Data processing method, device, electronic device and storage medium

Abstract

The present disclosure discloses a data processing method, apparatus, electronic device, and storage medium. The method comprises: obtaining an object to be processed; retrieving candidate objects that match the object to be processed from a set of objects including standard objects; and determining a target object corresponding to the object to be processed from the candidate objects using a recognition model. This technical solution improves the accuracy of identifying standard objects for the object to be processed by first retrieving the candidate objects and then reordering them using the recognition model.

Description

Technical Field

[0001] The present disclosure relates to the technical field of computer technology, and in particular to a data processing method, device, electronic device, and storage medium. Background Art

[0002] In the era of big data, data standardization is an indispensable task in all fields. Data standardization aims to find a corresponding standard representation for the various different representations of the same object. Taking clinical terminology in the medical field as an example, there are often hundreds or even thousands of different ways to write the same diagnosis, surgery, medication, examination, test, symptom, etc. Without clinical terminology standardization, it is difficult to conduct subsequent statistical analysis of relevant information such as patient medical records. In the process of clinical terminology standardization, the clinical names in case documents are usually matched with standardized clinical terms in a standard knowledge base for semantic similarity. However, due to the wide variety of ways different doctors express the same term in case documents, a single matching model is unlikely to achieve good results. Therefore, how to achieve a more effective data standardization process is one of the technical problems that technicians in related fields need to solve. Summary of the Invention

[0003] Embodiments of the present disclosure provide a data processing method, apparatus, electronic device, and computer-readable storage medium.

[0004] In a first aspect, an embodiment of the present disclosure provides a data processing method, which includes:

[0005] Get the object to be processed;

[0006] Retrieving a candidate object that matches the object to be processed from an object set including a standard object;

[0007] A recognition model is used to determine a target object corresponding to the object to be processed from the candidate objects.

[0008] Furthermore, retrieving candidate objects that match the object to be processed from an object set including standard objects includes:

[0009] Acquire a first object set; the first object set includes a first standard object;

[0010] The first object set is retrieved using the object to be processed to obtain a first candidate object that matches the object to be processed.

[0011] Furthermore, retrieving candidate objects that match the object to be processed from an object set including standard objects includes:

[0012] Acquire a second object set; the second object set includes an original object and a second standard object corresponding to the original object;

[0013] The second object set is retrieved using the object to be processed to obtain an original object that matches the object to be processed, and a second standard object corresponding to the original object is determined as a second candidate object.

[0014] Furthermore, determining a target object corresponding to the object to be processed from the candidate objects using a recognition model includes:

[0015] Processing the object to be processed and the current candidate object using a feature representation model in the recognition model to obtain correlation features between the object to be processed and the current candidate object;

[0016] The target object is determined using the correlation feature.

[0017] Furthermore, determining the target object by using the correlation feature includes:

[0018] Performing dimensionality reduction processing on the correlation features using a multi-layer perceptron in the recognition model;

[0019] Processing the reduced-dimensional correlation features using a normalization model in the recognition model to obtain a correlation between the candidate object and the object to be processed;

[0020] The target object is determined according to the correlation.

[0021] Furthermore, using the object to be processed to retrieve the first object set to obtain a first candidate object that matches the object to be processed includes:

[0022] Calculating the similarity between the object to be processed and the first standard object in the first object set;

[0023] The first candidate object is determined according to the similarity.

[0024] Furthermore, using the object to be processed to retrieve the second object set to obtain an original object that matches the object to be processed includes:

[0025] Calculating the similarity between the object to be processed and the original object;

[0026] The original object that matches the object to be processed is determined according to the similarity.

[0027] In a second aspect, an embodiment of the present invention provides a data processing device, comprising:

[0028] An acquisition module, configured to acquire an object to be processed;

[0029] A retrieval module is configured to retrieve a candidate object that matches the object to be processed from an object set including a standard object;

[0030] The determination module is configured to determine the target object corresponding to the object to be processed from the candidate objects by using a recognition model.

[0031] The functions can be implemented by hardware or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the functions.

[0032] In one possible design, the apparatus includes a memory and a processor. The memory is configured to store one or more computer instructions that enable the apparatus to perform the corresponding method, and the processor is configured to execute the computer instructions stored in the memory. The apparatus may also include a communication interface for communicating with other devices or a communication network.

