Information matching model training, information matching method and device

Abstract

The present invention discloses information matching model training, information matching method and device, and relates to the field of computer technology. A specific implementation of the method includes: determining multiple training information groups, wherein the training information groups include two pieces of training information and matching labels of the training information groups; generating corresponding feature score vectors for the training information groups according to the two pieces of training information; and training a pre-set model to be trained using the feature score vectors and matching labels to obtain an information matching model. This implementation can effectively reduce the operating cost of information matching.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a training of an information matching model, an information matching method and a device. Background Art

[0002] Information matching is the use of algorithms to calculate the similarity of two pieces of information. It is a core application in various current dialogue systems. For example, many systems can match questions with a large number of samples and give corresponding responses based on the matching results.

[0003] In the process of implementing the present invention, the inventors found that there are at least the following problems in the prior art:

[0004] Existing information matching methods require human resources to maintain and expand a large number of similar questions, resulting in high operating costs. Summary of the invention

[0005] In view of this, an embodiment of the present invention provides an information matching model training, information matching method and device, which completes information matching through the information matching model without requiring a large amount of human resources to maintain and expand similar sentences, thereby effectively reducing operating costs.

[0006] To achieve the above object, according to one aspect of an embodiment of the present invention, a method for training an information matching model is provided, comprising:

[0007] Determine a plurality of training information groups, wherein the training information groups include two pieces of training information and matching labels of the training information groups;

[0008] Generating a corresponding feature score vector for the training information group according to the two pieces of training information;

[0009] Using the feature score vector and the matching label, training a preset model to be trained;

[0010] According to the training results, an information matching model is generated.

[0011] Preferably, generating a corresponding feature score vector for the training information group includes:

[0012] Calculating semantic feature scores for the two pieces of training information in the training information group;

[0013] Calculating entity feature scores and attribute feature scores corresponding to the training information group;

[0014] A feature score vector corresponding to the training information group is generated by using the entity feature score, the attribute feature score and the semantic feature score.

[0015] Preferably,

[0016] Before calculating the entity feature score and the attribute feature score corresponding to the training information group, the method further comprises: generating a corresponding entity feature vector and an attribute feature vector for each segment of the training information;

[0017] The step of calculating the entity feature score and the attribute feature score corresponding to the training information group includes: calculating the entity feature score and the attribute feature score corresponding to the training information group according to the entity feature vector and the attribute feature vector corresponding to each segment of the training information.

[0018] Preferably,

[0019] The training information is annotated with entity labels and attribute labels;

[0020] Generating a corresponding entity feature vector and an attribute feature vector for each segment of the training information, including:

[0021] Determining entity features and attribute features included in the training information according to the entity label and the attribute label;

[0022] According to a preset first entity dictionary, the entity feature is converted into a corresponding entity feature vector;

[0023] According to a preset first attribute dictionary, the attribute feature is converted into a corresponding attribute feature vector.

[0024] Preferably, calculating the semantic feature scores for the two pieces of training information in the training information group includes:

[0025] Calculating the semantic feature vector corresponding to each segment of the training information in the training information group;

[0026] A distance score between the semantic feature vectors corresponding to two pieces of the training information is determined, and the distance score is used as a first semantic feature score.

[0027] Preferably, calculating the semantic feature scores for the two pieces of training information in the training information group includes:

[0028] A second semantic feature score is calculated according to the length of each segment of training information in the training information group and a preset length.

[0029] Preferably, calculating the semantic feature scores for the two pieces of training information in the training information group includes:

[0030] Determining a maximum matching length between the two pieces of training information in the training information group;

[0031] A third semantic feature score is calculated according to the maximum matching length and the length of each piece of training information in the training information group.

[0032] Preferably, the training method of the information matching model further comprises: normalizing the original training information;

[0033] The determining of the plurality of training information groups includes: determining the plurality of training information groups according to the normalized result.

[0034] Preferably, the determining of a plurality of training information groups comprises:

[0035] When the first training information is received, a preset number of segments of second training information are preliminarily screened out for the first training information from a preset information library;

[0036] For each segment of the second training information, execute:

[0037] A training information group is formed by using the second training information, the first training information, and a matching label determined by the first training information and the second training information.

[0038] In a second aspect, an embodiment of the present invention provides an information matching method implemented based on an information matching model trained in any of the above embodiments, including:

[0039] When receiving the query information, generating corresponding feature score vectors for the query information and a plurality of to-be-matched information in a preset matching library respectively;

[0040] Inputting the feature score vector corresponding to the query information and the feature score vector corresponding to the information to be matched into the information matching model;

[0041] According to the matching result determined by the information matching model, matching information corresponding to the query information is determined from the plurality of information to be matched.

[0042] Preferably, the information matching method further comprises:

[0043] Constructing an inverted index for the preset matching library;

[0044] Using the inverted index, a preset number of to-be-matched information is screened out for the query information from the preset matching library;

[0045] The step of generating a corresponding feature score vector for the to-be-matched information in the preset matching library includes:

[0046] Generate corresponding feature score vectors for the preset number of filtered information to be matched.

[0047] Preferably, the step of generating corresponding feature score vectors for the query information and the information to be matched in the preset matching library respectively includes:

[0048] Calculate the corresponding semantic feature scores, entity feature scores and attribute feature scores for the query information and the information to be matched respectively;

[0049] A corresponding feature score vector is generated using the entity feature score, the attribute feature score, and the semantic feature score.

[0050] Preferably, the information matching method further comprises: generating corresponding entity feature vectors and attribute feature vectors for the query information and the information to be matched, respectively, according to the entity features and attribute features included in the query information and the entity features and attribute features included in the information to be matched;

[0051] The step of calculating the corresponding entity feature score and attribute feature score for the query information includes: calculating the entity feature score and attribute feature score corresponding to the query information according to the entity feature vector, the attribute feature vector corresponding to the query information, and the entity feature vector and attribute feature vector corresponding to the information to be matched.

