An entity mining method and device

Abstract

The application discloses an entity mining method and device, comprising: obtaining an unstructured entity data set; processing the unstructured entity data set to obtain a first candidate entity data set, a second candidate entity data set and a third candidate entity data set; mining the first candidate entity data set, the second candidate entity data set and the third candidate entity data set based on an entity mining rule to obtain an entity mining result. In this way, in the application, for medical entities that do not exist in a dictionary, the unstructured entity data set is mined using the entity mining rule to obtain the entity mining result. Thus, the entity mining result can be finally mined based on the entity mining rule, avoiding manual mining of entities, thereby improving the accuracy of entity mining.

Description

Technical Field

[0001] This application relates to the field of medical technology, and in particular to a method and apparatus for excavating physical objects. Background Technology

[0002] During the process of organizing medical clinical data, there will be multiple unstructured clinical data. Since there are multiple entities in the unstructured clinical data, none of which exist in the medical domain dictionary, it is necessary to mine the entities in the clinical data.

[0003] Currently, entity mining of unstructured clinical data is performed manually. The mined entities are then verified through various channels such as medical websites, books, guidelines, and reference documents, and reviewed by professionals before being expanded into the medical dictionary. However, this method has low accuracy and is not comprehensive enough for the ever-increasing amount of clinical data, and it also consumes a lot of human resources.

[0004] Therefore, improving the accuracy of entity mining is a key issue of concern to those skilled in the art. Summary of the Invention

[0005] To address the aforementioned problems, this application provides a method and apparatus for entity mining to improve the accuracy of entity mining. The embodiments of this application disclose the following technical solutions:

[0006] Firstly, this application discloses an entity mining method, including:

[0007] Obtain a collection of unstructured entity data;

[0008] The unstructured entity data set is processed to obtain a first candidate entity data set, a second candidate entity data set, and a third candidate entity data set;

[0009] Based on entity mining rules, the first candidate entity dataset, the second candidate entity dataset, and the third candidate entity dataset are mined to obtain entity mining results.

[0010] Optionally, processing the unstructured entity data set to obtain a first candidate entity data set, a second candidate entity data set, and a third candidate entity data set includes:

[0011] The unstructured entity data set is segmented to obtain the first candidate entity data set;

[0012] The unstructured entity data set is segmented into words, and the second candidate entity data set is obtained by combining the segmented entity data set with the first candidate entity data set.

[0013] Sequence prediction is performed on the unstructured entity data set, and the entity data set obtained by sequence prediction is combined with the first candidate entity data set and the second candidate entity data set to obtain the third candidate entity data set.

[0014] Optionally, after performing sequence prediction on the unstructured entity data set, the method further includes:

[0015] Obtain multiple candidate predicted entities and their respective probability values;

[0016] Multiple candidate predicted entities with probability values ​​greater than a first threshold are selected as the entity data set obtained from the sequence prediction.

[0017] Optionally, the step of mining the first candidate entity dataset, the second candidate entity dataset, and the third candidate entity dataset based on entity mining rules to obtain entity mining results includes:

[0018] Determine whether there are identical entities in the first candidate entity data set, the second candidate entity data set, and the third candidate entity data set. If there are, retain the identical entities.

[0019] Determine whether there is partial character overlap between different entities in the first candidate entity data set, the second candidate entity data set, and the third candidate entity data set. If so, retain the entity with the longest character. The entity mining result is constructed based on the retained identical entities and / or the entity with the longest character.

[0020] Optionally, after obtaining the unstructured entity data set, the method further includes:

[0021] The unstructured entity data set is subjected to character conversion processing.

[0022] Secondly, this application discloses a physical excavation device, comprising:

[0023] The acquisition module is used to acquire a collection of unstructured entity data.

[0024] The processing module is used to process the unstructured entity data set to obtain a first candidate entity data set, a second candidate entity data set, and a third candidate entity data set.

[0025] The mining module is used to mine the first candidate entity dataset, the second candidate entity dataset, and the third candidate entity dataset based on entity mining rules to obtain entity mining results.

[0026] Optionally, the processing module includes:

[0027] The first processing subunit is used to perform word segmentation on the unstructured entity data set to obtain the first candidate entity data set.