[0033] In a third aspect, an embodiment of the present disclosure provides an electronic device comprising a memory and a processor; wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the method described in any one of the above aspects.

[0034] In a fourth aspect, an embodiment of the present disclosure provides a computer-readable storage medium for storing computer instructions used by any of the above-mentioned devices, which includes computer instructions involved in executing the method described in any of the above-mentioned aspects.

[0035] The technical solutions provided by the embodiments of the present disclosure may have the following beneficial effects:

[0036] The disclosed embodiment first retrieves multiple candidate objects from an object set that closely match (or are more closely correlated with) the object to be processed. For example, the similarity between the object to be processed and the standard object can be calculated, and multiple standard objects with high similarity can be identified as candidate objects. The recognition model is then used to identify the target object corresponding to the object to be processed from the multiple candidate objects. By first retrieving and then reordering the candidate objects using the recognition model, the disclosed embodiment improves the accuracy of identifying the standard object for the object to be processed.

[0037] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Other features, objectives and advantages of the present disclosure will become more apparent through the following detailed description of non-limiting embodiments in conjunction with the accompanying drawings. In the accompanying drawings:

[0039] Figure 1 A flowchart showing a data processing method according to an embodiment of the present disclosure;

[0040] Figure 2 A schematic diagram illustrating an application of an embodiment of the present disclosure in a clinical terminology standardization scenario is shown;

[0041] Figure 3 A schematic structural diagram of a recognition model according to an embodiment of the present disclosure is shown;

[0042] Figure 4 It is a structural diagram of an electronic device suitable for implementing a data processing method according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0043] Hereinafter, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily implement them. In addition, for the sake of clarity, parts not related to the description of the exemplary embodiments are omitted in the accompanying drawings.

[0044] In the present disclosure, it should be understood that terms such as "include" or "have" are intended to indicate the presence of features, numbers, steps, actions, components, parts, or combinations thereof disclosed in the present specification, and are not intended to exclude the possibility that one or more other features, numbers, steps, actions, components, parts, or combinations thereof exist or are added.

[0045] It should also be noted that, in the absence of conflict, the embodiments and features of the embodiments of the present disclosure may be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0046] The details of the embodiments of the present disclosure are described in detail below through specific examples.

[0047] Figure 1 Flowchart showing a data processing method according to an embodiment of the present disclosure. Figure 1 As shown, the data processing method includes the following steps:

[0048] In step S101, the object to be processed is obtained;

[0049] In step S102, a candidate object matching the object to be processed is retrieved from an object set including a standard object;

[0050] In step S103, a recognition model is used to determine a target object corresponding to the object to be processed from the candidate objects.

[0051] In this embodiment, the object to be processed may be an object to be standardized, such as a term name in a related field, a text including characters, words, sentences, etc., a table, an image, etc. The preset standard may be a standard established in a related field, and the standard object may be a well-known standardized object in the related field. Taking clinical terminology in the medical field as an example, the object to be processed may be a clinical name filled in by a doctor in a clinical medical record, and the standard object may be a standardized name of a clinical name in a recognized clinical knowledge base, such as a surgical name in the "ICD9-2017 Union Clinical Edition".

[0052] Under normal circumstances, the relevant fields can formulate multiple standard objects in the industry to form a standard object set. For the object to be processed, the target object corresponding to the object to be processed can be obtained by matching the object to be processed with the formulated standard object set. The above matching process can be to use the representation features of the object to be processed (for example, for objects of the term name class, it can be word-level features) to match the representation features of multiple standard objects. For example, the matching target object can be determined by calculating the similarity between the object to be processed and multiple standard objects. However, when the representation features of the object to be processed are significantly different from the representation features of the corresponding target object, the target object corresponding to the object to be processed cannot be accurately identified using the above method. Therefore, in the data processing method proposed in the embodiment of the present disclosure, a retrieval means is first used to determine multiple candidate objects that match the object to be processed from an object set including multiple standard objects. The multiple candidate objects can be a preset number of standard objects in the object set with a high correlation with the object to be processed; then the recognition model is used to determine the target object corresponding to the object to be processed from the multiple candidate objects, for example, the target object that best matches the object to be processed is determined from the multiple candidate objects.