[0052] Preferably,

[0053] The information to be matched is marked with entity tags and attribute tags;

[0054] Generating corresponding entity feature vectors and attribute feature vectors for the query information and the information to be matched respectively, including:

[0055] Determine entity features and attribute features included in the information to be matched according to the entity tag and the attribute tag marked by the information to be matched;

[0056] According to a preset second entity dictionary, entity features in the query information and the information to be matched are respectively converted into corresponding entity feature vectors;

[0057] According to a preset second attribute dictionary, attribute features in the query information and the information to be matched are respectively converted into corresponding attribute feature vectors.

[0058] Preferably,

[0059] The information to be matched is configured with answer information;

[0060] The information matching method further comprises:

[0061] Outputting answer information of the matching information corresponding to the inquiry information to the user who initiated the inquiry information.

[0062] In a third aspect, an embodiment of the present invention provides a training device for an information matching model, comprising: a determination unit, a processing unit, and a training unit, wherein:

[0063] The determining unit is used to determine a plurality of training information groups, wherein the training information groups include two pieces of training information and matching tags of the training information groups;

[0064] The processing unit is used to generate a corresponding feature score vector for the training information group according to the two pieces of training information in the training information group determined by the determining unit;

[0065] The training unit is used to train a preset model to be trained using the feature score vector generated by the processing unit and the matching label; and generate an information matching model according to the training result.

[0066] Preferably,

[0067] The processing unit is used to calculate semantic feature scores for the two pieces of training information in the training information group; calculate the entity feature score and the attribute feature score corresponding to the training information group; and generate a feature score vector corresponding to the training information group using the entity feature score, the attribute feature score and the semantic feature score.

[0068] In a fourth aspect, an embodiment of the present invention provides an information matching device, comprising: a vector generating unit and a matching unit, wherein:

[0069] The vector generating unit is used to generate corresponding feature score vectors for the query information and the information to be matched in the preset matching library respectively when receiving the query information;

[0070] The matching unit is used to input the feature score vector corresponding to the query information generated by the vector generation unit and the feature score vector corresponding to the information to be matched into the information matching model; and determine the matching information corresponding to the query information from the information to be matched according to the matching result determined by the information matching model, wherein the information matching model is obtained based on the training method of the information matching model provided in any of the above embodiments.

[0071] One embodiment of the above invention has the following advantages or beneficial effects: Since the corresponding feature score vector and matching label of the training information group are generated for the training information group, the model to be trained is trained, wherein the feature score vector is obtained based on the training information and can better reflect the training information, then the matching relationship between the information can be reflected by the feature score vector and the matching label of the training information group. The information matching model obtained based on this can realize information matching without requiring a large amount of human resources to maintain and expand similar sentences, which can effectively reduce operating costs.

[0072] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention.

[0074] Figure 1 is a schematic diagram of the main process of the training method of the information matching model according to an embodiment of the present invention;

[0075] Figure 2 is a schematic diagram of the main process of generating a corresponding feature score vector for a training information group according to an embodiment of the present invention;

[0076] Figure 3 is a schematic diagram of the main process of generating corresponding entity feature vectors and attribute feature vectors for training information according to an embodiment of the present invention;

[0077] Figure 4 is a schematic diagram of the main process of the information matching method according to an embodiment of the present invention;

[0078] Figure 5 is a schematic diagram of main units of a training device for an information matching model according to an embodiment of the present invention;

[0079] Figure 6 is a schematic diagram of main units of an information matching device according to an embodiment of the present invention;

[0080] Figure 7 is an exemplary system architecture diagram to which embodiments of the present invention may be applied;

[0081] Figure 8 It is a schematic diagram of the structure of a computer system suitable for implementing a server of an embodiment of the present invention. DETAILED DESCRIPTION

[0082] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0083] Training information refers to text information or voice information used to train the information matching model, which is generally a complete sentence, such as what are the famous tourist attractions in City A, what are the delicacies in Region B, etc.

[0084] The structure of a training information group may be (training information 1, training information 2).

[0085] The matching label of the training information group is mainly used to indicate the matching situation between two pieces of training information in a training information group. If the two pieces of training information in a training information group match each other (that is, the meanings expressed by the two pieces of training information are the same or similar), the matching label is represented as 1. If the two pieces of training information in a training information group do not match each other (that is, the meanings expressed by the two pieces of training information are completely different), the matching label is represented as 0.

[0086] Figure 1 is a training method for an information matching model according to an embodiment of the present invention, such as Figure 1 As shown, the training method of the information matching model may include the following steps:

[0087] S101: Determine a plurality of training information groups, wherein the training information groups include two pieces of training information and matching labels of the training information groups;

[0088] S102: Generate a corresponding feature score vector for the training information group according to the two pieces of training information;

[0089] S103: Using the feature score vector and the matching label, train the pre-set model to be trained to obtain an information matching model.

[0090] The model to be trained may be a SVR support vector regression model.

[0091] The feature score vector refers to a vector composed of feature scores calculated based on the features of the training information.

[0092] exist Figure 1In the embodiment shown, the model to be trained is trained by generating a corresponding feature score vector and a matching label of the training information group for the training information group, wherein the feature score vector is obtained based on the training information and can better reflect the training information, then the matching relationship between the information can be reflected by the feature score vector and the matching label of the training information group. The information matching model obtained based on this can realize information matching without requiring a large amount of human resources to maintain and expand similar sentences, which can effectively reduce operating costs.

