger word predecessors L and trigger word successors R are found in the sentence, and the content between adjacent L and R is extracted as a candidate entity.

[0020] According to the candidate entity screening method of at least one embodiment of the present disclosure, in step S3, the entities annotated by the model are a set of entities annotated using a model trained using a training set and a development set.

[0021] According to the candidate entity screening method of at least one embodiment of the present disclosure, the part-of-speech tagging marks the word by its own meaning and context information, and is completed by a machine learning algorithm, and the machine learning method algorithm includes a hidden Markov model or a conditional random field.

[0022] According to the candidate entity screening method of at least one embodiment of the present disclosure, N is an integer between 2 and 5.

[0023] According to another aspect of the present disclosure, a candidate entity screening device is provided, comprising:

[0024] A word library construction module, wherein the word library construction module constructs a trigger word library, an entity internal word formation rule library, and a standard semantic part-of-speech rule library based on the training set;

[0025] A candidate entity set acquisition module, wherein the candidate entity set acquisition module uses the trigger words in the trigger word library to perform entity extraction on the medical text to obtain a candidate entity set;

[0026] An entity judgment module, which compares the entities marked by the model with the entities in the candidate entity set and performs module operations according to the judgment result;

[0027] A word body semantic similarity set calculation module, when the entity judgment module determines that the entity annotated by the model is completely consistent with the entity in the candidate entity set, the word body semantic similarity set calculation module performs part-of-speech tagging on the completely consistent entity, and uses the edit distance to calculate the semantic similarity set between the part-of-speech tagging combination of the entity and each part-of-speech combination in the entity internal word formation rule library;

[0028] A semantic similarity threshold judgment module, wherein the semantic similarity threshold judgment module judges whether there is a value greater than the semantic similarity threshold in the semantic similarity set calculated by the word semantic similarity set calculation module: if there is, the entity is correct, and the entity screening ends;

[0029] A key semantic segment forming module, when the entity judgment module determines that the entity marked by the model is not completely consistent with the entity in the candidate entity set, the key semantic segment forming module combines the entity in the candidate entity set and the trigger word into a key semantic segment, and performs part-of-speech tagging on the key semantic segment;

[0030] A module for determining entities to be processed, wherein the module calculates the semantic similarity between each key semantic segment and an existing standard semantic segment in a standard semantic part-of-speech rule library using edit distance, and selects entities contained in the N key semantic segments with the highest similarity as entities to be processed;

[0031] The module for determining the candidate entity with the highest similarity performs part-of-speech tagging on the entity to be processed generated by the module for determining the entity to be processed, calculates the semantic similarity between the part-of-speech tagging combination of the entity to be processed and the standard entity rule base using the edit distance, and determines the candidate entity with the highest similarity.

[0032] According to another aspect of the present disclosure, a readable storage medium is provided, wherein the readable storage medium stores a computer program, wherein the computer program is used for a processor to execute the candidate entity screening method of any embodiment of the present disclosure.

[0033] According to another aspect of the present disclosure, an electronic device is provided, including a processor and a readable storage medium, wherein the readable storage medium stores execution instructions, and the processor executes the execution instructions in the readable storage medium, so that the processor executes the candidate entity screening method of any embodiment of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The accompanying drawings illustrate exemplary embodiments of the present disclosure and together with the description serve to explain the principles of the present disclosure. These drawings are included to provide a further understanding of the present disclosure and are incorporated in and constitute a part of this specification.

[0035] Figure 1 It is a flowchart of a candidate entity screening method according to one embodiment of the present disclosure.

[0036] Figure 2 It is a schematic block diagram of the structure of a candidate entity screening device according to an embodiment of the present disclosure.

[0037] Figure 3 It is a schematic block diagram of the structure of an electronic device with a candidate entity screening device according to an embodiment of the present disclosure.

[0038] Description of Reference Numerals

[0039] 1000 Electronic equipment

[0040] 1001 vocabulary building blocks

[0041] 1002 Candidate Entity Set Acquisition Module

[0042] 1003 Entity Judgment Module

[0043] 1004 Word semantic similarity set calculation module

[0044] 1005 Semantic Similarity Threshold Judgment Module

[0045] 1006 Key semantic segment building blocks

[0046] 1007 Entity determination module to be processed

[0047] 1008 Module for determining the candidate entity with the highest similarity

[0048] 1100 Bus

[0049] 1200 processor

[0050] 1300 Memory

[0051] 1400 Other circuits. DETAILED DESCRIPTION

[0052] The present disclosure is further described in detail below in conjunction with the accompanying drawings and implementations. It is understood that the specific implementations described herein are only used to explain the relevant content, rather than to limit the present disclosure. It should also be noted that, for ease of description, only the parts related to the present disclosure are shown in the accompanying drawings.

