Text selection method and device, electronic equipment and readable storage medium

Abstract

The present disclosure provides a text selection method and device, electronic equipment and readable storage medium, which are related to the technical fields of artificial intelligence such as natural language processing, deep learning and cloud service. The text selection method comprises: obtaining a target semantic unit and a to-be-processed text; selecting a first candidate text from the to-be-processed text according to the target semantic unit; obtaining a prediction result of the first candidate text using a first classification model, selecting a second candidate text corresponding to the prediction result and the target semantic unit as the first candidate text, and obtaining a prediction result of the second candidate text using a plurality of second classification models according to a sentence segmentation result of the second candidate text, and selecting a target text corresponding to the prediction result and the target semantic unit as the second candidate text. The present disclosure can achieve the purpose of selecting a target text that matches the target semantic unit more closely, thereby improving the accuracy of the selected target text and reducing the cost and complexity of selecting the target text.

Description

Technical Field

[0001] This disclosure relates to the field of data processing technology, specifically to artificial intelligence technologies such as natural language processing, deep learning, and cloud services. It provides a text selection method, apparatus, electronic device, and readable storage medium. Background Technology

[0002] Existing technologies require the use of labeled text with defined types and / or domains to train a text classification model, enabling the trained model to classify the type and / or domain of the input text. Therefore, the classification accuracy of the trained model heavily depends on the accuracy of the text data used for training.

[0003] The usual approach is to write rules to select texts that can be used to train a text classification model from coarsely labeled or unlabeled texts. This results in high cost and complexity for text selection, and the accuracy of the selected texts is also reduced due to the limitations of the written rules. Summary of the Invention

[0004] According to a first aspect of this disclosure, a text selection method is provided, comprising: acquiring a target semantic unit and a text to be processed; selecting a first candidate text from the text to be processed based on the target semantic unit; obtaining a prediction result of the first candidate text using a first classification model, and selecting the first candidate text whose prediction result corresponds to the target semantic unit as a second candidate text; obtaining a prediction result of the second candidate text using multiple second classification models based on the sentence segmentation result of the second candidate text, and selecting the second candidate text whose prediction result corresponds to the target semantic unit as the target text.

[0005] According to a second aspect of this disclosure, a text selection apparatus is provided, comprising: an acquisition unit for acquiring a target semantic unit and a text to be processed; a first selection unit for selecting a first candidate text from the text to be processed based on the target semantic unit; a second selection unit for obtaining a prediction result of the first candidate text using a first classification model, and selecting the first candidate text whose prediction result corresponds to the target semantic unit as a second candidate text; and a third selection unit for obtaining a prediction result of the second candidate text using multiple second classification models based on the sentence segmentation result of the second candidate text, and selecting the second candidate text whose prediction result corresponds to the target semantic unit as the target text.

[0006] According to a third aspect of this disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method as described above.

[0007] According to a fourth aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are used to cause the computer to perform the method described above.

[0008] According to a fifth aspect of this disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method described above.

[0009] As can be seen from the above technical solutions, this disclosure achieves the purpose of selecting target text that is more matched with the target semantic unit by using different classification models to predict the first candidate text and the second candidate text respectively. This can improve the accuracy of the selected target text, reduce the cost and complexity of selecting target text, and thus improve the accuracy of the text classification model trained based on the target text.

[0010] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0011] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0012] Figure 1 This is a schematic diagram based on the first embodiment of the present disclosure;

[0013] Figure 2 This is a schematic diagram according to the second embodiment of the present disclosure;

[0014] Figure 3 This is a schematic diagram according to the third embodiment of the present disclosure;

[0015] Figure 4 This is a block diagram of an electronic device used to implement the text selection method of the embodiments of this disclosure. Detailed Implementation

[0016] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and mechanisms are omitted in the following description.

[0017] Figure 1 This is a schematic diagram based on the first embodiment of this disclosure. (See diagram below.) Figure 1 As shown, the text selection method in this embodiment specifically includes the following steps:

[0018] S101. Obtain the target semantic unit and the text to be processed;

[0019] S102. Select a first candidate text from the text to be processed based on the target semantic unit;

[0020] S103. Use the first classification model to obtain the prediction result of the first candidate text, and select the first candidate text corresponding to the prediction result and the target semantic unit as the second candidate text;

[0021] S104. Based on the sentence segmentation results of the second candidate text, multiple second classification models are used to obtain the prediction results of the second candidate text. The second candidate text corresponding to the prediction results of the target semantic unit is selected as the target text.

[0022] The text selection method in this embodiment selects a first candidate text from the text to be processed based on the target semantic unit. Then, it uses different classification models to predict the first candidate text and the second candidate text respectively. This achieves the purpose of selecting a target text that is more consistent with the target semantic unit, which can improve the accuracy of the selected target text, reduce the cost and complexity of selecting the target text, and thus improve the accuracy of the text classification model trained based on the target text.

[0023] The target semantic unit obtained by executing S101 in this embodiment can be a word or a phrase. When executing S101 in this embodiment, the word or phrase input by the input terminal can be used as the target semantic unit, or the word or phrase selected by the input terminal on the network can be used as the target semantic unit.

[0024] The target semantic unit obtained by executing S101 in this embodiment can be any type and / or any domain of words or phrases. For example, the type can be emotional type, sensitive type, etc., and the domain can be economic domain, technology domain, etc. The number of target semantic units can be one or multiple. Different target semantic units can correspond to different types and / or domains, or they can correspond to the same type and / or domain.

[0025] In addition, when performing S101 to obtain the target semantic unit, this embodiment may also adopt the following method: obtain historical target text, which is the target text corresponding to the target semantic unit obtained previously; obtain candidate semantic units from the historical target text, and obtain the statistical value of the obtained candidate semantic units; take the candidate semantic units whose statistical values ​​meet the preset requirements as the target semantic units.

