Text matching method, training method, device, equipment and medium

Abstract

The present disclosure provides a text matching method, which can be applied to the fields of artificial intelligence technology and text processing technology. The text matching method includes: obtaining the text to be matched and the editing attribute information corresponding to the text to be matched; generating first encoded information based on the semantic feature information of the text to be matched and the target text; generating second encoded information based on the editing attribute information corresponding to the text to be matched; and inputting the first encoded information and the second encoded information into a text matching model to output the matching result of the text to be matched and the target text. The present disclosure also provides a training method, device, equipment, medium and program product for the text matching model.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, specifically to the field of text processing technology, and more specifically to a text matching method, a training method, an apparatus, a device, a medium, and a program product. Background Art

[0002] During the process of text entry, users will enter the names of subjects according to their own operation experience or work habits. After different users enter text, they will summarize the text they have entered respectively to form a summarized text.

[0003] In the process of implementing the concept of the present disclosure, the inventors found that there are at least the following technical problems in the related art: there will be a situation where the entered text is different from the standard text, so relevant personnel need to match the entered text with the standard text, resulting in low matching accuracy and matching efficiency. Summary of the Invention

[0004] In view of the above problems, the present disclosure provides a text matching method, a training method, an apparatus, a device, a medium, and a program product.

[0005] According to a first aspect of the present disclosure, there is provided a text matching method, including:

[0006] Obtain the text to be matched and the editing attribute information corresponding to the text to be matched;

[0007] Generate first coding information based on the semantic feature information of the text to be matched and the target text;

[0008] Generate second coding information based on the editing attribute information corresponding to the text to be matched;

[0009] Input the first coding information and the second coding information into a text matching model, and output the matching result of the text to be matched and the target text.

[0010] According to an embodiment of the present disclosure, the above target text includes multiple;

[0011] The above text matching method further includes:

[0012] Determine the target text that matches the text to be matched based on the matching results of the text to be matched and each of the above target texts.

[0013] According to an embodiment of the present disclosure, generating first coding information based on the semantic feature information of the text to be matched and the target text includes:

[0014] Encode the text to be matched to generate text coding information to be matched;

[0015] Encode the above target text to generate target text encoding information;

[0016] Perform semantic fusion encoding on the above text encoding information to be matched and the above target text encoding information to generate the above first encoding information.

[0017] According to an embodiment of the present disclosure, the editing attribute information corresponding to the above text to be matched includes at least one of the following:

[0018] The editing time information for the above text to be matched, the user name information of the user who edits the above text to be matched, the user affiliation information of the user who edits the above text to be matched, and the user permission level information of the user who edits the above text to be matched.

[0019] According to an embodiment of the present disclosure, the editing attribute information corresponding to the above text to be matched includes multiple items;

[0020] Generating second encoding information based on the editing attribute information corresponding to the above text to be matched includes:

[0021] Encode each of the multiple editing attribute information items corresponding to the above text to be matched to generate multiple editing attribute encoding information;

[0022] Perform fusion encoding on the multiple above editing attribute encoding information to generate the above second encoding information.

[0023] A second aspect of the present disclosure provides a method for training a text matching model, including:

[0024] Determine a sample text to be matched and sample editing attribute information corresponding to the above sample text to be matched;

[0025] Generate first sample encoding information based on the semantic feature information of the above sample text to be matched and the sample target text, wherein the above sample text to be matched is associated with the sample target text;

[0026] Generate second sample encoding information based on the sample editing attribute information corresponding to the above sample text to be matched;

[0027] Use the above first sample encoding information and the above second sample encoding information to train an untrained initial text matching model to obtain a text matching model for use in the above method.

[0028] According to an embodiment of the present disclosure, the method for training the above text matching model further includes:

[0029] Obtain an initial sample text to be matched and the number of editing times corresponding to the above initial sample text to be matched;

[0030] Obtain a sample target text, where the sample target text is associated with the initial sample text to be matched;

[0031] Based on the semantic similarity between the initial sample text to be matched and the sample target text, and the number of editing times corresponding to the initial sample text to be matched, determine the detection result of the initial sample text to be matched;

[0032] Based on the detection result of the initial sample text to be matched, determine the sample text to be matched.

[0033] The third aspect of the present disclosure provides a text matching device, including:

[0034] An acquisition module for acquiring the text to be matched and the editing attribute information corresponding to the text to be matched;

[0035] A first encoding module for generating first encoding information based on the semantic feature information of the text to be matched and the target text;

[0036] A second encoding module for generating second encoding information based on the editing attribute information corresponding to the text to be matched;

[0037] A text matching module for inputting the first encoding information and the second encoding information into a text matching model and outputting the matching result between the text to be matched and the target text.

[0038] The fourth aspect of the present disclosure provides a training device for a text matching model, including:

[0039] A determination module for determining the sample text to be matched and the sample editing attribute information corresponding to the sample text to be matched;

[0040] A first sample encoding module for generating first sample encoding information based on the semantic feature information of the sample text to be matched and the sample target text, where the sample text to be matched is associated with the sample target text;

[0041] A second sample encoding module for generating second sample encoding information based on the sample editing attribute information corresponding to the sample text to be matched;

[0042] A training module for training an untrained initial text matching model using the first sample encoding information and the second sample encoding information to obtain the text matching model used in the above method.

[0043] The fifth aspect of the present disclosure provides an electronic device, including: one or more processors; a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors are caused to execute the above text matching method or the training method of the above text matching model.

[0044] The sixth aspect of the present disclosure further provides a computer-readable storage medium, on which executable instructions are stored, and when the instructions are executed by a processor, the processor is caused to execute the above text matching method or the training method of the above text matching model.

