Text style transfer method, device and electronic device enhanced with user prior knowledge

Abstract

The present invention discloses a method, device and electronic device for text style transfer enhanced by user prior knowledge, the method comprising: obtaining user prior knowledge data, and building a preset neural network model based on the user prior knowledge data; obtaining user input data, and mapping the user input data to the preset neural network model to generate a target text content representation different from the expression of the user input data; obtaining a user focus area representation, and combining the target text content representation and the user focus area representation to generate the target text content. The present invention can not only realize the conversion of text style, but also improve the relevance of the converted text content with user requirements.

Description

Technical Field

[0001] The present invention relates to the technical field of natural language processing, and in particular to a text style transfer method, device and electronic device enhanced with user prior knowledge. Background Art

[0002] The goal of text style transfer is to rewrite the original text into a new text with a different style while keeping the semantic content of the text unchanged. The meaning of text style is very broad, for example, converting a sentence with negative emotions such as "we sit down and we got some really slow and lazy service" into a sentence with positive emotions such as "we sit down and we got some great and quick service". In addition, formality or informality, stance, tense, etc. can all be considered as the style of the text. In the task of text style transfer, it is usually difficult to collect paired data. It is difficult to know which positive sentence a negative sentence should correspond to. Therefore, it is difficult to directly train the model with supervised learning, but to use unsupervised methods to achieve automatic learning from unpaired text data. Existing unsupervised text style transfer methods mainly include two categories: methods that separate style and then integrate content, and methods that do not separate style and content. Existing unsupervised text style transfer methods mainly consider the input text itself, and use deep learning methods to perform style transfer, ignoring the unique expression styles of different users. Therefore, the generated text is often monotonous and boring, and it is difficult to generate attractive content that meets user needs. Summary of the invention

[0003] The object of the present invention is to provide a text style transfer method, device and electronic device enhanced with user prior knowledge that better meets user requirements.

[0004] In order to solve the above technical problems, a technical solution adopted by the present invention is to provide a text style transfer method enhanced with user prior knowledge, the method comprising:

[0005] Acquire user prior knowledge data, and construct a preset neural network model based on the user prior knowledge data;

[0006] Acquire user input data, and map the user input data to the preset neural network model to generate a target text content representation that is different from the expression of the user input data;

[0007] The user's focus area representation is obtained, and the target text content representation and the user's focus area representation are combined to generate the target text content.

[0008] In order to solve the above technical problem, another technical solution adopted by the present invention is: to provide a text style transfer device enhanced by user prior knowledge, including a unit for executing the text style transfer method enhanced by user prior knowledge as described above.

[0009] To solve the above technical problems, another technical solution adopted by the present invention is: to provide an electronic device, including a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus.

[0010] Memory is used to store computer programs.

[0011] The processor is used to implement the steps of the text style transfer method enhanced by user prior knowledge as described above when executing the program stored in the memory.

[0012] To solve the above technical problem, another technical solution adopted by the present invention is: a computer-readable storage medium having a computer program stored thereon, and when the computer program is processed and executed, the steps of the text style transfer method enhanced with user prior knowledge as described above are implemented.

[0013] The present invention discloses a method, device and electronic device for text style transfer enhanced by user prior knowledge. The method comprises: obtaining user prior knowledge data, and constructing a preset neural network model based on the user prior knowledge data; obtaining user input data, and mapping the user input data to the preset neural network model to generate a target text content representation different from the expression of the user input data; obtaining a user focus area representation, and combining the target text content representation and the user focus area representation to generate the target text content. The present invention learns the user prior knowledge data to construct a preset neural network model based on the user prior knowledge data. After the user data is input, the expression of the target text content generated changes, wherein the changed expression can be selected according to the required expression method, such as the expression commonly used by the user, and by embedding the user focus area representation, the generated target text content not only changes in style, but also generates the target text content in the user's desired focus area, thereby improving the relevance of the converted text content to the user's requirements. The present invention can serve multiple fields such as business, and provide technical support for decision makers to understand information in a timely manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the invention. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0015] Figure 1 It is a flowchart of a text style transfer method enhanced by user prior knowledge according to an embodiment of the present invention.

