Method, device, computer equipment and storage medium for determining text topic

Abstract

The present invention relates to artificial intelligence, and provides a method, device, computer equipment and storage medium for determining a text topic. The method comprises: dividing a target text into a plurality of text sentences according to a preset sentence segmentation rule; inputting each of the text sentences into a preset Bert model to obtain a sentence vector set composed of vectors of each of the text sentences; determining the cosine similarity between each summary sentence and each text sentence according to the sentence vector set, and obtaining a summary sentence similarity set including the cosine similarity between each summary sentence and each text sentence; determining the sentence importance value of each summary sentence according to the summary sentence similarity set; and determining the text topic of the target text from a plurality of summary sentences according to the sentence importance value. The sentence vector obtained by the Bert model in the embodiment of the present invention combines the semantics of the sentence context, so the accuracy of topic extraction can be improved by determining the text topic using the sentence vector obtained by the Bert model.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method, device, computer equipment and storage medium for determining a text topic. Background Art

[0002] General articles, especially research reports, will include two parts: summary text and main text. In order to allow readers to quickly see the focus of the article, in addition to the summary text, the article will also have a theme. The theme is generally part of the summary, such as one or more sentences in the summary. However, many authors of articles do not indicate the theme. At this time, in order to automatically identify the theme of the article, generally, the keywords of the article are searched, and then the sentences are scored according to the number of keywords in the sentences. Finally, one or more sentences with the highest scores are selected as the theme of the article.

[0003] However, this method ignores the meaning of key sentences in the context, and the extraction of topics is particularly rigid, resulting in low accuracy of topic extraction. Summary of the invention

[0004] The embodiments of the present invention provide a method, an apparatus, a computer device and a storage medium for determining a text topic, which can improve the accuracy of text topic extraction.

[0005] In a first aspect, an embodiment of the present invention provides a method for determining a text topic, which includes:

[0006] Divide the target text into multiple text sentences according to the preset sentence segmentation rules;

[0007] Input each of the text sentences into a preset Bert model to obtain a sentence vector set consisting of the vectors of each of the text sentences;

[0008] Determine the cosine similarity between the summary sentence in the text sentence and each of the text sentences according to the sentence vector set, and obtain a summary sentence similarity set;

[0009] Determining the sentence importance value of each summary sentence according to the summary sentence similarity set;

[0010] A text topic of the target text is determined from the plurality of summary sentences according to the sentence importance values.

[0011] In a second aspect, an embodiment of the present invention further provides a device for determining a text topic, which includes:

[0012] A sentence division unit, used for dividing the target text into multiple text sentences according to a preset sentence division rule;

[0013] An input unit, used to input each of the text sentences into a preset Bert model to obtain a sentence vector set composed of vectors of each of the text sentences;

[0014] A first determining unit is used to determine the cosine similarity between the summary sentence in the text sentence and each of the text sentences according to the sentence vector set, so as to obtain a summary sentence similarity set;

[0015] A second determining unit, configured to determine a sentence importance value of each summary sentence according to the summary sentence similarity set;

[0016] The third determining unit is used to determine the text topic of the target text from the plurality of summary sentences according to the sentence importance values.

[0017] In a third aspect, an embodiment of the present invention further provides a computer device, which includes a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the above method when executing the computer program.

[0018] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, wherein the storage medium stores a computer program, wherein the computer program includes program instructions, and the program instructions can implement the above method when executed by a processor.

[0019] The embodiment of the present invention provides a method, device, computer equipment and storage medium for determining a text topic. The method includes: dividing a target text into multiple text sentences according to a preset sentence segmentation rule; inputting each of the text sentences into a preset Bert model to obtain a sentence vector set composed of vectors of each of the text sentences; determining the cosine similarity between the summary sentence in the text sentence and each of the text sentences according to the sentence vector set to obtain a summary sentence similarity set; determining the sentence importance value of each summary sentence according to the summary sentence similarity set; and determining the text topic of the target text from multiple summary sentences according to the sentence importance value. The sentence vector obtained by the Bert model in the embodiment of the present invention combines the semantics of the sentence context, so the accuracy of topic extraction can be improved by determining the text topic using the sentence vector obtained by the Bert model. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0021] Figure 1A schematic diagram of an application scenario of a method for determining a text topic provided by an embodiment of the present invention;

[0022] Figure 2 A schematic diagram of a flow chart of a method for determining a text topic provided by an embodiment of the present invention;

[0023] Figure 3 A schematic diagram of a sub-flow of a method for determining a text theme provided by an embodiment of the present invention;

[0024] Figure 4 A schematic diagram of another sub-flow of the method for determining a text theme provided by an embodiment of the present invention;

[0025] Figure 5 A schematic diagram of another sub-flow of the method for determining a text theme provided by an embodiment of the present invention;

[0026] Figure 6 A schematic block diagram of a device for determining a text topic provided by an embodiment of the present invention;

[0027] Figure 7 A schematic block diagram of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0030] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0031] It should be further understood that the term "and / or" used in the present description and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0032] The embodiments of the present invention provide a method, an apparatus, a computer device and a storage medium for determining a text topic.

[0033] The executor of the method for determining the text theme may be the device for determining the text theme provided in an embodiment of the present invention, or a computer device integrating the device for determining the text theme, wherein the device for determining the text theme may be implemented in hardware or software, the computer device may be a terminal or a server, and the terminal may be a smart phone, a tablet computer, a PDA, or a laptop computer, etc.

[0034] See also Figure 1 , Figure 1 Schematic diagram of an application scenario of the method for determining a text theme provided by an embodiment of the present invention. The method for determining a text theme is applied to Figure 1 In the computer device 10 in the method, the computer device 10 divides the target text into multiple text sentences according to a preset sentence segmentation rule; inputs each of the text sentences into a preset Bert model to obtain a sentence vector set composed of vectors of each of the text sentences; determines the cosine similarity between the summary sentence in the text sentence and each of the text sentences according to the sentence vector set to obtain a summary sentence similarity set; determines the sentence importance value of each summary sentence according to the summary sentence similarity set; and determines the text theme of the target text from the multiple summary sentences according to the sentence importance values.