[0028] The second processing subunit is used to perform word segmentation on the unstructured entity data set, and obtain the second candidate entity data set based on the word segmented entity data set and the first candidate entity data set.

[0029] The third processing subunit is used to perform sequence prediction on the unstructured entity data set, and to obtain the third candidate entity data set by combining the entity data set obtained from the sequence prediction with the first candidate entity data set and the second candidate entity data set.

[0030] Optional, also includes:

[0031] The acquisition module is used to acquire multiple candidate predicted entities and the probability values ​​corresponding to the multiple candidate predicted entities respectively;

[0032] The selection module is used to select multiple candidate predicted entities whose probability values ​​are greater than a first threshold as the entity data set obtained from the sequence prediction.

[0033] Optionally, the mining module includes:

[0034] The first judgment subunit is used to determine whether there are identical entities in the first candidate entity data set, the second candidate entity data set, and the third candidate entity data set. If there are, the identical entities are retained.

[0035] The second judgment subunit is used to determine whether there is partial character overlap between different entities in the first candidate entity data set, the second candidate entity data set, and the third candidate entity data set. If so, the entity with the longest character is retained. The composition module is used to construct the entity mining result based on the retained identical entities and / or the entity with the longest character.

[0036] Optional, also includes:

[0037] The conversion processing module is used to perform character conversion processing on the unstructured entity data set.

[0038] Thirdly, embodiments of this application provide an electronic device, including:

[0039] Memory, used to store computer programs;

[0040] A processor is used to implement the steps of the entity mining method described above when executing the computer program.

[0041] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described entity mining method.

[0042] Compared with the prior art, this application has the following beneficial effects:

[0043] This application first obtains a set of unstructured entity data, then processes this set to obtain a first candidate entity data set, a second candidate entity data set, and a third candidate entity data set. Finally, it mines these three sets of entity data based on entity mining rules to obtain entity mining results. Thus, in this application, for medical entities not existing in the dictionary, entity mining rules can be used to mine the unstructured entity data set to obtain entity mining results. Therefore, entity mining rules can ultimately yield entity mining results, avoiding manual entity mining and thus improving the accuracy of entity mining. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 A flowchart illustrating an entity mining method provided in this application embodiment;

[0046] Figure 2 This is a schematic diagram of the structure of a physical excavation device provided in an embodiment of this application;

[0047] Figure 3 This is a schematic diagram of an electronic device structure provided in an embodiment of this application. Detailed Implementation

[0048] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0049] It should be noted that the entity excavation method and apparatus provided in this application are for the field of medical technology. The above is only an example and does not limit the application field of the method and apparatus provided in this application.

[0050] As described earlier, during the data processing of medical clinical data, numerous unstructured clinical data points are encountered. Since these unstructured clinical data contain multiple entities that are not found in medical lexicons, entity mining is necessary. Currently, entity mining of unstructured clinical data is performed manually. The mined entities are verified through various channels such as medical websites, books, guidelines, and reference documents, and then reviewed by professionals before being added to medical lexicons. However, this method suffers from low accuracy and incomplete identification for the continuously generated clinical data, and it is also resource-intensive. Therefore, improving the accuracy of entity mining is a key concern for those skilled in the art.

[0051] To address the problems encountered in existing technologies, this application first obtains a set of unstructured entity data, then processes this set to obtain a first candidate entity data set, a second candidate entity data set, and a third candidate entity data set. Finally, it mines these three sets of entity data based on entity mining rules to obtain entity mining results. Thus, in this application, for medical entities not existing in the dictionary, entity mining rules can be used to mine the unstructured entity data set to obtain entity mining results. Therefore, entity mining rules can ultimately yield entity mining results, avoiding manual entity mining and improving the accuracy of entity mining.

[0052] The technical terms that may appear in this application will be explained below.

[0053] Unstructured data refers to data with irregular or incomplete structures, lacking a predefined data model, and inconvenient to represent using two-dimensional logical tables in a database. It includes all formats of office documents, text, images, XML, HTML, various reports, images, and audio / video information.