[0045] The seventh aspect of the present disclosure further provides a computer program product, including a computer program, and when the computer program is executed by a processor, the above text matching method is implemented or the training method of the above text matching model is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Through the following description of the embodiments of the present disclosure with reference to the drawings, the above content and other objects, features and advantages of the present disclosure will become clearer. In the drawings:

[0047] Figure 1 Schematically shows an application scenario diagram of the text matching method and device according to an embodiment of the present disclosure;

[0048] Figure 2 Schematically shows a flowchart of the text matching method according to an embodiment of the present disclosure;

[0049] Figure 3 Schematically shows a flowchart of generating first coding information according to an embodiment of the present disclosure;

[0050] Figure 4 Schematically shows an application scenario diagram of the text matching method according to an embodiment of the present disclosure;

[0051] Figure 5 Schematically shows a flowchart of the training method of the text matching model according to an embodiment of the present disclosure;

[0052] Figure 6 Schematically shows an application scenario diagram of determining a sample text to be matched according to an embodiment of the present disclosure;

[0053] Figure 7 Schematically shows a structural block diagram of the text matching device according to an embodiment of the present disclosure;

[0054] Figure 8 Schematically shows a structural block diagram of the training device of the text matching model according to an embodiment of the present disclosure; and

[0055] Figure 9A block diagram of an electronic device suitable for implementing a text matching method and a training method of a text matching model according to an embodiment of the present disclosure is schematically shown. Detailed implementation manners

[0056] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present disclosure. In the following detailed description, for the sake of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, obviously, one or more embodiments may be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present disclosure.

[0057] The terms used herein are merely for describing specific embodiments and are not intended to limit the present disclosure. The terms "including", "comprising", etc. used herein indicate the presence of the described features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0058] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those of ordinary skill in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0059] In the case of using expressions such as "at least one of A, B, and C, etc.", generally, it should be interpreted according to the meaning commonly understood by those of ordinary skill in the art (for example, "a system having at least one of A, B, and C" should include, but is not limited to, a system having only A, only B, only C, having A and B, having A and C, having B and C, and / or having A, B, and C, etc.).

[0060] During the process of text entry, the user will enter the name of the subject according to their own operation experience or work habits. After different users enter the text, they will summarize the text they have entered respectively to facilitate the formation of the summarized text.

[0061] In the process of implementing the concept of the present disclosure, the inventors found that there are at least the following technical problems in the related art: there may be a difference between the entered text and the standard text, so relevant personnel need to match the entered text with the standard text, resulting in low matching accuracy and matching efficiency.

[0062] The present disclosure provides a text matching method, a training method, a device, a device, a medium, and a program product. The text matching method includes: obtaining a text to be matched and editing attribute information corresponding to the text to be matched; generating first encoded information based on semantic feature information of the text to be matched and a target text; generating second encoded information based on the editing attribute information corresponding to the text to be matched; and inputting the first encoded information and the second encoded information into a text matching model to output a matching result of the text to be matched and the target text.

[0063] According to an embodiment of the present disclosure, generating first encoded information based on semantic feature information of the text to be matched and a target text can enable the first encoded information to fuse the semantic feature information of the matching subject and the target text. Generating second encoded information based on the editing attribute information corresponding to the text to be matched can enable the second encoded information to include feature information of one or more editing attribute information. Inputting the first encoded information and the second encoded information into a text matching model can fully fuse the semantic feature information of the matching subject and the target text, and at the same time comprehensively consider the feature information of the editing attribute information corresponding to the text to be matched, thereby improving the accuracy of the matching result of the text to be matched and the target text. Moreover, using the text matching method to determine the matching result of the text to be matched and the target text can shorten the time for determining the matching result compared with manually matching the text to be matched and the target text, thereby improving the matching efficiency.

[0064] In the technical solution of the present disclosure, the acquisition, storage, and application of the user's personal information involved all comply with the provisions of relevant laws and regulations, take necessary confidentiality measures, and do not violate public order and good customs.

[0065] Figure 1 Schematically shows an application scenario diagram of the text matching method and device according to an embodiment of the present disclosure.

[0066] As Figure 1 shown, the application scenario 100 according to this embodiment may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 is used to provide a medium for a communication link between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.

[0067] Users can use the terminal devices 101, 102, 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications may be installed on the terminal devices 101, 102, 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only for example).

[0068] The terminal devices 101, 102, and 103 can be various electronic devices with a display screen and supporting web browsing, including but not limited to smartphones, tablet computers, laptop computers, desktop computers, and the like.

[0069] The server 105 can be a server that provides various services, such as a background management server (only for example) that supports the websites browsed by users using the terminal devices 101, 102, and 103. The background management server can analyze and process data such as user requests received, and feedback the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices.

[0070] It should be noted that the text matching method provided by the embodiments of the present disclosure can generally be executed by the server 105. Correspondingly, the text matching device provided by the embodiments of the present disclosure can generally be set in the server 105. The text matching method provided by the embodiments of the present disclosure can also be executed by a server or a server cluster different from the server 105 and capable of communicating with the terminal devices 101, 102, 103 and / or the server 105. Correspondingly, the text matching device provided by the embodiments of the present disclosure can also be set in a server or a server cluster different from the server 105 and capable of communicating with the terminal devices 101, 102, 103 and / or the server 105.

[0071] It should be understood that Figure 1 the numbers of the terminal devices, the network, and the server in

[0072] are merely illustrative. According to the implementation requirements, there can be any number of terminal devices, networks, and servers. Figure 1 The following will be based on Figures 2 to 4 the described scenario, and will describe in detail the text matching method of the disclosed embodiments through

[0073] Figure 2 FIG. schematically shows a flowchart of the text matching method according to an embodiment of the present disclosure.

[0074] As Figure 2 shown, the text matching method of this embodiment includes operation S210 to operation S240.

[0075] In operation S210, obtain the text to be matched and the editing attribute information corresponding to the text to be matched.

[0076] In operation S220, generate first coding information based on the semantic feature information of the text to be matched and the target text.

[0077] In operation S230, generate second coding information based on the editing attribute information corresponding to the text to be matched.