[0053] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in the present disclosure can be combined with each other. The technical solution of the present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0054] Unless otherwise specified, the exemplary embodiments / embodiments shown will be understood as providing exemplary features of various details of some ways in which the technical concept of the present disclosure can be implemented in practice. Therefore, unless otherwise specified, the features of the various embodiments / embodiments can be combined, separated, interchanged and / or rearranged without departing from the technical concept of the present disclosure.

[0055] The use of cross-hatching and / or shading in the accompanying drawings is generally used to make the boundaries between adjacent components clear. As such, unless otherwise specified, the presence or absence of cross-hatching or shading does not convey or indicate any preference or requirement for the specific materials, material properties, dimensions, proportions, commonalities between the components shown, and / or any other characteristics, attributes, properties, etc. of the components. In addition, in the accompanying drawings, the sizes and relative sizes of the components may be exaggerated for clarity and / or descriptive purposes. When the exemplary embodiments can be implemented differently, the specific process sequence can be performed in a different order than described. For example, two successively described processes can be performed substantially simultaneously or in an order opposite to the described order. In addition, the same figure numbers represent the same components.

[0056] When a component is referred to as being "on" or "over," "connected to," or "coupled to" another component, the component may be directly on, directly connected to, or directly coupled to the other component, or intervening components may be present. However, when a component is referred to as being "directly on," "directly connected to," or "directly coupled to" another component, there are no intervening components. For this purpose, the term "connected" may refer to a physical connection, an electrical connection, etc., with or without intervening components.

[0057] The terms used herein are for the purpose of describing specific embodiments, and are not intended to be restrictive. As used herein, unless the context clearly indicates otherwise, the singular forms "one (kind, person)" and "said (the)" are also intended to include plural forms. In addition, when the terms "comprise" and / or "include" and their variations are used in this specification, it is explained that there are stated features, integral bodies, steps, operations, parts, assemblies and / or their groups, but it is not excluded that there are or add one or more other features, integral bodies, steps, operations, parts, assemblies and / or their groups. It should also be noted that, as used herein, the terms "substantially", "approximately" and other similar terms are used as approximate terms and not as degree terms, so that they are used to explain the inherent deviations of the measured values, calculated values ​​and / or the values ​​provided that will be recognized by those of ordinary skill in the art.

[0058] Figure 1 It is a flowchart of a candidate entity screening method according to one embodiment of the present disclosure.

[0059] like Figure 1 As shown, the candidate entity screening method disclosed in the present invention includes:

[0060] S1: Build trigger word library, entity internal word formation rule library and standard semantic part-of-speech rule library based on the training set;

[0061] S2: Use the trigger words in the trigger word library to extract entities from medical texts to obtain candidate entity sets;

[0062] S3: Compare the entities marked by the model with the entities in the candidate entity set to determine whether they are completely consistent: if they are completely consistent, proceed to step S4; otherwise, proceed to step S6;

[0063] S4: Perform part-of-speech tagging on the entities that are completely consistent in step S3 based on the standard semantic part-of-speech rule library, and use the edit distance to calculate the semantic similarity set between the part-of-speech tagging combination of the entity and each part-of-speech combination in the entity's internal word formation rule library;

[0064] S5: Based on a preset semantic similarity threshold, determine whether there is a value greater than the semantic similarity threshold in the semantic similarity set: if there is, the entity is correct, and the entity screening ends;

[0065] S6: Entities and trigger words in the candidate entity set are combined into key semantic segments, and part-of-speech tags are performed on the key semantic segments respectively;

[0066] S7: using the edit distance to calculate the semantic similarity between each key semantic segment and the existing standard semantic segments in the standard semantic part-of-speech rule library, and taking the entities contained in the N key semantic segments with the highest similarity as the entities to be processed;

[0067] S8: Perform part-of-speech tagging on the entity to be processed, use the edit distance to calculate the semantic similarity between the part-of-speech tag combination of the entity to be processed and the standard entity rule library, and determine the candidate entity with the highest similarity.