[0035] Figure 2 FIG. 1 is a flow chart of a method for determining a text topic provided by an embodiment of the present invention. Figure 2 As shown, the method includes the following steps S110-150.

[0036] S110: Divide the target text into multiple text sentences according to a preset sentence segmentation rule.

[0037] The target text includes the abstract text and the main text. The target text is the text that needs to be subject-extracted, for example, a research report crawled from the official WeChat public account of a financial institution. The sentence segmentation rule is to segment the text according to the sentence segmentation symbols, where the sentence segmentation symbols include: ".", "?", "!" and "\n". The segmented text sentences are sorted according to the order in the target text.

[0038] In some embodiments, since there may be a lot of information that is meaningless to topic extraction in the acquired target text, in order to reduce the impact of this information on topic extraction and improve the accuracy of topic extraction, in some embodiments, step S110 includes: acquiring the target text, performing data cleaning on the target text to obtain a cleaned text, and then dividing the cleaned text into multiple text sentences according to the sentence segmentation rules. For example, the text to be cleaned is determined from the target text according to a preset text template to be cleaned, and then the text to be cleaned in the target text is deleted, wherein the text template to be cleaned includes a reference information template, a risk warning template, a disclaimer template, a legal statement template, an objective description template of the market, etc., which is used to clean the reference information, risk warning, disclaimer, legal statement, and objective description of the market in the original text.

[0039] S120, inputting each of the text sentences into a preset Bert model to obtain a sentence vector set composed of the vectors of each of the text sentences.

[0040] In this embodiment, in order to combine the meaning of the context of each output text sentence to obtain a more accurate sentence vector expression, this embodiment inputs each text sentence into a preset bidirectional encoder (BidirectionalEncoder Representations from Transformer, Bert) model.

[0041] Among them, the Bert model is a natural language model released by Google. It is a neural network model with an attention mechanism. It can personalize the encoding of sentences according to the context of sentences and words. Therefore, according to the Bert model, a sentence vector that combines the semantics of the context can be obtained, making the vector expression more accurate.

[0042] Specifically, the text sentences are input into the preset Bert model and its downstream sentence relevance task to obtain the vector expression of each sentence, and then the vectors of each sentence are assembled in order to obtain a sentence vector set.

[0043] In some embodiments, in order to calculate the relevance between two sentences, a 784-dimensional vector is used to represent the sentence information, thereby obtaining the relevance between the two sentences.

[0044] S130. Determine the cosine similarity between the summary sentence in the text sentence and each of the text sentences according to the sentence vector set, and obtain a summary sentence similarity set.

[0045] Specifically, the cosine similarity between each summary sentence in the sentence vector set and each text sentence in the target text is calculated, and a summary sentence set including the cosine similarity between each summary sentence and each text sentence is obtained.

[0046] In some embodiments, the summary sentence similarity set is represented by a matrix B m*n Represents, where n represents the number of all text sentences in the target text and m represents the number of all sentences in the summary text.

[0047] S140: Determine a sentence importance value of each summary sentence according to the summary sentence similarity set.

[0048] See also Figure 3 In some embodiments, specifically, step S140 includes:

[0049] S141 . Determine a first summary similarity of each summary sentence according to the summary sentence similarity set and a preset similarity threshold.

[0050] That is, the cosine similarities in the summary sentence similarity set that are greater than or equal to the preset similarity threshold are converted to 1, and the cosine similarities in the summary sentence similarity set that are less than the preset similarity threshold are converted to 0, so as to obtain a first summary similarity set.

[0051] At this time, when a cosine similarity in the summary sentence similarity set is greater than the preset similarity threshold, the two sentences corresponding to the cosine similarity are regarded as the same sentence. At this time, in order to improve the calculation efficiency, the value of the cosine similarity in the summary sentence similarity set that is greater than the preset similarity threshold is converted to 1, otherwise it is converted to 0. At this time, the set C of the first summary similarities corresponding to each summary sentence is obtained. m*n , C m*n Each row in corresponds to a summary sentence.

[0052] The preset similarity threshold may be 0.9, and the specific value may be adjusted according to the specific application scenario, so the preset similarity threshold is not limited at this time.

[0053] S142: Determine a sentence importance value of each summary sentence according to the first summary similarity of each summary sentence in the first summary similarity set.

[0054] In some embodiments, step S142 may directly convert C m*n Add the values ​​of each row in , and use the corresponding sum as the sentence importance value of the corresponding summary sentence.

[0055] See also Figure 4In some other embodiments, it is also necessary to determine the sentence importance value of the summary sentence in combination with the second summary similarity. In this case, specifically, step S142 includes:

[0056] S1421 . Determine a second summary similarity of each summary sentence according to the first summary similarity corresponding to each summary sentence and the sentence length corresponding to each summary sentence, to obtain a second summary similarity set.

[0057] That is, according to the first summary similarity of each summary sentence and the sentence length corresponding to each summary sentence, the mapping relationship between each summary sentence and all sentences in the full text is determined, and the second summary similarity set C′ composed of the second summary similarities of each summary sentence is obtained. m*n .

