Intelligent voice interaction system and method for disease and pest control consultation

Abstract

The invention discloses an intelligent voice interaction system and method for disease and pest control consultation, relates to the technical field of agricultural informatization, and solves the technical problem of low accuracy of disease and pest control consultation caused by lack of a real deep interaction process with a user in an existing disease and pest control consultation system. Questioning voice of the user is acquired; generating a recognition state according to the questioning voice; judging whether the identification state is a normal state or not; if yes, using a lookup table to convert the keyword to obtain a corresponding query keyword; querying in a current question text library based on the query keyword, and generating a normal answer text; if not, querying an abnormal answer text based on the recognition state; converting the answer text into a dialect answer text corresponding to the dialect attribute; the answer text comprises an abnormal answer text and a normal answer text; replying according to the answer voice corresponding to the dialect answer text; the interaction process between the system and the user is enhanced; and through multiple interactions, the accuracy of answers is improved.

Description

Technical Field

[0001] This application belongs to the technical field of agricultural informatization, and specifically relates to an intelligent voice interaction system and method for pest control consultation. Background Art

[0002] With the continuous deepening of the construction of the agricultural information service system, it has built a solid technical support for agricultural science and technology production. Agricultural technology consultation service, as a key way for farmers to obtain professional technical guidance in the process of agricultural production in a timely manner, is not only an important component of the digital agricultural service system construction, but also an indispensable important means in the agricultural technology promotion work.

[0003] Existing agricultural-related intelligent interaction systems have obvious limitations in pest and disease consultation; such systems usually only retrieve and match questions in the database based on the user's single query and answer; however, in actual applications, people engaged in agricultural or livestock husbandry work generally have a low educational level and face many difficulties in describing pest and disease problems. On the one hand, it is very difficult for them to accurately describe the key characteristics of pest and disease problems, especially when only through a single query, it is even more difficult to completely and accurately elaborate the relevant characteristics of pests and diseases. On the other hand, their descriptions of some pest and disease problems may be inaccurate. More critically, there are many types of agricultural pests and diseases with similar appearance characteristics but completely different treatment methods. Existing pest and disease consultation interaction systems often have difficulty giving accurate answers when faced with such similar situations. This limitation directly leads to a low accuracy of pest and disease consultation and cannot effectively meet the actual needs of agricultural practitioners.

[0004] Therefore, there is an urgent need to develop an intelligent voice interaction system and method for pest control consultation to improve the accuracy and practicality of pest and disease consultation. Summary of the Invention

[0005] This application aims to solve at least one of the technical problems existing in the prior art; for this purpose, this application proposes an intelligent voice interaction system and method for pest control consultation, which is used to solve the technical problem that the existing pest control consultation system lacks a real in-depth interaction process with users, resulting in a low accuracy of pest and disease consultation.

[0006] To achieve the above object, the first aspect of this application provides an intelligent voice interaction system for pest control consultation, including: a voice acquisition module, a voice recognition module, a knowledge query module, an answer module, and a database;

[0007] Voice acquisition module: Obtain the user's question voice through the user's Q&A device;

[0008] Speech recognition module: Obtain the voice of the question, input the voice of the question into the voice conversion and recognition model to obtain the dialect attribute, several keywords, and their corresponding confidence levels;

[0009] The knowledge query module includes a status recognition unit, an input conversion unit, a text query unit, and an answer conversion unit;

[0010] Status recognition unit: Obtain the recognition status of this voice based on the keywords and confidence levels; when the recognition status is the normal state, transmit the keywords to the input conversion unit; otherwise, query the abnormal answer text based on the recognition status and input it into the answer conversion unit;

[0011] Input conversion unit: Use the lookup table to convert the keywords to obtain the corresponding query keywords;

[0012] Text query unit: Query based on the query keywords in the current question text library and generate the normal answer text;

[0013] Answer conversion unit: Obtain the dialect attribute and the answer text, and convert the answer text into the dialect answer text corresponding to the dialect attribute; specifically, a text conversion model can be trained to convert the answer text; the answer text includes the abnormal answer text and the normal answer text;

[0014] Answer module: Reply according to the answer voice corresponding to the dialect answer text.

