CN115310426A - Poetry generation method and device, electronic equipment and readable storage medium - Google Patents

Abstract

The application discloses a poetry generating method, a poetry generating device, electronic equipment and a readable storage medium, which are applied to the technical field of natural language processing, wherein the poetry generating method comprises the following steps: acquiring poetry description information; generating a theme white language sequence corresponding to the target poetry according to the poetry description information and a first preset poetry generator; inputting the subject whispering text sequence into the first preset poetry generator to obtain a candidate poetry clause set corresponding to the target poetry, wherein the candidate poetry clause set is a set of candidate clauses describing the subject of the target poetry; and carrying out iterative adjustment on a candidate poetry clause sequence corresponding to the candidate poetry clause set according to a second preset poetry generator and a poetry clause processing command input by a user to obtain the target poetry. The poetry creation efficiency and the poetry quality are hardly taken into account.

Description

Poetry generation method, device, electronic device and readable storage medium

technical field

The present application relates to the technical field of natural language processing, and in particular to a poetry generation method, device, electronic equipment and readable storage medium.

Background technique

With the continuous development of artificial intelligence, deep learning technology has been applied to more and more fields. Taking poetry generation in the cultural field as an example, it can generate texts in a short time by virtue of a well-trained text generator based on deep learning technology. A large amount of text content, and then generate the poems that users need. However, due to the lack of learning signals that accurately quantify the theme of the poems, the poems generated based on the text generator are often not strong in theme. Fanglian is full of red in the county, and I went to the northwest of Qiongtai, but I suspected that I was already in the smoke." The theme of the first half is the tide, while the second half is described with people as the main body. Therefore, the current poetry generation system based on the text generator , the quality of the generated poems is poor, and if the method of artificial poetry creation is continued, the efficiency of poetry creation is usually low due to the influence of factors such as creative talent and creative quality, so it is currently difficult to balance the efficiency of poetry creation and the quality of poetry .

Contents of the invention

The main purpose of this application is to provide a poetry generation method, device, electronic equipment and readable storage medium, aiming to solve the technical problem in the prior art that it is difficult to balance the efficiency of poetry creation and the quality of poetry.

In order to achieve the above purpose, the present application provides a method for generating poetry, which includes:

Get poem description information;

Generate a theme vernacular sequence corresponding to the target poem according to the poem description information and the first preset poem generator;

The subject vernacular sequence is input to the first preset poem generator to obtain a set of candidate poem clauses corresponding to the target poem, wherein the set of candidate poem clauses is a set of candidate clauses describing the theme of the target poem gather;

According to the second preset poem generator and the poem clause processing command input by the user, iteratively adjust the candidate poem clause sequence corresponding to the candidate poem clause set to obtain the target poem.

In order to achieve the above purpose, the present application also provides a poetry generating device, which includes:

Obtaining module, used to obtain poem description information;

A generating module, configured to generate a theme vernacular sequence corresponding to the target poem according to the poem description information and the first preset poem generator;

An input module, configured to input the subject vernacular sequence into the first preset poem generator to obtain a set of candidate poem clauses corresponding to the target poem, wherein the set of candidate poem clauses describes the subject of the target poem The set of candidate clauses of ;

The adjustment module is configured to iteratively adjust the candidate poetry clause sequence corresponding to the candidate poetry clause set according to the second preset poetry generator and the poetry clause processing command input by the user, so as to obtain the target poem.

The present application also provides an electronic device, the electronic device comprising: a memory, a processor, and a program of the poetry generating method stored on the memory and operable on the processor, the poetry generating method When the program is executed by the processor, the steps of the above poetry generation method can be realized.

The present application also provides a computer-readable storage medium, the computer-readable storage medium stores a program for realizing the poetry generation method, and when the program of the poetry generation method is executed by a processor, the steps of the above-mentioned poetry generation method are realized .

The present application also provides a computer program product, including a computer program, and when the computer program is executed by a processor, the steps of the above poetry generation method are implemented.

The present application provides a poetry generation method, device, electronic equipment, and readable storage medium, that is, to obtain the poem description information; according to the poem description information and the first preset poem generator, generate the topic vernacular text corresponding to the target poem sequence; the subject vernacular sequence is input to the first preset poetry generator to obtain a set of candidate poetry clauses corresponding to the target poem, wherein the set of candidate poetry clauses is a candidate clause describing the theme of the target poem A set of sentences; according to the second preset poetry generator and the poetry clause processing command input by the user, iteratively adjust the candidate poetry clause sequence corresponding to the candidate poetry clause set to obtain the target poem. Since the set of candidate clauses describing the theme of poetry can be obtained through the first preset poetry generator, that is, the purpose of generating candidate clauses that highlight the theme of the target poetry can be achieved through the first preset poetry generator, and then through the second Preset the poem generator and the poetry clause processing command input by the user, adjust the candidate clauses interactively, and finally get the target poem corresponding to the target poem, that is, based on the poem description information of the target poem, in the first preset With the cooperation of the poem generator, the second preset poem generator and the user command, the purpose of obtaining the corresponding target verses of the poem description information is realized, because the first preset poem generator and the second preset poem generator have the ability to generate The characteristics of fast efficiency and low generation cost ensure that the generation process of the candidate clauses corresponding to the target verse is guaranteed, that is, the creation efficiency of the candidate clauses is improved. At the same time, due to the user’s relative to the text generator In other words, the pros and cons of candidate clauses can be evaluated more objectively, and then the target verses obtained after iterative adjustment can be more in line with the needs of the market, that is, through the human-computer interaction between the text generator and the user Iterative processing of candidate clauses in a cooperative manner can achieve the purpose of obtaining higher quality than the single processing of candidate clauses by text generators. Therefore, the purpose of balancing poetry creation efficiency and poetry quality is achieved, not only based on deep learning The technically trained text generator generates the poems that users need, so it overcomes the technical defect of the poetry generation system based on the text generator, which leads to poor thematicity of the generated poems. Therefore, it realizes the creation efficiency of poems and poetry quality.

Description of drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and together with the description serve to explain the principles of the application.

In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, In other words, other drawings can also be obtained from these drawings without paying creative labor.