[0078] In operation S240, the first encoded information and the second encoded information are input into a text matching model, and a matching result between the text to be matched and the target text is output.

[0079] According to an embodiment of the present disclosure, the text to be matched may include items entered in files such as reports, catalogs, etc. The editing attribute information corresponding to the text to be matched may include editing information generated for entering or modifying the text to be matched, such as the time information for entering the text to be matched, the user information for entering the text to be matched, etc. The target text may include standard items specified in relevant regulations or laws. For example, when the file is a financial statement, the text to be matched may include financial items entered in the financial statement, the target text may include standard financial items specified in relevant financial regulations or norms, and the editing attribute information corresponding to the text to be matched may include the user information for entering the financial items, etc.

[0080] It should be understood that the text to be matched may be an item that has not undergone a modification operation, or an item formed after one or more modification operations.

[0081] According to an embodiment of the present disclosure, the matching result may indicate whether the text to be matched and the target text match. For example, the matching result may include a predicted probability value. It may be set that when the predicted probability value is greater than or equal to a predetermined threshold, the text to be matched and the target text match.

[0082] According to an embodiment of the present disclosure, generating the first encoded information based on the semantic feature information of the text to be matched and the target text can enable the first encoded information to fully integrate the semantic feature information of the text to be matched and the target text, providing a basis for subsequent matching of the text to be matched and the target text. The second encoded information generated based on the editing attribute information corresponding to the text to be matched can enable the editing information generated during the process of editing the text to be matched to be integrated into the second encoded information. Therefore, inputting the first encoded information and the second encoded information into the text matching model can fully consider the semantic feature information of the text to be matched and the target text, as well as the editing information generated during the process of entering the text to be matched, and output a matching result between the text to be matched and the target text to indicate whether the text to be matched and the target text match, thereby improving the accuracy of the matching.

[0083] Figure 3 Schematically shows a flowchart of generating the first encoded information according to an embodiment of the present disclosure.

[0084] As Figure 3 shown, in operation S220, generating the first encoded information based on the semantic feature information of the text to be matched and the target text may include operations S310 to S330.

[0085] In operation S310, the text to be matched is encoded to generate the encoded information of the text to be matched.

[0086] In operation S320, the target text is encoded to generate the encoded information of the target text.

[0087] In operation S330, semantic fusion encoding is performed on the encoded information of the text to be matched and the encoded information of the target text to generate the first encoded information.

[0088] According to an embodiment of the present disclosure, a network model based on a neural network can be used to perform semantic fusion encoding on the encoded information of the text to be matched and the encoded information of the target text. For example, a word2vec model, a recurrent neural network model, etc. can be used to perform semantic fusion encoding on the encoded information of the text to be matched and the encoded information of the target text.

[0089] According to an embodiment of the present disclosure, generating the first encoded information based on the semantic feature information of the text to be matched and the target text can enable the first encoded information to fully integrate the semantic feature information of the text to be matched and the target text, so as to avoid losing the semantic feature information of the text to be matched and the target text, and provide a basis for subsequent matching of the text to be matched and the target text.

[0090] According to an embodiment of the present disclosure, the editing attribute information corresponding to the text to be matched may include at least one of the following:

[0091] The editing time information for the text to be matched, the user name information of the user who edits the text to be matched, the user affiliation information of the user who edits the text to be matched, and the user permission level information of the user who edits the text to be matched.

[0092] According to an embodiment of the present disclosure, the editing time information for the text to be matched may be the time point when the text to be matched is edited, such as the time point when the text to be matched is initially entered, the time point when the text to be matched is modified each time, but is not limited thereto, and may also be the time point when the text to be matched is last modified, etc. Those skilled in the art can select the editing time information for the text to be matched according to actual needs.

[0093] According to an embodiment of the present disclosure, the user name information of the user who edits the text to be matched may include the identification information of the user who edits the text to be matched, such as the user's name, the user's account name, etc. The user affiliation information of the user who edits the text to be matched may include the identification information of the institution or organization to which the user who edits the text to be matched belongs, such as the name of the institution or organization, but is not limited thereto, and may also include the account information of the institution or organization, etc.

[0094] According to an embodiment of the present disclosure, the user permission level information for editing the text to be matched may include the editing permission level information of the organization or institution to which the user belongs for the text to be matched. Editing permissions may include, for example, input permissions, modification permissions, etc. The editing permission level may be determined according to the level of the institution or organization to which the user belongs. Different editing permission levels may be sorted from high to low. For example, it may be determined that XX Branch has the input permission level, XY Regional Company has the primary modification permission level. For example, XY Regional Company can modify the subjects entered by lower-level branches; XXY Head Office has the advanced modification permission level, that is, it can modify the subjects entered by XX Branch or the subjects modified by XY Regional Company. That is, the sorting of the editing permission levels from high to low may be the advanced modification permission level, the primary modification permission level, and the input permission level.

[0095] It should be noted that the probability of the subject obtained by editing through an organization or institution with a high editing permission level matching the standard subject is relatively high. The second coding information generated based on the user permission level information for editing the text to be matched may incorporate the editing permission level of the institution or organization to which the user belongs. Therefore, by using the text matching model to process the first coding information and the second coding information, the matching result of the text to be matched and the target text can improve the matching accuracy.

[0096] According to an embodiment of the present disclosure, the editing attribute information corresponding to the text to be matched may include multiple items.

[0097] In operation S230, generating the second coding information based on the editing attribute information corresponding to the text to be matched may include:

[0098] Encoding each of the multiple editing attribute information items corresponding to the text to be matched separately to generate multiple editing attribute coding information; and performing a fusion coding on the multiple editing attribute coding information to generate the second coding information.

[0099] According to an embodiment of the present disclosure, in the case where the editing attribute information includes multiple items, the second coding information generated based on the multiple editing attribute information may incorporate the multiple editing attribute information, so as to comprehensively consider the editing information generated for inputting or modifying the text to be matched, and avoid affecting the accuracy of the matching result of the subsequent text to be matched and the target text due to the loss of editing attribute information.