[0026] In other words, this embodiment obtains target semantic units by using historical target text, making the obtained target semantic units more accurate, thereby improving the accuracy of the selected target text.

[0027] In this embodiment, the statistical value obtained by performing S101 can be at least one of the following: Term Frequency (TF), Inverse Document Frequency (IDF), and Term Frequency-Inverse Document Frequency (TF-IDF).

[0028] In this embodiment, when performing S101 to obtain candidate semantic units from historical target text, a word mining method based on mutual information and / or cohesion can be used to mine words and / or phrases from historical target text as candidate semantic units; alternatively, a word segmentation method can be used to use the word segmentation results of historical target text as candidate semantic units.

[0029] In this embodiment, when executing S101, word mining and word segmentation can be used simultaneously. The results of word mining and word segmentation can be used as candidate semantic units, thereby improving the comprehensiveness of candidate semantic units obtained from historical target text.

[0030] In this embodiment, after obtaining candidate semantic units from historical target text in S101, the semantic units that belong to stop words in the candidate semantic units can be removed, which can avoid the problem of reducing the accuracy of the selected target text caused by using stop words as target semantic units.

[0031] In this embodiment, when executing S101 to select candidate semantic units whose statistical values ​​meet preset requirements as target semantic units, candidate semantic units with TF values ​​greater than a first preset threshold can be selected as target semantic units, candidate semantic units with IDF values ​​greater than a second preset threshold can be selected as target semantic units, and candidate semantic units with TF-IDF values ​​greater than a third preset threshold can also be selected as target semantic units.

[0032] In this embodiment, when executing S101, the above three statistical values ​​of the candidate semantic units can also be obtained, and then the candidate semantic units that simultaneously meet the above three preset requirements can be used as the target semantic units, thereby further improving the accuracy of the obtained target semantic units.

[0033] In this embodiment, when executing S101 to select candidate semantic units whose statistical values ​​meet preset requirements as target semantic units, the following method can also be adopted: obtain the attribute information of historical target text, which can be the type of historical target text (e.g., the type containing sentiment words, the type containing sensitive words, etc.) and / or the field to which the historical target text belongs (e.g., the financial field, the technology field, etc.); determine the set of semantic units corresponding to the attribute information; and select candidate semantic units whose statistical values ​​meet preset requirements and do not appear in the set of semantic units as target semantic units.

[0034] In this embodiment, a correspondence between different attribute information and different semantic unit sets will be established in advance. In each different semantic unit set, there is at least one non-target semantic unit corresponding to different attribute information.

[0035] In other words, in addition to the statistical values ​​of candidate semantic units, this embodiment also combines the set of semantic units corresponding to historical target text to obtain target semantic units from historical target text, which can improve the accuracy of the obtained target semantic units.

[0036] The text to be processed obtained by executing S101 in this embodiment can be a sentence, a paragraph, or a passage. When executing S101 in this embodiment, the sentence, paragraph, or passage input by the input terminal can be used as the text to be processed, or the sentence, paragraph, or passage selected by the input terminal on the network can be used as the text to be processed. The number of texts to be processed obtained by executing S101 in this embodiment can be one or more.

[0037] After obtaining the text to be processed in step S101, this embodiment can also perform preprocessing on the text to be processed; the preprocessing in this embodiment includes simplified / traditional character conversion, case conversion, full-width / half-width character conversion, and replacing consecutive numbers with special symbols. <num>)wait.

[0038] In addition, in order to further improve the accuracy of the determined target text, this embodiment does not perform preprocessing such as removing stop words, punctuation marks, and special symbols on the text to be processed when performing S101 preprocessing.

[0039] In this embodiment, after executing S101 to obtain the target semantic unit and the text to be processed, S102 is executed to select the first candidate text from the text to be processed based on the target semantic unit. The number of first candidate texts selected by executing S102 in this embodiment can be one or more.

[0040] In this embodiment, when performing S102 to select the first candidate text from the text to be processed based on the target semantic unit, the optional implementation method is as follows: for each text to be processed, obtain the number of times the target semantic unit appears in the text to be processed; if it is determined that the number of occurrences obtained is greater than a preset number threshold, the text to be processed is taken as the first candidate text.

[0041] In other words, this embodiment selects the texts with more occurrences of the target semantic unit as the first candidate texts based on the frequency of occurrence of the target semantic unit in each text to be processed. This can improve the accuracy of the selected first candidate texts and avoid selecting texts with fewer occurrences of the target semantic unit as the first candidate texts.

[0042] In addition, after selecting the first candidate text in S102, this embodiment can also use preset tags to annotate the first candidate text. The preset tags are the type and / or domain corresponding to the target semantic unit, such as emotion type, sensitivity type, technology domain, financial domain, etc.

[0043] It is understood that if the number of first candidate texts selected in S102 of this embodiment is one, manual verification can be used to determine whether the first candidate text is the target text.

[0044] In this embodiment, after executing S102 to select the first candidate text from the text to be processed, S103 is executed to obtain the prediction result of the first candidate text using the first classification model, and the first candidate text corresponding to the prediction result and the target semantic unit is selected as the second candidate text; the number of second candidate texts selected in S103 in this embodiment can be one or more.