[0093] In one embodiment of the present invention, Figure 2 As shown, generating a corresponding feature score vector for a training information group may include the following steps:

[0094] S201: Calculating semantic feature scores for two pieces of training information in the training information group;

[0095] S202: Calculate entity feature scores and attribute feature scores corresponding to the training information group;

[0096] S203: Generate a feature score vector corresponding to the training information group using the entity feature score, the attribute feature score and the semantic feature score.

[0097] Generally speaking, the features of training information (sentences) mainly include semantic features, entity features and attribute features. That is, the training information (sentences) can be more accurately reflected through semantic features, entity features and attribute features, among which the semantic features are mainly used to represent the semantics expressed by the training information. For entities and attributes, for example, for the training information: What are the famous tourist attractions in City A, the entity is City A, and the attribute is tourist attractions. In order to facilitate the training of the model to be trained, in an embodiment of the present invention, the corresponding semantic feature scores, entity feature scores and attribute feature scores are calculated for the semantic features, entity features and attribute features, respectively. Therefore, the training information can be more accurately expressed through the semantic feature scores, entity feature scores and attribute feature scores.

[0098] Among them, generating the feature score vector corresponding to the training information group mainly involves splicing the semantic feature scores, entity feature scores, and attribute feature scores obtained from the two pieces of training information. Accordingly, one expression of the feature score vector obtained by splicing can be (semantic feature score, entity feature score, attribute feature score). It is worth noting that the semantic feature score, entity feature score, and attribute feature score can be arbitrarily spliced ​​to form a feature score vector (for example, another expression of the feature score vector obtained by splicing can be (entity feature score, semantic feature score, attribute feature score)), and it is only necessary to ensure that the expression of the feature score vectors corresponding to the multiple training information groups used to train the same model to be trained is consistent.

[0099] Through the above process, not only the training information is digitized and the complexity of the training information is simplified, but also the above-mentioned entity feature scores, the attribute feature scores and the semantic feature scores can more realistically reflect their corresponding training information. Correspondingly, the feature score vector can also more realistically reflect the training information group, which not only enables the training to proceed smoothly, but also ensures the accuracy of the training results.

[0100] In one embodiment of the present invention, before calculating the entity feature score and attribute feature score corresponding to the training information group, it may further include: generating a corresponding entity feature vector and attribute feature vector for each segment of training information in the training information; accordingly, the step of calculating the entity feature score and attribute feature score corresponding to the training information group includes: calculating the entity feature score and attribute feature score corresponding to the training information group based on the entity feature vector and attribute feature vector corresponding to each segment of training information.

[0101] The entity feature score corresponding to the training information group is calculated by using the entity feature vectors corresponding to the two pieces of training information in the training information group;

[0102] The attribute feature score corresponding to the training information group is calculated by using the attribute feature vectors corresponding to the two pieces of training information in the training information group.

[0103] The following calculation formula (1) can be used to calculate the entity feature score or the attribute feature score respectively.

[0104] Calculation formula (1):

[0105]

[0106] Among them, when F i When representing the entity feature score corresponding to the training information group i, A i1 A is the entity feature vector corresponding to a piece of training information in training information group i. i2 Represents the entity feature vector corresponding to another piece of training information in the training information group i; when F i When representing the attribute feature score corresponding to the training information group i, A i1 A represents the attribute feature vector corresponding to a piece of training information in training information group i. i2 Characterizes the attribute feature vector corresponding to another piece of training information in training information group i.

[0107] The above process realizes the calculation of entity feature scores and attribute feature scores of the training information group.

[0108] The above entity feature scores and attribute feature scores can reflect the matching situation between two pieces of training information in the training information group to a certain extent.

[0109] In one embodiment of the present invention, the training information is annotated with entity tags and attribute tags; accordingly, Figure 3 As shown, generating a corresponding entity feature vector and an attribute feature vector for each segment of training information may include the following steps:

[0110] S301: Determine entity features and attribute features included in the training information according to the entity tags and attribute tags;

[0111] S302: According to a preset first entity dictionary, the entity feature is converted into a corresponding entity feature vector;

[0112] S303: According to a preset first attribute dictionary, the attribute feature is converted into a corresponding attribute feature vector.

[0113] It is worth noting that there is no strict order between the above step S302 and step S303.

[0114] The first entity dictionary may be an existing entity dictionary, or may be an entity dictionary composed of all entities in all training data groups.

[0115] The first attribute dictionary may be an existing attribute dictionary, or may be an attribute dictionary composed of all attributes in all training data groups.

[0116] Entity labels and attribute labels are feature labels set to clearly distinguish entities and attributes in training information. For example, training information: What are the famous tourist attractions in City A? The corresponding entity is City A, and the attribute is tourist attractions. Then City A can be marked (T) and tourist attractions can be marked (X). City A can also be marked with one color and tourist attractions with another color to achieve feature labeling of entities and attributes.

[0117] The specific implementation method for converting entity features into corresponding entity feature vectors according to the preset first entity dictionary is as follows: according to the number of entities included in the first entity dictionary, the number of elements included in the entity feature vector is determined, and according to the position of the entity features in the first entity dictionary, the elements at the corresponding positions in the entity feature vector are assigned 1, and the other elements are assigned 0. For example, if the number of entities included in the first entity dictionary is 50, and the position of entity feature a in the first entity dictionary is the 50th, then the entity feature vector (50 elements in total) obtained is: (0, 0, 0, ..., 1).

[0118] The specific implementation method for converting the attribute feature into the corresponding attribute feature vector according to the preset first attribute dictionary is as follows: according to the number of attributes included in the first attribute dictionary, the number of elements included in the attribute feature vector is determined, and according to the position of the attribute feature in the first attribute dictionary, the elements at the corresponding positions in the attribute feature vector are assigned 1, and the other elements are assigned 0. For example, the number of entities included in the first attribute dictionary is 30, and the position of the attribute feature b in the first attribute dictionary is the 10th, then the attribute feature vector (30 elements in total) obtained is: (0, 0, 0, 0, 0, 0, 0, 0, 1, 0, ..., 0).