[0100] Figure 4 Schematically shows an application scenario diagram of the text matching method according to an embodiment of the present disclosure.

[0101] Such as Figure 4As shown, the text to be matched can be "derivative financial assets", and the target text can be "derivative financial assets". After encoding the text to be matched, the encoded information 411 of the text to be matched can be generated. After encoding the target text, the encoded information 421 of the target text can be generated. Using the first fusion encoding layer 441 to perform semantic fusion encoding on the encoded information 411 of the text to be matched and the encoded information 421 of the target text, the first encoded information 451 can be generated. In this embodiment, the first fusion encoding layer 441 can be a neural network layer constructed based on word2vec, and the first encoded information 451 output by the first fusion encoding layer 441 can be the matrix vector A.

[0102] The editing attribute information corresponding to the text to be matched may include: the editing time information for the text to be matched, the user name information for editing the text to be matched, the user affiliation information for editing the text to be matched, and the user permission level information for editing the text to be matched. The editing time information for the text to be matched may be the time information for editing and generating the text to be matched "derivative financial assets". The user name information may be the account name information of the user. The user affiliation information may be the account name information of the organization to which the user belongs. The user permission level information may be the editing permission level information of the organization or institution to which the user belongs for the text to be matched. The editing permission level may be, for example, the high-level modification permission level.

[0103] Encoding each item in the editing attribute information corresponding to the text to be matched can generate the editing attribute encoded information 431, 432, 433, 434. Using the second fusion encoding layer 442 to perform fusion encoding on the editing attribute encoded information 431, 432, 433, 434, the second encoded information 452 can be generated. In this embodiment, the second fusion encoding layer 442 can be a neural network layer constructed based on word2vec, and the second encoded information 452 output by the second fusion encoding layer 442 can be the matrix vector B.

[0104] Inputting the first encoded information 451 and the second encoded information 452 into the text matching model 460 can output the matching result 470 of the text to be matched and the target text. In this embodiment, the text matching model 460 can be constructed based on the weight parameter matrix. After the first encoded information 451 and the second encoded information 452 are input into the text matching model 460, the matching result 470 can be obtained based on formula (1).

[0105] P = A * M * B T ; (1)

[0106] In Formula (1), the first encoded information 451 can be represented as a matrix vector A, the second encoded information 452 can be represented as a matrix vector B, the weight parameter matrix can be represented as M, and the matching result 470 can be represented as P. P can be a predicted probability value, where 0 < P ≤ 1. The predetermined threshold can be set to 0.7. When P ≥ 0.7, the matching result 470 can characterize that the text to be matched matches the target text.

[0107] According to an embodiment of the present disclosure, by performing semantic fusion encoding on the text encoding information to be matched and the target text encoding information, the generated first encoded information can fuse the semantic feature information of the text to be matched and the target text. Based on the editing attribute information corresponding to the text to be matched, the second encoded information can be generated, and the second encoded information can contain the feature information of multiple editing attribute information. Inputting the first encoded information and the second encoded information into the text matching model can fully fuse the semantic feature information of the matching subject and the target text, and at the same time comprehensively consider the feature information of the editing attribute information corresponding to the text to be matched, thereby improving the accuracy of the matching result between the text to be matched and the target text.

[0108] According to an embodiment of the present disclosure, there can be multiple target texts.

[0109] The text matching method may further include:

[0110] Based on the matching results between the text to be matched and each target text, determine the target text that matches the text to be matched.

[0111] According to an embodiment of the present disclosure, the matching result between the text to be matched and the target text can be the predicted probability value that the text to be matched matches the target text. Therefore, when there are multiple target texts, the matching results between the text to be matched and each target text can be compared, that is, multiple predicted probability values can be compared, and it is determined that the target text corresponding to the matching result with the largest predicted probability value matches the text to be matched, so that the target text that matches the text to be matched can be quickly determined from multiple target texts to improve the efficiency of text matching.

[0112] Figure 5 Schematically shows a flowchart of a training method for a text matching model according to an embodiment of the present disclosure.

[0113] As Figure 5 shown, the training method of the text matching model in this embodiment includes operation S510 to operation S540.

[0114] In operation S510, determine the sample text to be matched and the sample editing attribute information corresponding to the sample text to be matched.

[0115] In operation S520, based on the semantic feature information of the sample text to be matched and the sample target text, a first sample encoding information is generated, where the sample text to be matched is associated with the sample target text.

[0116] In operation S530, based on the sample editing attribute information corresponding to the sample text to be matched, a second sample encoding information is generated.

[0117] In operation S540, the untrained initial text matching model is trained using the first sample encoding information and the second sample encoding information to obtain a text matching model for use in the above text matching method.

[0118] According to an embodiment of the present disclosure, the sample text to be matched may include the subjects entered into the sample file, and the sample file may include reports, catalogs, etc. The sample editing attribute information corresponding to the sample text to be matched may include the sample editing information generated for entering or modifying the sample text to be matched, such as the time information for entering the sample text to be matched, the user information for entering the sample text to be matched, etc. The sample target text may include the standard subjects specified in relevant norms or regulations. The sample target text may be the target text that matches the sample text to be matched. For example, in the case where the sample text is a financial statement, the sample text to be matched may include the financial subjects entered into the financial statement, the sample target text may be the standard financial subjects specified in relevant financial regulations or norms, and the sample editing attribute information corresponding to the sample text to be matched may include the user information for editing the financial subjects, etc.

[0119] It should be understood that the sample text to be matched may be a subject without any modification operations, or a subject formed after one or more modification operations.

[0120] According to an embodiment of the present disclosure, the sample text to be matched may be encoded to generate sample text to be matched encoding information, and the sample target text may be encoded to generate sample target text encoding information. Semantic fusion encoding is performed on the sample text to be matched encoding information and the sample target text encoding information to generate the first sample encoding information.