Fig. 1 is the schematic flow chart of the first embodiment of the poetry generation method of the present application;

Fig. 2 is the schematic flow chart of the second embodiment of the poetry generation method of the present application;

Fig. 3 is the schematic diagram of the device of the third embodiment of the poetry generation method of the present application;

FIG. 4 is a schematic diagram of the device structure of the hardware operating environment involved in the poetry generation method in the embodiment of the present application.

The realization of the purpose, functions and advantages of the present application will be further described in conjunction with the embodiments and with reference to the accompanying drawings.

Detailed ways

In order to make the above objects, features and advantages of the present invention more obvious and understandable, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Apparently, the described embodiments are only some of the embodiments of the present invention, but not all of them. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

Embodiment one

First of all, it should be understood that in the field of natural language processing technology, especially in poetry creation, both humans and machines have unique advantages. The poetry generator based on deep learning can generate poetry with high efficiency. It has universal subjectivity, that is, different people have different judgments on the quality of different quality poems, but they have the same judgment standard for poems with large quality differences. For example, two poems are generated based on the poem generator A "Looking for old friends in ten miles of streets and alleys, looking for bosom friends in high mountains and flowing water" and B "Today is a good day, with a big sun living in my heart", the generator cannot make an objective evaluation of the quality of poems A and B, but if it is created and The evaluation process is all carried out manually, which will make the creation efficiency low. At the same time, poetry pays attention to logic, emotion and artistic conception. Create a feeling that the theme is not prominent. At present, there is a lack of corresponding learning signals for the abstract concept of the theme, which also leads to the theme of the poems generated by the text generator usually not meeting people's expectations, so , the current poetry generation method has the technical defects of low creation efficiency of manual creation and poor creation quality of poetry generator creation. Therefore, there is an urgent need for a method that takes both poetry creation efficiency and poetry quality into account.

The embodiment of the present application provides a method for generating a poem. In the first embodiment of the method for generating a poem of the present application, referring to FIG. 1 , the method for generating a poem includes:

Step S10, obtaining poem description information;

Step S20, generating a theme vernacular sequence corresponding to the target poem according to the poem description information and the first preset poem generator;

In this embodiment, it should be noted that the poetry generation method is set in a poetry generation system, and the poetry generation system can be installed on smart devices such as mobile phones and computers, and the target poems are generated by users through the poetry generation system. The poems that meet the expected needs are obtained. The poem description information is used to describe the user's needs for poems. When the user needs to create a poem, the poem description information can be input in the poem generation system of the smart device, and then based on the user and The simple interaction of the poetry generation system generates poems that meet the needs. In an implementable way, assume that the poetry generation system is installed on the mobile terminal of the mobile phone, and the user inputs "generate a five-character verse to bless the future" in the interactive interface of the mobile phone. , the poetry generation system uses "generate a five-character verse that blesses the future" as the poem description information, and generates poems with the theme of "bless the future" and "five-character verse".

In addition, it should be noted that the first preset poem generator is used to generate sentence-by-sentence poems that conform to the small sentence specification of poetry and that each small sentence corresponds to a prominent theme. For example, in an implementable manner, The user inputs a plurality of keywords as the poem description information, wherein the poem description information includes "the genre of the poem is regulated poetry", then the first preset poem generator generates vernacular clauses corresponding to the user's keywords by generating the specified number of sentences in the poem, where , each vernacular clause is used to describe the theme of the poem, and then the corresponding candidate poetic clause is generated according to each vernacular clause, where the vernacular clause is used to represent the clause that meets the theme of the poem, and the candidate poetry clause is used to represent the clause that meets the theme of the poem and genre clauses, for example, assuming that the genre of poetry is a heptadic verse, then the number of sentences of the vernacular clause and the candidate poetry clause is 8, and the subject vernacular sequence is the input of the first preset poetry generator parameter, used to characterize the total vernacular clauses conforming to the theme of the poem, wherein the total vernacular clauses are the sum of each of the vernacular clauses, for example, in an implementable manner, it is assumed that the poem description information includes "The genre of poetry is five-character quatrains", "u ₁ ", "u ₂ ", and "u ₃ ", that is, the number of the vernacular clauses is 4 (k=4), and the vernacular clauses are "e ₁ " , "e ₂ ", "e ₃ ", and "e ₄ ", then the topic vernacular sequence can be (u ₁ ,u ₂ ,u ₃ ,#,e ₁ ,e ₂ ,e ₃ ,e ₄ ), where "#" is the separator in the middle of the sequence.

As an example, steps S10 to S20 include: obtaining the poem description information of the target poem input by the user; and generating a theme vernacular sequence corresponding to the target poem according to the poem description information and the first preset poem generator.

Wherein, the poem description information includes the description information of the number of poems and the poem theme description information, and the step of generating the topic vernacular sequence corresponding to the target poem according to the poem description information and the first preset poem generator includes:

Step A10, according to the description information of the number of verses, determine the target number of verses of the target poem;

Step A20, according to the poem theme description information, determine the poem theme sequence of the target poem;

Step A30, according to the candidate clause processing command input by the user, iteratively input the poem topic sequence in the first preset poem generator until the candidate topic sequence corresponding to the candidate vernacular clause sequence containing the target number of verses is obtained ;

Step A40, preprocessing the candidate topic sequence to obtain the topic vernacular sequence.

In this embodiment, it should be noted that the description information of the number of poems is used to describe the number of poem clauses of the target poem, and the description information of the theme of poetry is used to describe the number of poem clauses of the target poem. The theme of the poem can specifically be poetry keywords, etc. For example, in an implementable manner, it is assumed that the description information of the number of poems is "4", and the poem theme information is two key words "farewell" and "friends". words, and then the target poem is a five-character quatrain with the theme of "sending off friends", wherein, the description information of the number of poems is the number of target poems, and the poem theme sequence is determined by the description information of the poem theme after initialization Sequence, for example, assuming that the poem topic description information is poetry keywords “u ₄ ” and “u ₅ ”, then the poem topic sequence is (u ₁ , u ₂ , #).