[0053] The above-mentioned recognition model can be pre-trained. The training data can include sample objects and standard objects corresponding to the sample objects, which can be included in the object set. The standard objects corresponding to the sample objects in the training data can be obtained through manual annotation. Utilizing the annotation relationship between the sample objects and the standard objects corresponding to the sample objects, a recognition model can be trained to identify the target object corresponding to the object to be processed from multiple candidate objects.

[0054] The disclosed embodiment first retrieves multiple candidate objects from an object set that closely match (or are more closely correlated with) the object to be processed. For example, the similarity between the object to be processed and the standard object can be calculated, and multiple standard objects with high similarity can be identified as candidate objects. The recognition model is then used to identify the target object corresponding to the object to be processed from the multiple candidate objects. By first retrieving and then reordering the candidate objects using the recognition model, the disclosed embodiment improves the accuracy of identifying the standard object for the object to be processed.

[0055] In an optional implementation of this embodiment, step S102, i.e., the step of retrieving a candidate object matching the object to be processed from an object set including a standard object, further includes the following steps:

[0056] Acquire a first object set; the first object set includes a plurality of first standard objects;

[0057] The first object set is retrieved using the object to be processed to obtain a first candidate object that matches the object to be processed.

[0058] In this optional implementation, the first object set may be a standard object set developed in a related field. A first standard object with a high correlation with the object to be processed is retrieved from the first standard object set through a search method and determined as the first candidate object. In some embodiments, the correlation between the object to be processed and the first standard object in the first standard object set can be determined by calculating the similarity between the object to be processed and the first standard object in the first standard object set. In some embodiments, the first standard objects in the first standard object set can be sorted according to similarity, and the first standard objects ranked higher can be determined as the first candidate objects. In other embodiments, the first standard objects with a similarity higher than a preset value can be determined as the first candidate objects.

[0059] In an optional implementation of this embodiment, step S102, i.e., the step of retrieving a candidate object matching the object to be processed from an object set including a standard object, further includes the following steps:

[0060] Acquire a second object set; the second object set includes an original object and a second standard object corresponding to the original object;

[0061] The second object set is retrieved using the object to be processed to obtain an original object that matches the object to be processed, and a second standard object corresponding to the original object is determined as a second candidate object.

[0062] In this optional implementation, the original object may be a standardized object, and the second standard object may be an object that is standardized from the original object. The original objects in the second object set and the second standard objects corresponding to the original objects may be sample objects and the standard objects corresponding to the sample objects in part or all of the training data used to train the recognition model. Of course, it is understandable that the original objects in the second object set and the second standard objects corresponding to the original objects may not be sample objects and the standard objects corresponding to the sample objects in the training data. The original objects and the second standard objects corresponding to the original objects may be historical annotation data collected through various means, in which the standardized correspondence between the original objects and the second standard objects is annotated according to the standards formulated in the relevant field.

[0063] After obtaining the second object set, the original object that matches the object to be processed can be determined by searching the second object set, and then the second standard object corresponding to the matched original object can be determined as the second candidate object. According to the description above, the original object that matches the object to be processed can be one or more original objects that have a high correlation with the object to be processed. The correlation between the two can be determined by similarity. For details, please refer to the description in the above embodiment, which will not be repeated here. In this embodiment, since the object to be processed and the original object are both unstandardized objects, they may be closer in representation, and the second standard object corresponding to the original object is a known standardized object. Therefore, the object to be processed is matched with the original object, and then the second standard object corresponding to the matched original object is used as the second candidate object of the object to be processed. Compared with directly using the object to be processed to match the second standard object in the second object set, the accuracy is higher.

[0064] In some embodiments, the first object set may include some or all standardized objects developed in the relevant field; the second standard objects in the second object set may be standardized objects corresponding to the original objects in the first object set. Taking clinical terminology in the medical field as an example, the first object set may include some or all standardized surgical names specified in the "ICD9-2017 Union Medical College Clinical Edition".

[0065] It should be noted that, in some embodiments, the first object set and the second object set can be combined to obtain candidate objects for the object to be processed. That is, the first candidate object can be obtained by matching the object to be processed with the first standard object in the first object set, and the object to be processed can also be matched with the original object in the second object set, and then the second candidate object can be obtained by using the matched original object. Both the first candidate object and the second candidate object can be used as candidates for the object to be processed and input into the recognition model for re-ranking. In this way, candidate objects with a wider coverage can be obtained, which not only overcomes the problem of not being able to obtain candidate objects by searching the first object set alone when the representation features of the object to be processed and the standard object are significantly different, but also overcomes the problem of not being able to obtain candidate objects due to incomplete data in the second object set.

