阅读说明：本技术 数学应用题解答模型的处理方法、系统和存储介质 (Method, system and storage medium for processing mathematic application problem solution model ) 是由 肖菁 樊纬江 陈寅 曹阳 于 2021-01-29 设计创作，主要内容包括：本发明公开了一种数学应用题解答模型的处理方法、系统和存储介质,方法包括以下步骤：获取若干个原始数学应用题；对所述原始数学应用题的名称进行回译操作,得到第一数据集；判断所述第一数据集内的文本是否为问句,若是,则将所述文本添加于正常序列；反之,则将所述文本添加于非正常序列；对正常序列和非正常序列依次进行扩充和分词；将分词后的序列和原始数学应用题作为训练集对数学应用题解答模型进行训练,并在训练完成后,更新神经网络模型的参数。本发明可降低回译导致语句缺失带来的影响,提高数学应用题解答模型的训练精度,以提高应用了数学应用题解答模型的题型搜索软件的泛化能力和准确度。本发明可广泛应用于模型训练技术领域。(The invention discloses a processing method, a system and a storage medium of a mathematical application question answering model, wherein the method comprises the following steps: acquiring a plurality of original mathematic application questions; performing a back translation operation on the name of the original mathematic application question to obtain a first data set; judging whether the text in the first data set is a question or not, and if so, adding the text to a normal sequence; otherwise, adding the text to the abnormal sequence; sequentially expanding and segmenting the normal sequence and the abnormal sequence; and training the mathematical application question solution model by taking the sequence after word segmentation and the original mathematical application question as a training set, and updating the parameters of the neural network model after the training is finished. The invention can reduce the influence caused by sentence deletion caused by the translation, and improve the training precision of the mathematic application question answering model so as to improve the generalization ability and the accuracy of the question type searching software applying the mathematic application question answering model. The invention can be widely applied to the technical field of model training.)

1. A processing method of a mathematical application problem solution model is characterized by comprising the following steps:

acquiring a plurality of original mathematic application questions;

performing a back translation operation on the name of the original mathematic application question to obtain a first data set;

judging whether the text in the first data set is a question or not, and if so, adding the text to a normal sequence; otherwise, adding the text to the abnormal sequence;

sequentially expanding and segmenting the normal sequence and the abnormal sequence;

and training the mathematical application question solution model by taking the sequence after word segmentation and the original mathematical application question as a training set, and updating the parameters of the neural network model after the training is finished.

2. The method of claim 1, further comprising the following steps after said step of obtaining a plurality of original mathematical application questions:

and converting the original mathematical application questions into a first text sequence corresponding to a preset exchange format.

3. The method of claim 1, wherein said back-translating the name of the original mathematical application to obtain the first data set comprises:

converting the name of the original math application question into a character string type;

translating the names of the original mathematical application questions after type conversion into first text information of a plurality of types of preset languages;

the first text information of the preset language is translated back into second text information of the language corresponding to the original mathematic application question;

and taking the second text information as a first data set.

4. The method of claim 2, wherein said expanding and segmenting normal sequences and abnormal sequences in sequence comprises:

synonym replacement is carried out on non-keywords of the mathematical problem of the normal sequence;

performing data enhancement of a preset level on an original mathematic application question corresponding to the abnormal sequence;

and performing word segmentation on the replaced normal sequence and the data-enhanced abnormal sequence.

5. The method of claim 4, further comprising the steps of:

and converting the sequence after word segmentation into a second text sequence corresponding to the preset exchange format.

6. The method according to claim 5, wherein the training of the mathematical application problem solution model using the sequence after word segmentation and the original mathematical application problem as a training set and updating parameters of the neural network model after the training is completed comprises:

inputting the first text sequence and the second text sequence into a neural network model in a mathematical application problem solution model as a training set, and calculating a loss function;

calculating a word vector loss gradient according to the loss function;

adding disturbance to a word vector corresponding to a word vector loss gradient with a preset size;

calculating a loss function and a word vector loss gradient of the neural network model added with the disturbance;

accumulating the word vector loss gradient after adding disturbance and the word vector loss gradient before adding disturbance to serve as a target word vector loss gradient;

and after the training is finished, updating the parameters of the mathematical application problem solution model according to the vector loss gradient of the target word.

