Sample manuscript generation method and device based on large language model, equipment and medium

Abstract

The invention provides a sample manuscript generation method and device based on a large language model, equipment and a medium. The method comprises the following steps: acquiring initial detected material segmented words corresponding to detected material voice data; the initial detected material word segmentation is obtained by performing word segmentation on detected material text data corresponding to the detected material voice data; determining the initial detected material segmented words of which the detected material segmented word features meet corresponding preset numerical value conditions as target detected material segmented words; if the number of the target detected material segmented words is smaller than the preset segmented word number, a preset numerical value condition corresponding to the detected material segmented word features is adjusted; skipping to determine the initial detected material segmented words with the detected material segmented word features meeting the corresponding preset numerical value conditions as target detected material segmented words, and continuing execution until the number of the target detected material segmented words is larger than or equal to the preset segmented word number; and generating a sample manuscript according to the finally determined target detected material segmented words by using the large language model. Through the mode of the invention, the generation quality of the sample manuscript can be improved, and then the accuracy of the voiceprint identification result is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of voiceprint identification, and in particular to a sample document generation method, device, equipment and medium based on a large language model. Background Art

[0002] Voiceprint identification technology refers to the technology that determines whether it is the same speaker through human voice information including pronunciation characteristics, speech pattern characteristics, etc. The basic materials for voiceprint identification can be divided into test voice and sample voice. Test voice refers to audio data such as call records and phone recordings of the target person, and sample voice refers to audio data recorded on-site based on sample documents by the person to be identified who may be the target person. There needs to be a sufficient number of identical syllables between the test voice and the sample voice, and the pronunciation method, speaking speed, sound intensity and tone are the same or similar. Therefore, the content of the sample document is crucial to voiceprint identification technology.

[0003] Most of the existing sample documents are fixed texts or texts that are the same as the sample content. However, when the sample document is a fixed text, it cannot be guaranteed that the sample voice and the sample voice have a sufficient number of identical syllables. When the sample document and the sample voice have the same content, the person to be identified may feel nervous, frustrated, depressed, etc. when reading it, or may deliberately disguise himself when reading it, resulting in changes in voice characteristics, which makes voiceprint identification more difficult. Therefore, the existing method of generating sample documents will lead to inaccurate voiceprint identification results. Summary of the invention

[0004] In view of this, the purpose of this application is to provide a sample document generation method, device, equipment and medium based on a large language model, which can improve the generation quality of sample documents and thus improve the accuracy of voiceprint identification results.

[0005] In a first aspect, an embodiment of the present application provides a method for generating a sample document based on a large language model, the method comprising:

[0006] Obtaining initial test material segmentation words corresponding to the test material voice data; the initial test material segmentation words are obtained by segmenting the test material text data corresponding to the test material voice data;

[0007] Determine the initial sample segmentation whose sample segmentation features meet the corresponding preset numerical conditions as the target sample segmentation;

[0008] If the number of target sample segmentation words is less than the preset number of segmentation words, the preset numerical conditions corresponding to the sample segmentation features are adjusted; and the process jumps to determining the initial sample segmentation words whose sample segmentation features meet the corresponding preset numerical conditions as the target sample segmentation words, and continues to execute until the number of target sample segmentation words is greater than or equal to the preset number of segmentation words;

[0009] Using a large language model, sample documents are generated based on the final target sample word segmentation.

[0010] In a possible implementation, the sample word segmentation features include the resonance peak quantity features, average energy features, and single word duration features of the speech segment corresponding to the initial sample word segmentation in the sample speech data.

[0011] In a possible implementation, adjusting the preset numerical conditions corresponding to the sample segmentation features includes:

[0012] The segmentation feature with the highest preset adjustment priority among the segmentation features of the sample material that have been adjusted the least number of times is determined as the target segmentation feature of the sample material;

[0013] Adjust the preset numerical conditions corresponding to the target sample segmentation features.

[0014] In a possible implementation, if the target sample word segmentation feature is a resonance peak quantity feature, adjusting the preset numerical condition corresponding to the target sample word segmentation feature includes:

[0015] The minimum value in the numerical range corresponding to the preset numerical condition is reduced; wherein the numerical value corresponding to the adjusted preset numerical condition is within the preset adjustment numerical range corresponding to the resonance peak quantity feature.

[0016] In a possible implementation, if the target sample word segmentation feature is an average energy feature, adjusting the preset numerical condition corresponding to the target sample word segmentation feature includes:

[0017] Reduce the minimum value in the numerical range corresponding to the preset numerical condition; wherein the numerical range corresponding to the adjusted preset numerical condition is within the preset adjustment numerical range corresponding to the average energy characteristic.

