Method, device, electronic device and storage medium for generating speech
By identifying and correlating keywords in the target text and generating highly targeted marketing speeches, the problem of insufficient speech by novice marketers is solved and the marketing effect is improved.
Patent Information
- Application Number
- CN202011073081.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-10-09
- Publication Date
- 2024-12-13
- Estimated Expiration
- 2040-10-09
AI Technical Summary
The lack of experience and marketing speech of novices leads to poor marketing results. The existing technology mainly relies on manual summary and teaching of marketing experts, and cannot meet the needs of different marketing scenarios, resulting in insufficient speech.
By obtaining feature information in the target text, identifying keywords, and storing them with reference verbal text into the verbal library, deep learning models are used to extract and generate highly targeted marketing verbal techniques.
It realizes automated mining and enriches marketing speech, provides highly targeted marketing materials, solves the problem of insufficient speech, and improves marketing effectiveness.
Smart Images
Figure CN112214578B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of artificial intelligence / blockchain technology, and in particular to a speech generation method, device, electronic device and computer-readable storage medium. Background Art
[0002] With the development of the economy and Internet technology, the demand for marketing (such as insurance marketing) is growing, and the scale of marketers (such as insurance agents) is becoming larger and larger, especially novice marketers. However, novice marketers face the problem of lack of experience and low marketing level when selling to customers. At present, the relevant technology mainly relies on marketing experts to manually summarize the experience, solidify the experience in the form of rules, and pass it on to novice marketers through lectures or classes. However, since different marketing scenarios require different marketing strategies or marketing words, there are fewer marketing words that can be called through traditional marketing learning methods, which affects the marketing effect.
[0003] It can be seen from this that there is currently a problem of marketing effectiveness being affected by insufficient marketing language. Summary of the invention
[0004] In view of this, the embodiments of the present application provide a method, device, electronic device and computer-readable storage medium for generating marketing words to solve the current problem of insufficient marketing words affecting marketing effectiveness.
[0005] A first aspect of an embodiment of the present application provides a method for generating speech, the method comprising:
[0006] Obtain target text for the first type of business;
[0007] Acquire a reference speech text from the target text, the reference speech text being a text in the target text containing first characteristic information, the first characteristic information being characteristic information of the first type of service;
[0008] According to the first feature information, identifying at least one keyword in the reference speech text;
[0009] The first speech reference data is stored in a target speech library corresponding to the first type of business, wherein the first speech reference data includes an association relationship between the at least one keyword and the reference speech text.
[0010] In a possible implementation of the first aspect, obtaining a reference speech text from the target text includes:
[0011] The target text is input into the speech recognition model, and the reference speech text is output.
[0012] In a possible implementation of the first aspect, before inputting the target text into a speech recognition model to output the reference speech text, the method further includes:
[0013] The text content containing the target type of speech is used as a training sample, and deep learning is used for training to obtain the speech recognition model.
[0014] In a possible implementation of the first aspect, identifying at least one keyword in the reference speech text according to the first feature information includes:
[0015] Acquire, according to the first characteristic information, a preset keyword set associated with the first type of service;
[0016] A keyword in the reference speech text that belongs to the preset keyword set is determined as the at least one keyword.
[0017] In a possible implementation manner of the first aspect, after storing the first speech technique reference data in a target speech technique library corresponding to the first type of service, the method further includes:
[0018] Get the first keyword entered by the user;
[0019] According to the first keyword, a first reference speech text is called from the target speech library.
[0020] In a possible implementation of the first aspect, calling a first reference speech text from the target speech library according to the first keyword includes:
[0021] According to the first keyword, searching the target speech library for a second keyword whose matching degree with the first keyword is greater than a preset matching degree;
[0022] Obtain a first reference speech text associated with the second keyword in the target speech library.
[0023] In a possible implementation of the first aspect, after calling a first reference speech text from the target speech library according to the first keyword, the method further includes: displaying the first reference speech text.