7. The method according to claim 6, wherein said adding perturbation to the word vector corresponding to the word vector loss gradient with a predetermined size is specifically:

and adding disturbance step by step to the word vectors corresponding to the word vector loss gradient with the preset size, and setting the size of the disturbance radius as a preset value.

8. A system for processing a problem solution model for mathematical applications, comprising:

the acquisition module is used for acquiring a plurality of original math application questions;

the retranslation operation module is used for performing retranslation operation on the name of the original mathematic application question to obtain a first data set;

the judging module is used for judging whether the text in the first data set is a question or not, and if so, adding the text to a normal sequence; otherwise, adding the text to the abnormal sequence;

the expansion and word segmentation module is used for sequentially expanding and segmenting the normal sequence and the abnormal sequence;

and the training and updating module is used for training the mathematical application question solution model by taking the sequence after word segmentation and the original mathematical application question as a training set, and updating the parameters of the neural network model after the training is finished.

9. A system for processing a problem solution model for mathematical applications, comprising:

at least one memory for storing a program;

at least one processor configured to load the program to perform the method of processing a mathematical application problem solution model according to any one of claims 1 to 7.

10. A storage medium in which a processor-executable program is stored, wherein the processor-executable program, when executed by a processor, is for performing the processing method of the mathematical application problem solution model according to any one of claims 1 to 7.

Technical Field

The invention relates to the technical field of model training, in particular to a processing method, a system and a storage medium of a mathematical application problem solution model.

Background

With the development of information education, topic type search software is produced. The current topic type searching software has rare current data sets due to the particularity of Chinese mathematical topics and the complexity of labeling, a neural network model is easy to be over-fitted so that the generalization capability is not strong, and the accuracy is not high due to the influence of strange disturbance. Data enhancement can be performed to a certain extent on a limited data set, however, the complexity of natural language and the protection of semantic logic, translation may cause a problem of sentence loss, entity vocabulary replacement also needs to perform targeted replacement in the face of different problems, and currently, targeted data enhancement is hardly performed on a mathematic subject data set. In summary, the generalization capability and accuracy of the current topic search software are not strong.

Disclosure of Invention

The present invention is directed to solving at least one of the problems of the prior art. Therefore, the invention provides a processing method, a system and a storage medium for a mathematical application question answering model, which can effectively improve the generalization ability and accuracy of question type searching software.

According to a first aspect of the invention, the processing method of the problem solution model applied to mathematics comprises the following steps:

acquiring a plurality of original mathematic application questions;

performing a back translation operation on the name of the original mathematic application question to obtain a first data set;

judging whether the text in the first data set is a question or not, and if so, adding the text to a normal sequence; otherwise, adding the text to the abnormal sequence;

sequentially expanding and segmenting the normal sequence and the abnormal sequence;

and training the mathematical application question solution model by taking the sequence after word segmentation and the original mathematical application question as a training set, and updating the parameters of the neural network model after the training is finished.

The processing method of the mathematical application problem solution model according to the embodiment of the invention at least has the following beneficial effects: in the embodiment, after the name of the original mathematic application question is subjected to retranslation operation, whether the text in the first data set is a question is judged, the question text is added to the normal sequence, the non-question text is added to the abnormal sequence, then the normal sequence and the abnormal sequence are sequentially expanded and participled, the participled sequence and the original mathematic application question are used as a training set to train the mathematic application question answering model, and after the training is finished, the parameters of the neural network model are updated, so that the influence caused by sentence loss due to retranslation is reduced, the training precision of the mathematic application question answering model is improved, and the generalization capability and the accuracy of the question type search software using the mathematic application question answering model are improved.

According to some embodiments of the present invention, after the step of obtaining a plurality of original mathematical application questions, the method further comprises the following steps:

and converting the original mathematical application questions into a first text sequence corresponding to a preset exchange format.

According to some embodiments of the present invention, the translating back the name of the original math application topic to obtain a first data set includes:

converting the name of the original math application question into a character string type;

translating the names of the original mathematical application questions after type conversion into first text information of a plurality of types of preset languages;

the first text information of the preset language is translated back into second text information of the language corresponding to the original mathematic application question;

and taking the second text information as a first data set.