[0018] In a possible implementation, if the target sample word segmentation feature is a single word duration feature, adjusting the preset numerical condition corresponding to the target sample word segmentation feature includes:

[0019] Reduce the minimum value in the numerical range corresponding to the preset numerical condition and / or increase the maximum value in the numerical range corresponding to the preset numerical condition; wherein the numerical range corresponding to the adjusted preset numerical condition is within the preset adjustment numerical range corresponding to the single-word duration feature, and the numerical range corresponding to the preset numerical condition before adjustment is included in the numerical range corresponding to the adjusted preset numerical condition.

[0020] In a possible implementation, a large language model is used to generate a sample document based on the final determined target sample word segmentation, including:

[0021] Obtaining the sample document generation conditions input by the generating user; the sample document generation conditions include the relationship conditions between the sample document and the target sample material segmentation, and / or the relationship conditions between the sample document and the target sample material segmentation in the context of the sample material text data;

[0022] The target sample word segmentation and sample document generation conditions are input into the large language model to obtain a sample document; the sample document meets the sample document generation conditions.

[0023] In a second aspect, an embodiment of the present application further provides a sample document generation device based on a large language model, the device comprising:

[0024] An acquisition module is used to acquire initial sample segmentation words corresponding to the sample voice data; the initial sample segmentation words are obtained by segmenting the sample text data corresponding to the sample voice data;

[0025] A determination module, used to determine the initial sample segmentation whose sample segmentation features meet the corresponding preset numerical conditions as the target sample segmentation;

[0026] An adjustment module is used to adjust the preset numerical conditions corresponding to the sample segmentation features if the number of target sample segmentation features is less than the preset number of segmentation features; and jump to determine the initial sample segmentation whose sample segmentation features meet the corresponding preset numerical conditions as the target sample segmentation, so as to continue to execute until the number of target sample segmentation features is greater than or equal to the preset number of segmentation features;

[0027] The generation module is used to use the large language model to generate sample documents based on the final target sample word segmentation.

[0028] In a possible implementation, the sample word segmentation features include the resonance peak quantity features, average energy features, and single word duration features of the speech segment corresponding to the initial sample word segmentation in the sample speech data.

[0029] In a possible implementation, the adjustment module is specifically used to determine the segmentation feature with the highest preset adjustment priority among the sample segmentation features that have been adjusted the least number of times as the target sample segmentation feature; and adjust the preset numerical conditions corresponding to the target sample segmentation feature.

[0030] In a possible implementation, if the target sample word segmentation feature is a resonance peak quantity feature, the adjustment module is specifically used to reduce the minimum value in the numerical range corresponding to the preset numerical condition; wherein the numerical value corresponding to the adjusted preset numerical condition is within the preset adjustment numerical range corresponding to the resonance peak quantity feature.

[0031] In a possible implementation, if the target sample segmentation feature is an average energy feature, the adjustment module is specifically used to reduce the minimum value in the numerical range corresponding to the preset numerical condition; wherein the numerical range corresponding to the adjusted preset numerical condition is within the preset adjustment numerical range corresponding to the average energy feature.

[0032] In a possible implementation, if the target sample word segmentation feature is a single-word duration feature, the adjustment module is specifically used to reduce the minimum value in the numerical range corresponding to the preset numerical condition and / or increase the maximum value in the numerical range corresponding to the preset numerical condition; wherein the numerical range corresponding to the adjusted preset numerical condition is within the preset adjusted numerical range corresponding to the single-word duration feature, and the numerical range corresponding to the preset numerical condition before the adjustment is included in the numerical range corresponding to the adjusted preset numerical condition.

[0033] In one possible implementation, a generation module is specifically used to obtain sample document generation conditions input by a generating user; the sample document generation conditions include relationship conditions between the sample document and a target sample material segmentation, and / or relationship conditions between the sample document and the target sample material segmentation in the context of the sample material text data; the target sample material segmentation and the sample document generation conditions are input into a large language model to obtain a sample document; the sample document satisfies the sample document generation conditions.

[0034] In a third aspect, an embodiment of the present application further provides an electronic device, comprising: a processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the storage medium communicate through the bus, and the processor executes the machine-readable instructions to perform the steps of the sample document generation method based on a large language model as described in any one of the first aspects.

[0035] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for generating a sample document based on a large language model as described in any one of the first aspects are executed.