[0045] Specifically, in this embodiment, when executing S103 to obtain the prediction result of the first candidate text using the first classification model, the optional implementation method is as follows: according to the first preset number of parts, the selected first candidate text is divided into multiple text sets, each text set containing the same number of first candidate texts; one text set is selected as the first prediction text set, and the remaining text set is used as the first training text set; the first classification model is trained according to the first training text set, and the prediction result of each first candidate text in the first prediction text set is obtained using the trained first classification model; it is determined whether the prediction result of all first candidate texts has been obtained. If not, the process proceeds to the step of dividing the selected first candidate text into multiple text sets according to the first preset number of parts, until it is determined that the prediction result of all first candidate texts has been obtained.

[0046] In other words, this embodiment trains a first classification model based on a first training text set obtained by dividing the first candidate text, and then uses the trained first classification model to predict the first prediction text set to obtain the prediction results of all first candidate texts. This achieves the goal of obtaining more accurate prediction results of the first candidate texts when the number of first candidate texts is small.

[0047] For example, if the first preset number of copies is 5, and the number of first candidate texts is 10; in this embodiment, when executing S103, two non-repeating first candidate texts are used as a text set, so that each of the 5 text sets contains 2 first candidate texts.

[0048] In this embodiment, when selecting the first candidate text corresponding to the prediction result and the target semantic unit as the second candidate text in S103, the first candidate text whose prediction result is the type and / or domain of the target semantic unit can be used as the second candidate text, or the first candidate text whose prediction result is consistent with the preset label can be used as the second candidate text. In addition, this embodiment can also use the type and / or domain of the target semantic unit as the preset label to label the second candidate text.

[0049] In this embodiment, when executing S103, using the first classification model to obtain the prediction result of the first candidate text, and selecting the first candidate text corresponding to the prediction result and the target semantic unit as the second candidate text, the following may also be included: determining the number of first texts based on the texts that were not selected in the first candidate texts; if it is determined that the determined number of first texts is greater than a first number threshold, proceeding to the step of using the first classification model to obtain the prediction result of the first candidate text, until the determined number of first texts is less than or equal to the first number threshold.

[0050] In other words, this embodiment can repeatedly execute the process of predicting the first candidate text according to the first classification model based on the number of texts determined by the number of texts that were not selected in the first candidate text, thereby avoiding the problem of too few second candidate texts due to the unstable classification effect of the first classification model, and further improving the accuracy of the selected second candidate texts.

[0051] The first classification model used in S103 of this embodiment is a neural network model that can classify long texts (e.g., the entire document), such as the fastText model. This embodiment does not limit the specific type of the first classification model.

[0052] In this embodiment, after executing S103 to select the first candidate text corresponding to the prediction result and the target semantic unit as the second candidate text, S104 is executed to obtain the prediction result of the second candidate text using multiple second classification models based on the sentence segmentation result of the second candidate text, and the second candidate text corresponding to the prediction result and the target semantic unit is selected as the target text. The number of target texts selected in S104 of this embodiment can be one or more.

[0053] The second classification model used in S104 of this embodiment is a neural network model that can classify short texts (such as parts of a document), such as bert-base, ernie-1.0, albert-base models, etc. This embodiment does not limit the number or specific type of the second classification model used; the model types of the first classification model and the second classification model are not the same.

[0054] Therefore, in this embodiment, when executing S104 to obtain the prediction result of the second candidate text using the second classification model, it is necessary to first segment the second candidate text into sentences (e.g., 256 sentences, that is, each sentence contains 256 characters or words), and then use the second classification model to obtain the prediction result of the second candidate text based on the sentence segmentation results.

[0055] Specifically, in this embodiment, when executing S104 to obtain the prediction result of the second candidate text using multiple second classification models based on the sentence segmentation result of the second candidate text, the optional implementation method can be as follows: divide the second candidate text into multiple text sets according to the second preset number of parts, with each text set containing the same number of second candidate texts; select one text set as the second prediction text set, and use the remaining text sets as the second training text set; train multiple second classification models based on the sentence segmentation result of each second candidate text in the second training text set; use the trained multiple second classification models to obtain the prediction result of each second candidate text in the second prediction text set based on the sentence segmentation result of each second candidate text in the second prediction text set; determine whether the prediction result of all second candidate texts has been obtained; if not, proceed to the step of dividing the selected second candidate text into multiple text sets according to the second preset number of parts, until it is determined that the prediction result of all second candidate texts has been obtained.

[0056] In other words, this embodiment trains multiple second classification models based on the second training text set obtained by dividing the second candidate texts, and then uses the trained multiple second classification models to predict the second prediction text set to obtain the prediction results of all second candidate texts. This achieves the goal of obtaining more accurate prediction results of the second candidate texts when the number of second candidate texts is small.

[0057] For example, if the first preset number of copies is 5, and the number of second candidate texts is 10; in this embodiment, when executing S104, two non-repeating second candidate texts are used as a text set, so that each of the 5 text sets contains 2 second candidate texts.

[0058] In this embodiment, when executing S104 to obtain the prediction results of each second candidate text in the second prediction text set based on the sentence segmentation results of each second candidate text in the second prediction text set using multiple trained second classification models, an optional implementation method can be adopted as follows: For each second candidate text in the second prediction text set, the sentence segmentation results of the second candidate text are input into multiple second classification models respectively; for each second classification model, the candidate prediction result of the second candidate text corresponding to the second classification model is obtained based on the sentence segmentation prediction result output by the second classification model for each sentence segmentation result; and the prediction result of the second candidate text is obtained based on the candidate prediction result of the second candidate text corresponding to each second classification model.

[0059] In this embodiment, when executing S104 for each second classification model, based on the sentence prediction results output by the second classification model for each sentence result, and obtaining the candidate prediction results of the second candidate text corresponding to the second classification model, the number of different sentence prediction results can be obtained. Then, the sentence prediction results with the largest number can be used as the candidate prediction results of the second candidate text corresponding to the second classification model. Alternatively, sentence prediction results with a number exceeding a preset number threshold can also be used as the candidate prediction results of the second candidate text corresponding to the second classification model.