[0058] At this time, it is necessary to obtain the sentence length of each summary sentence and the number of elements whose element value is 1 in the first summary similarity of each summary sentence; then for each first element in the first summary similarity of each summary sentence, multiply the first element by the corresponding sentence length and divide by the corresponding number of elements to obtain a second summary similarity set consisting of the second summary similarities of each summary sentence, wherein the first element is an element in the first summary similarity, including 0 or 1; specifically as follows:

[0059] In this embodiment, C m*n After adding up the rows, the value of each row (the number of 1s in the first summary similarity set corresponding to each summary sentence) is the number of similarities between the corresponding summary sentence and some sentences in the full text (the number of elements). The number of elements is set to h m ,h m is an integer. At this time, specifically, step S1421 is:

[0060] Let the length of the summary sentence m be s m , at this time, the mapping relationship between the summary sentence and the text sentence in the target text is According to the mapping relationship, the mapping matrix C′ is obtained. m*n , c ij C m*n The value of the i-th row and j-th column in .

[0061] S1422: Determine a sentence importance value of each summary sentence according to the second summary similarity set.

[0062] See also Figure 5 In some embodiments, if the target text may contain multiple arguments, specifically, step S1422 includes:

[0063] S14221. Determine the cosine similarity between each text sentence according to the sentence vector set to obtain a text sentence similarity set.

[0064] Specifically, the similarities between the sentence vectors in the sentence vector set are determined, and the similarities between each sentence in the target text and all the text sentences in the target text are obtained, thereby obtaining a text sentence similarity set containing the pairwise similarities between all the sentences in the target text.

[0065] In some embodiments, the text sentence similarity set is represented by the matrix B n*n Represents, where n represents the number of all sentences in the target text.

[0066] S14222. Clustering the text sentence similarity set according to a clustering algorithm to obtain text sentence similarity subsets of multiple clusters.

[0067] In some embodiments, since it is uncertain how many arguments exist in the target text, and the corresponding topic should be extracted for each argument, this embodiment needs to cluster the target text to obtain one or more clusters containing multiple sentences, one cluster corresponding to one argument. At this time, it is necessary to cluster the text sentence similarity set according to the clustering algorithm to obtain text sentence similarity subsets of multiple clusters.

[0068] Specifically, we can consider B n*n is a graph network, B n*n Each element in represents the weight edge of the network, and the B n*n Cluster into several clusters. In some embodiments, the text sentence similarity subset of each cluster after clustering is B′ m′*n′ , m′ is the number of abstracts in the corresponding cluster, and n′ is the number of all sentences in the corresponding cluster. In other embodiments, the text sentence similarity subset of each cluster after clustering is B′ n′*n′ .

[0069] In some embodiments, the classification condition of clusters in the clustering algorithm is that the cosine similarity between sentences is greater than or equal to 0.7, and a cluster must have at least three data points.

[0070] S14223. For each cluster, determine the text importance value of each sub-text sentence according to the sum of the cosine similarities corresponding to each sub-text sentence in the text sentence similarity subset, and obtain a sub-text sentence importance value set corresponding to each cluster.

[0071] The sub-text sentences are text sentences of the corresponding cluster, and the sub-text sentence importance value set D n′*1 Contains the text importance value of each subtext sentence in the corresponding cluster.

[0072] Specifically, in some embodiments, when the text sentence similarity subset of each cluster after clustering is B′ m′*n′ At this time, we need to m′*n′ Add the values ​​of each column in and then transpose to get D n′*1 .

[0073] In some other embodiments, when the text sentence similarity subset of each cluster after clustering is B′ n′*n′ , then we need to change B′ n′*n′ Add the values ​​of each row in to get D n′*1 .

[0074] S14224. Search the second summary similarity set for the subtext sentences of each cluster according to the subtext sentences of each cluster to obtain a second summary similarity subset.

[0075] Specifically, we need to convert C′ m*n The sentences that do not belong to the cluster are deleted, and the second summary similarity corresponding to the subtext sentences of the cluster is retained to obtain the second summary similarity subset of the corresponding cluster. In this embodiment, C′ m*n′ The second summary similarity subset represents the cluster, where n′ represents the total number of sentences in the cluster.

[0076] S14225 . Determine the sentence importance value of each summary sentence corresponding to each cluster according to the sub-text sentence importance value set corresponding to each cluster and the second summary similarity subset corresponding to each cluster.

[0077] Specifically, for each cluster, each second element in the second summary similarity of each summary sentence is obtained, and the second element is multiplied by the text importance value corresponding to the second element in the text importance value set to obtain multiple element importance values; then, for each summary sentence, the corresponding element importance values ​​are accumulated to obtain the sentence importance values ​​of each summary sentence corresponding to each cluster, wherein the second element is the element in the second summary similarity.

[0078] That is, for each cluster, the importance value of each summary sentence is calculated separately. The specific calculation formula is: E m*1 =C′ m*n′ *D n′*1 , where E m*1 It is a set of important values ​​of each summary sentence in a cluster, representing the importance of each summary sentence in this cluster.

[0079] In some other embodiments, if the target text has only one argument by default, then step S1422 includes:

[0080] a. Determine the cosine similarity between each text sentence according to the sentence vector set to obtain a text sentence similarity set.

[0081] This step is similar to step S14221 and will not be described in detail here.

[0082] b. Determine the sum of the cosine similarities corresponding to each text sentence in the text sentence similarity set as the text importance value of each text sentence, and obtain a text importance value set including the text importance values ​​of each text sentence;

[0083] That is, the cosine similarities related to a text sentence in the text sentence similarity set are added, and the sum obtained is the text importance value of the text sentence. The above processing is performed on each text sentence in the target text to obtain a text importance value set containing the text importance value of each text sentence; at this time, the text importance value set is represented by the matrix D n*1 Logo.

[0084] c. Determine the sentence importance value of each summary sentence according to the second summary similarity set and the text importance value set.

[0085] In this embodiment, for each second element in the second summary similarity of each summary sentence, the second element is multiplied by the text importance value corresponding to the second element in the text importance value set to obtain multiple element importance values; then, for each summary sentence, the corresponding element importance values ​​are accumulated to obtain the sentence importance value of each summary sentence.