[0024] A second aspect of an embodiment of the present application provides a speech generation device, the device comprising an acquisition unit, a processing unit and a storage unit;
[0025] The acquisition unit is used to acquire the target text of the first type of business;
[0026] The processing unit is configured to obtain a reference speech text from the target text obtained by the obtaining unit, wherein the reference speech text is a text in the target text containing first characteristic information, and the first characteristic information belongs to characteristic information of the first type of service;
[0027] The processing unit is further configured to identify at least one keyword in the reference speech text according to the first feature information;
[0028] The storage unit is used to store the first speech reference data processed by the processing unit into a target speech library corresponding to the first type of business, and the first speech reference data includes an association relationship between the at least one keyword and the reference speech text.
[0029] In a possible implementation of the second aspect, the processing unit is specifically configured to input the target text into a speech recognition model, and output the reference speech text.
[0030] In a possible implementation of the second aspect, the processing unit is further used to use the text content containing the target type of speech as a training sample and adopt a deep learning method to perform training before inputting the target text into the speech recognition model to obtain the reference speech text, so as to obtain the speech recognition model.
[0031] In a possible implementation of the second aspect, the device further includes a display unit. In this case, the acquisition unit is further configured to acquire a first keyword input by a user after the storage unit stores the first speech reference data in a target speech library corresponding to the first type of business; based on the first keyword, search the target speech library for a second keyword whose matching degree with the first keyword is greater than a preset matching degree; and acquire a first reference speech text associated with the second keyword in the target speech library. The display unit is configured to display the first reference speech text acquired by the acquisition unit.
[0032] A third aspect of an embodiment of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the electronic device, wherein the processor implements the steps of the speech generation method provided in the first aspect when executing the computer program.
[0033] The fourth aspect of the embodiments of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of the speech generation method provided in the first aspect are implemented.
[0034] The method, device, electronic device and computer-readable storage medium for generating speech provided by the embodiments of the present application have the following beneficial effects:
[0035] In the embodiment of the present application, the reference speech text in the target text can be obtained, and the keywords in the reference speech text can be extracted, and then the keywords can be associated with the reference speech text and stored in the target speech library, so that the reference speech text (such as insurance marketing speech) can be automatically mined from existing public articles (such as insurance soft articles) to provide marketing speech materials for marketing personnel. This method can enrich the speech and avoid the problem of insufficient marketing speech in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0037] Figure 1 This is a flowchart of a method for generating speech provided by an embodiment of the present application;
[0038] Figure 2 This is a flowchart of a method for generating speech provided by another embodiment of the present application;
[0039] Figure 3 This is a specific implementation flow chart of a speech generation method provided by another embodiment of the present application;
[0040] Figure 4 This is a flowchart of a method for generating speech provided by another embodiment of the present application;
[0041] Figure 5 This is a flowchart for implementing a training speech recognition model in a speech generation method provided in another embodiment of the present application;
[0042] Figure 6 This is a flowchart of a method for generating speech provided in another embodiment of the present application;
[0043] Figure 7 This is a structural block diagram of a speech generation device provided by an embodiment of the present application;
[0044] Figure 8 It is a structural block diagram of a speech generation device provided by another embodiment of the present application;
[0045] Fig. 9 It is a structural block diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0046] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0047] At present, when marketers promote certain businesses to customers (such as insurance sales, education and learning, etc.), they may not be able to use targeted language to promote the corresponding business to customers due to the lack of business-related marketing language. Therefore, they cannot achieve the desired marketing goals, thus affecting the marketing effect.
[0048] In view of this, the embodiment of the present application provides a speech generation method, speech generation device, electronic device and computer-readable storage medium, which can obtain the target text of the first type of business; obtain the reference speech text from the target text, the reference speech text is the text containing the first characteristic information in the target text, and the first characteristic information belongs to the characteristic information of the first type of business; according to the first characteristic information, identify at least one keyword in the reference speech text; store the first speech reference data in the target speech library corresponding to the first type of business, and the first speech reference data includes the association relationship between the at least one keyword and the reference speech text. Through this scheme, the reference speech text in the target text can be identified by the speech recognition model, and the keywords in the reference speech text can be extracted, and then the keywords can be associated with the reference speech text and stored in the target speech library, so that the reference speech text (such as insurance marketing speech) can be automatically mined from the existing public articles (such as insurance soft articles) to provide marketing speech materials for marketing personnel. This method can enrich the speech and avoid the problem that the marketing speech in the prior art is not sufficient.