[0036] The embodiment of the present application provides a method, device, equipment and medium for generating a sample document based on a large language model, the method comprising: obtaining an initial sample segmentation corresponding to the sample voice data; the initial sample segmentation is obtained after segmenting the sample text data corresponding to the sample voice data; determining the initial sample segmentation whose sample segmentation feature satisfies the corresponding preset numerical condition as the target sample segmentation; if the number of target sample segmentations is less than the preset number of segmentations, adjusting the preset numerical condition corresponding to the sample segmentation feature; and jumping to determining the initial sample segmentation whose sample segmentation feature satisfies the corresponding preset numerical condition as the target sample segmentation, to continue execution until the number of target sample segmentations is greater than or equal to the preset number of segmentations; using the large language model, generating a sample document according to the target sample segmentation finally determined. Through the method of the present application, the generation quality of the sample document can be improved, thereby improving the accuracy of the voiceprint identification results. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application 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.

[0038] Figure 1 A flow chart of a sample document generation method based on a large language model provided in an embodiment of the present application is shown;

[0039] Figure 2 A flowchart of another sample document generation method based on a large language model provided in an embodiment of the present application is shown;

[0040] Figure 3 A schematic diagram of the structure of a sample document generation device based on a large language model provided in an embodiment of the present application is shown;

[0041] Figure 4 A schematic structural diagram of an electronic device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0042] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the drawings in the present application only serve the purpose of explanation and description and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn in real proportion. The flowchart used in this application shows the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowchart can be implemented out of sequence, and the steps without logical context can be reversed in order or implemented simultaneously. In addition, those skilled in the art can add one or more other operations to the flowchart under the guidance of the content of the present application, or remove one or more operations from the flowchart.

[0043] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0044] In order to enable those skilled in the art to use the content of this application, the following implementation is provided in conjunction with a specific application scenario, "the field of voiceprint identification technology". For those skilled in the art, the general principles defined herein can be applied to other embodiments and application scenarios without departing from the spirit and scope of this application. Although this application is mainly described around the "field of voiceprint identification technology", it should be understood that this is only an exemplary embodiment.

[0045] It should be noted that the term "comprising" will be used in the embodiments of the present application to indicate the existence of the features declared thereafter, but does not exclude the addition of other features.

[0046] A sample document generation method based on a large language model provided in an embodiment of the present application is described in detail below.

[0047] Reference Figure 1 As shown, it is a flow chart of a sample document generation method based on a large language model provided in an embodiment of the present application. The specific execution process of the sample document generation method based on a large language model is as follows:

[0048] S101, obtaining initial sample word segments corresponding to the sample voice data; the initial sample word segments are obtained by segmenting the sample text data corresponding to the sample voice data.

[0049] S102: Determine the initial sample segmentation whose sample segmentation features satisfy corresponding preset numerical conditions as the target sample segmentation.

[0050] S103. If the number of target sample segmentations is less than the preset number of segmentations, the preset numerical conditions corresponding to the sample segmentation features are adjusted; and the process jumps to determining the initial sample segmentation whose sample segmentation features meet the corresponding preset numerical conditions as the target sample segmentation, and continues to execute until the number of target sample segmentations is greater than or equal to the preset number of segmentations.

[0051] S104. Using the large language model, generate sample documents based on the final target sample word segmentation.

[0052] The embodiment of the present application provides a method for generating a sample document based on a large language model, the method comprising: obtaining an initial sample segmentation corresponding to the sample voice data; the initial sample segmentation is obtained after segmenting the sample text data corresponding to the sample voice data; determining the initial sample segmentation whose sample segmentation feature satisfies the corresponding preset numerical condition as the target sample segmentation; if the number of target sample segmentations is less than the preset number of segmentations, adjusting the preset numerical condition corresponding to the sample segmentation feature; and jumping to determining the initial sample segmentation whose sample segmentation feature satisfies the corresponding preset numerical condition as the target sample segmentation, to continue execution until the number of target sample segmentations is greater than or equal to the preset number of segmentations; using the large language model, generating a sample document according to the target sample segmentation finally determined. Through the method of the present application, the generation quality of the sample document can be improved, thereby improving the accuracy of the voiceprint identification results.

[0053] The following is a description of the exemplary steps of the present application embodiment:

[0054] S101, obtaining initial sample word segments corresponding to the sample voice data; the initial sample word segments are obtained by segmenting the sample text data corresponding to the sample voice data.

[0055] In the implementation manner of the present application, the sample folder contains sample voice data; the sample voice data is converted into text through ASR (Automatic Speech Recognition) technology to obtain sample text data; the sample text data is segmented to obtain initial sample segmentations; the initial sample segmentations can be words or phrases.