[0121] According to an embodiment of the present disclosure, a network model based on a neural network may be used to perform semantic fusion encoding on the sample text to be matched encoding information and the sample target text encoding information. For example, a word2vec model, a recurrent neural network model, etc. may be used to perform semantic fusion encoding on the sample text to be matched encoding information and the sample target text encoding information.

[0122] According to an embodiment of the present disclosure, generating first sample coding information based on the semantic feature information of the sample text to be matched and the sample target text can enable the first sample coding information to fully integrate the semantic feature information of the sample text to be matched and the sample target text, so as to avoid losing the semantic feature information of the sample text to be matched and the sample target text.

[0123] According to an embodiment of the present disclosure, the sample editing attribute information corresponding to the sample text to be matched may include at least one of the following:

[0124] The editing time information for the sample text to be matched, the user name information of the user who edits the sample text to be matched, the user affiliation information of the user who edits the sample text to be matched, and the user permission level information of the user who edits the sample text to be matched.

[0125] According to an embodiment of the present disclosure, the editing time information for the sample text to be matched may be the time point when the sample text to be matched is edited, such as the time point when the sample text to be matched is initially entered, the time point when the sample text to be matched is modified each time, but is not limited thereto, and may also be the time point when the sample text to be matched is last modified, etc. Those skilled in the art can select the editing time information for the sample text to be matched according to actual needs.

[0126] According to an embodiment of the present disclosure, the user name information of the user who edits the sample text to be matched may include the identification information of the user who edits the sample text to be matched, such as the name of the user, the account name of the user, etc. The user affiliation information of the user who edits the sample text to be matched may include the identification information of the institution or organization to which the user who edits the sample text to be matched belongs, such as the name of the institution or organization, but is not limited thereto, and may also include the account information of the institution or organization, etc.

[0127] According to an embodiment of the present disclosure, the user permission level information of the user who edits the sample text to be matched may include the editing permission level information of the organization or institution to which the user belongs for the sample text to be matched. The editing permissions may include, for example, input permissions, modification permissions, etc. The editing permission level may be determined according to the level of the institution or organization to which the user belongs. Different editing permission levels can be sorted from high to low. For example, it can be determined that XX branch company has the input permission level, XY regional company has the primary modification permission level, that is, XY regional company can modify the subjects entered by lower-level branch companies; XXY head office has the high-level modification permission level, that is, it can modify the subjects entered by XX branch company or the subjects modified by XY regional company. That is, the sorting of the editing permission levels from high to low can be the high-level modification permission level, the primary modification permission level, and the input permission level.

[0128] It should be noted that the probability of the subject obtained by editing through an organization or institution with a high editing permission level matching the standard text is relatively high. The second sample coding information generated based on the user permission level information of the text to be matched in the editing sample can integrate the editing permission level of the organization or institution to which the user belongs.

[0129] According to an embodiment of the present disclosure, an untrained initial text matching model can be constructed based on a weight parameter matrix. The first sample coding information and the second sample coding information can be input into the initial text matching model constructed based on the weight parameter matrix, and a predicted matching result can be output. The weight parameters of the weight parameter matrix can be iteratively adjusted based on the predicted matching result and the matching result between the text to be matched in the sample and the target text, so as to obtain a trained text matching model.

[0130] According to an embodiment of the present disclosure, the first sample coding information can be generated based on the semantic feature information of the text to be matched in the sample and the sample target text, so that the first sample coding information can fully integrate the semantic feature information of the text to be matched in the sample and the sample target text, so as to avoid losing the semantic feature information of the text to be matched in the sample and the sample target text. The second sample coding information generated based on the sample editing attribute information corresponding to the text to be matched in the sample can integrate the editing information generated during the process of editing the text to be matched in the sample into the second sample coding information. Therefore, by using the first sample coding information and the second sample coding information to train the initial text matching model, the initial text matching model can better learn the semantic feature information of the text to be matched in the sample and the sample target text, as well as the editing information generated during the process of editing the text to be matched in the sample, so that the trained text matching model can be used in the above text matching method to improve the accuracy of the matching result of the text to be matched and the target matching subject.

[0131] According to an embodiment of the present disclosure, the training method of the text matching model may further include:

[0132] Obtain the initial sample text to be matched and the number of editing times corresponding to the initial sample text to be matched; obtain the sample target text, where the sample target text is associated with the initial sample text to be matched; determine the detection result of the initial sample text to be matched based on the semantic similarity between the initial sample text to be matched and the sample target text and the number of editing times corresponding to the initial sample text to be matched; determine the text to be matched in the sample based on the detection result of the initial sample text to be matched.

[0133] According to an embodiment of the present disclosure, the initial sample text to be matched can be a subject that has not been modified, or a subject obtained after being modified one or more times. The number of editing times corresponding to the initial sample text to be matched can be the number of times the initial sample text to be matched is modified.

[0134] According to an embodiment of the present disclosure, the detection result of the initial sample to-be-matched text may be the number 1 or 0. When the detection result is 1, the initial sample to-be-matched text type may be determined as the sample to-be-matched text. When the detection result is 0, the initial sample to-be-matched text may be deleted or marked to avoid using the initial sample to-be-matched text as the sample to-be-matched text, so as to screen the sample to-be-matched text and accelerate the training speed of the text matching model.

[0135] According to an embodiment of the present disclosure, the cosine similarity distance between the initial sample to-be-matched text and the sample target text may be determined, and the cosine similarity distance may be used to characterize the semantic similarity between the initial sample to-be-matched text and the sample target text.

[0136] For example, the initial sample to-be-matched text and the sample target text are respectively encoded to generate the initial sample to-be-matched text encoding information and the sample target text encoding information. A neural network model based on word2vec is used to process the initial sample to-be-matched text encoding information and the sample target text encoding information to generate semantic similarity encoding information, and the semantic similarity encoding information may characterize the semantic similarity between the initial sample to-be-matched text and the sample target text.