According to some embodiments of the invention, said expanding and tokenizing the normal sequence and the abnormal sequence in sequence comprises:

synonym replacement is carried out on non-keywords of the mathematical problem of the normal sequence;

performing data enhancement of a preset level on an original mathematic application question corresponding to the abnormal sequence;

and performing word segmentation on the replaced normal sequence and the data-enhanced abnormal sequence.

According to some embodiments of the invention, further comprising the steps of:

and converting the sequence after word segmentation into a second text sequence corresponding to the preset exchange format.

According to some embodiments of the present invention, the training the mathematical application question solution model by using the sequence after word segmentation and the original mathematical application question as a training set, and updating parameters of the neural network model after the training is completed includes:

inputting the first text sequence and the second text sequence into a neural network model in a mathematical application problem solution model as a training set, and calculating a loss function;

calculating a word vector loss gradient according to the loss function;

adding disturbance to a word vector corresponding to a word vector loss gradient with a preset size;

calculating a loss function and a word vector loss gradient of the neural network model added with the disturbance;

accumulating the word vector loss gradient after adding disturbance and the word vector loss gradient before adding disturbance to serve as a target word vector loss gradient;

and after the training is finished, updating the parameters of the mathematical application problem solution model according to the vector loss gradient of the target word.

According to some embodiments of the present invention, the adding the perturbation to the word vector corresponding to the word vector loss gradient with the preset size specifically includes:

and adding disturbance step by step to the word vectors corresponding to the word vector loss gradient with the preset size, and setting the size of the disturbance radius as a preset value.

A system for processing a mathematically applied problem solution model according to an embodiment of a second aspect of the present invention comprises:

the acquisition module is used for acquiring a plurality of original math application questions;

the retranslation operation module is used for performing retranslation operation on the name of the original mathematic application question to obtain a first data set;

the judging module is used for judging whether the text in the first data set is a question or not, and if so, adding the text to a normal sequence; otherwise, adding the text to the abnormal sequence;

the expansion and word segmentation module is used for sequentially expanding and segmenting the normal sequence and the abnormal sequence;

and the training and updating module is used for training the mathematical application question solution model by taking the sequence after word segmentation and the original mathematical application question as a training set, and updating the parameters of the neural network model after the training is finished.

A processing system for mathematically applying a problem solution model according to an embodiment of a third aspect of the present invention includes:

at least one memory for storing a program;

at least one processor, configured to load the program to perform the processing method of the mathematical application problem solution model described in the embodiment of the first aspect.

A storage medium according to an embodiment of a fourth aspect of the present invention stores therein a program executable by a processor, and the program executable by the processor is configured to perform the processing method of the mathematical application problem solution model described in the embodiment of the first aspect when the program is executed by the processor.

Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.

Drawings

The invention is further described with reference to the following figures and examples, in which:

FIG. 1 is a flow chart of a method for processing a mathematically applied problem solution model according to an embodiment of the present invention;

FIG. 2 is an exemplary diagram of text classification in one embodiment;

FIG. 3 is a simplified diagram of an embodiment;

FIG. 4 is a line graph of impact factors for a sequence of characters according to one embodiment;

FIG. 5 is a flow diagram of disturbance addition according to an embodiment.

Detailed Description

Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the accompanying drawings are illustrative only for the purpose of explaining the present invention, and are not to be construed as limiting the present invention.

In the description of the present invention, the meaning of a plurality is one or more, the meaning of a plurality is two or more, and the above, below, exceeding, etc. are understood as excluding the present numbers, and the above, below, within, etc. are understood as including the present numbers. If the first and second are described for the purpose of distinguishing technical features, they are not to be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated or implicitly indicating the precedence of the technical features indicated.

In the description of the present invention, unless otherwise explicitly defined, terms such as set, etc. should be broadly construed, and those skilled in the art can reasonably determine the specific meanings of the above terms in the present invention in combination with the detailed contents of the technical solutions.

In the description of the present invention, reference to the description of the terms "one embodiment," "some embodiments," "an illustrative embodiment," "an example," "a specific example," or "some examples," etc., means that a particular feature or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

Referring to fig. 1, an embodiment of the present invention provides a method for processing a problem solution model for mathematical application, and the embodiment may be applied to a server or a background controller of various types of problem search software.