[0016] Figure 2 It is a schematic diagram of a sub-process of a text style transfer method enhanced by user prior knowledge according to an embodiment of the present invention.

[0017] Figure 3 It is a schematic diagram of a sub-process of a text style transfer method enhanced by user prior knowledge according to an embodiment of the present invention.

[0018] Figure 4 It is a schematic diagram of a sub-process of a text style transfer method enhanced by user prior knowledge according to an embodiment of the present invention.

[0019] Figure 5 It is a schematic diagram of a sub-process of a text style transfer method enhanced by user prior knowledge according to an embodiment of the present invention.

[0020] Figure 6 It is a schematic diagram of a sub-process of a text style transfer method enhanced by user prior knowledge according to an embodiment of the present invention.

[0021] Figure 7 It is a schematic diagram of a sub-process of a text style transfer method enhanced by user prior knowledge according to an embodiment of the present invention.

[0022] Figure 8 It is a schematic diagram of the module structure of an electronic device according to an embodiment of the present invention.

[0023] Fig. 9 It is a schematic diagram of the module structure of a text style transfer device enhanced by user prior knowledge according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Similar component numbers in the drawings represent similar components. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0026] It should also be understood that the terms used in this description of the embodiments of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the embodiments of the present invention. As used in the description of the embodiments of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms of "a", "an" and "the" are intended to include plural forms.

[0027] Please refer to Figure 1 , Figure 1 It is a flowchart of a text style transfer method enhanced by user prior knowledge according to an embodiment of the present invention.

[0028] The present invention provides a text style transfer method enhanced by user prior knowledge, the method comprising:

[0029] S101: Acquire user prior knowledge data, and construct a preset neural network model based on the user prior knowledge data.

[0030] Understandably, the user prior knowledge data is some text content, style / expression, and areas of interest that the user has previously paid attention to, etc. The preset neural network model is a long short-term memory neural network LSTM. LSTM, the full name of which is Long Short Term Memory, is a special recursive neural network used to solve the long-term dependency problem that is prevalent in general recursive neural networks. Using LSTM can effectively transmit and express information in long time series without causing useful information from a long time ago to be ignored or forgotten. At the same time, LSTM can also solve the gradient disappearance / explosion problem in RNN.

[0031] S102: Obtain user input data, and map the user input data to the preset neural network model to generate a target text content representation that is different from the expression of the user input data.

[0032] It can be understood that the user input data is the data that the user wants to convert in style / expression. After mapping it to the preset neural network model, it is processed by the preset neural network model, and the preset neural network model generates a vector representation related to the user input data based on the previous representation learning results.

[0033] S103: Acquire the user's focus area representation, and generate the target text content by combining the target text content representation and the user's focus area representation.

[0034] Understandably, the user's focus area is represented as what is learned when the preset neural network model is constructed based on the user's prior knowledge data.

[0035] The present invention learns the user's prior knowledge data and constructs a preset neural network model based on the user's prior knowledge data. After the user data is input, the expression of the target text content generated changes, wherein the changed expression can be selected according to the required expression method, such as the user's commonly used expression method, and by embedding the user's focus area representation, the generated target text content not only changes in style, but also generates target text content in the user's desired focus area, thereby improving the relevance of the converted text content to the user's requirements. The present invention can serve multiple fields such as business, and provide technical support for decision makers to understand information in a timely manner.

[0036] Please refer to Figure 2 , Figure 2 It is a schematic diagram of a sub-process of a text style transfer method enhanced by user prior knowledge according to an embodiment of the present invention.

[0037] Furthermore, the method of acquiring user prior knowledge data and constructing a preset neural network model based on the user prior knowledge data of the present invention includes:

[0038] S201: Construct an initial neural network model.

[0039] It can be understood that the initial neural network model can be constructed in a common manner.

[0040] S202: Using an initial neural network model, learning the textual expression of the user's prior knowledge data is performed to obtain a specific expression.