[0137] It should be noted that when the number of characters in the initial sample to-be-matched text and the sample target text is too large, the initial sample to-be-matched text and the sample target text may be segmented respectively to generate the word-level information of multiple initial sample to-be-matched texts and the word-level information of multiple sample target texts, and the initial sample to-be-matched text encoding information and the sample target text encoding information are generated based on the word-level information of multiple initial sample to-be-matched texts and the word-level information of multiple sample target texts.

[0138] For example, the initial sample to-be-matched text may be "derivative financial assets". Segmenting "derivative financial assets" may generate the word-level information "derivative", "finance", and "assets". Encoding the word-level information "derivative", "finance", and "assets" respectively generates the first word-level encoding information W 1 、the second word-level encoding information W 2 、the third word-level encoding information W 3 . The initial sample to-be-matched text encoding information is generated using formula (2).

[0139] W = sum(W 1 , W 2 , W 3 ) / 3 (2)

[0140] In formula (2), the initial sample to-be-matched text encoding information is represented as W, and the first word-level encoding information is represented as W 1, the second - level word encoding information is represented as W 2 , the third - level word encoding information is represented as W 3 . It should be understood that the sample target text encoding information can be generated using the same method as formula (2) to reduce the calculation time for determining the detection result later.

[0141] According to an embodiment of the present disclosure, the number of editing times corresponding to the initial sample text to be matched can be encoded to generate editing - time encoding information. The editing - time encoding information and semantic - similarity encoding information are processed using a logistic classifier to generate a detection result of 1 or 0, so as to determine the sample text to be matched according to the detection result.

[0142] Figure 6 Schematically shows an application scenario diagram for determining the sample text to be matched according to an embodiment of the present disclosure.

[0143] As Figure 6 shown, the initial sample text to be matched and the sample target text are encoded respectively, and the initial sample text - to - be - matched encoding information 611 and the sample target text encoding information 621 can be generated. The semantic - similarity encoding layer constructed based on word2vec is used to process the initial sample text - to - be - matched encoding information 611 and the sample target text encoding information 621 to generate semantic - similarity encoding information 641. The semantic - similarity encoding information 641 can represent the semantic similarity between the initial sample text to be matched and the sample target text.

[0144] The number of editing times corresponding to the initial sample text to be matched 611 can be encoded to generate editing - time encoding information 612. The editing - time encoding information 612 and the semantic - similarity encoding information 641 are input into the logistic classifier 650 to generate a detection result 660 for the initial sample text to be matched. The detection result 660 can include 1 or 0, so as to determine the sample text to be matched according to the detection result. When the detection result 660 is 1, the initial sample text type can be determined as the sample text to be matched. When the detection result 660 is 0, the initial sample text to be matched can be deleted or marked to avoid using the initial sample text to be matched as the sample text to be matched, thereby screening the sample text to be matched and accelerating the training speed of the text - matching model.

[0145] Based on the above - mentioned text - matching method, the present disclosure also provides a text - matching device. The following will be combined with Figure 7 to describe this device in detail.

[0146] Figure 7 Schematically shows a structural block diagram of the text - matching device according to an embodiment of the present disclosure.

[0147] AsFigure 7 As shown, the text matching device 700 of this embodiment includes an acquisition module 710, a first encoding module 720, a second encoding module 730, and a text matching module 740.

[0148] The acquisition module 710 is configured to acquire the text to be matched and the editing attribute information corresponding to the text to be matched.

[0149] The first encoding module 720 is configured to generate first encoding information based on the semantic feature information of the text to be matched and the target text.

[0150] The second encoding module 730 is configured to generate second encoding information based on the editing attribute information corresponding to the text to be matched.

[0151] The text matching module 740 is configured to input the first encoding information and the second encoding information into a text matching model, and output the matching result of the text to be matched and the target text.

[0152] According to an embodiment of the present disclosure, there may be multiple target texts.

[0153] The text matching method may further include a target text matching module.

[0154] The target text matching module is configured to determine the target text that matches the text to be matched based on the matching results of the text to be matched and each target text.

[0155] According to an embodiment of the present disclosure, the first encoding module may include: a first encoding sub-module, a second encoding sub-module, and a semantic fusion encoding sub-module.

[0156] The first encoding sub-module is configured to encode the text to be matched to generate text-to-be-matched encoding information.

[0157] The second encoding sub-module is configured to encode the target text to generate target text encoding information.

[0158] The semantic fusion encoding sub-module is configured to perform semantic fusion encoding on the text-to-be-matched encoding information and the target text encoding information to generate first encoding information.

[0159] According to an embodiment of the present disclosure, the editing attribute information corresponding to the text to be matched may include at least one of the following: editing time information for the text to be matched, user name information for editing the text to be matched, user affiliation information for editing the text to be matched, and user privilege level information for editing the text to be matched.

[0160] According to an embodiment of the present disclosure, there may be multiple items of editing attribute information corresponding to the text to be matched.

[0161] The second encoding module may include: a third encoding sub-module and a fusion encoding sub-module.

[0162] The third encoding sub-module is configured to encode each of a plurality of edit attribute information items corresponding to the text to be matched, generating a plurality of edit attribute encoding information.

[0163] The fusion encoding sub-module is configured to perform fusion encoding on the plurality of edit attribute encoding information, generating second encoding information.

[0164] Based on the above text matching model training method, the present disclosure also provides a training apparatus for a text matching model. The following will be combined with Figure 8 to describe this apparatus in detail.

[0165] Figure 8 A structural block diagram of a training apparatus for a text matching model according to an embodiment of the present disclosure is schematically shown.

[0166] As Figure 8 shown, the training apparatus 800 for the text matching model of this embodiment may include a determination module 810, a first sample encoding module 820, a second sample encoding module 830, and a training module 840.

[0167] The determination module 810 is configured to determine a sample text to be matched and sample edit attribute information corresponding to the sample text to be matched.