[0086] Specifically, the sentence importance value of each summary sentence can be determined using the calculation formula of the sentence importance value. The calculation formula of the sentence importance value is: m*1 =C' m*n *D n*1 At this time, E m*1 It is a set of important values ​​of each summary sentence in the target text, including the sentence importance value of each summary sentence.

[0087] S150: Determine the theme of the target text from the summary sentence according to the sentence importance value.

[0088] In some embodiments, when the target text is clustered to obtain multiple clusters, step S150 includes: obtaining the number of clusters of the clusters; determining the number of topics according to a preset rule for determining the number of topics and the number of clusters; and then selecting the summary sentences with the largest number of topics and the largest sentence importance values ​​from the multiple summary sentences as the text topics of the corresponding cluster based on the sentence importance values ​​of each summary sentence corresponding to each cluster.

[0089] For example, if the number of clusters is z, the number of topics is: g = max(1, 4-z).

[0090] The topic extraction is performed for each cluster. At this time, according to the E corresponding to each cluster m*1 Determine the g summary sentences with the highest values ​​in each cluster as the topics of the corresponding cluster.

[0091] In other embodiments, if the target text is not clustered, then according to a preset number of topics g, g summary sentences with the largest sentence importance values ​​can be obtained from multiple summary sentences as the topics of the target text.

[0092] For ease of understanding, a specific embodiment is provided below:

[0093] The abstract information of the article is a1, and the information of the main text is a2, then the cosine similarity matrix C is calculated m*n , m represents the number of sentences in the abstract a1, and n represents the number of all sentences in the article, that is, the sum of the number of sentences in a1 and the number of sentences in a2.

[0094] C m*n Element C ij Values ​​greater than 0.9 become 1, and the rest become 0. m*n The number of elements in each row that are equal to 1 is h m ; Considering that longer sentences may contain more information, we set the weights of the m sentences in C. The longer the sentence, the greater the weight. Let C m*n The length of the sentence corresponding to the middle line is s m , the weighted value of each sentence is Thus we get the matrix C′ m*n .

[0095] Calculate the similarity between all sentences in the article and get B. Since B contains all the sentences in the article, which may contain multiple arguments, it is necessary to cluster the single sentences into clusters through a clustering algorithm, and then divide B into multiple B′.

[0096] For each B′, add the elements of each row to get D n′ , represents the importance of each sentence in the article, and n' represents the number of sentences in the sub-cluster after clustering. The weight of the important information in this cluster mapped to each sentence in the summary is E m*1 =C' m*n′ *D n′ , where m represents the number of sentences in the abstract and n′ represents the number of sentences in the subcluster.

[0097] Determine E m*1 The first g sentences are key sentences, g = max(1, 4-z), where z is the number of clusters obtained by clustering.

[0098] In the prior art, a probabilistic method called Latent Dirichlet Allocation (LDA) can also be used. This method can find article keywords and sentences, but it requires specifying the number of article topics. If the number of topics is inconsistent with the preset number, the extracted article topics will be ambiguous.

[0099] This embodiment does not need to specify the number of topics in advance, but automatically determines the number of target texts according to the spectral clustering method, and then determines the number of topics according to the number of target texts, without the risk of ambiguity in extracting topics.

[0100] In some embodiments, after determining the number of topics g of the target text, the first g summary sentences with the last importance value of the target sentence are obtained from the summary text as the main topic of the corresponding target text.

[0101] In other embodiments, if the number of topics of the target text does not need to be determined beforehand, a default (preset) number of topics may be used to extract a default number of text topics from the summary text according to the importance value of the target sentence.

[0102] In the prior art, if you want to obtain the theme of a text through the Bert model, you first need to train the model, where the training method is: 1. Through the Bert model, convert each sentence of the article into a vector form to form a matrix X; 2. Manually mark each sentence with 0 or 1. If the sentence is the theme sentence of the article, it is marked as 1, otherwise it is 0, so as to form a list Y. After that, first, X and Y are input into the recursive neural network LSTM. This is because LSTM has forward memory. The algorithm will not only focus on the current sentence, but also consider the previous sentences, and truly consider the annotation results from the entire article. Then, set the activation function to the Sigmod function. Finally, the model can be obtained after parameter optimization. However, using this method to train the model, data annotation takes a lot of time, and the accuracy of the model depends on the amount of original data and the quality of annotation.

[0103] This embodiment does not require model training and can also extract topics from the text, which reduces a lot of annotation time and solves the problem that manual annotation is limited by personal consciousness, resulting in a lot of important information not being annotated, resulting in low coverage of the model's output results.

[0104] In summary, this embodiment divides the target text into multiple text sentences according to a preset sentence segmentation rule; inputs each of the text sentences into a preset Bert model to obtain a sentence vector set composed of vectors of each of the text sentences; determines the cosine similarity between the summary sentence in the text sentence and each of the text sentences according to the sentence vector set to obtain a summary sentence similarity set; determines the sentence importance value of each summary sentence according to the summary sentence similarity set; and determines the text theme of the target text from the multiple summary sentences according to the sentence importance value. The sentence vector obtained by the Bert model in the embodiment of the present invention combines the semantics of the sentence context, so by using the sentence vector obtained by the Bert model to determine the text theme, the accuracy of theme extraction can be improved.

[0105] It should be noted that the embodiments of the present invention can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0106] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0107] Figure 6 is a schematic block diagram of a device for determining a text topic provided by an embodiment of the present invention. Figure 6 As shown, corresponding to the above text theme determination method, the present invention also provides a text theme determination device. The text theme determination device includes a unit for executing the above text theme determination method, and the device can be configured in a desktop computer, a tablet computer, a laptop computer, etc. Specifically, please refer to Figure 6 The device for determining the text topic includes a sentence unit 601, an input unit 602, a first determining unit 603, a second determining unit 604 and a third determining unit 605.