[0068] In the real data annotation process, judgments are usually made in combination with the contextual semantic information of the entity. These semantic information are called trigger words. The steps to annotate the data are to first lock the location of the entity through the trigger word, and then use the entity's word formation features or external knowledge to assist in determining the entity type to complete the data annotation. The trigger word library is constructed by segmenting the sentences containing entities in the training set, and then constructing a trigger word library according to the contextual information of the entity after segmentation. For example, the sentence "feeling unwell, admitted to the hospital for observation with cerebral infarction", then the trigger word representation of the entity "cerebral infarction" is: admitted to the hospital with [cerebral infarction], and the trigger word pair of the entity "cerebral infarction" is "admitted to the hospital with xxx".

[0069] The entity internal word formation rule library is constructed by tagging entities in the training set with parts of speech, and based on the results of part of speech tagging, the entity internal word formation rule library is constructed for the entity. Taking surgical entities as an example, surgical entities are usually composed of "approach + (range) part + disease nature + procedure", where the approach and range are not fixed, but the part, disease nature and procedure are generally complete, thus forming a surgical entity. The word formation characteristics of the approach, part, disease nature and procedure can be determined through the part of speech tagging results, and then the word formation rules of surgical entities can be obtained. From the results of part-of-speech tagging for surgical entities, we can see that words indicating approaches such as "lower, left, inside" are usually represented by directional words f, words indicating parts such as "prostate, bladder" are usually represented by noun n, and words indicating the nature of the disease such as "lesions, fractures, tumors" are usually represented by biological-related words gb, disease-related words nhd or noun n. Finally, words indicating surgical procedures such as "resection, open reduction and internal fixation, vaporization" are composed of gb, gb|f|gb, v|ng and other combinations of biological / directional words / gerund words.

[0070] The standard semantic part-of-speech rule library defines the combination of trigger words and entities as standard semantic segments, performs lexical analysis on key semantic segments, and builds a standard semantic part-of-speech rule library based on the lexical analysis results. The standard semantic part-of-speech rule library can conduct in-depth analysis of entities and contexts at the part-of-speech level, providing rich semantic information for determining model-annotated entities. Combining the part-of-speech tagging results can merge multiple modes into one mode, providing effective information for subsequent verification of the accuracy of model-annotated entities.

[0071] Use trigger words to perform entity extraction on medical text to obtain a candidate entity set. For a given sentence in a medical text, find the locations of all trigger word predecessors L and trigger word successors R in the sentence, and extract the content between adjacent L and R as candidate entities.

[0072] In step S3, the model obtained by training the training set and the development set is used to annotate the text to obtain an entity set; the entities annotated by the model are compared with the entities in the candidate entity set to determine whether they are completely consistent: if they are completely consistent, proceed to step S4, otherwise proceed to step S6.

[0073] Perform part-of-speech tagging on the entities that are completely consistent in step S3, and use the edit distance to calculate the semantic similarity set of the part-of-speech tagging combination of the entity and the part-of-speech combination in the entity's internal word formation rule library. Among them, the part-of-speech tagging marks the word's part-of-speech by the word's own meaning and context information, which can be completed by a machine learning algorithm. The machine learning method algorithm includes a hidden Markov model, a conditional random field, etc. Based on a pre-set semantic similarity threshold, determine whether there is a value greater than the semantic similarity threshold in the semantic similarity set: if it exists, the entity is judged to be correct, and the entity screening is ended.

[0074] In the case where the entities marked by the model are not completely consistent with the entities in the candidate entity set, the entities in the candidate entity set and the trigger words are combined into key semantic segments, and the key semantic segments are tagged with parts of speech respectively. The semantic similarity between each key semantic segment and the existing standard semantic segments in the standard semantic part-of-speech rule library is calculated using the edit distance, and the entities contained in the N key semantic segments with the highest similarity are taken as the entities to be processed. Among them, the value of N can be an integer between 2 and 5, and the preferred value of N is 3. Part-of-speech tagging is performed on the entity to be processed, and the semantic similarity between the part-of-speech tagging combination of the entity to be processed and the standard entity rule library is calculated using the edit distance to determine the candidate entity with the highest similarity.

[0075] Since the preceding and succeeding words of a trigger word may appear multiple times in a sentence, it is very easy to obtain long texts containing noise when extracting entities through conventional regular matching, which reduces the quality of entities in the entity candidate set. The entity screening mechanism of the above method can double-check the context information and internal word formation characteristics of candidate entities from the perspective of part-of-speech tagging. On the one hand, it can determine the accuracy of the model's annotated entities, and on the other hand, it significantly improves the quality of candidate entities, laying the foundation for the next step of entity denoising.