In addition, it should be noted that the candidate clause processing command is used to process the poetry candidate clauses of the target poem, the candidate vernacular clause sequence is the sequence corresponding to the candidate vernacular clause, the The candidate topic sequence is a sequence of candidate topics of the target poem, wherein the candidate topic is a vernacular topic generated by the poetry generation system based on the poem description information. For example, in an implementable manner, it is assumed that the The genre of the target poem is five-character quatrains, among which, the poetry keywords “u ₇ ”, “u ₈ ”, “u ₉ ” and “u ₁₀ ” input by the user, and the theme sequence of the poems are (u ₇ ,u ₈ , u ₉ , u ₁₀ ,#) When the poetry theme sequence corresponding to the poetry keyword is input to the first preset poetry generator, the first preset poetry generator will generate a plurality of candidate vernacular sentences, wherein , the candidate vernacular can be "r ₁ ", "r ₂ ", "r ₃ "... "r _n ", then (r ₁ ), (r ₂ ), (r ₃ )...or r _n can be It is used to represent candidate vernacular clause sequences. If the target vernacular clauses finally selected are “r ₁ ”, “r ₂ ”, “r ₃ ” and “r ₄ ”, the candidate topic sequence can be (u ₇ , u ₈ ,u ₉ ,u ₁₀ ,#,e ₇ ,e ₈ ) or (u ₇ ,u ₈ ,u ₉ ,u ₁₀ ,#,e ₇ ,e ₈ ,e ₉ ), etc.

As an example, steps A10 to A40 include: using the number represented by the verse quantity description information as the target verse quantity of the target poem; splicing the description sequence corresponding to the poem theme description information into the target poem Poetry theme sequence, wherein, the description sequence is used to characterize the sequence corresponding to the description information, and the description information may be one or more; according to the candidate clause processing command input by the user, in the first preset poem The generator iteratively inputs the poem topic sequence until it obtains the candidate topic sequence corresponding to the candidate vernacular clause sequence containing the target number of verses; inserts the preset sequence spacer into the candidate topic sequence to obtain the topic vernacular sequence , wherein, the preset sequence delimiter may be "&".

In an implementable manner, after receiving the poem description information input by the user, the poetry generation system first obtains the poem theme sequence through initialization, and then iteratively inputs the poem theme sequence to the first preset poem generator to obtain the preset The number of candidate vernacular clauses, and then based on manual judgment whether the candidate vernacular clauses meet the quality effect, and then the candidate vernacular clauses that meet the quality effect are processed, and the corresponding candidate vernacular clause sequence is inserted into the poetry topic sequence, until the poem The candidate topic sequence corresponding to the topic sequence includes the candidate vernacular clause sequence of the number of poetry clauses corresponding to the description information of the number of verses, and finally space the candidate topic sequence (by adding sequence spacers) to obtain the topic vernacular sequence.

Wherein, the candidate clause processing command input by the user includes a candidate clause detection command and a candidate clause modification command, and according to the candidate clause processing command input by the user, iteratively input all The poem topic sequence until the candidate topic sequence corresponding to the candidate vernacular clause sequence containing the number of target verses is obtained includes:

Step B10, inputting the poem theme sequence into the first preset poem generator to generate a set of candidate vernacular clauses;

Step B20, according to the candidate clause detection command, detect whether there are candidate vernacular clauses in the candidate vernacular clause set;

Step B30, if the candidate vernacular clauses do not exist, return to the execution step: input the poem theme sequence into the first preset poetry generator to generate a set of candidate vernacular clauses;

Step B40, if the candidate vernacular clause exists, modify the candidate vernacular clause according to the candidate clause modification command to obtain the target vernacular clause;

Step B50, iteratively updating the poetry theme sequence according to the target vernacular clause until the updated poetry theme sequence has the number of candidate vernacular clause sequences of the target verse.

In this embodiment, it should be noted that the candidate clause detection command is used to detect whether the candidate vernacular clause meets the user's quality and effect requirements for the target poem, that is, to detect whether the candidate vernacular clause is The target vernacular clause, the target vernacular clause is a clause that satisfies the user's quality and effect requirements for the target poem, the candidate vernacular clause set is a set of candidate vernacular clauses describing the subject of the target poem, and the candidate vernacular clause The sentence modification command is used to modify the candidate vernacular clauses, and the specific degree of modification is determined by the user. For example, in an implementable manner, assuming that the poetry keyword input by the user is "hope", the candidate vernacular clauses are "Although there are many obstacles in the road ahead, we will still ride the long wind and break through the thousands of miles of waves, hang up the cloud sails, cross the sea, and reach the ideal shore." High sails across the vast sea", in which, the process from the candidate vernacular clause to the target vernacular clause is to improve the user's satisfaction with the vernacular clause.

As an example, steps B10 to B50 include: inputting the poem theme sequence into the first preset poem generator to generate a plurality of candidate vernacular clauses; according to the candidate clause detection command input by the user, Detect whether there are candidate vernacular clauses in the candidate vernacular clause set; if it is detected that the candidate vernacular clause does not exist in the candidate vernacular clause set, then return to the execution step: input the poetry theme sequence into the first preset A poetry generator is set to generate a set of candidate vernacular clauses; if it is detected that the candidate vernacular clauses exist in the set of candidate vernacular clauses, the candidate vernacular clauses are modified according to the modification command of the candidate clauses input by the user , to obtain the target vernacular clause; iteratively inserting the target vernacular clause sequence corresponding to the target vernacular clause into the poem topic sequence to obtain a candidate topic sequence having the number of candidate vernacular clause sequences of the target verse.

Step S30, input the subject vernacular sequence into the first preset poem generator to obtain a set of candidate poem clauses corresponding to the target poem, wherein the set of candidate poem clauses are candidates for describing the theme of the target poem a collection of clauses;

Step S40, according to the second preset poetry generator and the poetry clause processing command input by the user, iteratively adjust the candidate poetry clause sequence corresponding to the candidate poetry clause set to obtain the target poem.

In this embodiment, it should be noted that the set of candidate poetry clauses is a set of candidate sentences describing the subject of the target poetry, that is, the set of candidate poetry clauses is a set of multiple candidate poetry clauses , the second preset poetry generator is used to optimize poetry clauses sentence by sentence, so as to generate poems that meet user needs. If the clause is five characters, the second preset poem generator can optimize each character of each candidate poem clause, and the specific character is determined by the interaction process between the user and the poem generating system.

As an example, steps S30 to S40 include: inputting the subject vernacular sequence into the first preset poem generator to obtain a set of candidate poem clauses corresponding to the target poem, wherein the set of candidate poem clauses To describe the set of candidate clauses of the target poetry theme; according to the second preset poetry generator and the poetry clause processing command input by the user, iteratively adjust the sequence of candidate poetry clauses corresponding to the set of candidate poetry clauses, The target poem is obtained, wherein the poem clause processing commands are used to process the candidate poem clauses.