[0108] The sentence division unit 601 divides the target text into a plurality of text sentences according to a preset sentence division rule;

[0109] An input unit 602 inputs each of the text sentences into a preset Bert model to obtain a sentence vector set consisting of vectors of each of the text sentences;

[0110] A first determining unit 603 is used to determine the cosine similarity between the summary sentence in the text sentence and each of the text sentences according to the sentence vector set, to obtain a summary sentence similarity set;

[0111] A second determining unit 604 is used to determine the sentence importance value of each summary sentence according to the summary sentence similarity set;

[0112] The third determining unit 605 is configured to determine the text topic of the target text from the plurality of summary sentences according to the sentence importance values.

[0113] In some embodiments, the second determining unit 604 is specifically configured to:

[0114] The cosine similarities in the summary sentence similarity set that are greater than or equal to a preset similarity threshold are converted to 1, and the cosine similarities in the summary sentence similarity set that are less than the preset similarity threshold are converted to 0, to obtain a first summary similarity set;

[0115] The sentence importance value of each summary sentence is determined according to the first summary similarity of each summary sentence in the first summary similarity set.

[0116] In some embodiments, when the second determining unit 604 executes the step of determining the sentence importance value of each summary sentence according to the first summary similarity of each summary sentence in the first summary similarity set, it is further specifically configured to:

[0117] Obtain the sentence length of each summary sentence, and obtain the number of elements whose element value is 1 in the first summary similarity of each summary sentence;

[0118] For each first element in the first summary similarity of each summary sentence, multiply the first element by the corresponding sentence length and divide by the corresponding number of elements to obtain a second summary similarity set consisting of the second summary similarities of each summary sentence;

[0119] The sentence importance value of each summary sentence is determined according to the second summary similarity set.

[0120] In some embodiments, when the second determining unit 604 performs the step of determining the sentence importance value of each summary sentence according to the second summary similarity set, it is further specifically configured to:

[0121] Determine the cosine similarity between each text sentence according to the sentence vector set to obtain a text sentence similarity set;

[0122] Determine the sum of the cosine similarities corresponding to each text sentence in the text sentence similarity set as the text importance value of each text sentence, and obtain a text importance value set including the text importance values ​​of each text sentence;

[0123] For each second element in the second summary similarity of each summary sentence, multiply the second element by the text importance value corresponding to the second element in the text importance value set to obtain a plurality of element importance values;

[0124] For each summary sentence, the corresponding element importance values ​​are accumulated to obtain the sentence importance value of each summary sentence.

[0125] In some embodiments, after determining the cosine similarity between the text sentences according to the sentence vector set to obtain the text sentence similarity set, the second determining unit 604 further performs:

[0126] Clustering the text sentence similarity set according to a clustering algorithm to obtain text sentence similarity subsets of multiple clusters;

[0127] At this time, when the second determining unit 604 performs the step of determining the sum of the cosine similarities corresponding to each text sentence in the text sentence similarity set as the text importance value of each text sentence to obtain a text importance value set including the text importance values ​​of each text sentence, it is further specifically used to:

[0128] For each cluster, the text importance value of each sub-text sentence is determined according to the sum of the cosine similarities corresponding to each sub-text sentence in the text sentence similarity subset, and a sub-text sentence importance value set corresponding to each cluster is obtained, wherein the sub-text sentence is the text sentence of the corresponding cluster, and the sub-text sentence importance value set contains the text importance value of each sub-text sentence in the corresponding cluster.

[0129] At this time, in some embodiments, when the second determining unit 604 performs the step of determining the sentence importance value of each summary sentence according to the second summary similarity set and the text importance value of each text sentence, it is further specifically configured to:

[0130] According to the subtext sentences of each cluster, searching the second summary similarity of the subtext sentences of each cluster from the second summary similarity set to obtain a second summary similarity subset;

[0131] The sentence importance values ​​of the summary sentences corresponding to the clusters are determined according to the subtext sentence importance value sets corresponding to the clusters and the second summary similarity subsets corresponding to the clusters.

[0132] In some embodiments, the third determining unit 605 is specifically configured to:

[0133] Obtaining the number of clusters of the cluster;

[0134] Determine the number of topics according to a preset rule for determining the number of topics and the number of clusters;

[0135] According to the sentence importance values ​​of the summary sentences corresponding to the clusters, the summary sentences with the maximum number of topics and the largest sentence importance values ​​are selected from the multiple summary sentences as the text topics of the corresponding cluster.

[0136] It should be noted that those skilled in the art can clearly understand that the specific implementation process of the above-mentioned text topic determination device and each unit can refer to the corresponding description in the aforementioned method embodiment, and for the convenience and brevity of description, they will not be repeated here.

[0137] The above-mentioned text theme determination device can be implemented in the form of a computer program. The computer program can be used in Figure 7 Runs on the computer device shown.

[0138] See also Figure 7 , Figure 7 700 is a schematic block diagram of a computer device provided in an embodiment of the present invention. The computer device 700 may be a terminal or a server, wherein the terminal may be an electronic device with communication functions such as a smart phone, a tablet computer, a laptop computer, a desktop computer, a personal digital assistant, and a wearable device. The server may be an independent server or a server cluster composed of multiple servers.

[0139] See also Figure 7 The computer device 700 includes a processor 702 , a memory and a network interface 705 connected via a system bus 701 , wherein the memory may include a non-volatile storage medium 703 and an internal memory 704 .

[0140] The non-volatile storage medium 703 can store an operating system 7031 and a computer program 7032. The computer program 7032 includes program instructions, and when the program instructions are executed, the processor 702 can execute a method for determining a text theme.

[0141] The processor 702 is used to provide computing and control capabilities to support the operation of the entire computer device 700 .