[0076] Figure 2 A candidate entity screening device according to an embodiment of the present disclosure is shown, comprising:

[0077] The word library construction module 1001 constructs a trigger word library, an entity internal word formation rule library, and a standard semantic part-of-speech rule library based on the training set;

[0078] A candidate entity set acquisition module 1002, wherein the candidate entity set acquisition module 1002 uses trigger words to extract entities from medical text to obtain a candidate entity set;

[0079] Entity judgment module 1003, the entity judgment module 1003 compares the entities marked by the model with the entities in the candidate entity set, and performs module operations according to the judgment result;

[0080] The word body semantic similarity set calculation module 1004, when the entity judgment module 1003 determines that the entity annotated by the model is completely consistent with the entity in the candidate entity set, the word body semantic similarity set calculation module 1004 performs part-of-speech tagging on the completely consistent entity, and uses the edit distance to calculate the semantic similarity set between the part-of-speech tagging combination of the entity and each part-of-speech combination in the entity internal word formation rule library;

[0081] The semantic similarity threshold determination module 1005 determines whether there is a value greater than the semantic similarity threshold in the semantic similarity set calculated by the word semantic similarity set calculation module 1004: if there is, the entity is correct, and the entity screening ends;

[0082] The key semantic segment forming module 1006, when the entity judging module 1003 determines that the entity marked by the model is not completely consistent with the entity in the candidate entity set, the key semantic segment forming module 1006 forms a key semantic segment with the entity in the candidate entity set and the trigger word, and performs part-of-speech tagging on the key semantic segment;

[0083] The to-be-processed entity determination module 1007 calculates the semantic similarity between each key semantic segment and the existing standard semantic segments in the standard semantic part-of-speech rule library using the edit distance, and takes the entities contained in the N key semantic segments with the highest similarity as the to-be-processed entities;

[0084] The module 1008 for determining the candidate entity with the highest similarity performs part-of-speech tagging on the entity to be processed generated by the module for determining the entity to be processed, calculates the semantic similarity between the part-of-speech tagging combination of the entity to be processed and the standard entity rule library using the edit distance, and determines the candidate entity with the highest similarity.

[0085] Figure 3 It is a schematic block diagram of the structure of an electronic device with a candidate entity screening device according to an embodiment of the present disclosure.

[0086] like Figure 3As shown, the electronic device 1000 may include a corresponding module for executing each or several steps in the above method. Therefore, each step or several steps in the above method may be executed by a corresponding module, and the electronic device 1000 may include one or more modules in these modules. The module may be one or more hardware modules specially configured to execute the corresponding steps, or implemented by a processor configured to execute the corresponding steps, or stored in a computer-readable medium for implementation by a processor, or implemented by some combination.

[0087] The hardware structure of the device can be implemented using a bus architecture. The bus architecture can include any number of interconnected buses and bridges, depending on the specific application and overall design constraints of the hardware. The bus 1100 connects various circuits including one or more processors 1200, memory 1300 and / or hardware modules together. The bus 1100 can also connect various other circuits 1400 such as peripherals, voltage regulators, power management circuits, external antennas, etc.

[0088] The bus 1100 may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Component (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the figure only uses one connecting line, but does not mean that there is only one bus or one type of bus.

[0089] In the description of this specification, the description with reference to the terms "one embodiment / method", "some embodiments / methods", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment / method or example are included in at least one embodiment / method or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment / method or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments / methods or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments / methods or examples described in this specification and the features of the different embodiments / methods or examples, unless they are contradictory.

[0090] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0091] Those skilled in the art should understand that the above embodiments are only for the purpose of clearly illustrating the present disclosure, and are not intended to limit the scope of the present disclosure. For those skilled in the art, other changes or modifications may be made based on the above disclosure, and these changes or modifications are still within the scope of the present disclosure.