[0054] Chinese word segmentation: This involves dividing a sequence of Chinese characters into individual words. Word segmentation is the process of recombinizing a continuous sequence of Chinese characters into a word sequence according to certain rules. Existing word segmentation methods can be divided into three main categories: string matching-based segmentation methods, understanding-based segmentation methods, and statistical segmentation methods.

[0055] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0056] Method Implementation Examples

[0057] The following embodiment illustrates an entity mining method provided in this application.

[0058] See Figure 1 The figure is a flowchart of an entity mining method provided in an embodiment of this application, as shown below. Figure 1 As shown, the method includes:

[0059] S101: Obtain a set of unstructured entity data.

[0060] In this step, the first step is to obtain a medical clinical data set, which is an unstructured clinical data set. The entities in this medical clinical data set do not exist in the medical domain dictionary, so entity mining is required.

[0061] In one feasible implementation, after obtaining the unstructured entity data set, the data set is preprocessed. The preprocessing operation includes uniformly converting illegal characters and full-width / half-width characters in Chinese and English within the data set. The following example illustrates the unstructured entity data set, which can be represented as: "Three years ago, the patient experienced dull pain in the chest and lower sternum without any obvious cause. There was no sense of impending death. The pain radiated outwards, forcing the patient to stop moving. The pain was persistent, lasting 10 minutes, and subsided after 30 minutes of rest. It was accompanied by profuse sweating and numbness in three fingers."

[0062] S102: Process the unstructured entity data set to obtain a first candidate entity data set, a second candidate entity data set, and a third candidate entity data set.

[0063] In this step, after obtaining the unstructured entity data set, it undergoes various processing methods to obtain multiple candidate entity data sets. These methods include word segmentation, word grouping, and sequence prediction.

[0064] In another feasible implementation, the preprocessed unstructured entity data set can also be subjected to diversified processing.

[0065] Specifically, firstly, an open-source Chinese word segmentation toolkit is used to add a medical domain dictionary. Then, the unstructured entity data set is processed using Chinese word segmentation, directly yielding the first candidate entity data set, as shown in Table 1. (Using the unstructured entity data set in the above example, this solution only shows the entity results needed for demonstration purposes and is not intended to limit this application.)

[0066] Table 1 Data Table of First Candidate Entities

[0067] chest organ 13 14 lower segment of bone organ 17 19 Dull pain Symptoms 20 21 pain Symptoms 28 29 Great Khan Symptoms 66 67 numbness Symptoms 72 74

[0068] The entity name includes the name of the entity obtained after processing the unstructured entity data set; the entity category includes the category corresponding to each entity; the index start position includes the position number of the first Chinese character of the entity in the unstructured entity data set; and the index end position includes the position number of the last Chinese character of the entity in the unstructured entity data set. (This description also applies to Tables 2, 3, and 4 below.)

[0069] Furthermore, the unstructured entity data set is segmented using regular expressions. This segmentation primarily handles information such as numbers, letters, and time values ​​within the text. The processing results are shown in Table 2. The segmented entity data set is then combined with the first candidate entity data set obtained in the previous steps to generate a second candidate entity data set, as shown in Table 3. In this step, entity recombination rules are primarily used to generate the second candidate entity data set. These rules are mainly used for concatenating discontinuous entities and recombinating numbers, letters, and entities. For example, in this scheme, the entity recognition results are "front chest / organ", "and / O", "back / organ", and "pain / symptom". After applying entity recombination rules, the result becomes "front chest and back pain / symptom", thus generating the second candidate entity data set.

[0070] Table 2 Word Segmentation Processing Table

[0071] 3 value 3 3 Year unit 5 5 10 value 49 50 min unit 51 53 30 value 57 58 minute unit 59 60 3 value 71 71

[0072] Table 3. Data Table of Second Candidate Entities

[0073] chest organ 13 14 Dull pain in the lower part of the bone position 17 21 pain Symptoms 28 29 Great Khan Symptoms 66 67 numbness in 3 fingers Symptoms 71 74