[0056] Among them, the voice data of the inspection sample can be audio data such as call records, phone recordings, etc. provided by the reporter with the suspect in the fields of telecommunications fraud, Internet fraud, etc.

[0057] S102: Determine the initial sample segmentation whose sample segmentation features satisfy corresponding preset numerical conditions as the target sample segmentation.

[0058] In the embodiment of the present application, each initial sample word segmentation corresponds to one or more sample word segmentation features; each sample word segmentation feature corresponds to a preset numerical condition. The sample word segmentation features corresponding to each initial sample word segmentation may include the resonance peak quantity feature, average energy feature, single word duration feature, repetition count feature, etc. of the speech segment corresponding to the initial sample word segmentation in the sample speech data.

[0059] Here, resonance peaks are areas in the sound spectrum where energy is relatively concentrated. They reflect the physical characteristics of the vocal tract and are the main determinants of sound quality. The average energy can be obtained by calculating the value of each sampling point in the speech segment corresponding to the initial sample word segmentation, and then accumulating these values ​​and dividing them by the number of samples. The single-word duration feature refers to the ratio of the duration of the speech segment corresponding to the initial sample word segmentation to the number of words in the initial sample word segmentation; for example: the number of words in the initial sample word segmentation "Hello" is 3, and the duration of the speech segment corresponding to the initial sample word segmentation is 180 milliseconds, then the single-word duration feature is 60 milliseconds. The number of repetitions refers to the number of times the speech segment corresponding to the initial sample word segmentation appears in the entire sample speech data.

[0060] In addition, the preset numerical condition corresponding to the resonance peak quantity feature is a numerical range, such as [3, 4], etc.; the preset numerical condition corresponding to the average energy feature is a numerical range, such as [-20 decibels, 0 decibels]; the preset numerical condition corresponding to the single word duration feature is a numerical range, such as [60 milliseconds, 90 milliseconds]; the preset numerical condition corresponding to the repetition number feature is a numerical range.

[0061] Specifically, an initial sample segmentation whose sample segmentation features satisfy corresponding preset numerical conditions is determined as a target sample segmentation, including: for each initial sample segmentation, counting the number of sample segmentation features that satisfy the corresponding preset numerical conditions among all sample segmentation features corresponding to the initial sample segmentation; if the number is greater than or equal to the preset number, determining the initial sample segmentation as the target sample segmentation.

[0062] S103. If the number of target sample segmentations is less than the preset number of segmentations, the preset numerical conditions corresponding to the sample segmentation features are adjusted; and the process jumps to determining the initial sample segmentation whose sample segmentation features meet the corresponding preset numerical conditions as the target sample segmentation, and continues to execute until the number of target sample segmentations is greater than or equal to the preset number of segmentations.

[0063] In the implementation manner of the present application, the more the number of target sample segmentations, the better the quality of the subsequently generated sample documents; the fewer the number of target sample segmentations, the worse the quality of the subsequently generated sample documents; when the number of target sample segmentations is less than the preset number of segmentations, it means that the preset numerical conditions are set unreasonably and need to be adjusted; and based on the adjusted preset numerical conditions, the target sample segmentations are determined again to increase the number of target sample segmentations, thereby improving the quality of the subsequently generated sample documents.

[0064] In addition, each sample word segmentation feature corresponds to a preset adjustment value range, and the preset numerical condition corresponding to each sample word segmentation feature is within the preset adjustment value range.

[0065] Here, the preset adjustment value range corresponding to the resonance peak quantity feature can be determined according to the sample document generation field. The preset adjustment value range corresponding to the resonance peak quantity feature is different for different sample document generation fields. For example: in the judicial field, the preset adjustment value range of the resonance peak quantity is stipulated to be a maximum of 6 and a minimum of 3.

[0066] Here, the preset adjustment value range corresponding to the average energy feature can be calculated based on the historical average energy data, in decibels; the minimum and maximum values ​​of all historical average energies can be respectively determined as the minimum and maximum values ​​in the preset adjustment value range, and other methods can also be used, which are not limited here.

[0067] Here, the preset adjustment value range corresponding to the single-word duration feature can be determined based on the monosyllable duration, where the monosyllable duration refers to the duration of a syllable; the product of the maximum number of syllables contained in a word and the monosyllable duration is taken as the maximum value of the preset adjustment value range; the product of the minimum number of syllables contained in a word and the monosyllable duration is taken as the minimum value of the preset adjustment value range.