[0066] In an optional implementation of this embodiment, step S103, i.e., the step of determining a target object corresponding to the object to be processed from the candidate objects using a recognition model, further includes the following steps:

[0067] Processing the object to be processed and the current candidate object using a feature representation model in the recognition model to obtain correlation features between the object to be processed and the current candidate object;

[0068] The target object is determined using the correlation feature.

[0069] In this optional implementation, the recognition model may include a feature representation model for processing the object to be processed and the candidate objects to obtain correlation features corresponding to the object to be processed and the candidate objects. The correlation features can be used to characterize the correlation between the object to be processed and the candidate objects. The correlation features can be used to determine the correlation between the object to be processed and the candidate objects. For example, the candidate object with the highest correlation can be determined as the target object corresponding to the object to be processed.

[0070] In some embodiments, for term name objects, text objects, and the like, a language model can be used to extract semantic correlation features between the object to be processed and candidate objects. This semantic correlation feature can then be used to determine the correlation between the object to be processed and the candidate objects. Existing language models such as BERT, ESIM (Enhanced Sequential Inference Model), BiMPM (Bilateral Multi-perspective Matching), and MwAN (Multiway Attention Networks) can be used.

[0071] In an optional implementation of this embodiment, the step of determining the target object using the correlation feature further includes the following steps:

[0072] Performing dimensionality reduction processing on the correlation features using a multi-layer perceptron in the recognition model;

[0073] Processing the reduced-dimensional correlation features using a normalization model in the recognition model to obtain a correlation between the candidate object and the object to be processed;

[0074] The target object is determined according to the correlation.

[0075] In this optional implementation, the correlation features typically output by the feature representation model may be vector features of several hundred dimensions. In order to be able to normalize the correlation features using a normalization model, the correlation features can first be reduced in dimensionality using a multi-layer perceptron. For example, the correlation features can be mapped into 2-dimensional vector features, and then input into a normalization model such as a softmax function for normalization, so as to obtain a correlation score between the object to be processed and the candidate object. The higher the score, the higher the correlation between the object to be processed and the candidate object, and the more likely the candidate object is the target object corresponding to the object to be processed. For example, the candidate object with the highest correlation score can be determined as the target object of the object to be processed.

[0076] In an optional implementation of this embodiment, the step of using the object to be processed to search the first object set to obtain a first candidate object that matches the object to be processed further includes the following steps:

[0077] Calculating the similarity between the object to be processed and the first standard object in the first object set;

[0078] The first candidate object is determined according to the similarity.

[0079] In this optional implementation, during the search process, similarity can be calculated to determine whether the object to be processed matches the first standard object. After calculating the similarity between the object to be processed and each first standard object, a preset number of first standard objects with the highest similarity can be determined as first candidate objects, or first standard objects with similarity greater than a preset value can be determined as first candidate objects.

[0080] Taking the term name class as an example, the TF-IDF feature can be used to calculate the similarity between the object to be processed and each standard object in the standard object set. For example, the similarity calculation formula can be expressed as follows:

[0081]

[0082] Where q is the object to be processed, d is the current standard object, v(q) is the TF-IDF feature vector of the object to be processed q, and v(d) is the TF-IDF feature vector of the standard object d.

[0083] Of course, it is understandable that the above similarity calculation formula is only an example and can also be calculated by other methods.

[0084] In addition, when the object to be processed is another type of object, such as text, image, table, etc., the correlation between the object to be processed and the standard object can be determined using an existing correlation calculation method of the related type, and no specific limitation is made here.

[0085] In an optional implementation of this embodiment, the step of using the object to be processed to retrieve the second object set to obtain the original object that matches the object to be processed further includes the following steps:

[0086] Calculate the similarity between the object to be processed and the original object.

[0087] The original object that matches the object to be processed is determined according to the similarity.

[0088] In this optional implementation, similarity can be calculated to determine whether the object to be processed matches the original object. After calculating the similarity between the object to be processed and each original object, the second standard objects corresponding to a preset number of original objects with the highest similarity can be determined as the first candidate objects, or the second standard objects corresponding to original objects with similarity greater than a preset value can be determined as the first candidate objects.