Wherein, the poetry clause processing command input by the user includes a poetry clause detection command and a poetry clause evaluation command, and according to the second preset poem generator and the poetry clause processing command input by the user, the candidate poetry The clause sequence of the candidate poem corresponding to the clause set is adjusted iteratively, and the steps of obtaining the target poem include:

Step C10, according to the set of candidate poem clauses, determine the number of target verses of the target poem;

Step C20, according to the poetry clause detection command, detect whether there are candidate poetry clauses in the candidate poetry clause set;

Step C30, if it does not exist, return to the execution step: input the subject vernacular sequence into the first preset poem generator to obtain a set of candidate poem clauses corresponding to the target poem, wherein the set of candidate poem clauses is a collection of candidate clauses describing the subject of said target poem;

Step C40, if it exists, according to the poetry clause evaluation command and the candidate poetry clauses, iteratively update the candidate poetry clause sequence until the updated candidate poetry clause sequence contains the target number of poems The target sequence of verses;

Step C50, converting the target verse sequence in the updated candidate poetry clause sequence into the target verse sequence.

As an example, steps C10 to C50 include: using the number of candidate poem clauses in the candidate poem clause set as the number of target poems in the target poem; detecting the candidate poem clauses according to the poetry clause detection command Whether there are candidate poetry clauses in the sentence set, wherein, the poetry clause detection command is used to detect whether the candidate poetry clauses meet the user's quality effect requirements for the target poetry; if it is detected that the candidate poetry clause set does not exist, the Candidate poetry clauses, then return to the execution step: input the subject vernacular sequence into the first preset poetry generator to obtain a set of candidate poetry clauses corresponding to the target poem, wherein the set of candidate poetry clauses is description A collection of candidate clauses of the target poetry theme; if it is detected that the candidate poetry clauses exist in the candidate poetry clause set, then according to the poetry clause evaluation command and the candidate poetry clauses, the candidate The poetry clause sequence is iteratively updated until the updated candidate poetry clause sequence contains the target verse sequence of the target verse number; the target verse sequence in the updated candidate poetry clause sequence is converted into the target verse sequence .

Wherein, according to the poetry clause evaluation command and the candidate poetry clauses, the sequence of candidate poetry clauses is iteratively updated until the updated sequence of candidate poetry clauses contains the target verses of the target verse number The sequence of steps includes:

Step D10: Evaluate and edit the candidate poetry clauses according to the first poem clause evaluation command, and obtain the candidate poems after evaluation and editing;

Step D20, according to the second poem clause evaluation command, determine whether the candidate poem is the target poem;

Step D30, if yes, update the clause sequence of the candidate poems according to the candidate poems;

Step D40, if not, then determine the candidate verse sequence corresponding to the candidate verse clause sequence;

Step D50, processing the candidate verse sequence according to the clause sequence numbers of the candidate verse clauses to obtain the processed candidate verse sequence;

Step D60, input the processed candidate poem sequence into the second preset poem generator to obtain the target poem sentence, and return to the execution step: according to the poem clause detection command, detect whether the candidate poem clause set is There are candidate poetry clauses.

In this embodiment, it should be noted that the first poetry clause evaluation command is used to feed back the user's evaluation result on the poetry clause, and the second poetry clause evaluation command is used to feed back the user's evaluation on the overall poem As a result, the clause sequence numbers are used to identify the positions of different candidate poem clauses in the target poem, and the candidate poem sequence is formed by splicing each candidate poem clause sequence, for example, in an implementable In the method, it is assumed that after the topic vernacular sequence is input into the first preset poetry generator, the candidate poetry clauses are "p ₁ ", "p ₂ ", "p ₃ "..."p _n ", according to the first A poem clause evaluation command to obtain the evaluated and edited candidate verse "s _k ", and then insert the sequence corresponding to the evaluated and edited candidate verse into the subject vernacular sequence, until the target sentence inserted into the subject vernacular sequence The quantity is consistent with the quantity represented by the verse quantity description information.

As an example, step D10 to step D60 include: according to the first poem clause evaluation command input by the user, perform evaluation and editing on the candidate poem clauses to obtain the candidate poem sentences after evaluation; according to the second poem clause input by the user, Clause evaluation command to determine whether the candidate verse is the target verse; if it is determined that the candidate verse is the target verse, the candidate verse sequence corresponding to the candidate verse is spliced into the candidate verse clause sequence; if determined If the candidate verse is not the target verse, then determine the candidate verse sequence corresponding to the candidate verse clause sequence; process the candidate verse sequence according to the clause sequence number of the candidate verse clause to obtain a processed the candidate verse sequence after processing; input the processed candidate verse sequence to the second preset poetry generator to obtain the target verse sentence, and return to the execution step: detect the candidate poetry according to the poetry clause detection command Whether there are candidate poem clauses in the clause set.

Wherein, before the step of obtaining poem description information, the poem generation method also includes:

Step E10, obtaining the poetry keyword sequence corresponding to the poetry training corpus and the poetry topic sequence corresponding to the translation training corpus, wherein the poetry training corpus of the poetry training corpus and the translation training corpus of the translation training corpus are one by one correspond;

Step E20, splicing together the poetry training corpus sequence, the poetry keyword sequence and the poetry theme sequence into a poetry training corpus sequence;

Step E30, splicing and adjusting the poetry training corpus sequence according to the first preset splicing method to obtain the first training corpus sequence;

Step E40, according to the first training corpus sequence, train the first generator to be trained to obtain a first preset poetry generator.

In this embodiment, it should be noted that the poetry training corpus is a collection of poetry training corpora, and the poetry training corpus is used to characterize poems randomly extracted from the cloud database by the poetry generation system, and the translation training corpus is A collection of translations of the poetry training corpus, the poetry keyword sequence is a sequence corresponding to the poetry keywords extracted from the poetry training corpus by the poetry generation system, and the poetry theme sequence is used to characterize the translation training corpus of the translation training corpus Describe the sequence corresponding to the topic, the first preset splicing method is the splicing sequence of keywords-theme-poetry clauses, the first training corpus sequence is used to train the first generator to be trained, so The first generator to be trained can be GTP-2 or the like.