[0142] The internal memory 704 provides an environment for the operation of the computer program 7032 in the non-volatile storage medium 703. When the computer program 7032 is executed by the processor 702, the processor 702 can execute a method for determining a text theme.

[0143] The network interface 705 is used to communicate with other devices over the network. Figure 7The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device 700 to which the solution of the present invention is applied. The specific computer device 700 may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0144] The processor 702 is used to run the computer program 7032 stored in the memory to implement the following steps:

[0145] Divide the target text into multiple text sentences according to the preset sentence segmentation rules;

[0146] Input each of the text sentences into a preset Bert model to obtain a sentence vector set consisting of the vectors of each of the text sentences;

[0147] Determine the cosine similarity between the summary sentence in the text sentence and each of the text sentences according to the sentence vector set, and obtain a summary sentence similarity set;

[0148] Determining the sentence importance value of each summary sentence according to the summary sentence similarity set;

[0149] A text topic of the target text is determined from the plurality of summary sentences according to the sentence importance values.

[0150] In one embodiment, when the processor 702 implements the step of determining the sentence importance value of each summary sentence according to the summary sentence similarity set, the processor 702 specifically implements the following steps:

[0151] The cosine similarities in the summary sentence similarity set that are greater than or equal to a preset similarity threshold are converted to 1, and the cosine similarities in the summary sentence similarity set that are less than the preset similarity threshold are converted to 0, to obtain a first summary similarity set;

[0152] The sentence importance value of each summary sentence is determined according to the first summary similarity of each summary sentence in the first summary similarity set.

[0153] In one embodiment, when the processor 702 implements the step of determining the sentence importance value of each summary sentence according to the first summary similarity of each summary sentence in the first summary similarity set, the processor 702 specifically implements the following steps:

[0154] Obtain the sentence length of each summary sentence, and obtain the number of elements whose element value is 1 in the first summary similarity of each summary sentence;

[0155] For each first element in the first summary similarity of each summary sentence, multiply the first element by the corresponding sentence length and divide by the corresponding number of elements to obtain a second summary similarity set consisting of the second summary similarities of each summary sentence;

[0156] The sentence importance value of each summary sentence is determined according to the second summary similarity set.

[0157] In one embodiment, when the processor 702 implements the step of determining the sentence importance value of each summary sentence according to the second summary similarity set, the processor 702 specifically implements the following steps:

[0158] Determine the cosine similarity between each text sentence according to the sentence vector set to obtain a text sentence similarity set;

[0159] Determine the sum of the cosine similarities corresponding to each text sentence in the text sentence similarity set as the text importance value of each text sentence, and obtain a text importance value set including the text importance values ​​of each text sentence;

[0160] For each second element in the second summary similarity of each summary sentence, multiply the second element by the text importance value corresponding to the second element in the text importance value set to obtain a plurality of element importance values;

[0161] For each summary sentence, the corresponding element importance values ​​are accumulated to obtain the sentence importance value of each summary sentence.

[0162] In one embodiment, after implementing the step of determining the cosine similarity between each text sentence according to the sentence vector set to obtain the text sentence similarity set, the processor 702 further implements the following steps:

[0163] Clustering the text sentence similarity set according to a clustering algorithm to obtain text sentence similarity subsets of multiple clusters;

[0164] The step of determining the sum of the cosine similarities corresponding to each text sentence in the text sentence similarity set as the text importance value of each text sentence, and obtaining a text importance value set including the text importance values ​​of each text sentence, comprises:

[0165] For each cluster, the text importance value of each sub-text sentence is determined according to the sum of the cosine similarities corresponding to each sub-text sentence in the text sentence similarity subset, and a sub-text sentence importance value set corresponding to each cluster is obtained, wherein the sub-text sentence is the text sentence of the corresponding cluster, and the sub-text sentence importance value set contains the text importance value of each sub-text sentence in the corresponding cluster.

[0166] In one embodiment, when the processor 702 implements the step of determining the sentence importance value of each summary sentence according to the second summary similarity set, the processor 702 specifically implements the following steps:

[0167] According to the subtext sentences of each cluster, searching the second summary similarity of the subtext sentences of each cluster from the second summary similarity set to obtain a second summary similarity subset;

[0168] The sentence importance values ​​of the summary sentences corresponding to the clusters are determined according to the subtext sentence importance value sets corresponding to the clusters and the second summary similarity subsets corresponding to the clusters.

[0169] In one embodiment, when the processor 702 implements the step of determining the text topic of the target text from the plurality of summary sentences according to the sentence importance values, the processor 702 specifically implements the following steps:

[0170] Obtaining the number of clusters of the cluster;

[0171] Determine the number of topics according to a preset rule for determining the number of topics and the number of clusters;

[0172] According to the sentence importance values ​​of the summary sentences corresponding to the clusters, the summary sentences with the maximum number of topics and the largest sentence importance values ​​are selected from the multiple summary sentences as the text topics of the corresponding cluster.

[0173] It should be understood that in the embodiment of the present invention, the processor 702 may be a central processing unit (CPU), and the processor 702 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0174] It can be understood by those skilled in the art that all or part of the processes in the method for implementing the above embodiment can be completed by instructing the relevant hardware through a computer program. The computer program includes program instructions, and the computer program can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the embodiment of the above method.

[0175] Therefore, the present invention also provides a storage medium. The storage medium may be a computer-readable storage medium. The storage medium stores a computer program, wherein the computer program includes program instructions. When the program instructions are executed by a processor, the processor executes the following steps:

[0176] Divide the target text into multiple text sentences according to the preset sentence segmentation rules;

[0177] Input each of the text sentences into a preset Bert model to obtain a sentence vector set consisting of the vectors of each of the text sentences;

[0178] Determine the cosine similarity between the summary sentence in the text sentence and each of the text sentences according to the sentence vector set, and obtain a summary sentence similarity set;

[0179] Determining the sentence importance value of each summary sentence according to the summary sentence similarity set;

[0180] A text topic of the target text is determined from the plurality of summary sentences according to the sentence importance values.