[0119] In one embodiment of the present invention, calculating semantic feature scores for two pieces of training information in the training information group may include six implementation methods:

[0120] Implementation method 1: Calculate the semantic feature vector corresponding to each piece of training information in the training information group; determine the distance score between the semantic feature vectors corresponding to two pieces of training information, and use the distance score as the first semantic feature score.

[0121] The following calculation formula (2) is used to calculate the semantic feature values ​​corresponding to the words in the training information, and the semantic feature vector corresponding to the training information is determined by using the semantic feature values ​​corresponding to all the words included in the training information and a preset word dictionary;

[0122] Calculation formula (2):

[0123]

[0124] Among them, Y ijk represents the semantic feature value corresponding to word k in the jth training information in training information group i; k Represents the frequency of occurrence of word k in all training information groups; n ij represents the total number of words included in the jth training information in the training information group i; Z represents the total amount of training information; Z k Represents the amount of training information including word k.

[0125] The preset word dictionary may be composed of words included in all training data sets, or an existing word dictionary may be selected.

[0126] Using the semantic feature values ​​corresponding to all the words included in the training information and the preset word dictionary, the semantic feature vector corresponding to the training information can be determined by determining the number of elements of the semantic feature vector according to the total number of words included in the preset word dictionary, and assigning the semantic feature value of the word to the element at the corresponding position in the semantic feature vector according to the position of the word in the word dictionary in the training information. For example, the word dictionary includes 100 words, and the 10 words included in a training information correspond to the first 10 words in the word dictionary in order, then the semantic feature vector corresponding to the training information is: (Y ij1 , Y ij2 , …, Y ij10 , 0, 0, ..., 0) (i.e., the first 10 elements in the semantic feature vector correspond to the semantic feature values ​​of the words included in the training information, and the last 90 are 0).

[0127] The specific implementation method of determining the distance score between the semantic feature vectors corresponding to two pieces of training information can be:

[0128] The distance score between the semantic feature vectors corresponding to the two pieces of training information is calculated using the following calculation formula (3);

[0129]

[0130] Among them, f i Represents the distance score between two training information in training information group i, B i1 Represents the semantic feature vector corresponding to a piece of training information in training information group i, B i2 Represents the semantic feature vector corresponding to another piece of training information in training information group i.

[0131] Implementation method 2: Calculate the second semantic feature score according to the length of each piece of training information in the training information group and a preset length.

[0132] The second semantic feature score is calculated specifically as follows: the length of each training information segment is divided by the preset length to obtain the second semantic feature score. For example, if the length of a training information segment is 10 characters and the maximum length is set to 30 characters, then the second semantic feature score is 0.33.

[0133] In this implementation, a training information group includes two second semantic feature scores. Accordingly, the expression corresponding to the feature score vector can be (second semantic feature score 1, second semantic feature score 2, entity feature score, attribute feature score).

[0134] Implementation method three: determine the maximum matching length between two pieces of training information in the training information group; calculate the third semantic feature score according to the maximum matching length and the length of each piece of training information in the training information group.

[0135] The calculation of the third semantic feature score is specifically: the maximum matching length between two pieces of training information in the training information group is divided by the length of each piece of training information in the training information group. For example, the length of training information 1 is 10 words, the length of training information 2 is 15 words, and the maximum matching length of the two training information is 5 words. Then the third semantic feature score corresponding to training information 1 is 0.5, and the third semantic feature score corresponding to training information 2 is 0.33. Accordingly, the expression corresponding to the feature score vector can be (third semantic feature score 1, third semantic feature score 2, entity feature score, attribute feature score).

[0136] Implementation method 4: Implementation method 1 is concatenated with implementation method 2. Accordingly, the expression corresponding to the feature score vector can be (first semantic feature score, second semantic feature score 1, second semantic feature score 2, entity feature score, attribute feature score).

[0137] Implementation method five: Implementation method one and implementation method three are spliced ​​together. Accordingly, the expression corresponding to the feature score vector can be (first semantic feature score, third semantic feature score 1, third semantic feature score 2, entity feature score, attribute feature score).

[0138] Implementation method six: Implementation method one, implementation method two and implementation method three are spliced ​​together. Accordingly, the expression corresponding to the feature score vector can be (first semantic feature score, second semantic feature score 1, second semantic feature score 2, third semantic feature score 1, third semantic feature score 2, entity feature score, attribute feature score).

[0139] In one embodiment of the present invention, the training method of the information matching model may further include: normalizing the original training information; and accordingly, determining multiple training information groups, including: determining multiple training information groups according to the normalized result. The above normalization process can ensure the consistency of information, thereby further improving the accuracy of the training result.

[0140] The normalization process refers to converting several words that express the same content or the same position into the same word. For example, if word A, word B, and word C all express the same content or the same position, word B and word C can be converted into word A.

[0141] In one embodiment of the present invention, determining multiple training information groups may include: when receiving first training information, initially screening out a preset number of segments of second training information for the first training information from a preset information library; for each segment of the second training information, executing: using the second training information, the first training information, and matching tags determined by the first training information and the second training information to form a training information group.

[0142] Specifically, the method of preliminarily screening out a preset number of segments of second training information for the first training information can be as follows: preliminarily screening out a preset number of segments of second training information that are closest to the first training information for the first training information using the inverted index of the preset information library. The preset number of segments is, for example, 20 segments.

[0143] The above process can effectively ensure that the matching labels corresponding to multiple training information groups are evenly distributed, avoid excessive number of non-matching matching labels, and further ensure the accuracy of the training results.