[0089] The method for calculating the similarity between the object to be processed and the original object can refer to the method for calculating the similarity between the object to be processed and the first standard object in the above embodiment, which will not be repeated here.

[0090] Figure 2 FIG. 1 shows an application diagram of an embodiment of the present disclosure in a clinical terminology standardization scenario. Figure 2As shown, the coding file may include all standard words formulated by relevant departments for clinical medicine (corresponding to the first standard object in the embodiment of the present disclosure) and the coding ID assigned to the standard word, and the annotation file may include the collected surgical original words (corresponding to the original object in the embodiment of the present disclosure) and the standard word corresponding to the surgical original word (corresponding to the second standard object in the embodiment of the present disclosure). A retrieval index can be pre-established for the coding file and the annotation file. After receiving the surgical original word to be standardized (i.e., the surgical name filled in by the doctor in the case document) input by the user from the client, the first candidate standard word and the second candidate standard word (corresponding to the candidate object in the embodiment of the present disclosure) are retrieved from the coding file and the annotation file respectively; for each of the first candidate standard word and the second candidate standard word, the word sequence S consisting of the candidate standard word and the surgical original word is input into the recognition model for scoring, and the target standard word corresponding to the surgical original word to be standardized is output based on the score given by the recognition model, that is, the candidate answer with the final score is determined as the target standard word. The standardization of the surgical original word obtained by the embodiment of the present disclosure can not only be provided to hospitals, but also can be used as the original word standardization of health records or physical examination records, and can provide patients with standardized guidance lists, etc., to guide the standardization of patient descriptions of symptoms, etc.

[0091] The following uses the Lucene search tool and the Transformer framework as examples to explain in detail the process of generating candidate answers and the process of scoring candidate answers.

[0092] Lucene is a tool for full-text search and retrieval. Its default sorting method is based on TF-IDF and the vector space model, which allows it to easily and quickly find results that are similar in text to the search phrase (i.e., the original surgical term). The search process is as follows:

[0093] 1) Given a surgical term q and a search term d in the index (either a standard term in the encoding file or a surgical term in the annotation file), calculate the TF-IDF features for the surgical term q and the search term d, and represent them as v(q) and v(d) respectively.

[0094] 2) Calculate the similarity score of q and d through the vector space model:

[0095]

[0096] 3) Obtain multiple candidate answers from the encoded file and the annotated file according to the similarity scores.

[0097] After obtaining the candidate standard words through Lucene retrieval, the recognition model based on the Transformer framework scores the candidate standard words.

[0098] Figure 3FIG. 1 shows a schematic diagram of a structure of a recognition model according to an embodiment of the present disclosure. Figure 3 As shown in the figure, the Transformer framework includes a BERT encoder and a decoder composed of a multi-layer perceptron and a softmax function. During the scoring process, according to the specifications of the BERT model, the original surgical word to be processed and the current candidate answer are segmented and arranged into the form of "[CLS] original surgical word [SEP] standard word [SEP]" and input into the BERT encoder. That is, given the original surgical word and candidate answers According to the BERT specification, it is spliced into a sequence Input the sequence S to the BERT encoder, and take the output vector V at CL as the vector feature representation of S, V = BERT (S); input V to the multi-layer perceptron, which converts it into a 2D vector P, P = W T V; P is normalized through the Softmax operation, and the first dimension is taken as the probability Prob between 0 and 1. It should be noted that during the training process of the above recognition model, the model parameters can be optimized by minimizing the cross entropy loss function (based on Prob).

[0099] like Figure 3 As shown in the figure, the original surgical term is "vertebroplasty" and the standardized term is "percutaneous vertebroplasty." Therefore, the input to the BERT encoder is "[CLS] percutaneous vertebroplasty [SEP] vertebroplasty [SEP]." The BERT encoder's output, the vector feature representation at the position "[CLS]," is then input into a multi-layer perceptron to produce a 2D vector. This 2D vector is then normalized to a score between 0 and 1 after passing through the Softmax function.

[0100] The following are embodiments of the apparatus disclosed herein, which can be used to execute embodiments of the method disclosed herein.