[0181] In one embodiment, when the processor executes the program instructions to implement the step of determining the sentence importance value of each summary sentence according to the summary sentence similarity set, the processor specifically implements the following steps:

[0182] The cosine similarities in the summary sentence similarity set that are greater than or equal to a preset similarity threshold are converted to 1, and the cosine similarities in the summary sentence similarity set that are less than the preset similarity threshold are converted to 0, to obtain a first summary similarity set;

[0183] The sentence importance value of each summary sentence is determined according to the first summary similarity of each summary sentence in the first summary similarity set.

[0184] In one embodiment, when the processor executes the step of determining the sentence importance value of each summary sentence according to the first summary similarity of each summary sentence in the first summary similarity set, the processor specifically implements the following steps:

[0185] Obtain the sentence length of each summary sentence, and obtain the number of elements whose element value is 1 in the first summary similarity of each summary sentence;

[0186] For each first element in the first summary similarity of each summary sentence, multiply the first element by the corresponding sentence length and divide by the corresponding number of elements to obtain a second summary similarity set consisting of the second summary similarities of each summary sentence;

[0187] The sentence importance value of each summary sentence is determined according to the second summary similarity set.

[0188] In one embodiment, when the processor executes the step of determining the sentence importance value of each summary sentence according to the second summary similarity set, the processor specifically implements the following steps:

[0189] Determine the cosine similarity between each text sentence according to the sentence vector set to obtain a text sentence similarity set;

[0190] Determine the sum of the cosine similarities corresponding to each text sentence in the text sentence similarity set as the text importance value of each text sentence, and obtain a text importance value set including the text importance values ​​of each text sentence;

[0191] For each second element in the second summary similarity of each summary sentence, multiply the second element by the text importance value corresponding to the second element in the text importance value set to obtain a plurality of element importance values;

[0192] For each summary sentence, the corresponding element importance values ​​are accumulated to obtain the sentence importance value of each summary sentence.

[0193] In one embodiment, after executing the step of determining the cosine similarity between each text sentence according to the sentence vector set to obtain a text sentence similarity set, the processor further implements the following steps:

[0194] Clustering the text sentence similarity set according to a clustering algorithm to obtain text sentence similarity subsets of multiple clusters;

[0195] At this time, in one embodiment, when the processor executes the step of determining the sum of the cosine similarities corresponding to each text sentence in the text sentence similarity set as the text importance value of each text sentence to obtain a text importance value set including the text importance values ​​of each text sentence, the processor specifically implements the following steps:

[0196] For each cluster, the text importance value of each sub-text sentence is determined according to the sum of the cosine similarities corresponding to each sub-text sentence in the text sentence similarity subset, and a sub-text sentence importance value set corresponding to each cluster is obtained, wherein the sub-text sentence is the text sentence of the corresponding cluster, and the sub-text sentence importance value set contains the text importance value of each sub-text sentence in the corresponding cluster.

[0197] In one embodiment, when the processor executes the step of determining the sentence importance value of each summary sentence according to the second summary similarity set, the processor specifically implements the following steps:

[0198] According to the subtext sentences of each cluster, searching the second summary similarity of the subtext sentences of each cluster from the second summary similarity set to obtain a second summary similarity subset;

[0199] The sentence importance values ​​of the summary sentences corresponding to the clusters are determined according to the subtext sentence importance value sets corresponding to the clusters and the second summary similarity subsets corresponding to the clusters.

[0200] In one embodiment, when the processor executes the step of determining the text topic of the target text from the plurality of summary sentences according to the sentence importance values, the processor specifically implements the following steps:

[0201] Obtaining the number of clusters of the cluster;

[0202] Determine the number of topics according to a preset rule for determining the number of topics and the number of clusters;

[0203] According to the sentence importance values ​​of the summary sentences corresponding to the clusters, the summary sentences with the maximum number of topics and the largest sentence importance values ​​are selected from the multiple summary sentences as the text topics of the corresponding cluster.

[0204] The storage medium may be a USB flash drive, a mobile hard disk, a read-only memory (ROM), a magnetic disk, or an optical disk, etc., which are computer-readable storage media that can store program codes.

[0205] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0206] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of each unit is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0207] The steps in the method of the embodiment of the present invention can be adjusted in order, combined and deleted according to actual needs. The units in the device of the embodiment of the present invention can be combined, divided and deleted according to actual needs. In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0208] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, terminal, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention.