As an example, step E10 to step E40 include: obtaining the poetry keyword sequence corresponding to the poetry training corpus and the poetry theme sequence corresponding to the translation training corpus, wherein the poetry training corpus of the poetry training corpus and the translation The translation training corpus of the training corpus corresponds one-to-one; the poetry training corpus sequence, the poetry keyword sequence and the poetry theme sequence are jointly spliced into the poetry training corpus sequence; the poetry training corpus sequence, the poetry keyword sequence The sequence and the poetry theme sequence are jointly spliced into a poetry training corpus sequence; according to the method of splicing keywords-theme-poetry clauses as the splicing order, the poetry training corpus sequence is spliced and adjusted to obtain the first training corpus sequence.

In a practicable manner, assume that the poetry training corpus is {s ₁ , s ₂ , s ₃ , s ₄ }, where the subscript “k” of s _k is used to represent the poetry subscript corresponding to the poetry training corpus The position of the sentence in the poem, and then extract the keywords of the poetry training corpus to form the poetry keyword sequence (u ₁ , u ₂ , u ₃ ... u _s ), where the subscript "s" of u _s is used to represent the translation training The importance level of the poetry keywords corresponding to the corpus, that is, keywords with higher importance levels are higher in the corresponding keyword sequence, and the extraction method can be based on translation or completely based on manual extraction, and then Obtain the corresponding translation topic sequence (e ₁ , e ₂ , e ₃ ... e _k ) after processing and editing the translation training corpus, and stitch it into the first training corpus sequence (u ₁ , u ₂ , u ₃ …u _s ,#,e ₁ ,e ₂ ,e ₃ …e _k ,&,s ₁ ,s ₂ ,s ₃ …s _k ), and then train the first generator to be trained based on the first training corpus sequence generator to get the first preset poem generator.

The embodiment of the present application provides a poetry generation method, that is, to obtain the poem description information; according to the poem description information and the first preset poem generator, generate the theme vernacular sequence corresponding to the target poem; convert the theme vernacular sequence Input to the first preset poetry generator to obtain a set of candidate poem clauses corresponding to the target poem, wherein the set of candidate poem clauses is a set of candidate clauses describing the theme of the target poem; according to the second preset It is assumed that the poetry generator and the poetry clause processing command input by the user perform iterative adjustment on the candidate poetry clause sequence corresponding to the candidate poetry clause set to obtain the target poem. Since the set of candidate clauses describing the theme of poetry can be obtained through the first preset poetry generator, that is, the purpose of generating candidate clauses that highlight the theme of the target poetry can be achieved through the first preset poetry generator, and then through the second Preset the poem generator and the poetry clause processing command input by the user, adjust the candidate clauses interactively, and finally get the target poem corresponding to the target poem, that is, based on the poem description information of the target poem, in the first preset With the cooperation of the poem generator, the second preset poem generator and the user command, the purpose of obtaining the corresponding target verses of the poem description information is realized, because the first preset poem generator and the second preset poem generator have the ability to generate The characteristics of fast efficiency and low generation cost ensure that the generation process of the candidate clauses corresponding to the target verse is guaranteed, that is, the creation efficiency of the candidate clauses is improved. At the same time, due to the user’s relative to the text generator In other words, the pros and cons of candidate clauses can be evaluated more objectively, and then the target verses obtained after iterative adjustment can be more in line with the needs of the market, that is, through the human-computer interaction between the text generator and the user Iterative processing of candidate clauses in a cooperative manner can achieve the purpose of obtaining higher quality than the single processing of candidate clauses by text generators. Therefore, the purpose of balancing poetry creation efficiency and poetry quality is achieved, not only based on deep learning The technically trained text generator generates the poems that users need, so it overcomes the technical defect of the poetry generation system based on the text generator, which leads to poor thematicity of the generated poems. Therefore, it realizes the creation efficiency of poems and poetry quality.

Embodiment two

Further, referring to FIG. 2 , in another embodiment of the present application, for content that is the same as or similar to the first embodiment above, reference may be made to the introduction above, and details will not be repeated hereafter. On this basis, after the step of splicing together the poetry training corpus sequence, the poetry keyword sequence and the poetry theme sequence into a poetry training corpus sequence, the poetry generation method further includes:

Step F10, splicing and adjusting the poetry training corpus sequence according to the second preset splicing method to obtain a second training corpus sequence;

Step F20, performing mask processing on the second training corpus sequence to obtain a target training corpus sequence;

Step F30, according to the target training corpus sequence, train the second generator to be trained to obtain a second preset poetry generator.

In this embodiment, it should be noted that the second preset splicing method is used to characterize the splicing method with theme-poetry clauses as splicing order, wherein different sequences are separated by intermediate spacers, in one possible In an implementation manner, the second training corpus sequence may be (e ₁ , e ₂ , e ₃ ... e _k , &, s ₁ , s ₂ , s ₃ ... s _k ), and the target training corpus sequence is the second training corpus sequence after masking processing by a preset training algorithm, for example, assuming that the target training corpus sequence is (e ₁ , e ₂ , e ₃ ...... e _k ,&, s ₁ , s ₂ , s ₃ ... s _k ^* ), then s _k ^* is the poem clause for masking. In order to be consistent with the actual application, the masking position can be set as k ₁ , k ₂ ... k _m , where "m" is represented by In order to represent the number of occlusions, 1≤m≤M, the value of "M" is: 4 for five-character poems, and 5 for seven-character poems. In order to ensure the operating efficiency of the poetry generation system and the user's psychology, this algorithm only applies to the current clause s _k ^* Perform mask processing, where the following cross-entropy loss function is used when training the generator to be trained with the target training corpus sequence:

由于此算法输入中的某一部分始终保持不变，进而在保证了机器翻译优势的基础上，采用了随机遮挡解码，可实现对某一诗歌子句中不相邻的词语进行掩码处理。Among them, s is, D ₁ is the poetry training corpus, k is the clause sequence number, K is the number of target verses, i is the character of the occluded poetry clause, m is the occluded number, softmax(G ₂ (x _k ), k _i ) is to calculate the softmax function for the occluded character at k _i position,

Because a certain part of the input of this algorithm remains unchanged, and on the basis of ensuring the advantages of machine translation, random occlusion decoding is used to realize the mask processing of non-adjacent words in a certain poem clause.