In the application process, the embodiment includes the following steps:

and S11, acquiring a plurality of original math application questions. After the original mathematic application questions are obtained, a plurality of original mathematic application questions are converted into a first text sequence corresponding to a preset exchange format and then stored. The preset exchange format can be a JSON format, and the JSON format is a lightweight data exchange format. The first text sequence includes "topic name", "topic number", "segmentation sequence", "answer equation", and "result", respectively.

And S12, performing a back translation operation on the name of the original math application question to obtain a first data set.

In some embodiments, step S12 may be implemented by:

the name of the original math application topic is converted into a string type. Translating the names of the original mathematic application questions with the converted types into first text information of a plurality of types of preset languages; several classes of predetermined languages may be languages with high global universal language usage. The translation process may use existing translation software for translation. And then the first text information of the preset language is translated back into second text information of the language corresponding to the original mathematic application question, and finally the second text information is used as a first data set. In the step, by means of translation back, grammar and lexical method are all transformed to a certain extent under the condition of ensuring that the semantics are not changed, so that the data set is expanded on the premise of ensuring the quality.

Due to the problem of the length of part of the mathematical subjects, question missing occurs after translation, so that the translated sequence needs to be processed to avoid the damage of the model caused by the missing sentences.

S13, judging whether the text in the first data set is a question or not, and if so, adding the text to the normal sequence; otherwise, the text is added to the abnormal sequence. In this step, the judgment can be made by constructing a long-short term memory network trained by the natural language topic corpus data set.

And S14, sequentially expanding and segmenting the normal sequence and the abnormal sequence.

In some embodiments, step S14 may be implemented by:

synonym replacement is carried out on non-keywords of the mathematical problem of the normal sequence; performing data enhancement of a preset level on an original mathematic application question corresponding to the abnormal sequence; and then performing word segmentation on the replaced normal sequence and the data-enhanced abnormal sequence.

Specifically, for normal sequences, the statistics of the keywords of the mathematical problem can be performed by sending the statistics to a word frequency statistics device of a common weighting technology of information retrieval and data mining, and then the non-keywords, such as entity nouns irrelevant to the problem core solution, such as airplanes, automobiles, trees, and the like, are replaced by querying a synonym library. For an abnormal sequence, an original mathematic application question can be found back according to an original sequence number, and then low-level data enhancement is performed on the original mathematic application question, for example, a small amount of synonym replacement, random vocabulary insertion, exchange, deletion and the like.

Then, a Chinese word segmentation model is adopted, for example, a jieba is adopted to segment the processed normal sequence and the processed abnormal sequence, and the segmented sequence is converted into a second text sequence corresponding to a preset exchange format, namely, the segmented sequence is arranged in a data format of an original mathematic application topic.

And S15, training the mathematical application question solution model by taking the sequence after word segmentation and the original mathematical application question as a training set, and updating parameters of the neural network model after training.

Specifically, the countermeasure sample is used as an attack mode, the defensiveness of the trained model is enhanced when the trained model faces strange input, the trained model is applied in a large scale in the image field and is also applied to a certain extent in the aspect of natural language processing, the anti-noise capability and the generalization capability of the natural language model are enhanced due to the addition of disturbance, and the robustness and the accuracy of model classification or detection are improved.

In some embodiments, the step S15 can be implemented by:

inputting the first text sequence and the second text sequence into a neural network model in a mathematical application problem solution model as a training set, and calculating a forward loss function and gradient back propagation.

And then calculating word vector Loss embedding matrix gradient according to the Loss function Loss value, adding disturbance to word vector distribution corresponding to the word vector Loss gradient with the preset size, and setting the disturbance radius as the preset value. Specifically, since in a mathematical application topic, the position where the gradient is fastest to fall is often the vocabulary sequence associated with the topic solution, such as "how much worse" and the like, the present implementation sets the absolute value of the gradient to be its influence factor, and adds continuous perturbation to the direction of increasing loss, i.e., the embeding layer of the vocabulary sequence associated with the topic solution, in order to make the model more robust.

To keep the perturbations from being too large, closer to the optimum in the constraints, we divide the addition of the perturbations into k steps, typically set k to 3, set the perturbation radius, divide the step, and return if the radius is exceeded. The constraint space of the disturbance is shown in formula 1:

S＝r∈R^d:||r||₂≦ ε formula 1.