[0049] The execution subject of the speech generation method provided in the embodiment of the present application can be a speech generation device or an electronic device, or a functional module and / or functional entity in the speech generation device or electronic device that can implement the speech generation method. The specific can be determined according to actual use requirements, and the embodiment of the present application does not limit it. The speech generation method provided in the embodiment of the present application is exemplarily described below by taking the speech generation device as an example.
[0050] See also Figure 1 , Figure 1 The following is a flowchart of a method for generating speech provided by an embodiment of the present application. Figure 1 As shown, the speech generation method includes steps S11 to S14.
[0051] S11: Obtain target text of the first type of business.
[0052] Optionally, in step S11, the first category of business may be insurance marketing business, or may be children's education business, or may be employment recommendation business, or may be any other possible business, which may be determined based on actual usage requirements and is not limited in this embodiment of the present application.
[0053] Optionally, in step S11, the target text of the first type of service may be the content of a published article obtained from the Internet, or may be text obtained in any other possible manner.
[0054] For example, taking the first type of business as insurance marketing business as an example, the target text of the insurance marketing business can be a publicly available insurance soft article, or it can be any other text related to insurance sales. The specific text can be determined based on actual usage requirements, and the embodiments of the present application are not limited to this.
[0055] S12: Obtain a reference speech text from the target text, where the reference speech text is a text in the target text containing first characteristic information, where the first characteristic information belongs to characteristic information of the first type of business.
[0056] In the embodiment of the present application, the content of the target text relates to the first type of business, and the target text contains first characteristic information of the first type of business. The first characteristic information may include a preset keyword set associated with the first type of business.
[0057] In the embodiments of the present application, various possible methods may be used to obtain reference speech text from the target text. For example, a specific speech recognition model may be used to extract reference speech text from the target text.
[0058] Exemplary, combined Figure 1 ,like Figure 2 As shown, the above step S12 can be specifically implemented by the following step S12A.
[0059] S12A: Input the target text into the speech recognition model, and output the reference speech text.
[0060] In an embodiment of the present application, after obtaining the target text for the first type of business, the speech generation device can input the target text into a speech recognition model corresponding to the first type of business, identify the speech content of the target text through the speech recognition model, and obtain a reference speech text.
[0061] For example, taking the first category of business as insurance marketing business, the target text of the insurance marketing business can be input into the insurance marketing speech recognition type, and then the insurance marketing speech recognition type can be used to output the reference speech text belonging to the insurance marketing business, and the reference speech text contains insurance sales characteristic information.
[0062] In step S12, the speech recognition model corresponding to the first type of business may be a model obtained by performing deep learning training on a large number of samples containing the first feature information. The speech recognition model may be used to extract reference speech text for the first type of business.
[0063] For example, taking the first type of business as insurance marketing business, the first characteristic information is characteristic information belonging to the insurance marketing business, such as "illness", "huge medical expenses", "natural disasters and man-made disasters", "protection", etc.
[0064] As an embodiment of the present application, taking the target text as an insurance soft article as an example, Table 1 shows an exemplary result of inputting the target text into a speech classification model to obtain a reference speech text.
[0065] Table 1
[0066]
[0067] It should be noted that the target text can be one text or multiple texts, which can be determined according to actual usage requirements and is not limited to this.
[0068] Exemplarily, if multiple texts are input into the speech classification model, multiple reference speech texts can be obtained.
[0069] It should be noted that the reference speech text can be in the form of a sentence, multiple sentences, or a paragraph, which is not limited in the embodiments of the present application.
[0070] S13: Identify at least one keyword in the reference speech text according to the first feature information.
[0071] In step S13, various keyword extraction techniques may be used to identify and extract at least one keyword in the reference speech text according to the first feature information.
[0072] In an embodiment of the present application, since the content of the target text involves the first category of business, and the target text contains characteristic information of the first category of business, the characteristic information may correspond to a preset keyword set associated with the first category of business. Therefore, for the target text containing the first characteristic information, the preset keyword set can be obtained based on the first characteristic information, and further, the keywords in the reference speech text belonging to the preset keyword set can be determined as at least one keyword.
[0073] Exemplarily, the first characteristic information may include a large number of keywords (preset keyword set) related to the first type of business, and the keywords therein may be identified by comparing the first characteristic information with the reference speech text.