As an example, steps F10 to F30 include: splicing and adjusting the poetry training corpus sequence to obtain a second training corpus sequence in the manner of splicing the topic-poetry clauses in splicing order; The corpus sequence is masked to obtain the target training corpus sequence; according to the target training corpus sequence, the second generator to be trained is trained to obtain a second preset poetry generator, wherein the second generator to be trained can be Bidirectional generators such as BERT-2.

The embodiment of the present application provides a poetry generator training method, that is, according to the second preset splicing method, the sequence of the poetry training corpus is spliced and adjusted to obtain the second sequence of training corpus, and the sequence of the second training corpus is The sequence is masked to obtain a target training corpus sequence, and according to the target training corpus sequence, the second generator to be trained is trained to obtain a second preset poetry generator. Compared with training the second generator to be trained through the traditional text generation training method, the embodiment of the present application can randomly mask the preset number of characters of the current poetry clause, because only the current poetry clause Masking, and then can achieve the purpose of ensuring the efficiency of the training poetry generator, at the same time, because the algorithm can be used for the preset number of characters of the current poetry clause, that is, both adjacent to the current poetry clause Characters are masked, and the non-adjacent characters of the current poetry clause can also be masked, so it overcomes the long time-consuming mask processing of the non-adjacent characters of the current poetry clause in the prior art, which leads to The low efficiency of the training generator has technical defects, so it lays the foundation for the application effect of the poetry generator.

Embodiment Three

The embodiment of the present application also provides a poem generation device, referring to Figure 3, the poem generation device includes:

Obtaining module, used to obtain poem description information;

A generating module, configured to generate a theme vernacular sequence corresponding to the target poem according to the poem description information and the first preset poem generator;

An input module, configured to input the subject vernacular sequence into the first preset poem generator to obtain a set of candidate poem clauses corresponding to the target poem, wherein the set of candidate poem clauses describes the subject of the target poem The set of candidate clauses of ;

The adjustment module is configured to iteratively adjust the candidate poetry clause sequence corresponding to the candidate poetry clause set according to the second preset poetry generator and the poetry clause processing command input by the user, so as to obtain the target poem.

Optionally, the poetry clause processing command input by the user includes a poetry clause detection command and a poetry clause evaluation command, and the adjustment module is also used for:

Determining the number of target verses of the target poem according to the set of candidate verses;

According to the poetry clause detection command, detect whether there are candidate poetry clauses in the candidate poetry clause set;

If it does not exist, then return to the execution step: input the subject vernacular sequence into the first preset poem generator to obtain a set of candidate poem clauses corresponding to the target poem, wherein the set of candidate poem clauses is described A collection of candidate clauses describing the subject of the target poem;

If it exists, according to the poetry clause evaluation command and the candidate poetry clauses, iteratively update the sequence of candidate poetry clauses until the updated sequence of candidate poetry clauses contains the target verses of the target number of verses sequence;

converting the target verse sequence in the updated candidate verse clause sequence into the target verse sequence.

Optionally, the adjustment module is also used for:

According to the evaluation command of the first poetry clause, the candidate poetry clause is evaluated and edited to obtain the candidate poetry sentence after the evaluation and editing;

determining whether the candidate verse is a target verse according to the second verse clause evaluation command;

If so, then update the sequence of clauses of the candidate poems according to the candidate poems;

If not, then determine the candidate verse sequence corresponding to the candidate verse clause sequence;

Processing the candidate verse sequence according to the clause sequence numbers of the candidate verse clauses to obtain the processed candidate verse sequence;

Input the processed candidate verse sequence to the second preset poetry generator to obtain the target verse sentence, and return to the execution step: according to the poetry clause detection command, detect whether there are candidate poems in the candidate poetry clause set clause

Optionally, the poem description information includes the description information of the number of poems and the description information of the theme of the poem, and the generation module is also used for:

Determining the target number of lines of the target poem according to the description information of the number of lines;

Determining the poem theme sequence of the target poem according to the poem theme description information;

According to the candidate clause processing command input by the user, iteratively input the poem topic sequence in the first preset poetry generator until the candidate topic sequence corresponding to the candidate vernacular clause sequence containing the number of target verses is obtained;

The candidate topic sequence is preprocessed to obtain the topic vernacular sequence.

Optionally, the candidate clause processing command input by the user includes a candidate clause detection command and a candidate clause modification command, and the generating module is also used for:

inputting the poem theme sequence into the first preset poem generator to generate a set of candidate vernacular clauses;

According to the candidate clause detection command, detect whether there are candidate vernacular clauses in the candidate vernacular clause set;

If the candidate vernacular clauses do not exist, return to the execution step: input the poem theme sequence to the first preset poetry generator to generate a set of candidate vernacular clauses;

If the candidate vernacular clause exists, modify the candidate vernacular clause according to the candidate clause modification command to obtain the target vernacular clause;

According to the target vernacular clauses, the poetry theme sequence is iteratively updated until the updated poetry theme sequence has the number of candidate vernacular clause sequences of the target verse.

Optionally, the poetry generating device is also used for:

Obtain the poetry keyword sequence corresponding to the poetry training corpus and the poetry theme sequence corresponding to the translation training corpus, wherein the poetry training corpus of the poetry training corpus is in one-to-one correspondence with the translation training corpus of the translation training corpus;

Splicing together the poetry training corpus sequence, the poetry keyword sequence and the poetry theme sequence into a poetry training corpus sequence;

According to the first preset splicing method, splicing and adjusting the poetry training corpus sequence to obtain the first training corpus sequence;

According to the first training corpus sequence, the first generator to be trained is trained to obtain a first preset poetry generator.

Optionally, the poetry generating device is also used for:

According to the second preset splicing method, splicing and adjusting the poetry training corpus sequence to obtain the second training corpus sequence;

Perform mask processing on the second training corpus sequence to obtain a target training corpus sequence;

According to the target training corpus sequence, the second generator to be trained is trained to obtain a second preset poetry generator.

The poetry generation device provided by the present invention adopts the poetry generation method in the above-mentioned embodiments, and solves the technical problem that it is difficult to balance the efficiency of poetry creation and the quality of poetry. Compared with the prior art, the beneficial effect of the poem generating device provided by the embodiment of the present invention is the same as that of the poem generating method provided by the above embodiment, and other technical features in the poem generating device are the same as those disclosed by the method of the above embodiment The features are the same, and will not be repeated here.