[0041] S203: Using the initial neural network model to perform representation learning on the basic text content of the user's prior knowledge data to obtain a specific basic content representation.

[0042] Understandably, before step S203, the text representation separation process may be performed on the user prior knowledge data to generate separate basic text content and text expression.

[0043] S204: Acquire similar expression representations according to the specific expression representations.

[0044] Understandably, similarly, before step S204, the user prior knowledge data may be subjected to text representation separation processing to generate separate basic text content and text expression.

[0045] S205: Generate a specific text content representation according to the similar expression representation and the specific basic content representation.

[0046] Understandably, after obtaining a specific expression, a similar expression that is different from the specific expression but has a high similarity in basic text content is further obtained.

[0047] S206: Using an initial neural network model that can generate a specific text content representation as the preset neural network model.

[0048] Understandably, the specific basic content is represented as a vector representation representing the corresponding text content separated from the text content of the user input data, and the vector representation is expanded into the complete text content according to a similar expression, that is, according to an expression with a higher similarity.

[0049] Understandably, when data is imported into the preset neural network model, the preset neural network model will generate a corresponding vector bar representation.

[0050] Please refer to Figure 3 , Figure 3 It is a schematic diagram of a sub-process of a text style transfer method enhanced by user prior knowledge according to an embodiment of the present invention.

[0051] Furthermore, the present invention obtains user input data, maps the user input data to the preset neural network model, and generates a target text content representation that is different from the expression of the user input data:

[0052] S301: Obtain user input data.

[0053] It can be understood that the user input data includes basic text content and text expression.

[0054] S302: Map the user input data to the preset neural network model, and generate an input expression based on the specific expression.

[0055] It can be understood that, through the representation learning before the preset neural network model, the corresponding input expression representation can be generated based on the specific expression representation.

[0056] S303: Map the user input data to the preset neural network model, and generate an input basic content representation based on the specific basic content representation.

[0057] It can be understood that, through the representation learning before the preset neural network model, the corresponding input basic content representation can be generated based on the specific basic content representation.

[0058] S304: Obtain input similar expression representation according to the input expression representation.

[0059] It can be understood that according to step S205, after obtaining the input expression representation, an input similar expression representation with a high similarity to the input expression representation can be obtained from the neural network model, that is, a vector representation of an expression method similar to the expression method included in the user input data is obtained, and this vector representation is the input similar expression representation.

[0060] S305: Generate the target text content according to the input similar expression representation and the input basic content representation.

[0061] Understandably, according to the representation learning of the preset network model before, after obtaining the input similar expression representation and the input basic content representation, a text content with a high degree of similarity in expression can be generated accordingly, and the text content is used as the target text content. Among them, the input similar expression representation is obtained by the preset network model performing representation learning on the user's prior knowledge data. Therefore, the expression contained in the input similar expression representation conforms to the expression that the user is concerned about or the expression required by the user, that is, the generated target text content is the text content that conforms to the style / expression required by the user.

[0062] Specifically, the user input data x is separated into the input basic text content representation c(x, g src ) and the input text expression a(x, g src ). Use Glove to initialize the word vector and get a vector W with dimension n. i . Use the cosine distance formula to calculate the input expression and the input similar expression a′(g oth ) is calculated as follows:

[0063] s=sim(a(x,g src ), a'(g oth ))

[0064] The smaller the cosine distance, the greater the word correlation. Here, the input similarity expression with the largest correlation is taken as the representation. The long short-term memory neural network LSTM is used to respectively src ) and a′(g oth ) is encoded, and the specific encoding formula is as follows:

[0065] c(x,g src ) = {x1, x, …, x N};

[0066]

[0067]

[0068] h e =[h c ;ha ];

[0069] The last hidden state h obtained by mapping the input basic text content c The last hidden state h obtained by mapping is similar to the input expression a Splice and get the final target text content representation h e .

[0070] Please refer to Figure 4 , Figure 4 It is a schematic diagram of a sub-process of a text style transfer method enhanced by user prior knowledge according to an embodiment of the present invention.