[0168] The first sample encoding module 820 is configured to generate first sample encoding information based on semantic feature information of the sample text to be matched and a sample target text, where the sample text to be matched is associated with the sample target text.

[0169] The second sample encoding module 830 is configured to generate second sample encoding information based on the sample edit attribute information corresponding to the sample text to be matched.

[0170] The training module 840 is configured to train an untrained initial text matching model using the first sample encoding information and the second sample encoding information, obtaining a text matching model for use in the above text matching method.

[0171] According to an embodiment of the present disclosure, the training apparatus for a text matching model may further include: a first acquisition module, a second acquisition module, a detection module, and a text to be matched determination module.

[0172] The first acquisition module is configured to acquire an initial sample text to be matched and the number of editing times corresponding to the initial sample text to be matched.

[0173] The second acquisition module is configured to acquire a sample target text, where the sample target text is associated with the initial sample text to be matched.

[0174] A detection module, configured to determine a detection result of an initial sample text to be matched based on the semantic similarity between the initial sample text to be matched and a sample target text, and the number of editing times corresponding to the initial sample text to be matched.

[0175] A text to be matched determination module, configured to determine a text to be matched based on the detection result of the initial sample text to be matched.

[0176] According to an embodiment of the present disclosure, any plurality of modules among the acquisition module 710, the first encoding module 720, the second encoding module 730, the text matching module 740, the determination module 810, the first sample encoding module 820, the second sample encoding module 830, and the training module 840 may be combined and implemented in one module, or any one of them may be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules may be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present disclosure, at least one of the acquisition module 710, the first encoding module 720, the second encoding module 730, the text matching module 740, the determination module 810, the first sample encoding module 820, the second sample encoding module 830, and the training module 840 may be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on a substrate, a system in a package, an application specific integrated circuit (ASIC), or any other reasonable way of integrating or packaging circuits, etc., implemented by hardware or firmware, or implemented in any one of the three implementation manners of software, hardware, and firmware, or in an appropriate combination of any several of them. Alternatively, at least one of the acquisition module 710, the first encoding module 720, the second encoding module 730, the text matching module 740, the determination module 810, the first sample encoding module 820, the second sample encoding module 830, and the training module 840 may be at least partially implemented as a computer program module, which can execute corresponding functions when the computer program module is run.

[0177] Figure 9 Schematically shows a block diagram of an electronic device suitable for implementing a text matching method and a training method of a text matching model according to an embodiment of the present disclosure.

[0178] As Figure 9As shown, an electronic device 900 according to an embodiment of the present disclosure includes a processor 901, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 902 or a program loaded from a storage section 908 into a random access memory (RAM) 903. The processor 901 may include, for example, a general microprocessor (such as a CPU), an instruction set processor, and / or a related chipset, and / or a dedicated microprocessor (such as an application-specific integrated circuit (ASIC)), etc. The processor 901 may also include on-board memory for caching purposes. The processor 901 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0179] In the RAM 903, various programs and data required for the operation of the electronic device 900 are stored. The processor 901, ROM 902, and RAM 903 are connected to each other via a bus 904. The processor 901 performs various operations of the method flow according to an embodiment of the present disclosure by executing the programs in the ROM 902 and / or RAM 903. It should be noted that the program may also be stored in one or more memories other than the ROM 902 and RAM 903. The processor 901 may also perform various operations of the method flow according to an embodiment of the present disclosure by executing the programs stored in the one or more memories.

[0180] According to an embodiment of the present disclosure, the electronic device 900 may further include an input / output (I / O) interface 905, and the input / output (I / O) interface 905 is also connected to the bus 904. The electronic device 900 may further include one or more of the following components connected to the I / O interface 905: an input section 906 including a keyboard, a mouse, etc.; an output section 907 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc. and a speaker, etc.; a storage section 908 including a hard disk, etc.; and a communication section 909 including a network interface card such as a LAN card, a modem, etc. The communication section 909 performs communication processing via a network such as the Internet. A drive 910 is also connected to the I / O interface 905 as needed. A removable medium 911, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 910 as needed so that a computer program read from it can be installed into the storage section 908 as needed.

[0181] The present disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or may exist separately without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the one or more programs are executed, the method according to an embodiment of the present disclosure is implemented.

[0182] According to an embodiment of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium, and may include, for example, but not limited to: portable computer disks, hard disks, random access memories (RAMs), read-only memories (ROMs), erasable programmable read-only memories (EPROMs or flash memories), portable compact disk read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the above. In the present disclosure, the computer-readable storage medium may be any tangible medium that contains or stores a program, and the program may be used by or in combination with an instruction execution system, apparatus, or device. For example, according to an embodiment of the present disclosure, the computer-readable storage medium may include the above-described ROM 902 and / or RAM 903 and / or one or more memories other than ROM 902 and RAM 903.

[0183] An embodiment of the present disclosure further includes a computer program product, which includes a computer program that contains program code for executing the method shown in the flowchart. When the computer program product runs in a computer system, the program code is used to cause the computer system to implement the text matching method and the training method of the text matching model provided by the embodiments of the present disclosure.

[0184] When the computer program is executed by the processor 901, it executes the above functions defined in the system / apparatus of the embodiments of the present disclosure. According to an embodiment of the present disclosure, the above-described systems, apparatuses, modules, units, etc. may be implemented by computer program modules.

[0185] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices and magnetic storage devices. In another embodiment, the computer program may also be transmitted and distributed in the form of a signal on a network medium, and downloaded and installed through the communication part 909, and / or installed from the removable medium 911. The program code included in the computer program may be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0186] In such an embodiment, the computer program may be downloaded and installed from the network through the communication part 909, and / or installed from the removable medium 911. When the computer program is executed by the processor 901, it executes the above functions defined in the system of the embodiments of the present disclosure. According to an embodiment of the present disclosure, the above-described systems, devices, apparatuses, modules, sub-modules, etc. may be implemented by computer program modules.