[0074] Furthermore, in this step, a trained deep learning sequence labeling model is used to predict sequences in the unstructured entity data set, outputting multiple candidate predicted entities and their corresponding probability values. This involves ranking the candidate predicted entities and determining the probability of each ranking result. Then, multiple candidate predicted entities with probability values ​​greater than a first threshold are selected as the entity data set obtained from the sequence prediction. This means filtering based on a set threshold; candidate predicted entities with probabilities less than or equal to the first threshold are discarded, while those greater than the first threshold are retained. The first threshold can be manually preset, such as setting it to 90%, without specific limitations here. Finally, based on the entity data set obtained from the sequence prediction and the first and second candidate entity data sets obtained in the previous steps, a third candidate entity data set is generated, as shown in Table 4. It should also be noted that in this step, entity recombination rules are also used to generate the third candidate entity data set. Thus, by combining the deep learning sequence labeling model, the recall rate of entity recognition is increased, further improving the accuracy of entity recognition.

[0075] Table 4. Data Table of Third Candidate Entities

[0076] chest organ 13 14 Dull pain in the lower sternum Symptoms 16 21 Near-death experience Symptoms 24 26 pain Symptoms 28 29 Great Khan Symptoms 66 67 numbness in 3 fingers Symptoms 71 74

[0077] S103: Based on entity mining rules, mine the first candidate entity dataset, the second candidate entity dataset, and the third candidate entity dataset to obtain entity mining results.

[0078] In this step, it is first determined whether there are identical entities in the first, second, and third candidate entity data sets. If they exist, the identical entities are retained; otherwise, the determination process ends. In other words, it checks whether the same entity exists in the first, second, and third candidate entity data sets. If it does, it is output directly. For example, if the first, second, and third candidate entity data sets all contain the entities "chest," "pain," and "excessive sweating," then the output should be "chest," "pain," and "excessive sweating."

[0079] Then, it is determined whether there is partial character overlap between different entities in the first, second, and third candidate entity datasets. If so, the entity with the longest character is retained using the longest matching principle; otherwise, the determination process ends. In other words, it checks whether there is partial character overlap among entities in the first, second, and third candidate entity datasets. If so, the entity with the longest character is retained using the longest matching principle. For example, in the first candidate entity dataset, "lower segment of bone" and "dull pain"; in the second candidate entity dataset, "lower segment of bone dull pain"; and in the third candidate entity dataset, "lower segment of sternum dull pain". This scheme uses the longest matching principle for entity recognition, and the result is "lower segment of sternum dull pain". Similarly, for "numbness of fingers" in the first candidate entity dataset and "numbness of three fingers" in the second and third candidate entity datasets, the entity recognition result is "numbness of three fingers" according to the longest matching principle. Finally, the entity mining results are constructed based on the retained identical entities and the entity with the longest character.

[0080] In some feasible implementations, if entities in the first and second candidate entity data sets intersect with entities in the third candidate entity data set (partial character intersection), the proportion of the number of intersecting characters to the total number of characters in the multiple entities (the intersection part is counted once) is compared with a preset threshold. If it is greater than the preset threshold, the set of intersecting entity strings is taken and concatenated into a new entity, and its entity category is selected from the type with the larger proportion of the intersecting strings to the original entity. If it is less than or equal to the preset threshold, the entities in the first and second candidate entity data sets are retained, and the entities in the third candidate entity data set are discarded.

[0081] In addition, the obtained entity mining results can be further filtered from the location information and entity category information of existing entities. If the re-word formation strategy is satisfied, the re-word formation strategy is to repeat the judgment process in step S103. The purpose is to search for whether there are any mining omissions. If so, a new candidate entity list is output until no new entities are generated, and then the entity mining process ends.

[0082] As can be seen, this optional solution mainly describes how to improve the accuracy of entity mining. Specifically, in this optional solution, this application first obtains an unstructured entity data set, then processes the unstructured entity data set to obtain a first candidate entity data set, a second candidate entity data set, and a third candidate entity data set, and finally mines the first candidate entity data set, the second candidate entity data set, and the third candidate entity data set based on entity mining rules to obtain entity mining results.

[0083] In summary, in this embodiment, for medical entities not existing in the dictionary, entity mining rules are used to mine the unstructured entity data set, thus obtaining entity mining results. Therefore, entity mining results can be ultimately extracted based on entity mining rules, avoiding manual entity mining and improving the accuracy of entity mining.