[0060] For example, if a second candidate text contains 10 sentence results, and a second classification model outputs 3 sentence prediction results A, 2 sentence prediction results B, and 5 sentence prediction results C for the 10 sentence results, then in this embodiment, when executing S104, sentence prediction result C can be used as the candidate prediction result of the second candidate text for the second classification model.

[0061] In this embodiment, when executing S104 to obtain the prediction result of the second candidate text based on the candidate prediction results of each second classification model corresponding to the second candidate text, if it is determined that the multiple candidate prediction results obtained are consistent, the candidate prediction result can be used as the prediction result of the second candidate text.

[0062] For example, if the second classification model is Model 1, Model 2 and Model 3, if the candidate prediction result of the second candidate text is prediction result C for Model 1, prediction result C for Model 2 and prediction result C for Model 3, then prediction result C is taken as the prediction result of the second candidate text; if the candidate prediction result of Model 2 is prediction result B, then the second candidate text is discarded.

[0063] In addition, when executing S104, based on the sentence segmentation results of the second candidate text, using multiple second classification models to obtain the prediction results of the second candidate text, and selecting the second candidate text corresponding to the target semantic unit as the target text, this embodiment may also include the following: determining the number of second texts based on the texts that were not selected in the second candidate text; if the number of second texts is determined to be greater than a second number threshold, proceeding to the step of obtaining the prediction results of the second candidate text based on the sentence segmentation results of the second candidate text using multiple second classification models, and repeating this process until the determined number of second texts is less than or equal to the second number threshold.

[0064] In other words, this embodiment can repeatedly execute the process of predicting the second candidate texts based on multiple second classification models, according to the number of second texts determined by the number of unselected texts in the second candidate texts. This avoids the problem of too few target texts due to the unstable classification effect of the second classification model, and further improves the accuracy of the selected target texts.

[0065] After selecting the target text in S104, this embodiment can further process the target text, such as manual screening, screening by regular expressions or keywords, to further improve the accuracy of the screened target text.

[0066] In this embodiment, after selecting the target text in S104, the selected target text can be used as training text, and the type and / or domain of the target semantic unit corresponding to the target text can be used as a label to train the neural network model, so that the trained text classification model can output the type and / or domain of the text based on the input text.

[0067] Figure 2 This is a schematic diagram according to the second embodiment of the present disclosure. Figure 2 The flowchart illustrates the process of selecting the target text after selecting the first candidate text based on the target semantic unit in this embodiment: After obtaining the first candidate text, if the text to be processed has not been preprocessed, the first candidate text is preprocessed; the preprocessed first candidate text is used to train a first classification model, and the prediction result of the first candidate text is obtained based on the trained first classification model; a second candidate text is selected based on the prediction result of the first candidate text (this process can be repeated until the number of unselected first candidate texts is less than or equal to a first quantity threshold); the second candidate text is segmented into sentences, and multiple second classification models (second classification model 1, second classification model 2, and second classification model 3 in the figure) are trained based on the sentence segmentation results of the second candidate text, and the prediction result of the second candidate text is obtained using the trained multiple second classification models; if the prediction results output by the multiple second classification models are consistent, the second candidate text is used as the target text (this process can be repeated until the number of unselected second candidate texts is less than or equal to a second quantity threshold); the target text is used as a training sample to train a neural network model to obtain a text classification model.

[0068] Figure 3 This is a schematic diagram according to the third embodiment of this disclosure. (See diagram below.) Figure 3 As shown, the text selection device 300 of this embodiment includes:

[0069] Acquisition unit 301 is used to acquire the target semantic unit and the text to be processed;

[0070] The first selection unit 302 is used to select a first candidate text from the text to be processed according to the target semantic unit;

[0071] The second selection unit 303 is used to obtain the prediction result of the first candidate text using the first classification model, and select the first candidate text corresponding to the prediction result and the target semantic unit as the second candidate text;

[0072] The third selection unit 304 is used to obtain the prediction result of the second candidate text by using multiple second classification models based on the sentence segmentation result of the second candidate text, and select the second candidate text corresponding to the prediction result of the target semantic unit as the target text.

[0073] The target semantic unit acquired by the acquisition unit 301 can be a word or a phrase. The acquisition unit 301 can use the word or phrase input by the input terminal as the target semantic unit, or it can use the word or phrase selected by the input terminal on the network as the target semantic unit.

[0074] The target semantic unit acquired by the acquisition unit 301 can be any type and / or any domain of words or phrases. For example, the type can be emotion type, sensitivity type, etc., and the domain can be economic domain, technology domain, etc. The number of target semantic units can be one or multiple. Different target semantic units can correspond to different types and / or domains, or they can correspond to the same type and / or domain.

[0075] In addition, when acquiring target semantic units, the acquisition unit 301 may also adopt the following methods: acquire historical target text; obtain candidate semantic units from historical target text, and acquire statistical values ​​of the acquired candidate semantic units; and take candidate semantic units whose statistical values ​​meet preset requirements as target semantic units.

[0076] The statistical values ​​acquired by the acquisition unit 301 can be at least one of the following: term frequency (TF) value, inverse document frequency (IDF) value, and term frequency-inverse document frequency (TF-IDF) value.

[0077] When obtaining candidate semantic units from historical target text, the acquisition unit 301 can use word mining based on mutual information and / or cohesion to mine words and / or phrases from historical target text as candidate semantic units; or it can use word segmentation to use the word segmentation results of historical target text as candidate semantic units.