[0074] For example, taking the first category of business as insurance marketing business and the first characteristic information as characteristic information belonging to the insurance marketing business, at least one keyword in the reference speech text can be identified and extracted based on the characteristic information of the insurance marketing business, such as "illness", "huge medical expenses", "natural disasters and man-made disasters", "protection", etc.
[0075] Exemplarily, the above-mentioned keyword extraction technology (also called tag extraction technology) includes but is not limited to Textrank, LDA, topWords and other technologies.
[0076] In actual applications, a keyword can be a single word or a phrase, which can be determined based on actual usage requirements and is not limited to this.
[0077] For example, taking the target text as an insurance soft article, combined with the above Table 1, as shown in the following Table 2, after obtaining the reference speech text in the insurance soft article, the reference speech text can be identified to extract the insurance marketing keywords of the reference speech text.
[0078] Table 2
[0079]
[0080] S14: Storing first speech reference data in a target speech library corresponding to the first type of business, the first speech reference data including an association relationship between at least one keyword and a reference speech text.
[0081] In step S14, a correspondence between at least one keyword and a reference speech text can be established to achieve association between at least one keyword and the reference speech text, generate first speech reference data, and store the first speech reference data in a target speech library corresponding to the first type of business.
[0082] Taking marketing language as an example, the marketing language recognition model is used to obtain the marketing language in the article and the labels obtained through keyword extraction to form a pair of "label-language text" structures, and this result is stored in the marketing language library.
[0083] It can be understood that the target speech library can store a large number of reference speech sets for users to call according to their needs. This method can enrich the speech, avoiding the problem of insufficient speech in the prior art.
[0084] In combination with Table 1 and Table 2 above, as shown in Table 3 below, exemplary results of associating keywords with reference speech texts and storing them in the target speech library are shown.
[0085] Table 3
[0086]
[0087] Figure 3 The specific implementation flow diagram of the speech generation method is shown in FIG. Figure 3 As shown, the reference speech text in the target text can be obtained through the term recognition model, and the keywords in the reference speech text can be extracted, and then the keywords can be associated with the reference speech text to obtain the first speech reference data, and stored in the target speech library. In this way, reference speech texts (such as insurance marketing speech) can be automatically mined from existing public articles (such as insurance soft articles) to provide marketing speech materials for marketers.
[0088] Optionally, the above-mentioned target speech library can be stored in a speech generation device, or in any other possible electronic device (such as a mobile phone) for users (such as marketers) to use when they need to obtain the speech. For details, please refer to the description of the following method embodiment, which will not be repeated here.
[0089] It should be understood that keywords can be used as labels for reference speech texts, establish an association with the reference speech texts, and store them in the target speech library, so that users can call the target speech library and search for reference speech texts corresponding to the keywords or label information from the target speech library by entering keywords or label information when necessary, thereby improving the marketing success rate.
[0090] The speech generation method provided in the embodiment of the present application can obtain the reference speech text in the target text, extract the keywords in the reference speech text, and then associate the keywords with the reference speech text and store them in the target speech library, so that the reference speech text (such as insurance marketing speech) can be automatically mined from existing public articles (such as insurance soft articles) to provide marketing speech materials for marketing personnel. This method can enrich the speech and avoid the problem of insufficient marketing speech in the prior art.
[0091] See also Figure 4 , Figure 4 This is a flowchart of a method for generating speech provided by another embodiment of the present application. Figure 2 Corresponding to the embodiment, the speech generation method provided in this embodiment further includes step S15 before step S12A. The details are as follows:
[0092] S15: Using text content containing target type speech as training samples, deep learning is used for training to obtain a speech recognition model.
[0093] For example, the following is combined Figure 5 , taking insurance soft articles as an example, the model training process is described.
[0094] (1) Segmentation processing: After obtaining the insurance soft article (i.e., training sample), the insurance soft article can be split into multiple paragraphs.
[0095] Among them, the insurance soft article can be split into multiple paragraphs by splitting by line breaks.
[0096] (2) Marking the marketing language paragraphs: Marking the marketing language paragraphs in multiple paragraphs (e.g., manual marking).
[0097] Exemplarily, some paragraphs are manually annotated as marketing talk, for example, the marking Y represents whether it is marketing talk, Y=0 represents not marketing talk, and Y=1 represents marketing talk.