[0101] According to an embodiment of the present disclosure, a data processing device can be implemented as part or all of an electronic device through software, hardware, or a combination of both. The data processing device includes:

[0102] An acquisition module, configured to acquire an object to be processed;

[0103] A retrieval module is configured to retrieve a candidate object that matches the object to be processed from an object set including a standard object;

[0104] The determination module is configured to determine the target object corresponding to the object to be processed from the candidate objects by using a recognition model.

[0105] In an optional implementation of this embodiment, the retrieval module includes:

[0106] A first acquisition submodule is configured to acquire a first object set; the first object set includes a first standard object;

[0107] The first retrieval submodule is configured to use the object to be processed to retrieve the first object set to obtain a first candidate object that matches the object to be processed.

[0108] In an optional implementation of this embodiment, the retrieval module includes:

[0109] A second acquisition submodule is configured to acquire a second object set; the second object set includes original objects and second standard objects corresponding to the original objects;

[0110] The second retrieval submodule is configured to use the object to be processed to retrieve the second object set, obtain an original object matching the object to be processed, and determine a second standard object corresponding to the original object as a second candidate object.

[0111] In an optional implementation of this embodiment, the determining module includes:

[0112] an identification submodule configured to process the object to be processed and the current candidate object using a feature representation model in the identification model to obtain correlation features between the object to be processed and the current candidate object;

[0113] The first determination submodule is configured to determine the target object using the correlation feature.

[0114] In an optional implementation of this embodiment, the first determining submodule includes:

[0115] A dimensionality reduction submodule is configured to perform dimensionality reduction processing on the correlation features using a multi-layer perceptron in the recognition model;

[0116] a processing submodule configured to process the correlation features after dimensionality reduction using a normalization model in the recognition model to obtain a correlation between the candidate object and the object to be processed;

[0117] The second determining submodule is configured to determine the target object according to the correlation.

[0118] In an optional implementation of this embodiment, the first retrieval submodule includes:

[0119] a first calculation submodule, configured to calculate the similarity between the object to be processed and the first standard object in the first object set;

[0120] The third determining submodule is configured to determine the first candidate object according to the similarity.

[0121] In an optional implementation of this embodiment, the second retrieval submodule includes:

[0122] A second calculation submodule is configured to calculate the similarity between the object to be processed and the original object;

[0123] The fourth determining submodule is configured to determine the original object that matches the object to be processed according to the similarity.

[0124] The data processing device in this embodiment corresponds to the above-mentioned data processing method. For specific details, please refer to the above-mentioned description of the data processing method, which will not be repeated here.

[0125] Figure 4 It is a structural diagram of an electronic device suitable for implementing the data processing method according to an embodiment of the present disclosure.

[0126] like Figure 4 As shown, the electronic device 400 includes a processing unit 401, which can be implemented as a processing unit such as a CPU, a GPU, an FPGA, an NPU, etc. The processing unit 401 can perform various processes in the embodiment of any of the above methods of the present disclosure according to a program stored in a read-only memory (ROM) 402 or a program loaded from a storage portion 408 into a random access memory (RAM) 403. In the RAM 403, various programs and data required for the operation of the electronic device 400 are also stored. The processing unit 401, the ROM 402, and the RAM 403 are connected to each other via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0127] The following components are connected to the I / O interface 405: an input section 406 including a keyboard, a mouse, and the like; an output section 407 including devices such as a cathode ray tube (CRT), a liquid crystal display (LCD), and a speaker; a storage section 408 including a hard disk; and a communication section 409 including a network interface card such as a LAN card or a modem. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the I / O interface 405 as needed. Removable media 411, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is installed in the drive 410 as needed, so that computer programs read therefrom can be installed into the storage section 408 as needed.

[0128] In particular, according to embodiments of the present disclosure, any of the methods described above with reference to the embodiments of the present disclosure may be implemented as a computer software program. For example, embodiments of the present disclosure include a computer program product comprising a computer program tangibly embodied on a computer-readable medium, the computer program comprising program code for executing any of the methods described in the embodiments of the present disclosure. In such embodiments, the computer program may be downloaded and installed from a network via the communication portion 409 and / or installed from the removable medium 411.

[0129] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the diagram or block diagram can represent a module, program segment or part of the code, and the module, program segment or part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, as well as the combination of boxes in the block diagram and / or flow chart, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or can be implemented using a combination of dedicated hardware and computer instructions.