[0078] The acquisition unit 301 can also use both word mining and word segmentation methods simultaneously, using the word mining results and word segmentation results as candidate semantic units, thereby improving the comprehensiveness of the candidate semantic units obtained from the historical target text.

[0079] After obtaining candidate semantic units from the historical target text, the acquisition unit 301 can also remove semantic units that belong to stop words from the candidate semantic units, which can avoid the problem of reducing the accuracy of the selected target text caused by using stop words as target semantic units.

[0080] When the acquisition unit 301 selects candidate semantic units whose statistical values ​​meet preset requirements as target semantic units, it can select candidate semantic units whose TF values ​​are greater than the first preset threshold, candidate semantic units whose IDF values ​​are greater than the second preset threshold, or candidate semantic units whose TF-IDF values ​​are greater than the third preset threshold.

[0081] The acquisition unit 301 can also acquire the above three statistical values ​​of the candidate semantic units, and then take the candidate semantic units that simultaneously meet the above three preset requirements as the target semantic units, thereby further improving the accuracy of the acquired target semantic units.

[0082] When the acquisition unit 301 uses candidate semantic units whose statistical values ​​meet preset requirements as target semantic units, it can also use the following methods: acquire the attribute information of historical target text; determine the set of semantic units corresponding to the attribute information; and use candidate semantic units whose statistical values ​​meet preset requirements but do not appear in the set of semantic units as target semantic units.

[0083] In this embodiment, a correspondence between different attribute information and different semantic unit sets will be established in advance. In each different semantic unit set, there is at least one non-target semantic unit corresponding to different attribute information.

[0084] In other words, in addition to the statistical values ​​of candidate semantic units, the acquisition unit 301 also combines the set of semantic units corresponding to the historical target text to obtain target semantic units from the historical target text, which can improve the accuracy of the obtained target semantic units.

[0085] The text to be processed acquired by the acquisition unit 301 can be a sentence, a paragraph, or a passage. The acquisition unit 301 can use sentences, paragraphs, or passages input from the input terminal as text to be processed, or it can use sentences, paragraphs, or passages selected by the input terminal on the network as text to be processed. The number of texts to be processed acquired by the acquisition unit 301 can be one or more.

[0086] After acquiring the text to be processed, the acquisition unit 301 can also perform preprocessing on the text; the preprocessing in this embodiment includes simplified / traditional character conversion, case conversion, full-width / half-width character conversion, and replacing consecutive numbers with special symbols. <num>etc.

[0087] In addition, in order to further improve the accuracy of the determined target text, the preprocessing unit 301 does not perform preprocessing such as removing stop words, punctuation marks, special symbols, etc. when preprocessing the to-be-processed text.

[0088] After the target semantic unit and the to-be-processed text are obtained by the obtaining unit 301, the first selection unit 302 selects a first candidate text from the to-be-processed text according to the target semantic unit. The number of the first candidate text selected by the first selection unit 302 can be one or multiple.

[0089] The optional implementation manner that the first selection unit 302 selects the first candidate text from the to-be-processed text according to the target semantic unit can be as follows: for each to-be-processed text, the number of occurrences of the target semantic unit in the to-be-processed text is obtained; and in a case where it is determined that the obtained number of occurrences is greater than a preset number threshold, the to-be-processed text is selected as the first candidate text.

[0090] That is, the first selection unit 302 selects the to-be-processed text in which the target semantic unit occurs more frequently as the first candidate text according to the number of occurrences of the target semantic unit in each to-be-processed text, which can improve the accuracy of the selected first candidate text and avoid selecting the to-be-processed text in which the target semantic unit occurs less frequently as the first candidate text.

[0091] In addition, after the first selection unit 302 selects the first candidate text, the first selection unit 302 can also label the first candidate text using a preset label. The preset label is the type and / or field corresponding to the target semantic unit, such as an emotional type, a sensitive type, a technology field, a finance field, etc.

[0092] It can be understood that if the number of the first candidate text selected by the first selection unit 302 is one, a manual verification manner can be used to determine whether the first candidate text is the target text.

[0093] After the first selection unit 302 selects the first candidate text from the to-be-processed text, the second selection unit 303 obtains a prediction result of the first candidate text using a first classification model, and selects the first candidate text corresponding to the prediction result and the target semantic unit as a second candidate text. The number of the second candidate text selected by the second selection unit 303 can be one or multiple.

[0094] Specifically, when the second selection unit 303 obtains the prediction result of the first candidate text using the first classification model, the optional implementation method is as follows: divide the selected first candidate text into multiple text sets according to the first preset number of parts; select one text set as the first prediction text set, and use the remaining text set as the first training text set; train the first classification model according to the first training text set, and use the trained first classification model to obtain the prediction result of each first candidate text in the first prediction text set; determine whether the prediction result of all first candidate texts has been obtained, and if not, proceed to the step of dividing the selected first candidate text into multiple text sets according to the first preset number of parts, until it is determined that the prediction result of all first candidate texts has been obtained.

[0095] In other words, the second selection unit 303 trains the first classification model based on the first training text set obtained by dividing the first candidate text, and then uses the trained first classification model to predict the first prediction text set to obtain the prediction results of all the first candidate texts. This achieves the goal of obtaining more accurate prediction results of the first candidate texts when the number of first candidate texts is small.

[0096] When selecting the first candidate text whose prediction result corresponds to the target semantic unit as the second candidate text, the second selection unit 303 may select the first candidate text whose prediction result is the type and / or domain of the target semantic unit as the second candidate text, or select the first candidate text whose prediction result is consistent with the preset label as the second candidate text. In addition, the second selection unit 303 may also use the type and / or domain of the target semantic unit as the preset label to label the second candidate text.