[0098] (3) Feature extraction: extracting the semantic features of the speech paragraph.
[0099] Among them, methods for extracting semantic features include but are not limited to word2vec, bert, bag of word, etc.
[0100] It should be noted that the semantic features of the target paragraph include the basic features of the text, such as the length of the text, whether it contains names of people / places / time, features of adjacent paragraphs, etc.
[0101] For example, the semantic features of the paragraph are extracted as X, such as paragraph length, whether it contains names of people / places / time, features of adjacent paragraphs, etc. Therefore, X is a multidimensional feature, including dimensions such as X1, X2, and X3. For example, X1=23 means that the paragraph length is 23, X2=1 means that the paragraph does not contain names of people, and X2=0 means that the paragraph contains names of people.
[0102] (4) The results of (2) and (3) above are used to train the model and obtain a speech recognition model.
[0103] For example, the speech recognition model is a marketing speech recognition model, which is used to recognize paragraphs belonging to marketing speech from multiple paragraphs. Furthermore, the marketing speech recognition model can be used to recognize marketing speech from more paragraphs to be recognized.
[0104] For example, the results of (2) and (3) above can be used to form a training set (X, Y); a specific training model (such as the xgboost model) is used for training to obtain a speech recognition model, where X is the input of the model and Y is the training target. This model is used to classify more unclassified speech.
[0105] It should be noted that the embodiment of the present application may not limit the execution order of the above steps S11 and S15. That is, the embodiment of the present application may execute S11 first and then S15; it may also execute S15 first and then S11; it may also execute S11 and S15 at the same time. It is understandable that the above Figure 3 The example of executing S15 first and then executing S11 is used for illustration.
[0106] In this way, the speech recognition model for the first category of business is obtained through model training, so that the reference speech text for the first category of business can be accurately and effectively obtained through the speech recognition model for the first category of business.
[0107] The above describes possible implementations of the speech generation method provided in the embodiment of the present application. The following describes specific implementations of the target speech library obtained by the above speech generation method in practical applications. Figure 6 , Figure 6 This is a flowchart of a method for generating speech provided by another embodiment of the present application. Figure 1 Corresponding to the embodiment, the speech generation method provided in this embodiment further includes steps S16 to S19 after step S14. The details are as follows:
[0108] S16: Obtain the first keyword input by the user;
[0109] Optionally, the keyword input by the user can be a single word or a phrase.
[0110] Optionally, the speech generation device can obtain the first keyword input by the user through text, or can obtain the first keyword input by the user through voice, or can obtain the first keyword input by the user in any other possible input method. The specific method can be determined based on actual usage requirements and is not limited in the embodiments of the present application.
[0111] It should be noted that the above-mentioned users may be marketing personnel (such as insurance agents), or any other personnel who need to obtain relevant marketing scripts. The specific number may be determined based on actual usage requirements, and the embodiments of the present application are not limited thereto.
[0112] S17: searching, based on the first keyword, from the target speech library for a second keyword whose matching degree with the first keyword is greater than a preset matching degree;
[0113] S18: Obtain reference speech text associated with the second keyword in the target speech library;
[0114] Among them, the first keyword can be compared with the keyword stored in the target speech library. If the matching degree between the first keyword and the second keyword stored in the target speech library is greater than the preset matching degree, the reference speech text associated with the second keyword in the target speech library is obtained.
[0115] S19: Display the reference script text.
[0116] Among them, the reference speech text can also be called recommended speech text.
[0117] In this embodiment, after the reference speech text is acquired, it can be directly displayed to the user for reference through the display screen of the electronic device.
[0118] Exemplarily, referring to Table 4, assuming that the user inputs the first keyword "woman", the associated reference speech text can be obtained accordingly through the keyword.
[0119] Table 4
[0120]
[0121] In this way, users input keywords, obtain corresponding guidance text, and adjust the content of the conversation with customers according to the guidance text, which is helpful for novice agents in marketing.
[0122] The method provided in the embodiment of the present application can recommend words of speech to manual customer service based on a pre-generated reference word set, and the recommended words of speech in the Internet finance scenario can meet the needs of manual customer service.