[0144] It is worth noting that the number of training information groups is generally not less than 5,000.

[0145] Based on the feature score vectors and matching labels obtained in the above embodiments, a specific implementation method of training a preset model to be trained is as follows: input the preset model to be trained: training set T = {(x1, y1), (x2, y2), ..., (x N ,y N )}, where x m ∈R n (x m represents the feature score vector corresponding to the mth training information group, R n The set of feature score vectors corresponding to all training information groups), y m ∈{0,1}(y m represents the matching label corresponding to the mth training information group, 0 represents no match, 1 represents a match), m = 1, 2, ... N; Output: regression decision function.

[0146] The essence of the training process is to select appropriate support vector regression parameters for the SVR support vector regression model in an iterative manner to construct a regression decision function. The specific iterative process is consistent with the existing support vector regression iterative process and will not be repeated here.

[0147] like Figure 4 As shown, an embodiment of the present invention provides an information matching method implemented by an information matching model, and the information matching method may include the following steps:

[0148] S401: when receiving a query message, generating corresponding feature score vectors for the query message and a plurality of to-be-matched information in a preset matching library;

[0149] S402: Inputting the feature score vector corresponding to the query information and the feature score vector corresponding to the information to be matched into the information matching model;

[0150] S403: According to the matching result determined by the information matching model, matching information corresponding to the query information is determined from the plurality of information to be matched.

[0151] The information matching model used in the embodiment of the present invention may be obtained by the training method of the information matching model provided in any of the above embodiments.

[0152] The inquiry information may be in text form or in voice form.

[0153] The preset matching library may be a database of a question-answering system, and the matching library generally contains a large amount of inquiry information and answers matching the inquiry information.

[0154] The information to be matched refers to any piece of information stored in the matching library or any piece of information initially screened from the matching library.

[0155] The matching result is generally 1 or 0, where 1 indicates that the query information matches the information to be matched (i.e., the query information is similar or identical to the information to be matched); 0 indicates that the query information does not match the information to be matched (i.e., the query information is neither similar nor identical to the information to be matched).

[0156] Information matching based on the information matching model can effectively improve the efficiency and accuracy of information matching.

[0157] In one embodiment of the present invention, the information matching method may further include: constructing an inverted index for a preset matching library; using the inverted index, filtering out a preset number of information to be matched for the query information from the preset matching library; accordingly, the step of generating a corresponding feature score vector for the information to be matched in the preset matching library may include: generating a corresponding feature score vector for the filtered preset number of information to be matched.

[0158] A preset number of to-be-matched information is screened out for the query information by means of the above-mentioned inverted index, so that the initially screened to-be-matched information is most similar or close to the query information, thereby avoiding unnecessary calculations and effectively saving computing resources.

[0159] In one embodiment of the present invention, the steps of generating corresponding feature score vectors for the query information and the information to be matched in a preset matching library may include: calculating the corresponding semantic feature scores, entity feature scores and attribute feature scores for the query information and the information to be matched respectively; and generating corresponding feature score vectors using the entity feature scores, attribute feature scores and semantic feature scores.

[0160] The training process of the previous information matching model is: the information matching model is trained by generating a feature score vector based on the semantic feature scores calculated based on the semantic features of the training information (sentences), the entity feature scores calculated based on the entity features of the training information (sentences), and the attribute feature scores calculated based on the attribute features of the training information (sentences). Based on this, in the information matching process, the corresponding semantic feature scores, entity feature scores, and attribute feature scores are calculated for the query information and the information to be matched respectively; the corresponding feature score vector is generated using the entity feature scores, attribute feature scores, and semantic feature scores, so that the information matching results based on the information matching model are more accurate.

[0161] The corresponding feature score vector generated by using the entity feature score, the attribute feature score and the semantic feature score is consistent with the feature score vector expression corresponding to the training information group in the training process of the information matching model.

[0162] In one embodiment of the present invention, the information matching method further includes: generating corresponding entity feature vectors and attribute feature vectors for the query information and the information to be matched, respectively, based on the entity features and attribute features included in the query information and the entity features and attribute features included in the information to be matched; accordingly, the step of calculating the corresponding entity feature score and attribute feature score for the query information may include: calculating the entity feature score and attribute feature score corresponding to the query information based on the entity feature vector and attribute feature vector corresponding to the query information and the entity feature vector and attribute feature vector corresponding to the information to be matched.

[0163] The entity features and attribute features included in the query information may be marked by the user, or may be marked by the information matching device during the information matching process.

[0164] The generation of corresponding entity feature vectors and attribute feature vectors for the query information and the information to be matched can be implemented by using existing technologies or by using the solutions provided in the following embodiments.

[0165] The following calculation formula (4) can be used to calculate the entity feature score or attribute feature score corresponding to the query information.

[0166] Calculation formula (4):

[0167]

[0168] Among them, when R represents the entity feature score corresponding to the query information, G represents the entity feature vector corresponding to the query information, and C v represents the entity feature vector corresponding to the information to be matched v; when R represents the attribute feature score corresponding to the query information, G represents the attribute feature vector corresponding to the query information, and C vCharacterize the attribute feature vector corresponding to the information v to be matched.

[0169] In one embodiment of the present invention, the information to be matched is annotated with entity tags and attribute tags; accordingly, corresponding entity feature vectors and attribute feature vectors are generated for the query information and the information to be matched, respectively, including: determining the entity features and attribute features included in the information to be matched according to the entity tags and attribute tags annotated with the information to be matched; converting the entity features in the query information and the information to be matched into corresponding entity feature vectors according to a preset second entity dictionary; converting the attribute features in the query information and the information to be matched into corresponding attribute feature vectors according to a preset second attribute dictionary.

[0170] The second entity dictionary may be an existing entity dictionary, or an entity dictionary composed of entities included in a preset matching library.