[0097] When the second selection unit 303 obtains the prediction result of the first candidate text using the first classification model and selects the first candidate text corresponding to the target semantic unit as the second candidate text, it may also include the following: determining the number of first texts based on the texts that were not selected in the first candidate texts; if it is determined that the determined number of first texts is greater than a first number threshold, it proceeds to the step of obtaining the prediction result of the first candidate text using the first classification model until the determined number of first texts is less than or equal to the first number threshold.

[0098] In other words, the second selection unit 303 can repeatedly execute the process of predicting the first candidate text according to the first classification model based on the number of texts determined by the number of texts that were not selected in the first candidate text, thereby avoiding the problem of too few second candidate texts due to the unstable classification effect of the first classification model, and further improving the accuracy of the selected second candidate texts.

[0099] The first classification model used by the second selection unit 303 is a neural network model that can classify long texts (e.g., the entire document), such as the fastText model. This embodiment does not limit the specific type of the first classification model.

[0100] It is understandable that if the second selection unit 303 selects only one second candidate text, manual verification can be used to determine whether the second candidate text is the target text.

[0101] In this embodiment, after the second selection unit 303 selects the first candidate text corresponding to the prediction result of the target semantic unit as the second candidate text, the third selection unit 304 uses multiple second classification models to obtain the prediction result of the second candidate text based on the sentence segmentation result of the second candidate text, and selects the second candidate text corresponding to the prediction result of the target semantic unit as the target text; the number of target texts selected by the third selection unit 304 can be one or more.

[0102] The second classification model used by the third selection unit 304 is a neural network model that can classify short texts (such as parts of a document), such as bert-base, ernie-1.0, albert-base models, etc. This embodiment does not limit the number or specific type of the second classification model used; the model types of the first classification model and the second classification model are not the same.

[0103] Therefore, when the third selection unit 304 uses the second classification model to obtain the prediction result of the second candidate text, it needs to first segment the second candidate text into sentences, and then use the second classification model to obtain the prediction result of the second candidate text based on the segmentation result.

[0104] Specifically, when the third selection unit 304 obtains the prediction results of the second candidate text using multiple second classification models based on the sentence segmentation results of the second candidate text, the optional implementation method can be as follows: divide the second candidate text into multiple text sets according to the second preset number of parts; select one text set as the second prediction text set, and use the remaining text set as the second training text set; train multiple second classification models based on the sentence segmentation results of each second candidate text in the second training text set; use the trained multiple second classification models to obtain the prediction results of each second candidate text in the second prediction text set based on the sentence segmentation results of each second candidate text in the second prediction text set; determine whether the prediction results of all second candidate texts have been obtained; if not, proceed to the step of dividing the selected second candidate text into multiple text sets according to the second preset number of parts, until it is determined that the prediction results of all second candidate texts have been obtained.

[0105] In other words, the third selection unit 304 trains multiple second classification models based on the second training text set obtained by dividing the second candidate texts, and then uses the trained multiple second classification models to predict the second prediction text set to obtain the prediction results of all second candidate texts. This achieves the goal of obtaining more accurate prediction results of the second candidate texts when the number of second candidate texts is small.

[0106] When the third selection unit 304 obtains the prediction results of each second candidate text in the second prediction text set based on the sentence segmentation results of each second candidate text in the second prediction text set using multiple trained second classification models, the optional implementation method can be as follows: For each second candidate text in the second prediction text set, the sentence segmentation results of the second candidate text are input into multiple second classification models respectively; for each second classification model, the candidate prediction result of the second candidate text corresponding to the second classification model is obtained based on the sentence segmentation prediction result output by the second classification model for each sentence segmentation result; and the prediction result of the second candidate text is obtained based on the candidate prediction result of the second candidate text corresponding to each second classification model.

[0107] Specifically, when the third selection unit 304 obtains the candidate prediction results of the second candidate text corresponding to the second classification model based on the sentence prediction results output by the second classification model for each sentence result, it can obtain the number of different sentence prediction results, and then take the sentence prediction results with the largest number as the candidate prediction results of the second candidate text corresponding to the second classification model. It can also take the sentence prediction results with a number exceeding a preset number threshold as the candidate prediction results of the second candidate text corresponding to the second classification model.

[0108] When the third selection unit 304 obtains the prediction result of the second candidate text based on the candidate prediction results of each second classification model corresponding to the second candidate text, it may, if it is determined that the multiple candidate prediction results obtained are consistent, use the candidate prediction result as the prediction result of the second candidate text.

[0109] In addition, when the third selection unit 304 obtains the prediction results of the second candidate text by using multiple second classification models based on the sentence segmentation results of the second candidate text, and selects the second candidate text corresponding to the target semantic unit as the target text, it may also include the following: determining the number of second texts based on the unselected texts in the second candidate text; if the number of second texts is determined to be greater than the second number threshold, it switches to the step of obtaining the prediction results of the second candidate text by using multiple second classification models based on the sentence segmentation results of the second candidate text, and repeats this process until the determined number of second texts is less than or equal to the second number threshold.

[0110] In other words, the third selection unit 304 can repeatedly execute the process of predicting the second candidate texts based on multiple second classification models, according to the number of second texts determined by the number of unselected texts in the second candidate texts. This avoids the problem of too few target texts due to the unstable classification effect of the second classification model, and further improves the accuracy of the selected target texts.

[0111] In this embodiment, after the target text is selected by the third selection unit 304, the target text can be further processed, such as by manual screening, screening by regular expressions or keywords, etc., to further improve the accuracy of the screened target text.

[0112] After selecting the target text, the third selection unit 304 can also use the selected target text as training text, and use the type and / or domain of the target semantic unit corresponding to the target text as label to train the neural network model, so that the trained text classification model can output the type and / or domain of the text based on the input text.