[0068] Here, the maximum value of the preset adjustment value range corresponding to the repetition number feature can be calculated based on the number of characters in the initial sample segmentation and the number of characters in the sample text data; specifically, it can be: calculating the ratio of the number of characters in the sample text data to the number of characters in the initial sample segmentation, and rounding down the ratio to obtain the maximum value of the preset adjustment value range. The minimum value of the preset adjustment value range can be specified by the user.

[0069] Specifically, the preset numerical conditions corresponding to the word segmentation features are adjusted through the following steps:

[0070] Step 1: The word segmentation feature with the highest preset adjustment priority among the word segmentation features of the sample that have been adjusted the least number of times is determined as the target word segmentation feature of the sample.

[0071] In the implementation manner of the present application, the number of times adjusted refers to the number of times the sample segmentation feature has been adjusted during the process of generating the sample document this time. The smaller the impact of the sample segmentation feature on the quality of the sample document, the higher the corresponding preset adjustment priority.

[0072] Step 2: Adjust the preset numerical conditions corresponding to the target sample word segmentation features.

[0073] Furthermore, if the target sample word segmentation feature is a resonance peak quantity feature, the preset numerical condition corresponding to the target sample word segmentation feature is adjusted, including: reducing the minimum value in the numerical range corresponding to the preset numerical condition and keeping the maximum value unchanged; wherein the numerical value corresponding to the adjusted preset numerical condition is within the preset adjustment numerical range corresponding to the resonance peak quantity feature.

[0074] In the implementation manner of the present application, the more resonance peaks there are, the better the initial sample word segmentation is; however, the larger the minimum value in the numerical range corresponding to the preset numerical condition, the smaller the number of target sample word segments will be, so it is sufficient to reduce the minimum value in the numerical range corresponding to the preset numerical condition.

[0075] Furthermore, if the target sample word segmentation feature is an average energy feature, the preset numerical condition corresponding to the target sample word segmentation feature is adjusted, including: reducing the minimum value in the numerical range corresponding to the preset numerical condition, and keeping the maximum value in the numerical range corresponding to the preset numerical condition unchanged; wherein the numerical range corresponding to the adjusted preset numerical condition is within the preset adjustment numerical range corresponding to the average energy feature.

[0076] In the implementation manner of the present application, the greater the average energy, the better the initial sample segmentation; however, the larger the minimum value in the numerical range corresponding to the preset numerical condition, the smaller the number of target sample segmentations will be, so it is sufficient to reduce the minimum value in the numerical range corresponding to the preset numerical condition.

[0077] Furthermore, if the target sample word segmentation feature is a single-word duration feature, the preset numerical condition corresponding to the target sample word segmentation feature is adjusted, including: reducing the minimum value in the numerical range corresponding to the preset numerical condition and / or increasing the maximum value in the numerical range corresponding to the preset numerical condition; wherein the numerical range corresponding to the adjusted preset numerical condition is within the preset adjusted numerical range corresponding to the single-word duration feature, and the numerical range corresponding to the preset numerical condition before the adjustment is included in the numerical range corresponding to the adjusted preset numerical condition.

[0078] In the implementation manner of the present application, if the duration of a single word is too long or too short, it may be due to the disguise of the person to be identified. However, the smaller the numerical range corresponding to the preset numerical condition, the smaller the number of word segments of the target sample will be. Therefore, the minimum value in the numerical range corresponding to the preset numerical condition is reduced and / or the maximum value in the numerical range corresponding to the preset numerical condition is increased to expand the numerical range corresponding to the preset numerical condition.

[0079] Furthermore, if the target sample word segmentation feature is a repetition number feature, the preset numerical condition corresponding to the target sample word segmentation feature is adjusted, including: reducing the minimum value in the numerical range corresponding to the preset numerical condition; wherein the numerical range corresponding to the adjusted preset numerical condition is within the preset adjustment numerical range corresponding to the repetition number feature.

[0080] S104. Using the large language model, generate sample documents based on the final target sample word segmentation.

[0081] In the implementation mode of the present application, the target sample word segmentation can be directly input into the large language model to obtain a sample document. The type of the sample document can be a reading document or a dialogue document, and the user can set it according to actual needs; the user can also modify and manage the generated document (such as sharing, exporting or printing, etc.).

[0082] It is also possible to obtain sample document generation conditions generated by generating user input; the sample document generation conditions include relationship conditions between the sample document and the target sample material segmentation, and / or relationship conditions between the sample document and the target sample material segmentation in the context of the sample material text data; the target sample material segmentation and the sample document generation conditions are input into the large language model to obtain a sample document; the sample document satisfies the sample document generation conditions.