[0209] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for determining a text topic, characterized in that: include: Divide the target text into multiple text sentences according to the preset sentence segmentation rules; Input each of the text sentences into a preset Bert model to obtain a sentence vector set consisting of the vectors of each of the text sentences; Determine the cosine similarity between the summary sentence in the text sentence and each of the text sentences according to the sentence vector set, and obtain a summary sentence similarity set; Determining the sentence importance value of each summary sentence according to the summary sentence similarity set; Determining a text topic of the target text from a plurality of the summary sentences according to the sentence importance values; The determining the sentence importance value of each summary sentence according to the summary sentence similarity set includes: The cosine similarities in the summary sentence similarity set that are greater than or equal to a preset similarity threshold are converted to 1, and the cosine similarities in the summary sentence similarity set that are less than the preset similarity threshold are converted to 0, to obtain a first summary similarity set; Determining a sentence importance value of each summary sentence according to the first summary similarity of each summary sentence in the first summary similarity set; The determining the sentence importance value of each summary sentence according to the first summary similarity of each summary sentence in the first summary similarity set includes: Obtain the sentence length of each summary sentence, and obtain the number of elements whose element value is 1 in the first summary similarity of each summary sentence; For each first element in the first summary similarity of each summary sentence, multiply the first element by the corresponding sentence length and divide by the corresponding number of elements to obtain a second summary similarity set consisting of the second summary similarities of each summary sentence; Determining a sentence importance value of each summary sentence according to the second summary similarity set; The determining the sentence importance value of each summary sentence according to the second summary similarity set includes: Determine the cosine similarity between each text sentence according to the sentence vector set to obtain a text sentence similarity set; Determine the sum of the cosine similarities corresponding to each text sentence in the text sentence similarity set as the text importance value of each text sentence, and obtain a text importance value set including the text importance values ​​of each text sentence; For each second element in the second summary similarity of each summary sentence, multiply the second element by the text importance value corresponding to the second element in the text importance value set to obtain a plurality of element importance values; For each summary sentence, the corresponding element importance values ​​are accumulated to obtain the sentence importance value of each summary sentence; After determining the cosine similarity between the text sentences according to the sentence vector set and obtaining the text sentence similarity set, the method further includes: Clustering the text sentence similarity set according to a clustering algorithm to obtain text sentence similarity subsets of multiple clusters; The step of determining the sum of the cosine similarities corresponding to each text sentence in the text sentence similarity set as the text importance value of each text sentence, and obtaining a text importance value set including the text importance values ​​of each text sentence, comprises: For each cluster, the text importance value of each sub-text sentence is determined according to the sum of the cosine similarities corresponding to each sub-text sentence in the text sentence similarity subset, and a sub-text sentence importance value set corresponding to each cluster is obtained, wherein the sub-text sentence is the text sentence of the corresponding cluster, and the sub-text sentence importance value set contains the text importance value of each sub-text sentence in the corresponding cluster.

2. The method according to claim 1, characterized in that The determining the sentence importance value of each summary sentence according to the second summary similarity set includes: According to the subtext sentences of each cluster, searching the second summary similarity of the subtext sentences of each cluster from the second summary similarity set to obtain a second summary similarity subset; The sentence importance values ​​of the summary sentences corresponding to the clusters are determined according to the subtext sentence importance value sets corresponding to the clusters and the second summary similarity subsets corresponding to the clusters.

3. The method according to claim 2, characterized in that The step of determining the text topic of the target text from the plurality of summary sentences according to the sentence importance values ​​comprises: Obtaining the number of clusters of the cluster; Determine the number of topics according to a preset rule for determining the number of topics and the number of clusters; According to the sentence importance values ​​of the summary sentences corresponding to the clusters, the summary sentences with the maximum number of topics and the largest sentence importance values ​​are selected from the multiple summary sentences as the text topics of the corresponding cluster.

4. A device for determining a text topic, characterized in that: include: A sentence division unit, used for dividing the target text into multiple text sentences according to a preset sentence division rule; An input unit, used to input each of the text sentences into a preset Bert model to obtain a sentence vector set composed of vectors of each of the text sentences; A first determining unit is used to determine the cosine similarity between the summary sentence in the text sentence and each of the text sentences according to the sentence vector set, so as to obtain a summary sentence similarity set; A second determining unit, configured to determine a sentence importance value of each summary sentence according to the summary sentence similarity set; A third determining unit, configured to determine a text topic of the target text from a plurality of the summary sentences according to the sentence importance values; The second determining unit is specifically configured to: The cosine similarities in the summary sentence similarity set that are greater than or equal to a preset similarity threshold are converted to 1, and the cosine similarities in the summary sentence similarity set that are less than the preset similarity threshold are converted to 0, to obtain a first summary similarity set; Determining a sentence importance value of each summary sentence according to the first summary similarity of each summary sentence in the first summary similarity set; When the second determining unit executes the step of determining the sentence importance value of each summary sentence according to the first summary similarity of each summary sentence in the first summary similarity set, it is specifically used to: Obtain the sentence length of each summary sentence, and obtain the number of elements whose element value is 1 in the first summary similarity of each summary sentence; For each first element in the first summary similarity of each summary sentence, multiply the first element by the corresponding sentence length and divide by the corresponding number of elements to obtain a second summary similarity set consisting of the second summary similarities of each summary sentence; Determining a sentence importance value of each summary sentence according to the second summary similarity set; When the second determining unit executes the step of determining the sentence importance value of each summary sentence according to the second summary similarity set, it is specifically used to: Determine the cosine similarity between each text sentence according to the sentence vector set to obtain a text sentence similarity set; Determine the sum of the cosine similarities corresponding to each text sentence in the text sentence similarity set as the text importance value of each text sentence, and obtain a text importance value set including the text importance values ​​of each text sentence; For each second element in the second summary similarity of each summary sentence, multiply the second element by the text importance value corresponding to the second element in the text importance value set to obtain a plurality of element importance values; For each summary sentence, the corresponding element importance values ​​are accumulated to obtain the sentence importance value of each summary sentence; After the second determining unit determines the cosine similarity between the text sentences according to the sentence vector set to obtain the text sentence similarity set, it further performs: Clustering the text sentence similarity set according to a clustering algorithm to obtain text sentence similarity subsets of multiple clusters; When the second determining unit performs the step of determining the sum of the cosine similarities corresponding to each text sentence in the text sentence similarity set as the text importance value of each text sentence to obtain a text importance value set including the text importance values ​​of each text sentence, it is specifically used to: For each cluster, the text importance value of each sub-text sentence is determined according to the sum of the cosine similarities corresponding to each sub-text sentence in the text sentence similarity subset, and a sub-text sentence importance value set corresponding to each cluster is obtained, wherein the sub-text sentence is the text sentence of the corresponding cluster, and the sub-text sentence importance value set contains the text importance value of each sub-text sentence in the corresponding cluster.

5. A computer device, characterized in that: The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 3 when executing the computer program.

6. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the method according to any one of claims 1 to 3 can be implemented.

Images