[0113] The acquisition, storage, and application of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0114] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0115] like Figure 4 The diagram shown is a block diagram of an electronic device for a text selection method according to an embodiment of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0116] like Figure 4 As shown, device 400 includes a computing unit 401, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 402 or a computer program loaded from storage unit 408 into random access memory (RAM) 403. RAM 403 may also store various programs and data required for the operation of device 400. The computing unit 401, ROM 402, and RAM 403 are interconnected via bus 404. Input / output (I / O) interface 405 is also connected to bus 404.

[0117] Multiple components in device 400 are connected to I / O interface 405, including: input unit 406, such as keyboard, mouse, etc.; output unit 407, such as various types of displays, speakers, etc.; storage unit 408, such as disk, optical disk, etc.; and communication unit 409, such as network card, modem, wireless transceiver, etc. Communication unit 409 allows device 400 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0118] The computing unit 401 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 401 performs the various methods and processes described above, such as text selection methods. For example, in some embodiments, the text selection method may be implemented as a computer software program that is tangibly contained in a machine-readable medium, such as storage unit 408.

[0119] In some embodiments, part or all of the computer program may be loaded and / or installed on the device 500 via ROM 402 and / or communication unit 409. When the computer program is loaded into RAM 403 and executed by computing unit 401, one or more steps of the text selection method described above may be performed. Alternatively, in other embodiments, computing unit 401 may be configured to perform the text selection method by any other suitable means (e.g., by means of firmware).

[0120] Various implementations of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transferring data and instructions to the storage system, the at least one input device, and the at least one output device.

[0121] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable text selection device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0122] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0123] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for showing information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0124] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0125] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service system that addresses the management difficulties and weak business scalability inherent in traditional physical hosts and VPS (Virtual Private Server) services. Servers can also be servers for distributed systems or servers integrated with blockchain technology.

[0126] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0127] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.< / num> < / num>

Claims

1. A text selection method, comprising: Acquire the target semantic unit and multiple texts to be processed; Based on the target semantic unit, a first candidate text is selected from the plurality of texts to be processed, wherein the first candidate text is one or more of the plurality of texts to be processed; The prediction result of the first candidate text is obtained using the first classification model. The first candidate text whose prediction result is the type and / or domain of the target semantic unit is selected as the second candidate text. The prediction result is the type and / or domain of the first candidate text. Based on the sentence segmentation results of the second candidate text, multiple second classification models are used to obtain the prediction results of the second candidate text. The second candidate text whose prediction results correspond to the target semantic unit is selected as the target text. The step of obtaining the prediction result of the second candidate text using multiple second classification models based on the sentence segmentation result of the second candidate text includes: For each second candidate text, the sentence segmentation result of the second candidate text is input into the multiple second classification models respectively; For each second classification model, the candidate prediction result of the second candidate text corresponding to the second classification model is obtained based on the number of different sentence prediction results output by the second classification model for different sentence results. In response to the determination that the candidate prediction results of each second classification model are consistent with the second candidate text, the candidate prediction results are used as the prediction results of the second candidate text.

2. The method according to claim 1, wherein, The acquisition of the target semantic unit includes: Retrieve historical target text; Candidate semantic units are obtained from the historical target text, and statistical values ​​of the candidate semantic units are acquired. Candidate semantic units whose statistical values ​​meet preset requirements are selected as target semantic units.

3. The method according to claim 2, wherein, The step of using candidate semantic units whose statistical values ​​meet preset requirements as the target semantic unit includes: Obtain the attribute information of the historical target text; Determine the set of semantic units corresponding to the attribute information; Candidate semantic units whose statistical values ​​meet preset requirements and do not appear in the set of semantic units are taken as the target semantic units.

4. The method according to any one of claims 1-3, wherein, The step of selecting a first candidate text from the plurality of texts to be processed based on the target semantic unit includes: For each text to be processed, obtain the number of times the target semantic unit appears in that text; If the number of occurrences is determined to be greater than a preset threshold, the text to be processed is selected as the first candidate text.

5. The method according to any one of claims 1-4, wherein, The prediction result of the first candidate text obtained using the first classification model includes: Based on a first preset number of parts, the first candidate text is divided into multiple text sets; Select one text set as the first prediction text set, and use the remaining text sets as the first training text set. Text set; The first classification model is trained based on the first training text set, and the prediction results of each first candidate text in the first prediction text set are obtained using the trained first classification model. Determine whether prediction results for all first candidate texts have been obtained. If not, proceed to the step of dividing the first candidate texts into multiple text sets according to the first preset number of parts, until prediction results for all first candidate texts are determined.

6. The method according to any one of claims 1-5, wherein, The step of using a first classification model to obtain the prediction result of the first candidate text, and selecting the first candidate text whose prediction result corresponds to the type and / or domain of the target semantic unit as the second candidate text, includes: The number of first texts is determined based on the texts that were not selected from the first candidate texts; If it is determined that the number of the first texts is greater than the first quantity threshold, the process proceeds to the step of using the first classification model to obtain the prediction result of the first candidate texts, until the number of the first texts is less than or equal to the first quantity threshold.

7. The method according to any one of claims 1-6, wherein, The step of obtaining the prediction result of the second candidate text using multiple second classification models based on the sentence segmentation results of the second candidate text includes: According to the second preset number of parts, the second candidate text is divided into multiple text sets; Select one text set as the second prediction text set, and use the remaining text set as the second training text set; The multiple second classification models are trained based on the sentence segmentation results of each second candidate text in the second training text set. Based on the sentence segmentation results of each second candidate text in the second predicted text set, the prediction results of each second candidate text in the second predicted text set are obtained using multiple trained second classification models. Determine whether prediction results for all second candidate texts have been obtained. If not, proceed to the step of dividing the second candidate texts into multiple text sets according to the second preset number of parts, until prediction results for all second candidate texts are obtained.