Claims

1. A candidate entity screening method, characterized in that: The steps include: S1: Build trigger word library, entity internal word formation rule library and standard semantic part-of-speech rule library based on the training set; S2: Using the trigger words in the trigger word library to perform entity extraction on the medical text to obtain a candidate entity set; S3: Compare the entities marked by the model with the entities in the candidate entity set to determine whether they are completely consistent: if they are completely consistent, proceed to step S4; otherwise, proceed to step S6; S4: performing part-of-speech tagging on the entities that are completely consistent in step S3 based on the standard semantic part-of-speech rule library, and calculating the semantic similarity set between the part-of-speech tagging combination of the entity and each part-of-speech combination in the entity internal word formation rule library using the edit distance; S5: Based on a preset semantic similarity threshold, determine whether there is a value greater than the semantic similarity threshold in the semantic similarity set: if there is, the entity is correct, and the entity screening ends; S6: Entities and trigger words in the candidate entity set are combined into key semantic segments, and part-of-speech tagging is performed on the key semantic segments respectively; S7: using the edit distance to calculate the semantic similarity between each key semantic segment and the existing standard semantic segments in the standard semantic part-of-speech rule library, and taking the entities contained in the N key semantic segments with the highest similarity as the entities to be processed; and S8: performing part-of-speech tagging on the entity to be processed, calculating the semantic similarity between the part-of-speech tag combination of the entity to be processed and the standard entity rule library using the edit distance, and determining the candidate entity with the highest similarity; The trigger word library is constructed by segmenting the sentences containing entities in the training set, and constructing the trigger word library according to the context information of the entities after segmentation; The entity internal word formation rule library is constructed for the entity by performing part-of-speech tagging on the entity and according to the part-of-speech tagging result; The standard semantic part-of-speech rule library defines the combination of trigger words and entities as standard semantic segments, performs lexical analysis on key semantic segments, and constructs the standard semantic part-of-speech rule library according to the lexical analysis results; The part-of-speech tagging is performed by marking the word's part-of-speech based on the word's own meaning and context information, and is completed by a machine learning algorithm, wherein the machine learning algorithm includes a hidden Markov model or a conditional random field.

2. The candidate entity screening method according to claim 1, characterized in that: In step S2, for a given sentence of a medical text, the positions of all trigger word predecessors L and trigger word successors R are found in the sentence, and the content between adjacent L and R is extracted as a candidate entity.

3. The candidate entity screening method according to claim 1, characterized in that: In step S3, the entities annotated by the model are a set of entities annotated using the model trained by the training set and the development set.

4. A candidate entity screening device, characterized in that: include: A word library construction module, wherein the word library construction module constructs a trigger word library, an entity internal word formation rule library, and a standard semantic part-of-speech rule library based on the training set; The trigger word library is constructed by segmenting the sentences containing entities in the training set, and constructing the trigger word library according to the context information of the entities after segmentation; The entity internal word formation rule library is constructed for the entity by performing part-of-speech tagging on the entity and according to the part-of-speech tagging result; The standard semantic part-of-speech rule library defines the combination of trigger words and entities as standard semantic segments, performs lexical analysis on key semantic segments, and constructs the standard semantic part-of-speech rule library according to the lexical analysis results; The part-of-speech tagging is performed on the words by using the word's own meaning and context information, and is completed by a machine learning algorithm, wherein the machine learning algorithm includes a hidden Markov model or a conditional random field; A candidate entity set acquisition module, wherein the candidate entity set acquisition module uses trigger words to perform entity extraction on medical text to obtain a candidate entity set; An entity judgment module, which compares the entities marked by the model with the entities in the candidate entity set and performs module operations according to the judgment result; A word body semantic similarity set calculation module, when the entity judgment module determines that the entity annotated by the model is completely consistent with the entity in the candidate entity set, the word body semantic similarity set calculation module performs part-of-speech tagging on the completely consistent entity based on the standard semantic part-of-speech rule library, and uses the edit distance to calculate the semantic similarity set of the part-of-speech tagging combination of the entity and each part-of-speech combination in the entity internal word formation rule library; A semantic similarity threshold judgment module, wherein the semantic similarity threshold judgment module judges whether there is a value greater than the semantic similarity threshold in the semantic similarity set calculated by the word semantic similarity set calculation module: if there is, the entity is correct, and the entity screening ends; A key semantic segment forming module, when the entity judgment module determines that the entity marked by the model is not completely consistent with the entity in the candidate entity set, the key semantic segment forming module combines the entity in the candidate entity set and the trigger word into a key semantic segment, and performs part-of-speech tagging on the key semantic segment; A module for determining entities to be processed, wherein the module calculates the semantic similarity between each key semantic segment and an existing standard semantic segment in a standard semantic part-of-speech rule library using edit distance, and selects entities contained in the N key semantic segments with the highest similarity as entities to be processed; and The module for determining the candidate entity with the highest similarity performs part-of-speech tagging on the entity to be processed generated by the module for determining the entity to be processed, calculates the semantic similarity between the part-of-speech tagging combination of the entity to be processed and the standard entity rule library using the edit distance, and determines the candidate entity with the highest similarity.

5. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and the computer program is used for a processor to execute the candidate entity screening method according to any one of claims 1 to 3.

6. An electronic device, characterized in that: The method comprises a processor and a readable storage medium, wherein the readable storage medium stores execution instructions, and the processor executes the execution instructions in the readable storage medium, so that the processor executes the candidate entity screening method according to any one of claims 1 to 3.

Images