[0083] In addition, the sample document generation method based on the large language model provided in the embodiment of the present application can be applied to any application scenario related to voiceprint identification technology, such as voiceprint laboratory workstation products, etc., and can also be used as a separate technology.

[0084] Reference Figure 2 As shown, it is a flow chart of a sample document generation method based on a large language model provided in an embodiment of the present application. The exemplary steps of the embodiment of the present application are described below:

[0085] S201. Obtaining sample document generation conditions input by a generating user; the sample document generation conditions include relationship conditions between the sample document and target sample segmentation words, and / or relationship conditions between the sample document and the target sample segmentation words in the context of the sample text data.

[0086] Specifically, the relationship conditions between the sample document and the target sample segmentation include:

[0087] (1) Can the corresponding text between the "target text corresponding to the target sample word segmentation in the sample manuscript" and the "target sample word segmentation" be of the same syllable, the same tone, and different writing forms? For example, if the target sample word segmentation is "如果" (rú guǒ), if it can, then the target text is "如果" (rú guǒ); if it cannot, then the target text can be "茹果" (rú guǒ), "如裹" (rú guǒ), etc.

[0088] Here, a syllable in pinyin is a phonetic unit composed of an initial and a final; the tones in pinyin include: the first tone (yinping), with a high and flat sound when pronounced, such as "ā"; the second tone (yangping), with a sound rising from low to high when pronounced, such as "á". The third tone (shang sheng), with a sound first descending and then rising when pronounced, such as "ǎ". The fourth tone (qu sheng), with a sound descending from high to low when pronounced, such as "à".

[0089] (2) Can the corresponding text between the "target text corresponding to the target sample word segmentation in the sample manuscript" and the "target sample word segmentation" be of the same syllable and different tones? For example, if the target sample word segmentation is "如果" (rú guǒ), if it can, then the target text can be "如国" (rú guó), "入国" (rù guó), etc.; if it cannot, then the target text must be "如果" (rú guǒ).

[0090] Specifically, the relationship conditions between the sample manuscript and the context of the target sample word segmentation in the sample text data include:

[0091] (1) The number of the first identical characters in the corresponding positions between the "above text of the target text corresponding to the target sample word segmentation in the sample manuscript" and the "above text of the target sample word segmentation in the sample text data". For example, if the target text is "如果" (rú guǒ) and the original text in the sample text data is "我们如果真的成功了" (wǒ men rú guǒ zhēn de chéng gōng le); if the number of the first identical characters in the corresponding positions is 1, then the above text of the target text corresponding to the target sample word segmentation in the sample manuscript is "们" (men); and "们" (men) is in front of the target text corresponding to the target sample word segmentation, that is, "们如果" (men rú guǒ).

[0092] (2) The number of the second identical characters in the corresponding positions between the "below text of the target text corresponding to the target sample word segmentation in the sample manuscript" and the "below text of the target sample word segmentation in the sample text data". For example, if the target text is "如果" (rú guǒ) and the original text in the sample text data is "我们如果真的成功了" (wǒ men rú guǒ zhēn de chéng gōng le); if the number of the first identical characters in the corresponding positions is 2, then the below text of the target text corresponding to the target sample word segmentation in the sample manuscript is "真的" (zhēn de); and "真的" (zhēn de) is behind the target text corresponding to the target sample word segmentation, that is, "如果真的" (rú guǒ zhēn de).

[0093] S202. Input the target sample word segmentation and the sample manuscript generation conditions into the large language model to obtain the sample manuscript; the sample manuscript meets the sample manuscript generation conditions.

[0094] The embodiment of the present application provides another sample document generation method based on a large language model, which can generate sample documents based on sample document generation conditions, improve the quality of sample documents, reduce the difficulty of voiceprint identification work, and make the voiceprint identification results more accurate.

[0095] Based on the same inventive concept, the embodiments of the present application also provide a sample document generation device based on a large language model corresponding to the sample document generation method based on a large language model. Since the principle of solving the problem by the device in the embodiments of the present application is similar to the sample document generation method based on a large language model in the embodiments of the present application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be repeated.