[0071] Furthermore, the step of acquiring the user's focus area representation and combining the target text content representation and the user's focus area representation to generate the target text content includes:

[0072] S401: Using the preset neural network model to perform representation learning on the user focus area of ​​the user prior knowledge data to obtain a user focus area representation.

[0073] S402: Perform weight enhancement processing on the user's focus area representation.

[0074] S403: Decoding the target text content representation using the preset neural network model, and adding the user focus domain representation after weight enhancement processing during the decoding process to obtain the focus text content with enhanced domain features.

[0075] S404: Using the focused text content as the target text content.

[0076] Understandably, the target text content is represented as text content that meets the user's style / expression requirements. The user's focus area is represented by a preset neural network model learned based on the user's prior knowledge data, which represents the vector representation of the user's focus area or required area.

[0077] Specifically, the weight enhancement process is as follows: first, the target text content is semantically decoded using the preset neural network model to obtain each hidden state s when the target text content is decoded. t :

[0078]

[0079] in The augmented character vector representing the input in the decoding phase is calculated as follows:

[0080]

[0081] in, It is used to calculate the character scores contained in the target text content. It is obtained by calculating the tfidf of the characters and performing softmax normalization to obtain the basic weight α of the characters. t On this basis, the user's focus area V contained in the user's prior knowledge data is used to give the weight of the user's focus area a high bias term. The formalization is as follows:

[0082]

[0083] Please refer to Figure 5 , Figure 5 It is a schematic diagram of a sub-process of a text style transfer method enhanced by user prior knowledge according to an embodiment of the present invention.

[0084] Furthermore, the step of obtaining a similar expression according to the specific expression of the present invention includes:

[0085] S501: Obtain a preset expression method.

[0086] It can be understood that the preset expression is a style / expression that the user is interested in and is included in the user's prior knowledge data.

[0087] S502: Map the preset expression method to the preset neural network model to obtain a preset expression representation.

[0088] Understandably, the preset expression is learned by means of a preset neural network model.

[0089] S503: Obtain a specific expression.

[0090] S504: Perform a similarity operation on the specific expression and the preset expression to obtain a calculation result.

[0091] S505: Obtain the similar expression representation according to the calculation result.

[0092] Please refer to Figure 6 , Figure 6 It is a schematic diagram of a sub-process of a text style transfer method enhanced by user prior knowledge according to an embodiment of the present invention.

[0093] Furthermore, the step of generating a specific text content representation based on the similar expression representation and the specific basic content representation of the present invention includes:

[0094] S601: Acquire similar expression representation and specific basic content representation;

[0095] Understandably, similar expression representation: obtain the vector representation of the text expression that is closest to the text expression input by the user from the user's prior knowledge database, that is, select a text expression that is closest to but not the same as the text expression input by the user from the expressions that the user is concerned about.

[0096] S602: Mapping the similar expression representation to the preset neural network model.

[0097] S603: Based on the similar expression representation, the specific basic content representation is decoded using the preset neural network model to generate a similar text content representation that is different from the expression of the specific basic content.

[0098] S604: Using the similar text content representation as the specific text content representation.

[0099] Please refer to Figure 7 , Figure 7 It is a schematic diagram of a sub-process of a text style transfer method enhanced by user prior knowledge according to an embodiment of the present invention.

[0100] Furthermore, the present invention further includes, before obtaining the user's prior knowledge data and constructing a preset neural network model;

[0101] S701: Acquire user history data, and build a user prior knowledge database based on the user history data.

[0102] Understandably, a large amount of user historical data is obtained, which includes text content data, user-focused field data, and user-focused expression data, and a user prior knowledge database is constructed based on this data.

[0103] S702: Importing the user history data into the user prior knowledge database as the user prior knowledge data.

[0104] S703: Obtain user input data.

[0105] It is understandable that the user prior knowledge database stores not only the user historical data that is imported, but also when the user inputs new data, the new data will also be imported and used as historical data.