[0084] Device Examples

[0085] The following describes a physical excavation device provided in an embodiment of this application. The physical excavation device described below can be referred to in correspondence with the physical excavation method described above.

[0086] See Figure 2 The figure is a schematic diagram of the structure of a physical excavation device provided in an embodiment of this application, as shown below. Figure 2 As shown, the device includes:

[0087] Acquisition module 100 is used to acquire a set of unstructured entity data;

[0088] Processing module 200 is used to process the unstructured entity data set to obtain a first candidate entity data set, a second candidate entity data set, and a third candidate entity data set;

[0089] The mining module 300 is used to mine the first candidate entity data set, the second candidate entity data set, and the third candidate entity data set based on entity mining rules to obtain entity mining results.

[0090] Optionally, the processing module 200 includes:

[0091] The first processing subunit is used to perform word segmentation on the unstructured entity data set to obtain the first candidate entity data set.

[0092] The second processing subunit is used to perform word segmentation on the unstructured entity data set, and obtain the second candidate entity data set based on the word segmented entity data set and the first candidate entity data set.

[0093] The third processing subunit is used to perform sequence prediction on the unstructured entity data set, and to obtain the third candidate entity data set by combining the entity data set obtained from the sequence prediction with the first candidate entity data set and the second candidate entity data set.

[0094] Optional, also includes:

[0095] The acquisition module is used to acquire multiple candidate predicted entities and the probability values ​​corresponding to the multiple candidate predicted entities respectively;

[0096] The selection module is used to select multiple candidate predicted entities whose probability values ​​are greater than a first threshold as the entity data set obtained from the sequence prediction.

[0097] Optionally, the mining module 300 includes:

[0098] The first judgment subunit is used to determine whether there are identical entities in the first candidate entity data set, the second candidate entity data set, and the third candidate entity data set. If there are, the identical entities are retained.

[0099] The second judgment subunit is used to determine whether there is partial character overlap between different entities in the first candidate entity data set, the second candidate entity data set and the third candidate entity data set. If there is, the entity with the longest character is retained.

[0100] A composition module is used to construct the entity mining result based on the retained identical entities and / or the entity with the longest character.

[0101] Optional, also includes:

[0102] The conversion processing module is used to perform character conversion processing on the unstructured entity data set.

[0103] The physical excavation device provided in this application embodiment has the same beneficial effects as the physical excavation method provided in the above embodiments, and therefore will not be described again.

[0104] Electronic device examples

[0105] See Figure 3 This figure is a schematic diagram of an electronic device structure provided in an embodiment of this application, such as... Figure 3 As shown, it includes:

[0106] Memory 11 is used to store computer programs;

[0107] The processor 12 is configured to implement the steps of the entity mining method described in any of the above method embodiments when executing the computer program.

[0108] In this embodiment, the device can be an in-vehicle computer, a PC (Personal Computer), or a terminal device such as a smartphone, tablet computer, handheld computer, or portable computer.

[0109] The device may include a memory 11, a processor 12, and a bus 13. The memory 11 includes at least one type of readable storage medium, such as flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), magnetic storage, magnetic disk, optical disk, etc. In some embodiments, the memory 11 may be an internal storage unit of the device, such as the hard disk of the device. In other embodiments, the memory 11 may be an external storage device of the device, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, Flash Card, etc. Furthermore, the memory 11 may include both internal and external storage units of the device. The memory 11 can be used not only to store application software and various types of data installed on the device, such as program code for executing fault prediction methods, but also to temporarily store data that has been output or will be output.

[0110] In some embodiments, processor 12 may be a central processing unit (CPU), controller, microcontroller, microprocessor or other data processing chip, used to run program code stored in memory 11 or process data, such as program code for executing a fault prediction method.

[0111] This bus 13 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0112] Furthermore, the device may also include a network interface 14, which may optionally include a wired interface and / or a wireless interface (such as a Wi-Fi interface, a Bluetooth interface, etc.), typically used to establish communication connections between the device and other electronic devices.

[0113] Optionally, the device may also include a user interface 15, which may include a display, an input unit such as a keyboard, and optionally, a standard wired interface or a wireless interface.