[0096] Reference Figure 3 FIG. 1 is a schematic diagram of a sample document generation device based on a large language model provided in an embodiment of the present application. The sample document generation device based on a large language model includes:

[0097] The acquisition module 301 is used to acquire the initial sample segmentation words corresponding to the sample voice data; the initial sample segmentation words are obtained by segmenting the sample text data corresponding to the sample voice data;

[0098] A determination module 302 is used to determine the initial sample segmentation whose sample segmentation feature meets the corresponding preset numerical condition as the target sample segmentation;

[0099] The adjustment module 303 is used to adjust the preset numerical conditions corresponding to the sample segmentation features if the number of target sample segmentation features is less than the preset number of segmentation features; and jump to determine the initial sample segmentation whose sample segmentation features meet the corresponding preset numerical conditions as the target sample segmentation, so as to continue to execute until the number of target sample segmentation features is greater than or equal to the preset number of segmentation features;

[0100] The generation module 304 is used to generate a sample document based on the final determined target sample word segmentation using the large language model.

[0101] In a possible implementation, the sample word segmentation features include the resonance peak quantity features, average energy features, and single word duration features of the speech segment corresponding to the initial sample word segmentation in the sample speech data.

[0102] In a possible implementation, the adjustment module 303 is specifically used to determine the segmentation feature with the highest preset adjustment priority among the segmentation features of the sample material that have been adjusted the least number of times as the target segmentation feature of the sample material; and adjust the preset numerical conditions corresponding to the target segmentation feature of the sample material.

[0103] In a possible implementation, if the target sample word segmentation feature is a resonance peak quantity feature, the adjustment module 303 is specifically used to reduce the minimum value in the numerical range corresponding to the preset numerical condition; wherein the numerical value corresponding to the adjusted preset numerical condition is within the preset adjustment numerical range corresponding to the resonance peak quantity feature.

[0104] In a possible implementation, if the target sample segmentation feature is an average energy feature, the adjustment module 303 is specifically used to reduce the minimum value in the numerical range corresponding to the preset numerical condition; wherein the numerical range corresponding to the adjusted preset numerical condition is within the preset adjustment numerical range corresponding to the average energy feature.

[0105] In a possible implementation, if the target sample word segmentation feature is a single-word duration feature, the adjustment module 303 is specifically used to reduce the minimum value in the numerical range corresponding to the preset numerical condition and / or increase the maximum value in the numerical range corresponding to the preset numerical condition; wherein the numerical range corresponding to the adjusted preset numerical condition is within the preset adjusted numerical range corresponding to the single-word duration feature, and the numerical range corresponding to the preset numerical condition before the adjustment is included in the numerical range corresponding to the adjusted preset numerical condition.

[0106] In one possible implementation, the generation module 304 is specifically used to obtain sample document generation conditions input by the generation user; the sample document generation conditions include the relationship conditions between the sample document and the target sample material segmentation, and / or the relationship conditions between the sample document and the target sample material segmentation in the context of the sample material text data; the target sample material segmentation and the sample document generation conditions are input into the large language model to obtain a sample document; the sample document satisfies the sample document generation conditions.

[0107] The embodiment of the present application provides a sample document generation device based on a large language model, which includes: an acquisition module 301, which is used to obtain the initial sample segmentation corresponding to the sample voice data; the initial sample segmentation is obtained after segmenting the sample text data corresponding to the sample voice data; a determination module 302, which is used to determine the initial sample segmentation whose sample segmentation feature meets the corresponding preset numerical condition as the target sample segmentation; an adjustment module 303, which is used to adjust the preset numerical condition corresponding to the sample segmentation feature if the number of target sample segmentations is less than the preset number of segmentations; and jump to determine the initial sample segmentation whose sample segmentation feature meets the corresponding preset numerical condition as the target sample segmentation, to continue execution until the number of target sample segmentations is greater than or equal to the preset number of segmentations; a generation module 304, which is used to generate a sample document using a large language model according to the target sample segmentation finally determined. Through the method of the present application, the generation quality of the sample document can be improved, thereby improving the accuracy of the voiceprint identification results.

[0108] like Figure 4 As shown, an electronic device 400 provided in an embodiment of the present application includes: a processor 401, a memory 402 and a bus, the memory 402 stores machine-readable instructions executable by the processor 401, and when the electronic device is running, the processor 401 communicates with the memory 402 through the bus, and the processor 401 executes the machine-readable instructions to perform the steps of the sample document generation method based on the large language model as described above.

[0109] Specifically, the memory 402 and the processor 401 can be general-purpose memories and processors, which are not specifically limited here. When the processor 401 runs the computer program stored in the memory 402, the sample document generation method based on the large language model can be executed.

[0110] Corresponding to the above-mentioned sample document generation method based on a large language model, an embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned sample document generation method based on a large language model are executed.

[0111] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the system and device described above can refer to the corresponding process in the method embodiment, and will not be repeated in this application. In the several embodiments provided in this application, it should be understood that the disclosed system, device and method can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0112] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0113] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0114] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the information processing method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard drives, ROM, RAM, magnetic disks, or optical disks.