Embodiment Four

An embodiment of the present invention provides an electronic device. The electronic device includes: at least one processor; and a memory connected to the at least one processor in communication; wherein, the memory stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor. The processor executes, so that at least one processor can execute the device control method based on the electronic manual in the first embodiment above.

Referring now to FIG. 4 , it shows a schematic structural diagram of an electronic device suitable for implementing an embodiment of the present disclosure. The electronic equipment in the embodiment of the present disclosure may include but not limited to such as mobile phone, notebook computer, digital broadcast receiver, PDA (personal digital assistant), PAD (tablet computer), PMP (portable multimedia player), vehicle terminal (such as mobile terminals such as car navigation terminals) and fixed terminals such as digital TVs, desktop computers and the like. The electronic device shown in FIG. 4 is only an example, and should not limit the functions and scope of use of the embodiments of the present disclosure.

As shown in FIG. 4 , the electronic device may include a processing device 1001 (such as a central processing unit, a graphics processing unit, etc.), which may be loaded into a random access memory according to a program stored in a read-only memory (ROM) 1002 or from a storage device 1003. (RAM) 1004 to execute various appropriate actions and processing. In RAM 1004, various programs and data necessary for the operation of the electronic device are also stored. The processing device 1001 , ROM 1002 , and RAM 1004 are connected to each other via a bus 1005 . An input/output (I/O) interface 1006 is also connected to the bus.

Typically, the following systems can be connected to the I/O interface 1006: input devices 1007 including, for example, a touch screen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; including, for example, a liquid crystal display (LCD), speakers, an output device 1008 of a vibrator or the like; a storage device 1003 including, for example, a magnetic tape, a hard disk, or the like; and a communication device 1009 . A communication device may allow an electronic device to communicate with other devices wirelessly or by wire to exchange data. While an electronic device is shown with various systems in the figures, it should be understood that implementing or having all of the systems shown is not a requirement. More or fewer systems may alternatively be implemented or provided.

In particular, according to an embodiment of the present disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, where the computer program includes program codes for executing the methods shown in the flowcharts. In such an embodiment, the computer program may be downloaded and installed from the network via the communication means 1009, or from the storage means 1003, or from the ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the methods of the embodiments of the present disclosure are executed.

The electronic device provided by the present invention adopts the device control method based on the electronic instruction manual in the above-mentioned embodiments, and solves the technical problem of low efficiency of controlling the electronic instruction manual and the operation interface of the smart device to complete the pairing. Compared with the prior art, the beneficial effect of the electronic device provided by the embodiment of the present invention is the same as that of the device control method based on the electronic manual provided by the above embodiment, and other technical features of the electronic device are the same as those of the above embodiment method The disclosed features are the same and will not be repeated here.

It should be understood that various parts of the present disclosure may be implemented in hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics may be combined in any one or more embodiments or examples in an appropriate manner.

The above is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Anyone skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present invention. Should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be determined by the protection scope of the claims.

Embodiment five

This embodiment provides a computer-readable storage medium, which has computer-readable program commands stored thereon, and the computer-readable program commands are used to execute the poetry generation method in the first embodiment above.

The computer-readable storage medium provided by the embodiments of the present invention may be, for example, a USB flash drive, but is not limited to an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, system, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, electrical connections with one or more wires, portable computer diskettes, 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 device, magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program, and the program may be used by or in conjunction with the command execution system, system or device. Program code embodied on a computer readable storage medium may be transmitted by any appropriate medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

The above-mentioned computer-readable storage medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

The above-mentioned computer-readable storage medium carries one or more programs, and when the above-mentioned one or more programs are executed by the electronic device, the electronic device: acquires poem description information; according to the poem description information and the first preset poem generator , generating the subject vernacular sequence corresponding to the target poem; inputting the subject vernacular sequence to the first preset poetry generator to obtain a set of candidate poem clauses corresponding to the target poem, wherein the set of candidate poem clauses is description The set of candidate clauses of the target poetry theme; according to the second preset poetry generator and the poetry clause processing command input by the user, iteratively adjust the candidate poetry clause sequence corresponding to the candidate poetry clause set to obtain the Describe the target poem.

Computer program code for carrying out the operations of the present disclosure can be written in one or more programming languages, or combinations thereof, including object-oriented programming languages—such as Java, Smalltalk, C++, and conventional Procedural Programming Language - such as "C" or a similar programming language. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (such as through an Internet Service Provider). Internet connection).

The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, program segment, or portion of code that contains one or more logical functions for implementing specified executable command. It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or they may sometimes be executed in the reverse order, depending upon the functionality involved. It should also be noted that each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations, can be implemented by a dedicated hardware-based system that performs the specified functions or operations , or can be implemented with a combination of dedicated hardware and computer commands.

The modules involved in the embodiments described in the present disclosure may be implemented by software or by hardware. Wherein, the name of the module does not constitute a limitation of the unit itself under certain circumstances.

The computer-readable storage medium provided by the present invention stores computer-readable program commands for executing the above-mentioned poem generation method, which solves the technical problem that it is difficult to take into account both poetry creation efficiency and poetry quality. Compared with the prior art, the beneficial effect of the computer-readable storage medium provided by the embodiment of the present invention is the same as the beneficial effect of the poetry generating method provided by the above-mentioned embodiment, and will not be repeated here.

Embodiment six

The present application also provides a computer program product, including a computer program, and when the computer program is executed by a processor, the steps of the above poetry generation method are implemented.

The computer program product provided by this application solves the technical problem that it is difficult to balance the efficiency of poetry creation and the quality of poetry. Compared with the prior art, the beneficial effect of the computer program product provided by the embodiment of the present invention is the same as that of the method for generating poetry provided by the above embodiment, and will not be repeated here.

The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structures or equivalent process transformations made by using the description of the application and the accompanying drawings are directly or indirectly used in other related technical fields. , are all included in the patent processing scope of the present application in the same way.

Claims (10)

1. A poetry generating method, characterized in that the poetry generating method comprises:

acquiring poetry description information;

generating a theme white language sequence corresponding to the target poetry according to the poetry description information and a first preset poetry generator;

inputting the subject whispering text sequence into the first preset poetry generator to obtain a candidate poetry clause set corresponding to the target poetry, wherein the candidate poetry clause set is a set of candidate clauses describing the subject of the target poetry;

and iteratively adjusting the candidate poetry clause sequence corresponding to the candidate poetry clause set according to a second preset poetry generator and a poetry clause processing command input by a user to obtain the target poetry.