Epsilon is a preset disturbance radius; | r | non-conducting phosphor₂Epsilon is less than or equal to the constraint space of disturbance; | | non-woven hair₂For the L2 norm calculation, the square sum of each element of the vector is calculated and then the square root is calculated; s is a disturbance space; r^dRepresenting a D-dimensional vector.

If k is not the final step, calculating Loss after adding disturbance to obtain a gradient; if k is the final step, recovering the original gradient value and accumulating the final gradient to be used as a target word vector loss gradient; and the root restores the embedding layer to the initial value and carries out return transmission according to the gradient of the previous step to update the mathematical application problem solution model parameters.

The training is repeated until the training is finished.

In this embodiment, a smaller data set in the problem of the mathematical application problem is expanded to several times of the original scale by using a certain algorithm, and robustness and generalization of a confrontation training improvement model are introduced. Taking a certain topic in the data set as an example, the description of the specific example is taken as a flow example in most data streams:

step 1: the data enhancement is carried out on the data set of the mathematical problem, which can be specifically divided into the following steps:

step 1.1: storing the original mathematics application topic sequence in JSON format, such as:

step 1.2: using API interface of preset translation software to translate original _ text into English and Japanese, selecting language according to specific situation, and translating the language back into Chinese, such as

"two-year-old children in Zhenhai Yale School plant trees to one side of a road" how many meters the children grow into every 2meters (trees are planted on both sides of the road), and Finally, a total of 11trees are found, how many meters the road is long. Children have one tree every two meters (one at each end of the road). Finally, they found a total of 11 trees. How long this path is, it can be seen that semantic information is retained, but the expression and vocabulary are changed.

Step 1.3: the LSTM neural network is designed for text classification for question detection, where the data set comes from an open-source question detection game, the content includes questions of various tests and answers in the form of statement sentences, and label is used to indicate whether the test is a question, which is specifically shown in fig. 2.

Step 1.4: for longer questions identified as missing questions after translation back, such as: a company imports a batch of such goods, the tax is 6 ten thousand yuan, and the price value of the batch of goods is ten thousand yuan. "then send this kind of problem as abnormal sequence into the simple vocabulary replacement module, carry on a few synonyms replacement, insert, delete randomly, etc., choose synonyms forest expansion as synonyms library, set the random vocabulary transformation seed as 3, that is, the simple transformation in each sentence is not greater than 3, its transformation result is shown in fig. 3.

Step 1.5: for normal sequences, the normal sequences are sent to a word frequency statistics device based on TFIDF to carry out mathematical problem keyword statistics, wherein a TFIDF calculation formula is shown as a formula 2:

wherein, TFIDF can represent the key degree of each word in the mathematical topic, TF is the word frequency, which means the number of times of the occurrence of the vocabulary entry i in a certain class divided by the number of all vocabulary entries in the file j; IDF is the inverse file frequency, N is the total number of documents in the corpus, df_iIs the number of documents containing the word i.

And performing near word replacement on low-key-degree vocabularies such as entity nouns such as airplanes (0.000012) and earths (0.000028) by using a near word library, setting a replacement seed to be 3 similarly, indicating that the conversion of each sentence is not more than 3, and expanding the converted sentences into an original data set as expansion samples.

Step 1.6: and summarizing the two sequences, utilizing a Chinese word segmentation library jieba to segment the words of the extended samples, and then expanding the words into the extended samples according to the format of the original data set.

Step 2: introduction of math application problem oriented confrontation training in neural network

Step 2.1: and (3) normally calculating a forward Loss function and gradient back propagation when the data set obtained in the step (1) is used for the first training of the neural network model. Where the loss is calculated as a loss function in cross entropy.

Step 2.2: the gradient of an emboding matrix is obtained according to the Loss value, words in an original sequence are mapped back according to emboding, meanwhile, words with the largest influence on an answer result in a certain mathematical question are found according to the position where the gradient is descending fastest, continuous disturbance is added to an emboding layer of the word sequence, for example, fig. 4 is a broken line diagram of influence factors of a certain question character sequence drawn according to the gradient, it can be seen that large influence factors are obtained in 'the first month', 'the first batch' and the like, the large influence factors represent that the large influence factors have larger influence in question answering, and certain disturbance is added to vectors of the large influence words.