[0114] Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen, etc. The display may also be appropriately referred to as a screen or display unit, used to display information processed in the device and to display a visual user interface.

[0115] Figure 3 Only devices with components 11-15 are shown; those skilled in the art will understand that... Figure 3 The structure shown does not constitute a limitation on the device and may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0116] Readable storage medium embodiments

[0117] This application also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the steps of the entity mining method described in any of the above method embodiments. The storage medium may include various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0118] It should be noted that the entity mining method provided by this invention can be used in the field of medical technology. The above is merely an example and does not limit the application area of ​​the entity mining method provided by this invention.

[0119] It should also be noted that the "first" and "second" in the names such as "first" and "second" (if they exist) mentioned in the embodiments of this application are only used as name identifiers and do not represent the first and second in order.

[0120] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0121] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0122] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0123] The above provides a detailed description of the entity mining method and apparatus provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of this application.

Claims

1. A method for entity mining, characterized in that, include: Obtain a collection of unstructured entity data; The unstructured entity data set is processed to obtain a first candidate entity data set, a second candidate entity data set, and a third candidate entity data set; The process of processing the unstructured entity data set to obtain a first candidate entity data set, a second candidate entity data set, and a third candidate entity data set includes: The unstructured entity data set is segmented to obtain the first candidate entity data set; The unstructured entity data set is segmented into words, and the second candidate entity data set is obtained by combining the segmented entity data set with the first candidate entity data set. Sequence prediction is performed on the unstructured entity data set, and the entity data set obtained by sequence prediction is combined with the first candidate entity data set and the second candidate entity data set to obtain the third candidate entity data set; Based on entity mining rules, the first candidate entity dataset, the second candidate entity dataset, and the third candidate entity dataset are mined to obtain entity mining results.

2. The method according to claim 1, characterized in that, After performing sequence prediction on the unstructured entity data set, the method further includes: Obtain multiple candidate predicted entities and their respective probability values; Multiple candidate predicted entities with probability values ​​greater than a first threshold are selected as the entity data set obtained from the sequence prediction.

3. The method according to claim 1, characterized in that, The entity mining results obtained by mining the first candidate entity dataset, the second candidate entity dataset, and the third candidate entity dataset based on entity mining rules include: Determine whether there are identical entities in the first candidate entity data set, the second candidate entity data set, and the third candidate entity data set. If there are, retain the identical entities. Determine whether there is partial character overlap between different entities in the first candidate entity data set, the second candidate entity data set, and the third candidate entity data set. If so, retain the entity with the longest character. The entity mining result is constructed based on the retained identical entities and / or the entity with the longest character.

4. The method according to claim 1, characterized in that, After obtaining the unstructured entity data set, the method further includes: The unstructured entity data set is subjected to character conversion processing.

5. A physical excavation device, characterized in that, include: The acquisition module is used to acquire a collection of unstructured entity data. The processing module is used to process the unstructured entity data set to obtain a first candidate entity data set, a second candidate entity data set, and a third candidate entity data set. The processing module includes: The first processing subunit is used to perform word segmentation on the unstructured entity data set to obtain the first candidate entity data set. The second processing subunit is used to perform word segmentation on the unstructured entity data set, and obtain the second candidate entity data set based on the word segmented entity data set and the first candidate entity data set. The third processing subunit is used to perform sequence prediction on the unstructured entity data set, and to obtain the third candidate entity data set by combining the entity data set obtained from the sequence prediction with the first candidate entity data set and the second candidate entity data set. The mining module is used to mine the first candidate entity data set, the second candidate entity data set, and the third candidate entity data set based on entity mining rules to obtain entity mining results.

6. The apparatus according to claim 5, characterized in that, Also includes: The acquisition module is used to acquire multiple candidate predicted entities and the probability values ​​corresponding to the multiple candidate predicted entities respectively; The selection module is used to select multiple candidate predicted entities whose probability values ​​are greater than a first threshold as the entity data set obtained from the sequence prediction.

7. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the entity mining method as described in any one of claims 1 to 4.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the entity mining method as described in any one of claims 1 to 4.

Images