2. The poetry generating method as claimed in claim 1, wherein said poetry clause processing commands input by the user include poetry clause detecting commands and poetry clause evaluating commands,

the step of iteratively adjusting the candidate poetry clause sequence corresponding to the candidate poetry clause set according to a second preset poetry generator and a poetry clause processing command input by a user to obtain the target poetry comprises the following steps:

determining the number of target poetry sentences of the target poetry according to the candidate poetry sentence clause set;

detecting whether the candidate poetry clause set has a candidate poetry clause or not according to the poetry clause detection command;

if not, returning to the execution step: inputting the theme Chinese sequence to the first preset poetry generator to obtain a candidate poetry clause set corresponding to the target poetry, wherein the candidate poetry clause set is a set of candidate clauses describing the theme of the target poetry;

if the poetry clause sequence exists, carrying out iterative updating on the candidate poetry clause sequence according to the poetry clause evaluation command and the candidate poetry clauses until the updated candidate poetry clause sequence contains the target poetry clause sequence with the number of the target poetry clauses;

and converting the target verse sequence in the updated candidate verse clause sequence into the target verse.

3. The poetry generating method as claimed in claim 2, wherein said step of iteratively updating the sequence of the candidate poetry clauses according to the poetry clause evaluation command and the candidate poetry clauses until the updated sequence of the candidate poetry clauses includes the sequence of the target poetry clauses with the number of the target poetry clauses comprises:

evaluating and editing the candidate poetry clauses according to the first poetry clause evaluation command to obtain the evaluated and edited candidate poetry clauses;

determining whether the candidate poetry sentence is a target poetry sentence according to the second poetry clause evaluation command;

if yes, updating the candidate poetry clause sequence according to the candidate poetry clauses;

if not, determining a candidate poetry sentence sequence corresponding to the candidate poetry sentence sequence;

processing the candidate poetry sentence sequence according to the clause serial number of the candidate poetry clause to obtain a processed candidate poetry sentence sequence;

inputting the processed candidate poetry sentence sequence into the second preset poetry generator to obtain a target poetry sentence clause, and returning to the executing step: and detecting whether the candidate poetry clause set has the candidate poetry clauses or not according to the poetry clause detection command.

4. The poetry generating method as claimed in claim 1, wherein said poetry description information includes poetry quantity description information and poetry theme description information,

the step of generating the theme white language sequence corresponding to the target poetry according to the poetry description information and the first preset poetry generator comprises the following steps:

determining the target poetry sentence number of the target poetry according to the poetry sentence number description information;

determining a poetry theme sequence of the target poetry according to the poetry theme description information;

iteratively inputting the poetry theme sequence in the first preset poetry generator according to a candidate clause processing command input by a user until a candidate theme sequence corresponding to the candidate white clause sequence containing the target poetry sentence number is obtained;

and preprocessing the candidate theme sequence to obtain the theme white language sequence.

5. The poetry generating method according to claim 4, wherein said user-inputted candidate clause processing commands include a candidate clause detecting command and a candidate clause modifying command,

the step of iteratively inputting the poetry theme sequence in the first preset poetry generator according to the candidate clause processing command input by the user until obtaining the candidate theme sequence corresponding to the candidate white clause sequence containing the target poetry sentence number comprises the following steps:

inputting the poetry theme sequence into the first preset poetry generator to generate a candidate white-speech clause set;

detecting whether the candidate white-language clause set has a candidate white-language clause or not according to the candidate clause detection command;

if the candidate white clause does not exist, returning to the execution step: inputting the poetry theme sequence into the first preset poetry generator to generate a candidate white-language clause set;

if the candidate white-language clause exists, modifying the candidate white-language clause according to the candidate clause modification command to obtain a target white-language clause;

and iteratively updating the poetry theme sequence according to the target white words, until the updated poetry theme sequence has the candidate white words clause sequence with the target poetry clause number.

6. The poetry generating method as claimed in claim 1, wherein before said step of obtaining poetry descriptive information, said poetry generating method further comprises:

acquiring a poetry keyword sequence corresponding to a poetry training corpus and a poetry theme sequence corresponding to a translation training corpus, wherein poetry training corpuses of the poetry training corpus correspond to translation training corpuses of the translation training corpus one by one;

splicing a poetry training corpus sequence, the poetry keyword sequence and the poetry theme sequence into a poetry training corpus sequence;

splicing and adjusting the poetry training corpus sequence according to a first preset splicing mode to obtain a first training corpus sequence;

and training a first generator to be trained according to the first training corpus sequence to obtain a first preset poetry generator.

7. The method of generating poetry as claimed in claim 1, wherein after said step of concatenating a poetry training corpus sequence, said poetry keyword sequence and said poetry topic sequence into a poetry training corpus sequence, said method of generating poetry further comprises:

splicing and adjusting the poetry training corpus sequence according to a second preset splicing mode to obtain a second training corpus sequence;

performing mask processing on the second corpus sequence to obtain a target corpus sequence;

and training a second generator to be trained according to the target training corpus sequence to obtain a second preset poetry generator.

8. A poetry generating apparatus, characterized in that the poetry generating apparatus comprises:

the acquisition module is used for acquiring poetry description information;

the generating module is used for generating a theme white language sequence corresponding to the target poetry according to the poetry description information and a first preset poetry generator;

the input module is used for inputting the subject white language sequence into the first preset poetry generator to obtain a candidate poetry clause set corresponding to the target poetry, wherein the candidate poetry clause set is a set of candidate clauses describing the subject of the target poetry;

and the adjusting module is used for iteratively adjusting the candidate poetry clause sequence corresponding to the candidate poetry clause set according to a second preset poetry generator and a poetry clause processing command input by a user to obtain the target poetry.

9. An electronic device, characterized in that the electronic device comprises:

at least one processor; and the number of the first and second groups,

a memory communicatively coupled to the at least one processor; wherein,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the steps of the poetry generation method of any of claims 1 to 7.

10. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a program implementing a poetry generating method, which is executed by a processor to implement the steps of the poetry generating method as recited in any one of claims 1 to 7.

Images