[0123] In all embodiments of the present application, a target speech library corresponding to the first type of business is obtained based on the above-mentioned speech generation method. Specifically, the target speech library corresponding to the first type of business is composed of multiple first speech reference data, wherein each first speech reference data is obtained by combining at least one keyword with a reference speech text. Uploading the target speech library corresponding to the first type of business to the blockchain can ensure its security and fairness and transparency to users. The user device can download the target speech library corresponding to the first type of business from the blockchain to verify whether the target speech library corresponding to the first type of business has been tampered with.
[0124] The blockchain referred to in this example is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, and encryption algorithm. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.
[0125] See also Figure 7 , Figure 7 is a structural block diagram of a speech generation device provided in an embodiment of the present application. In this embodiment, the speech generation device includes various units for executing Figures 1 to 6Each step in the embodiment corresponding to the speech generation device in the embodiment. For details, please refer to Figures 1 to 6 as well as Figures 1 to 6 For the convenience of explanation, only the parts related to this embodiment are shown. Figure 7 , the speech generation device 40 includes an acquisition unit 41, a processing unit 42 and a storage unit 43;
[0126] The acquisition unit 41 is used to acquire the target text of the first type of business;
[0127] The processing unit 42 is used to obtain a reference speech text from the target text obtained by the obtaining unit 41, where the reference speech text is a text in the target text containing first characteristic information, where the first characteristic information belongs to the characteristic information of the first type of business;
[0128] The processing unit 42 is further configured to identify at least one keyword in the reference speech text according to the first feature information;
[0129] The storage unit 43 is used to store the first speech reference data processed by the processing unit 42 into the target speech library corresponding to the above-mentioned first type of business, and the first speech reference data includes the association relationship between the above-mentioned at least one keyword and the reference speech text.
[0130] In a possible implementation, the processing unit 42 is specifically configured to input the target text into a speech recognition model, and output the reference speech text.
[0131] In one possible implementation, the processing unit 42 is also used to use the text content containing the target type of speech as a training sample and adopt a deep learning method to perform training before inputting the target text into the speech recognition model and outputting the reference speech text to obtain the above-mentioned speech recognition model.
[0132] In a possible implementation, the processing unit 42 is specifically configured to obtain a preset keyword set associated with the first type of business based on the first characteristic information; and determine a keyword in the reference speech text that belongs to the preset keyword set as the at least one keyword.
[0133] In a possible implementation, the acquisition unit 41 is also used to obtain the first keyword input by the user after the storage unit 43 stores the first speech reference data in the target speech library corresponding to the first type of business; the processing unit 42 is also used to call the first reference speech text from the target speech library based on the first keyword obtained by the acquisition unit 41.
[0134] In one possible implementation, the processing unit 42 is specifically used to search for a second keyword from the target speech library based on the first keyword, the second keyword having a matching degree greater than a preset matching degree with the first keyword; and obtain a first reference speech text associated with the second keyword in the target speech library.
[0135] Combination Figure 7 ,like Figure 8 As shown, the speech generation device 40 also includes a display unit 44, and the display unit 44 is used to display the first reference speech text after the processing unit 42 calls the first reference speech text from the target speech library according to the first keyword.
[0136] The speech generation device provided in the embodiment of the present application can identify the reference speech text in the target text through the speech recognition model, extract the keywords in the reference speech text, and then associate the keywords with the reference speech text and store them in the target speech library, so that the reference speech text (such as insurance marketing speech) can be automatically mined from existing public articles (such as insurance soft articles) to provide marketing speech materials for marketing personnel. This method can enrich the speech and avoid the problem of insufficient marketing speech in the prior art.
[0137] Fig. 9 is a structural block diagram of an electronic device provided by another embodiment of the present application. Fig. 9 As shown, the electronic device 70 of this embodiment includes: a processor 71, a memory 72, and a computer program 73 stored in the memory 72 and executable on the processor 71, such as a program of the speech generation method. When the processor 71 executes the computer program 73, the steps in each embodiment of the speech generation method described above are implemented, such as Figure 1 Alternatively, the processor 71 executes the computer program 73 to implement the above Figure 7 For details on the functions of each unit in the corresponding embodiment, please refer to Figure 7 The relevant descriptions in the corresponding embodiments are not repeated here.