8. The method according to any one of claims 1-7, wherein, The step of using multiple second classification models to obtain prediction results for the second candidate text based on the sentence segmentation results of the second candidate text, and selecting the second candidate text whose prediction results correspond to the target semantic unit as the target text includes: The number of second texts is determined based on the texts that were not selected from the second candidate texts; If the number of the second text is determined to be greater than the second number threshold, the process proceeds to the step of using multiple second classification models to obtain the prediction result of the second candidate text based on the sentence segmentation result of the second candidate text, until the number of the second text is less than or equal to the second number threshold.

9. A text selection device, comprising: The acquisition unit is used to acquire the target semantic unit and multiple texts to be processed. The first selection unit is configured to select a first candidate text from the plurality of texts to be processed based on the target semantic unit, wherein the first candidate text is one or more of the plurality of texts to be processed; The second selection unit is used to obtain the prediction result of the first candidate text using the first classification model, and select the first candidate text whose prediction result is the type and / or domain of the target semantic unit as the second candidate text, wherein the prediction result is the type and / or domain of the first candidate text. The third selection unit is used to obtain the prediction result of the second candidate text by using multiple second classification models based on the sentence segmentation result of the second candidate text, and select the second candidate text corresponding to the prediction result of the target semantic unit as the target text; Specifically, when the third selection unit obtains the prediction result of the second candidate text using multiple second classification models based on the sentence segmentation result of the second candidate text, it performs the following: For each second candidate text, the sentence segmentation result of the second candidate text is input into the multiple second classification models respectively; For each second classification model, the candidate prediction result of the second candidate text corresponding to the second classification model is obtained based on the number of different sentence prediction results output by the second classification model for different sentence results. In response to the determination that the candidate prediction results of each second classification model are consistent with the second candidate text, the candidate prediction results are used as the prediction results of the second candidate text.

10. The apparatus according to claim 9, wherein, When acquiring the target semantic unit, the acquisition unit specifically performs the following: Retrieve historical target text; Candidate semantic units are obtained from the historical target text, and statistical values ​​of the candidate semantic units are acquired. Candidate semantic units whose statistical values ​​meet preset requirements are selected as target semantic units.

11. The apparatus according to claim 10, wherein, When the acquisition unit selects candidate semantic units whose statistical values ​​meet preset requirements as the target semantic units, it specifically performs the following: Obtain the attribute information of the historical target text; Determine the set of semantic units corresponding to the attribute information; Candidate semantic units whose statistical values ​​meet preset requirements and do not appear in the set of semantic units are taken as the target semantic units.

12. The apparatus according to any one of claims 9-11, wherein, When the first selection unit selects a first candidate text from the plurality of texts to be processed according to the target semantic unit, it specifically performs the following: For each text to be processed, obtain the number of times the target semantic unit appears in that text; If the number of occurrences is determined to be greater than a preset threshold, the text to be processed is selected as the first candidate text.

13. The apparatus according to any one of claims 9-12, wherein, When the second selection unit obtains the prediction result of the first candidate text using the first classification model, it specifically performs the following: Based on a first preset number of parts, the first candidate text is divided into multiple text sets; Select one text set as the first prediction text set, and use the remaining text set as the first training text set; The first classification model is trained based on the first training text set, and the prediction results of each first candidate text in the first prediction text set are obtained using the trained first classification model. Determine whether prediction results for all first candidate texts have been obtained. If not, proceed to the step of dividing the first candidate texts into multiple text sets according to the first preset number of parts, until prediction results for all first candidate texts are determined.

14. The apparatus according to any one of claims 9-13, wherein, When the second selection unit uses the first classification model to obtain the prediction result of the first candidate text, and selects the first candidate text whose prediction result corresponds to the type and / or domain of the target semantic unit as the second candidate text, the specific execution is as follows: The number of first texts is determined based on the texts that were not selected from the first candidate texts; If it is determined that the number of the first texts is greater than the first quantity threshold, the process proceeds to the step of using the first classification model to obtain the prediction result of the first candidate texts, until the number of the first texts is less than or equal to the first quantity threshold.

15. The apparatus according to any one of claims 9-14, wherein, When the third selection unit obtains the prediction result of the second candidate text using multiple second classification models based on the sentence segmentation result of the second candidate text, it specifically performs the following: According to the second preset number of parts, the second candidate text is divided into multiple text sets; Select one text set as the second prediction text set, and use the remaining text set as the second training text set; The multiple second classification models are trained based on the sentence segmentation results of each second candidate text in the second training text set. Based on the sentence segmentation results of each second candidate text in the second predicted text set, the prediction results of each second candidate text in the second predicted text set are obtained using multiple trained second classification models. Determine whether prediction results for all second candidate texts have been obtained. If not, proceed to the step of dividing the second candidate texts into multiple text sets according to the second preset number of parts, until prediction results for all second candidate texts are obtained.

16. The apparatus according to any one of claims 9-15, wherein, When the third selection unit selects the second candidate text as the target text by using multiple second classification models to obtain the prediction results of the second candidate text based on the sentence segmentation results of the second candidate text, the specific execution is as follows: The number of second texts is determined based on the texts that were not selected from the second candidate texts; If the number of the second text is determined to be greater than the second number threshold, the process proceeds to the step of using multiple second classification models to obtain the prediction result of the second candidate text based on the sentence segmentation result of the second candidate text, until the number of the second text is less than or equal to the second number threshold.

17. An electronic device comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-8.

18. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-8.

19. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1-8.

Images