[0106] S704: Map the user input data to the user prior knowledge database.

[0107] S705: Importing the user input data into the user prior knowledge data as the user prior knowledge data.

[0108] Please refer to Figure 8 , Figure 8It is a schematic diagram of the module structure of an electronic device according to an embodiment of the present invention.

[0109] The present invention further provides an electronic device, including a processor 801 , a communication interface 802 , a memory 803 and a communication bus 804 , wherein the processor 801 , the communication interface 802 and the memory 803 communicate with each other via the communication bus 804 .

[0110] The memory 803 is used to store computer programs.

[0111] The processor 801 is used to implement the steps of the text style transfer method as described above when executing the program stored in the memory.

[0112] Please refer to Fig. 9 , Fig. 9 It is a schematic diagram of the module structure of a text style transfer device enhanced by user prior knowledge according to an embodiment of the present invention.

[0113] The present invention also provides a text style transfer device 900 enhanced with user prior knowledge, comprising a unit for executing the text style transfer method as described above, including:

[0114] The first acquisition unit 901 acquires user prior knowledge data and constructs a preset neural network model.

[0115] The second acquisition unit 902 acquires user input data, and maps the user input data to the preset neural network model to generate a target text content representation that is different from the expression of the user input data.

[0116] The third acquisition unit 903 acquires the user's focus area representation, and generates the target text content by combining the target text content representation and the user's focus area representation.

[0117] In one embodiment, the step of obtaining user prior knowledge data and constructing a preset neural network model includes:

[0118] Build an initial neural network model;

[0119] Acquiring user prior knowledge data, wherein the user prior knowledge data includes basic text content and text expression;

[0120] Using an initial neural network model to perform representation learning on the textual expression of the user's prior knowledge data to obtain a specific expression representation;

[0121] Using an initial neural network model to perform representation learning on the basic text content of the user prior knowledge data to obtain a specific basic content representation;

[0122] According to the specific expression, obtaining a similar expression;

[0123] Generate a specific text content representation based on the similar expression representation and the specific basic content representation;

[0124] An initial neural network model that can generate a specific text content representation is used as the preset neural network model.

[0125] In one embodiment, the obtaining of user input content and mapping the user input content to the preset neural network model to generate a target text content representation different from the user input data expression includes:

[0126] Get user input data;

[0127] Mapping the user input data to the preset neural network model, and generating an input expression based on the specific expression;

[0128] Mapping the user input data to the preset neural network model, and generating an input basic content representation based on the specific basic content representation;

[0129] According to the input expression representation, obtaining an input similar expression representation;

[0130] The target text content is generated according to the input similar expression representation and the input basic content representation.

[0131] In one embodiment, the obtaining of the user's focus area representation and combining the target text content representation with the user's focus area representation to generate the target text content includes:

[0132] Get the target text content representation;

[0133] Obtaining the user's focus area representation;

[0134] Performing weight enhancement processing on the user's focus area representation;

[0135] Decoding the target text content representation using the preset neural network model, and adding the user's focus domain representation after weight enhancement processing during the decoding process to obtain the focus text content with enhanced domain features;

[0136] The focused text content is used as the target text content.

[0137] In one embodiment, obtaining a similar expression according to the specific expression includes:

[0138] Get the preset expression;

[0139] Mapping the preset expression mode to the preset neural network model to obtain a preset expression representation;

[0140] Get a specific expression representation;

[0141] Performing a similarity operation on the specific expression and the preset expression to obtain an operation result;

[0142] According to the operation result, the similar expression representation is obtained.

[0143] In one embodiment, generating a specific text content representation according to the similar expression representation and the specific basic content representation includes:

[0144] Obtaining similar expression representations and specific basic content representations;

[0145] Mapping the similar expression representation to the preset neural network model;

[0146] Based on the similar expression representation, the specific basic content representation is decoded using the preset neural network model to generate a similar text content representation that is different from the expression of the specific basic content;

[0147] The similar text content representation is used as the specific text content representation.