[0171] The second attribute dictionary may be an existing attribute dictionary, or an attribute dictionary composed of attributes included in a preset matching library.

[0172] Among them, according to the preset second entity dictionary, the entity features in the query information and the information to be matched are converted into corresponding entity feature vectors, and according to the preset second attribute dictionary, the attribute features in the query information and the information to be matched are converted into corresponding attribute feature vectors, respectively. This is consistent with the method provided in the previous embodiment of converting entity features into corresponding entity feature vectors according to the preset first entity dictionary and converting attribute features into corresponding attribute feature vectors according to the preset first attribute dictionary, and will not be repeated here.

[0173] In one embodiment of the present invention, the information to be matched is configured with answer information; accordingly, the information matching method may further include: outputting answer information of the matching information corresponding to the inquiry information to the user who initiated the inquiry information. This process can realize automatic question-and-answer service. Since the above information matching method can find matching information for the inquiry information more accurately, the answer information corresponding to the matching information can meet the user's needs.

[0174] like Figure 5 As shown, an embodiment of the present invention provides a training device 500 for an information matching model. The training device 500 for an information matching model may include: a determination unit 501, a processing unit 502, and a training unit 503, wherein:

[0175] A determination unit 501 is used to determine a plurality of training information groups, wherein the training information group includes two pieces of training information and matching tags of the training information group;

[0176] A processing unit 502 is used to generate a corresponding feature score vector for the training information group according to the two pieces of training information in the training information group determined by the determining unit 501;

[0177] The training unit 503 is used to train a preset model to be trained using the feature score vector and matching label generated by the processing unit 502 to obtain an information matching model.

[0178] In one embodiment of the present invention, the processing unit 502 is used to calculate semantic feature scores for two pieces of training information in the training information group; calculate entity feature scores and attribute feature scores corresponding to the training information group; and generate a feature score vector corresponding to the training information group using the entity feature scores, attribute feature scores, and semantic feature scores.

[0179] In one embodiment of the present invention, the processing unit 502 is used to generate a corresponding entity feature vector and an attribute feature vector for each segment of training information; and calculate the entity feature score and the attribute feature score corresponding to the training information group according to the entity feature vector and the attribute feature vector corresponding to each segment of training information.

[0180] In one embodiment of the present invention, the training information is annotated with entity tags and attribute tags; the processing unit 502 is used to determine the entity features and attribute features included in the training information based on the entity tags and the attribute tags; according to a preset first entity dictionary, the entity features are converted into corresponding entity feature vectors; according to a preset first attribute dictionary, the attribute features are converted into corresponding attribute feature vectors.

[0181] In one embodiment of the present invention, the processing unit 502 is used to calculate the semantic feature vector corresponding to each segment of training information in the training information group; determine the distance score between the semantic feature vectors corresponding to two segments of training information, and use the distance score as the first semantic feature score.

[0182] In one embodiment of the present invention, the processing unit 502 is configured to calculate the second semantic feature score according to the length of each segment of training information in the training information group and a preset length.

[0183] In one embodiment of the present invention, the processing unit 502 is used to determine the maximum matching length between two pieces of training information in the training information group; and calculate the three semantic feature scores according to the maximum matching length and the length of each piece of training information in the training information group.

[0184] In one embodiment of the present invention, the determining unit 501 is used to perform normalization processing on the original training information; and determine multiple training information groups according to the normalized result.

[0185] In one embodiment of the present invention, the determination unit 501 is used to, when receiving the first training information, initially screen out a preset number of second training information for the first training information from a preset information library; for each segment of the second training information, execute: using the second training information, the first training information, and the matching tags determined by the first training information and the second training information to form a training information group.

[0186] like Figure 6 As shown, the embodiment of the present invention provides an information matching device 600, including: a vector generating unit 601 and a matching unit 602, wherein:

[0187] The vector generating unit 601 is used to generate corresponding feature score vectors for the query information and the information to be matched in the preset matching library respectively when receiving the query information;

[0188] The matching unit 602 is used to input the feature score vector corresponding to the query information and the feature score vector corresponding to the information to be matched generated by the vector generation unit 601 into the information matching model; according to the matching result determined by the information matching model, the matching information corresponding to the query information is determined from the information to be matched, and the information matching model is obtained based on the training method of the information matching model provided in the above embodiment or the information matching model is obtained based on the training device of the information matching model provided in the above embodiment.

[0189] In one embodiment of the present invention, the information matching device 600 may further include: an index generating unit and a preliminary screening unit (not shown in the figure), wherein:

[0190] An index generation unit, used to construct an inverted index for a preset matching library;

[0191] A primary screening unit, used to screen a preset number of to-be-matched information for the query information from a preset matching library using an inverted index;

[0192] The vector generating unit 601 is used to generate corresponding feature score vectors for the screened preset number of to-be-matched information.

[0193] In one embodiment of the present invention, the vector generation unit 601 is used to calculate the corresponding semantic feature scores, entity feature scores and attribute feature scores for the query information and the information to be matched respectively; and generate the corresponding feature score vector using the entity feature scores, attribute feature scores and semantic feature scores.

[0194] In one embodiment of the present invention, the vector generation unit 601 is used to generate a corresponding entity feature vector and an attribute feature vector for the query information according to the entity features and attribute features included in the query information; and calculate the entity feature score and the attribute feature score corresponding to the query information according to the entity feature vector and the attribute feature vector corresponding to the query information.

[0195] In one embodiment of the present invention, the information to be matched is annotated with entity tags and attribute tags; the vector generation unit 601 is further used to determine the entity features and attribute features included in the information to be matched according to the entity tags and attribute tags annotated with the information to be matched; according to a preset second entity dictionary, the entity features in the information to be matched are converted into corresponding entity feature vectors; according to a preset second attribute dictionary, the attribute features in the information to be matched are converted into corresponding attribute feature vectors; according to the entity feature vector and attribute feature vector corresponding to the information to be matched, the entity feature score and attribute feature score corresponding to the information to be matched are calculated.