[0138] Exemplarily, the computer program 73 may be divided into one or more units, which are stored in the memory 72 and executed by the processor 71 to implement the solution of the present application. The one or more units may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program 73 in the electronic device 70.
[0139] Those skilled in the art will understand that Fig. 9 The electronic device 70 is merely an example and does not limit the electronic device 70 , and may include more or less components than those shown in the figure, or may combine certain components, or may include different components.
[0140] The processor 71 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.
[0141] The memory 72 may be an internal storage unit of the electronic device 70, such as a hard disk or memory of the electronic device 70. The memory 72 may also be an external storage device of the electronic device 70, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 70. Further, the memory 72 may also include both an internal storage unit of the electronic device 70 and an external storage device. The memory 72 is used to store computer programs and other programs and data. The memory 72 may also be used to temporarily store data that has been output or is to be output.
[0142] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.
Claims
1. A method for generating speech, characterized in that: include: Obtaining a target text for the first type of business, where the target text is a published article; Acquire a reference speech text from the target text, the reference speech text being a text in the target text containing first characteristic information, the first characteristic information being characteristic information of the first type of service; According to the first feature information, identifying at least one keyword in the reference speech text; storing first speech reference data in a target speech library corresponding to the first type of business, wherein the first speech reference data includes an association relationship between the at least one keyword and the reference speech text; The obtaining of a reference speech text from the target text comprises: Input the target text into a speech recognition model, and output the reference speech text; Before inputting the target text into the speech recognition model and outputting the reference speech text, the method further includes: Using text content containing target type speech as training samples, training is performed using a deep learning method to obtain the speech recognition model; The training process includes: Segmentation processing: After obtaining the training sample, the training sample is divided into multiple segments; Marking paragraphs: Mark the paragraphs of marketing language in multiple paragraphs; Feature extraction: extracting semantic features of speech paragraphs; The annotated marketing speech paragraphs and the semantic features of the marketing speech paragraphs are trained using the xgboost model to obtain a speech recognition model.
2. The method according to claim 1, characterized in that: The step of identifying at least one keyword in the reference speech text according to the first feature information includes: Acquire a preset keyword set corresponding to the first type of service according to the first characteristic information; A keyword in the reference speech text that belongs to the preset keyword set is determined as the at least one keyword.
3. The method according to claim 1, characterized in that After storing the first speech technique reference data in a target speech technique library corresponding to the first type of service, the method further includes: Get the first keyword entered by the user; According to the first keyword, a first reference speech text is called from the target speech library.
4. The method according to claim 3, characterized in that The calling a first reference speech text from the target speech library according to the first keyword includes: According to the first keyword, searching the target speech library for a second keyword whose matching degree with the first keyword is greater than a preset matching degree; Obtain a first reference speech text associated with the second keyword in the target speech library.
5. The method according to claim 3 or 4, characterized in that: After calling the first reference speech text from the target speech library according to the first keyword, the method further includes: The first reference script text is displayed.
6. A speech generating device, characterized in that: It includes an acquisition unit, a processing unit and a storage unit; The acquisition unit is used to acquire a target text for the first type of business, wherein the target text is a published article; The processing unit is configured to obtain a reference speech text from the target text obtained by the obtaining unit, wherein the reference speech text is a text in the target text containing first characteristic information, and the first characteristic information belongs to characteristic information of the first type of service; The processing unit is further configured to identify at least one keyword in the reference speech text according to the first feature information; The storage unit is used to store the first speech reference data processed by the processing unit into a target speech library corresponding to the first type of business, wherein the first speech reference data includes an association relationship between the at least one keyword and the reference speech text; The processing unit is specifically used to input the target text into the speech recognition model, and output the reference speech text; The processing unit is further configured to, before inputting the target text into the speech recognition model and outputting the reference speech text, use the text content containing the target type speech as a training sample and perform training in a deep learning manner to obtain the speech recognition model; The training process includes: Segmentation processing: After obtaining the training sample, the training sample is divided into multiple segments; Marking paragraphs: Mark the paragraphs of marketing language in multiple paragraphs; Feature extraction: extracting semantic features of speech paragraphs; The annotated marketing speech paragraphs and the semantic features of the marketing speech paragraphs are trained using the xgboost model to obtain a speech recognition model.
7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.
8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.