Step 2.3: the flow of adding perturbations is similar as described in the previous section, with the added perturbations shown in FIG. 5 for each train of models:

step 2.3.1: for the first training of the neural network model, a forward Loss function and gradient back propagation are normally calculated, and cross entropy is used as a Loss function.

Step 2.3.2: and obtaining the gradient of the embedding matrix according to the Loss value, and adding continuous disturbance to the characters with larger influence factors in the embedding layer.

Step 2.3.3: setting the disturbance step k as 3, and setting the disturbance radius as shown in formula 1:

S＝γ∈R^d:||r||₂≦ ε formula 1.

Epsilon is a preset disturbance radius; | r | non-conducting phosphor₂Epsilon is less than or equal to the constraint space of disturbance; | | non-woven hair₂For the L2 norm calculation, the square sum of each element of the vector is calculated and then the square root is calculated; s is a disturbance space; r^dRepresenting a D-dimensional vector.

Step 2.3.4: if k is not the final step, calculating Loss after adding disturbance to obtain a gradient; if k is the final step, the original gradient values are restored and the final gradient is accumulated.

Step 2.3.5: and restoring the embedding layer to the initial value, and returning according to the gradient of the previous step to update the model parameters.

Step 2.4: and continuing the fitting training until the training is finished.

Specifically, Math23K is used as an experimental data set, the size of the data set is 23163 mathematical problems, two languages (English and Japanese) are translated back in a data enhancement part, the sentence sequence for replacing the synonym by using the TFIDF word frequency statistics device accounts for about 75%, the sequence for simply replacing accounts for about 25%, and the enhanced data set has 82618 expansion problems; in the deep learning training phase, the embodiment uses a Seq2Seq model with greedy search as a reference, and respectively shows the results of five-fold cross validation on the Math23K data set by the two generalization promotion algorithms shown in table 1.

TABLE 1 Effect of the Algorithm

In summary, the embodiment can reduce the influence caused by sentence deletion due to translation, and improve the training precision of the mathematical application question answering model, thereby improving the generalization ability and accuracy of the question type search software to which the mathematical application question answering model is applied.

The embodiment of the invention provides a processing system of a mathematical application problem solution model, which comprises:

the acquisition module is used for acquiring a plurality of original math application questions;

the retranslation operation module is used for performing retranslation operation on the name of the original mathematic application question to obtain a first data set;

the judging module is used for judging whether the text in the first data set is a question or not, and if so, adding the text to a normal sequence; otherwise, adding the text to the abnormal sequence;

the expansion and word segmentation module is used for sequentially expanding and segmenting the normal sequence and the abnormal sequence;

and the training and updating module is used for training the mathematical application question solution model by taking the sequence after word segmentation and the original mathematical application question as a training set, and updating the parameters of the neural network model after the training is finished.

The content of the embodiment of the method of the invention is all applicable to the embodiment of the system, the function of the embodiment of the system is the same as the embodiment of the method, and the beneficial effect achieved by the embodiment of the system is the same as the beneficial effect achieved by the method.

The embodiment of the invention provides a processing system of a mathematical application problem solution model, which comprises:

at least one memory for storing a program;

at least one processor for loading the program to execute the processing method of the mathematical application problem solution model shown in fig. 1.

The content of the embodiment of the method of the invention is all applicable to the embodiment of the system, the function of the embodiment of the system is the same as the embodiment of the method, and the beneficial effect achieved by the embodiment of the system is the same as the beneficial effect achieved by the method.

An embodiment of the present invention provides a storage medium in which a processor-executable program is stored, which, when being executed by a processor, is used to perform the processing method of the mathematical application problem solution model shown in fig. 1.

The embodiment of the invention also discloses a computer program product or a computer program, which comprises computer instructions, and the computer instructions are stored in a computer readable storage medium. The computer instructions may be read by a processor of a computer device from a computer-readable storage medium, and the computer instructions executed by the processor cause the computer device to perform the method shown in fig. 1.

The embodiments of the present invention have been described in detail with reference to the accompanying drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the gist of the present invention. Furthermore, the embodiments of the present invention and the features of the embodiments may be combined with each other without conflict.