[0196] In one embodiment of the present invention, the information to be matched is configured with answer information; the information matching device 600 further comprises: an output unit (not shown in the figure), wherein:

[0197] The output unit is used to output the answer information of the matching information corresponding to the inquiry information to the user who initiated the inquiry information.

[0198] Figure 7 An exemplary system architecture 700 is shown to which the information matching model training method or information matching model training device or information matching method or information matching device according to the embodiments of the present invention can be applied.

[0199] like Figure 7 As shown, system architecture 700 may include terminal devices 701, 702, 703, network 704 and server 705. Network 704 is used to provide a medium for communication links between terminal devices 701, 702, 703 and server 705. Network 704 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0200] Users can use terminal devices 701, 702, and 703 to interact with server 705 through network 704 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 701, 702, and 703, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only examples).

[0201] The terminal devices 701 , 702 , and 703 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0202] The server 705 may be a server that provides various services, such as a background management server (only an example) that provides support for users to initially screen out corresponding second training information from the first training information provided by the terminal devices 701, 702, and 703 to form a training information group, or a background management server (only an example) that provides support for users to match query information provided by the terminal devices 701, 702, and 703 with query results. The background management server may analyze and process the received training information, query information, and other data, and feed back the processing results (such as query results - only an example) to the terminal device.

[0203] It should be noted that the information matching model training method or information matching method provided in the embodiment of the present invention is generally executed by the server 705 , and accordingly, the information matching model training device or information matching device is generally arranged in the server 705 .

[0204] It should be understood that Figure 7 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0205] Reference below Figure 8 , which shows a schematic diagram of the structure of a server computer system 800 suitable for implementing an embodiment of the present invention. Figure 8 The terminal device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0206] like Figure 8 As shown, the computer system 800 includes a central processing unit (CPU) 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage part 808 into a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the system 800 are also stored. The CPU 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0207] The following components are connected to the I / O interface 805: an input section 806 including a keyboard, a mouse, etc.; an output section 807 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, a modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the I / O interface 805 as needed. A removable medium 811, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 810 as needed, so that a computer program read therefrom is installed into the storage section 808 as needed.

[0208] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 809, and / or installed from the removable medium 811. When the computer program is executed by the central processing unit (CPU) 801, the above-mentioned functions defined in the system of the present invention are executed.

[0209] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0210] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order 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 or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0211] The units involved in the embodiments of the present invention may be implemented by software or hardware. The units described may also be arranged in a processor, for example, may be described as: a processor includes a determination unit, a processing unit, and a training unit. The names of these units do not constitute limitations on the units themselves in certain circumstances, for example, the determination unit may also be described as a "unit for determining a plurality of training information groups".

[0212] As another aspect, the present invention further provides a computer-readable medium, which may be included in the device described in the above embodiment; or may exist independently without being assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device includes: determining multiple training information groups, wherein the training information groups include two segments of training information and matching labels of the training information groups; generating corresponding feature score vectors for the training information groups based on the two segments of training information; and using the feature score vectors and matching labels to train a pre-set model to be trained to obtain an information matching model.

[0213] The above-mentioned computer-readable medium carries one or more programs. When the above-mentioned one or more programs are executed by a device, the device includes: calculating semantic feature scores for two pieces of training information in the training information group; calculating entity feature scores and attribute feature scores corresponding to the training information group; using the entity feature scores, attribute feature scores and semantic feature scores to generate a feature score vector corresponding to the training information group.

[0214] As another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiment; or may exist independently without being assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device includes: when receiving query information, generating corresponding feature score vectors for the query information and the information to be matched in the preset matching library; inputting the feature score vector corresponding to the query information and the feature score vector corresponding to the information to be matched into the information matching model; and determining the matching information corresponding to the query information from the information to be matched according to the matching result determined by the information matching model.

[0215] According to the technical solution of the embodiment of the present invention, the model to be trained is trained by generating a corresponding feature score vector and a matching label of the training information group for the training information group, wherein the feature score vector is obtained based on the training information and can better reflect the training information, then the matching relationship between the information can be reflected by the feature score vector and the matching label of the training information group. The information matching model obtained based on this can realize information matching without requiring a large amount of human resources to maintain and expand similar sentences, which can effectively reduce operating costs.

[0216] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A training method for an information matching model, characterized in that: include: Determine a plurality of training information groups, wherein the training information groups include two pieces of training information and matching labels of the training information groups; Generating a corresponding feature score vector for the training information group according to the two pieces of training information; Using the feature score vector and the matching label, training a preset model to be trained to obtain an information matching model; The training information refers to text information or voice information used to train the information matching model, and the matching label of the training information group is used to indicate the matching status between two pieces of the training information in a training information group; Generating a corresponding feature score vector for the training information group includes: Calculating semantic feature scores for the two pieces of training information in the training information group; Calculating entity feature scores and attribute feature scores corresponding to the training information group; A feature score vector corresponding to the training information group is generated by using the entity feature score, the attribute feature score and the semantic feature score.

2. The training method of the information matching model according to claim 1, characterized in that: Before calculating the entity feature score and the attribute feature score corresponding to the training information group, the method further comprises: generating a corresponding entity feature vector and an attribute feature vector for each segment of the training information; The step of calculating the entity feature score and the attribute feature score corresponding to the training information group includes: calculating the entity feature score and the attribute feature score corresponding to the training information group according to the entity feature vector and the attribute feature vector corresponding to each segment of the training information.