[0130] The units or modules described in the embodiments of the present disclosure may be implemented in software or hardware. The units or modules described may also be provided in a processor, and the names of these units or modules do not, in certain circumstances, limit the units or modules themselves.

[0131] As another aspect, the present disclosure further provides a computer-readable storage medium. This computer-readable storage medium may be included in the apparatus described in the above embodiments, or may be a standalone computer-readable storage medium not incorporated into the apparatus. The computer-readable storage medium stores one or more programs, which are used by one or more processors to execute the methods described in the present disclosure.

[0132] The above description is merely a preferred embodiment of the present disclosure and an illustration of the technical principles employed. Those skilled in the art should understand that the scope of the invention herein is not limited to the technical solutions formed by the specific combination of the above-mentioned technical features, but also encompasses other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the inventive concept. For example, a technical solution formed by replacing the above-mentioned features with (but not limited to) technical features with similar functions disclosed in this disclosure.

Claims

1. A data processing method, wherein: include: Get the object to be processed; Retrieving candidate objects that match the object to be processed from an object set including standard objects, wherein the object set includes a first object set and a second object set, the candidate objects include a first candidate object and a second candidate object, the first object set is used to represent the standard object set, the second object set includes original objects and second standard objects corresponding to the original objects, the first candidate object is determined based on standardized objects in the first object set, and the second candidate object is determined based on the original object in the second object set; Determine a target object corresponding to the object to be processed from the candidate objects using a recognition model; Among them, the method also includes: using the feature representation model in the recognition model to process the object to be processed and the current candidate object to obtain the correlation characteristics of the object to be processed and the current candidate object, wherein the correlation characteristics are used to determine the correlation between the object to be processed and the current candidate object; using the multi-layer perceptron and normalization model in the recognition model to process the correlation characteristics to determine the target object.

2. The method according to claim 1, wherein Retrieving a candidate object that matches the object to be processed from an object set including standard objects includes: Acquire the first object set; the first object set includes a first standard object; The first object set is retrieved using the object to be processed to obtain a first candidate object that matches the object to be processed.

3. The method according to claim 1 or 2, wherein: Retrieving a candidate object that matches the object to be processed from an object set including standard objects includes: Acquire the second object set; retrieve the second object set using the object to be processed, obtain the original object that matches the object to be processed, and determine the second standard object corresponding to the original object as the second candidate object.

4. The method according to claim 1, wherein Processing the correlation features using a multi-layer perceptron and a normalization model in the recognition model to determine the target object includes: Performing dimensionality reduction processing on the correlation features using the multi-layer perceptron; Processing the reduced-dimensional correlation features using the normalization model to obtain the correlation between the candidate object and the object to be processed; The target object is determined according to the correlation.

5. The method according to claim 2, wherein: Retrieving the first object set using the object to be processed to obtain a first candidate object that matches the object to be processed includes: Calculating the similarity between the object to be processed and the first standard object in the first object set; The first candidate object is determined according to the similarity.

6. The method according to claim 3, wherein: Retrieving the second object set using the object to be processed to obtain an original object that matches the object to be processed includes: Calculating the similarity between the object to be processed and the original object; The original object that matches the object to be processed is determined according to the similarity.

7. A data processing device, wherein: include: An acquisition module, configured to acquire an object to be processed; a retrieval module configured to retrieve candidate objects matching the object to be processed from an object set including standard objects, wherein the object set includes a first object set and a second object set, the candidate objects include a first candidate object and a second candidate object, the first object set is used to represent the standard object set, the second object set includes an original object and a second standard object corresponding to the original object, the first candidate object is determined based on a standardized object in the first object set, and the second candidate object is determined based on the original object in the second object set; a determination module configured to determine a target object corresponding to the object to be processed from the candidate objects using a recognition model; Among them, the determination module is also used to: use the feature representation model in the recognition model to process the object to be processed and the current candidate object to obtain the correlation characteristics of the object to be processed and the current candidate object, wherein the correlation characteristics are used to determine the correlation between the object to be processed and the current candidate object; use the multi-layer perceptron and normalization model in the recognition model to process the correlation characteristics to determine the target object.

8. An electronic device, wherein: comprising a memory and a processor; wherein, The memory is configured to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer instructions stored thereon, wherein: When the computer instructions are executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images