[0148] In one embodiment, before obtaining the user's prior knowledge data and building a preset neural network model, the method further includes:

[0149] Acquire user historical data, and build a user prior knowledge database based on the user historical data;

[0150] Obtaining the user input data;

[0151] The user input data is mapped to the user prior knowledge database.

[0152] The present invention also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is processed and executed, the steps of the text style transfer method as described above are implemented.

[0153] It should be noted that, in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0154] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A text style transfer method enhanced by user prior knowledge, characterized in that: The method comprises: Acquire user prior knowledge data, and construct a preset neural network model based on the user prior knowledge data; Obtaining user input data, and mapping the user input data to the preset neural network model, generating a target text content representation that is different from the user input data in an expression mode, including: obtaining user input data; mapping the user input data to the preset neural network model, and generating an input expression representation based on a specific expression representation; mapping the user input data to the preset neural network model, and generating an input basic content representation based on a specific basic content representation; obtaining an input similar expression representation based on the input expression representation; generating the target text content based on the input similar expression representation and the input basic content representation; Obtaining a user focus area representation, and combining the target text content representation and the user focus area representation to generate the target text content, including: using the preset neural network model to perform representation learning on the user focus area of ​​the user prior knowledge data to obtain the user focus area representation; performing weight enhancement processing on the user focus area representation; using the preset neural network model to perform decoding processing on the target text content representation, and adding the user focus area representation after weight enhancement processing during the decoding process to obtain the focus text content with enhanced domain features; and using the focus text content as the target text content.

2. The text style transfer method enhanced by user prior knowledge as claimed in claim 1, characterized in that: The obtaining of user prior knowledge data and constructing a preset neural network model based on the user prior knowledge data includes: Build an initial neural network model; Using an initial neural network model to perform representation learning on the textual expression of the user's prior knowledge data to obtain a specific expression representation; Using an initial neural network model to perform representation learning on the basic text content of the user prior knowledge data to obtain a specific basic content representation; According to the specific expression, obtaining a similar expression; Generate a specific text content representation based on the similar expression representation and the specific basic content representation; An initial neural network model that can generate a specific text content representation is used as the preset neural network model.

3. The text style transfer method enhanced by user prior knowledge as claimed in claim 1, characterized in that: The acquiring of similar expression representations according to the specific expression representations comprises: Get the preset expression; Mapping the preset expression mode to the preset neural network model to obtain a preset expression representation; Get a specific expression representation; Performing a similarity operation on the specific expression and the preset expression to obtain an operation result; According to the operation result, the similar expression representation is obtained.

4. The text style transfer method enhanced by user prior knowledge as claimed in claim 3, characterized in that: The generating of the specific text content representation according to the similar expression representation and the specific basic content representation comprises: Obtaining similar expression representations and specific basic content representations; Mapping the similar expression representation to the preset neural network model; Based on the similar expression representation, the specific basic content representation is decoded using the preset neural network model to generate a similar text content representation that is different from the expression of the specific basic content; The similar text content representation is used as the specific text content representation.

5. The text style transfer method enhanced by user prior knowledge as claimed in claim 4, characterized in that: Before obtaining the user prior knowledge data and constructing a preset neural network model based on the user prior knowledge data, it also includes: Acquire user historical data, and build a user prior knowledge database based on the user historical data; Importing the user history data into the user prior knowledge database as the user prior knowledge data; Get user input data; Mapping the user input data to the user prior knowledge database; The user input data imported into the user prior knowledge data is used as the user prior knowledge data.

6. A text style transfer device enhanced by user prior knowledge, characterized in that: The method comprises a unit for executing the text style transfer method enhanced by user prior knowledge as described in any one of claims 1 to 5.

7. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; Memory, used to store computer programs; The processor is used to implement the steps of the text style transfer method enhanced by user prior knowledge as described in any one of claims 1 to 5 when executing the program stored in the memory.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is processed and executed, the steps of the text style transfer method enhanced by user prior knowledge as described in any one of claims 1 to 5 are implemented.

Images