[0187] According to embodiments of the present disclosure, program code for executing the computer programs provided by the embodiments of the present disclosure can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. The programming languages include, but are not limited to, such as Java, C++, Python, the "C" language, or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user's computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (for example, by using an Internet service provider to connect through the Internet).

[0188] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram can represent a module, a program segment, or a part of code that 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 blocks may occur in a different order than marked in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram or flowchart, and combinations of blocks in the block diagram or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0189] Those skilled in the art can understand that the features recited in the various embodiments and / or claims of the present disclosure can be combined or / and combined in various ways, even if such combinations or combinations are not explicitly recited in the present disclosure. In particular, without departing from the spirit and teachings of the present disclosure, the features recited in the various embodiments and / or claims of the present disclosure can be combined and / or combined in various ways. All such combinations and / or combinations fall within the scope of the present disclosure.

[0190] The embodiments of the present disclosure have been described above. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of the present disclosure. Although the embodiments have been described separately above, this does not mean that the measures in each embodiment cannot be used advantageously in combination. The scope of the present disclosure is defined by the appended claims and their equivalents. Without departing from the scope of the present disclosure, those skilled in the art can make various substitutions and modifications, and these substitutions and modifications should fall within the scope of the present disclosure.

Claims

1. A text matching method, comprising: obtaining the text to be matched and the editing attribute information corresponding to the text to be matched; generating first coding information based on the semantic feature information of the text to be matched and the target text, wherein the target text includes standard financial subjects; generating second coding information based on the editing attribute information corresponding to the text to be matched; inputting the first coding information and the second coding information into a text matching model, and outputting the matching result of the text to be matched and the target text; wherein, the editing attribute information corresponding to the text to be matched includes at least one of the editing time information for the text to be matched, the user name information of the user who edits the text to be matched, the user affiliation information of the user who edits the text to be matched, and the user permission level information for editing the text to be matched.

2. The method according to claim 1, wherein, there are multiple target texts; the text matching method further comprises: determining the target text that matches the text to be matched based on the matching results of the text to be matched and each of the target texts.

3. The method according to claim 1, wherein, generating first coding information based on the semantic feature information of the text to be matched and the target text includes: encoding the text to be matched to generate text-to-be-matched coding information; encoding the target text to generate target text coding information; performing semantic fusion coding on the text-to-be-matched coding information and the target text coding information to generate the first coding information.

4. The method according to claim 1, wherein, the editing attribute information corresponding to the text to be matched includes multiple items; generating second coding information based on the editing attribute information corresponding to the text to be matched includes: encoding each of the multiple editing attribute information items corresponding to the text to be matched to generate multiple editing attribute coding information; performing fusion coding on the multiple editing attribute coding information to generate the second coding information.

5. A training method for a text matching model, comprising: determining a sample text to be matched and sample editing attribute information corresponding to the sample text to be matched; generating first sample coding information based on the semantic feature information of the sample text to be matched and the sample target text, wherein the sample text to be matched is associated with the sample target text, and wherein the sample target text includes standard financial subjects; generating second sample coding information based on the sample editing attribute information corresponding to the sample text to be matched; training an untrained initial text matching model using the first sample coding information and the second sample coding information to obtain the text matching model for the method according to any one of claims 1 to 4; Among them, the sample editing attribute information corresponding to the sample text to be matched includes at least one of the editing time information for the sample text to be matched, the user name information for editing the sample text to be matched, the user affiliation information for editing the sample text to be matched, and the user privilege level information for editing the sample text to be matched.

6. The training method according to claim 5, further comprises: obtaining an initial sample text to be matched and the number of editing times corresponding to the initial sample text to be matched; obtaining a sample target text, wherein the sample target text is associated with the initial sample text to be matched; determining a detection result of the initial sample text to be matched based on the semantic similarity between the initial sample text to be matched and the sample target text, and the number of editing times corresponding to the initial sample text to be matched; determining the sample text to be matched based on the detection result of the initial sample text to be matched.

7. A text matching device, comprises: an obtaining module, configured to obtain a text to be matched and the editing attribute information corresponding to the text to be matched; a first encoding module, configured to generate first encoding information based on the semantic feature information of the text to be matched and a target text, wherein the target text includes standard financial subjects; a second encoding module, configured to generate second encoding information based on the editing attribute information corresponding to the text to be matched; a text matching module, configured to input the first encoding information and the second encoding information into a text matching model, and output a matching result between the text to be matched and the target text; wherein the editing attribute information corresponding to the text to be matched includes at least one of the editing time information for the text to be matched, the user name information for editing the text to be matched, the user affiliation information for editing the text to be matched, and the user privilege level information for editing the text to be matched.

8. A training device for a text matching model, comprises: a determining module, configured to determine a sample text to be matched and the sample editing attribute information corresponding to the sample text to be matched; a first sample encoding module, configured to generate first sample encoding information based on the semantic feature information of the sample text to be matched and a sample target text, wherein the sample text to be matched is associated with the sample target text, and the sample target text includes standard financial subjects; a second sample encoding module, configured to generate second sample encoding information based on the sample editing attribute information corresponding to the sample text to be matched; a training module, configured to train an untrained initial text matching model by using the first sample encoding information and the second sample encoding information, and obtain the text matching model for the method according to any one of claims 1 to 4. Among them, the sample editing attribute information corresponding to the sample text to be matched includes at least one of the editing time information for the sample text to be matched, the user name information for editing the sample text to be matched, the user affiliation information for editing the sample text to be matched, and the user privilege level information for editing the sample text to be matched.

9. An electronic device, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors are caused to execute the method according to any one of claims 1 to 4, or execute the training method according to any one of claims 5 to 6.

10. A computer-readable storage medium having stored thereon executable instructions that, when executed by a processor, cause the processor to execute the method according to any one of claims 1 to 4, or execute the training method according to any one of claims 5 to 6.

11. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1 to 4, or implements the training method according to any one of claims 5 to 6.

Images