3. The training method of the information matching model according to claim 2, characterized in that: The training information is annotated with entity labels and attribute labels; Generating a corresponding entity feature vector and an attribute feature vector for each segment of the training information, including: Determining entity features and attribute features included in the training information according to the entity label and the attribute label; According to a preset first entity dictionary, the entity feature is converted into a corresponding entity feature vector; According to a preset first attribute dictionary, the attribute feature is converted into a corresponding attribute feature vector.

4. The training method of the information matching model according to claim 1, characterized in that: Calculating semantic feature scores for the two pieces of training information in the training information group includes: Calculating the semantic feature vector corresponding to each segment of the training information in the training information group; A distance score between the semantic feature vectors corresponding to two pieces of the training information is determined, and the distance score is used as a first semantic feature score.

5. The training method of the information matching model according to claim 1 or 4, characterized in that: Calculating semantic feature scores for the two pieces of training information in the training information group includes: Calculating a second semantic feature score according to the length of each segment of training information in the training information group and a preset length; and / or, Determining a maximum matching length between the two pieces of training information in the training information group; A third semantic feature score is calculated according to the maximum matching length and the length of each piece of training information in the training information group.

6. The training method of the information matching model according to claim 1, characterized in that: Further comprising: normalizing the original training information; The determining of the plurality of training information groups includes: determining the plurality of training information groups according to the normalized result.

7. The training method of the information matching model according to any one of claims 1 to 4 and 6, characterized in that: The determining of a plurality of training information groups comprises: When the first training information is received, a preset number of segments of second training information are preliminarily screened out for the first training information from a preset information library; For each segment of the second training information, execute: A training information group is formed by using the second training information, the first training information, and a matching label determined by the first training information and the second training information.

8. An information matching method, characterized in that: include: When receiving the query information, generating corresponding feature score vectors for the query information and a plurality of to-be-matched information in a preset matching library respectively; Inputting the feature score vector corresponding to the query information and the feature score vector corresponding to the information to be matched into an information matching model; the information matching model is obtained based on the training method of the information matching model described in any one of claims 1 to 7; According to the matching result determined by the information matching model, matching information corresponding to the query information is determined from the plurality of information to be matched.

9. The information matching method according to claim 8, characterized in that: Further including: Constructing an inverted index for the preset matching library; Using the inverted index, a preset number of to-be-matched information is screened out for the query information from the preset matching library; The step of generating a corresponding feature score vector for the to-be-matched information in the preset matching library includes: Generate corresponding feature score vectors for the preset number of filtered information to be matched.

10. The information matching method according to claim 8 or 9, characterized in that: The steps of generating corresponding feature score vectors for the query information and the information to be matched in the preset matching library respectively include: Calculate the corresponding semantic feature scores, entity feature scores and attribute feature scores for the query information and the information to be matched respectively; A corresponding feature score vector is generated using the entity feature score, the attribute feature score, and the semantic feature score.

11. The information matching method according to claim 10, characterized in that: Further comprising: generating corresponding entity feature vectors and attribute feature vectors for the query information and the information to be matched, respectively, according to the entity features and attribute features included in the query information and the entity features and attribute features included in the information to be matched; The step of calculating the corresponding entity feature score and attribute feature score for the query information includes: calculating the entity feature score and attribute feature score corresponding to the query information according to the entity feature vector, the attribute feature vector corresponding to the query information, and the entity feature vector and attribute feature vector corresponding to the information to be matched.

12. The information matching method according to claim 10, characterized in that: The information to be matched is marked with entity tags and attribute tags; Generating corresponding entity feature vectors and attribute feature vectors for the query information and the information to be matched respectively, including: Determine entity features and attribute features included in the information to be matched according to the entity tag and the attribute tag marked by the information to be matched; According to a preset second entity dictionary, entity features in the query information and the information to be matched are respectively converted into corresponding entity feature vectors; According to a preset second attribute dictionary, attribute features in the query information and the information to be matched are respectively converted into corresponding attribute feature vectors.

13. The information matching method according to any one of claims 8, 9, 11 and 12, characterized in that: The information to be matched is configured with answer information; The information matching method further comprises: Outputting answer information of the matching information corresponding to the inquiry information to the user who initiated the inquiry information.

14. A training device for an information matching model, characterized in that: include: A determination unit, a processing unit and a training unit, wherein: The determining unit is used to determine a plurality of training information groups, wherein the training information groups include two pieces of training information and matching tags of the training information groups; The processing unit is used to generate a corresponding feature score vector for the training information group according to the two pieces of training information in the training information group determined by the determining unit; The training unit is used to train a preset model to be trained using the feature score vector generated by the processing unit and the matching label to obtain an information matching model; The training information refers to text information or voice information used to train the information matching model, and the matching label of the training information group is used to indicate the matching status between two pieces of the training information in a training information group; The processing unit is used to calculate semantic feature scores for the two pieces of training information in the training information group; calculate the entity feature score and the attribute feature score corresponding to the training information group; and generate a feature score vector corresponding to the training information group using the entity feature score, the attribute feature score and the semantic feature score.

15. An information matching device, characterized in that: include: A vector generation unit and a matching unit, wherein, The vector generating unit is used to generate corresponding feature score vectors for the query information and the information to be matched in the preset matching library respectively when receiving the query information; The matching unit is used to input the feature score vector corresponding to the query information generated by the vector generation unit and the feature score vector corresponding to the information to be matched into an information matching model; and determine the matching information corresponding to the query information from the information to be matched according to the matching result determined by the information matching model, wherein the information matching model is obtained based on the training method of the information matching model described in any one of claims 1 to 7 or the information matching model is obtained based on the training device of the information matching model described in claim 14.

16. An electronic device, characterized in that: include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 13.

17. A computer readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 13 is implemented.

Images