[0115] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A sample document generation method based on a large language model, characterized in that: The method comprises: Acquire initial sample segmentation words corresponding to the sample voice data; the initial sample segmentation words are obtained by segmenting the sample text data corresponding to the sample voice data; Determine the initial sample segmentation whose sample segmentation features meet the corresponding preset numerical conditions as the target sample segmentation; If the number of the target sample segmentation words is less than the preset number of segmentation words, the preset numerical condition corresponding to the sample segmentation feature is adjusted; and the process jumps to the step of determining the initial sample segmentation word whose sample segmentation feature satisfies the corresponding preset numerical condition as the target sample segmentation word, so as to continue the process until the number of the target sample segmentation words is greater than or equal to the preset number of segmentation words; Using a large language model, sample documents are generated based on the final target sample word segmentation.

2. The method for generating sample documents based on a large language model according to claim 1, characterized in that: The sample word segmentation features include the resonance peak quantity features, average energy features and single word duration features of the speech segment corresponding to the initial sample word segment in the sample voice data.

3. The method for generating sample documents based on a large language model according to claim 2, characterized in that: The preset numerical conditions corresponding to the word segmentation features of the sample are adjusted, including: The segmentation feature with the highest preset adjustment priority among the segmentation features of the sample material that have been adjusted the least number of times is determined as the target segmentation feature of the sample material; Adjust the preset numerical conditions corresponding to the target sample word segmentation features.

4. The method for generating sample documents based on a large language model according to claim 3, characterized in that: If the target sample word segmentation feature is the resonance peak quantity feature, then the preset numerical condition corresponding to the target sample word segmentation feature is adjusted, including: The minimum value in the numerical range corresponding to the preset numerical condition is reduced; wherein the numerical value corresponding to the adjusted preset numerical condition is within the preset adjustment numerical range corresponding to the resonance peak quantity feature.

5. The method for generating sample documents based on a large language model according to claim 3, characterized in that: If the target sample word segmentation feature is the average energy feature, then the preset numerical condition corresponding to the target sample word segmentation feature is adjusted, including: Reduce the minimum value in the numerical range corresponding to the preset numerical condition; wherein the numerical range corresponding to the adjusted preset numerical condition is within the preset adjustment numerical range corresponding to the average energy characteristic.

6. The method for generating sample documents based on a large language model according to claim 3, characterized in that: If the target sample word segmentation feature is the single word duration feature, then the preset numerical condition corresponding to the target sample word segmentation feature is adjusted, including: Reduce the minimum value in the numerical range corresponding to the preset numerical condition and / or increase the maximum value in the numerical range corresponding to the preset numerical condition; wherein the numerical range corresponding to the adjusted preset numerical condition is within the preset adjustment numerical range corresponding to the single-word duration feature, and the numerical range corresponding to the preset numerical condition before adjustment is included in the numerical range corresponding to the adjusted preset numerical condition.

7. The method for generating sample documents based on a large language model according to claim 1, characterized in that: The method of using the large language model to generate sample documents according to the final determined target sample word segmentation includes: Obtaining sample document generation conditions input by a generating user; the sample document generation conditions include a relationship condition between the sample document and the target sample material segmentation, and / or a relationship condition between the sample document and the target sample material segmentation in the context of the sample material text data; The target sample word segmentation and the sample document generation condition are input into the large language model to obtain a sample document; the sample document satisfies the sample document generation condition.

8. A sample document generation device based on a large language model, characterized in that: The device comprises: An acquisition module is used to acquire initial sample segmentation words corresponding to the sample voice data; the initial sample segmentation words are obtained by segmenting the sample text data corresponding to the sample voice data; A determination module, used to determine the initial sample segmentation whose sample segmentation features meet the corresponding preset numerical conditions as the target sample segmentation; An adjustment module, for adjusting the preset numerical conditions corresponding to the sample segmentation features if the number of the target sample segmentation features is less than the preset number of segmentation features; and jumping to the step of determining the initial sample segmentation whose sample segmentation features satisfy the corresponding preset numerical conditions as the target sample segmentation, to continue execution until the number of the target sample segmentation features is greater than or equal to the preset number of segmentation features; The generation module is used to use the large language model to generate sample documents based on the final target sample word segmentation.

9. An electronic device, characterized in that: include: A processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of the sample document generation method based on a large language model as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for generating a sample document based on a large language model according to any one of claims 1 to 7 are executed.

Images