[0015] This application obtains the voice of the user's question; inputs the voice of the question into the voice conversion and recognition model to obtain the dialect attribute, several keywords, and their corresponding confidence levels; when the recognition status is the normal state, use the lookup table to convert the keywords to obtain the corresponding query keywords; query based on the query keywords in the current question text library and generate the normal answer text; otherwise, query the abnormal answer text based on the recognition status; convert the answer text into the dialect answer text corresponding to the dialect attribute; the answer text includes the abnormal answer text and the normal answer text; reply according to the answer voice corresponding to the dialect answer text; by setting multiple answer texts according to the actual question, the interaction process between the system and the user is strengthened; through multiple interactions, the accuracy of the answer is improved.

[0016] Preferably, the voice conversion and recognition model is obtained by training with an artificial intelligence model, including:

[0017] Obtain several voice data from the database, as well as their corresponding dialect attributes and several keywords; the several voice data include voices recorded in various dialects, the dialect attribute is the dialect type of the voice data; the keywords are the respective keywords after translating the voice data corresponding to the keywords into text; integrate the voice data, the voice data, and several keywords into training data and test data;

[0018] Train an artificial intelligence model using training data; test the trained artificial intelligence model using test data to obtain a speech conversion and recognition model with speech data as input and its corresponding dialect attributes, several keywords, and the confidence levels corresponding to the keywords as output; wherein, the artificial intelligence model includes a BP neural network model and an RBF neural network model.

[0019] Preferably, obtaining the recognition status of the current speech based on the keywords and confidence levels includes:

[0020] Obtain each keyword and its corresponding confidence level; sequentially determine whether the confidence level corresponding to each keyword is less than a set confidence level threshold; if so, mark the keyword as an abnormal keyword; if not, mark the keyword as a normal keyword;

[0021] Obtain the proportion of abnormal keywords; determine whether the proportion is greater than a set high abnormal proportion threshold; if so, set the recognition status to a high-noise state; if not, when the proportion is greater than a set low abnormal proportion threshold; set the recognition status to a low-accuracy state; otherwise, set the recognition status to a normal state.

[0022] Preferably, the abnormal response text includes a set high-noise response text and a low-accuracy response text; the high-noise response text corresponds to the high-noise state; the low-accuracy response text corresponds to the low-accuracy state.

[0023] Preferably, using a lookup table to convert the corresponding keyword to obtain a corresponding query keyword includes:

[0024] Obtain the scientific name corresponding to the keyword in the lookup table; mark the scientific name as the query keyword corresponding to the keyword; the lookup table is constructed from several scientific names and their corresponding common names.

[0025] Preferably, querying based on the query keyword in the current question text library and generating the normal response text includes:

[0026] Obtain several feature keywords of each prevention and control question in the current question text library; record the prevention and control question containing all query keywords as a matching prevention and control question;

[0027] Judge whether the matching prevention and control question is unique; if so, obtain the solution corresponding to the matching prevention and control question, and generate a question response text according to the solution; if not, obtain several feature keywords of each prevention and control question; remove the feature keywords that appear in all prevention and control questions, and mark the remaining feature keywords as the distinguishing keywords corresponding to the corresponding prevention and control questions; generate a supplementary inquiry text according to the distinguishing keywords; the normal response text includes the question response text and the supplementary inquiry text.

[0028] Preferably, generating the supplementary inquiry text according to the discriminative keywords includes:

[0029] Obtaining the discriminative keywords of each prevention and control problem and their corresponding keyword attributes;

[0030] Obtaining the number of discriminative keywords in the keyword attributes corresponding to each discriminative keyword; performing normalization processing on the number of each discriminative keyword to obtain a selection coefficient one corresponding to each keyword;

[0031] Obtaining the repetition rate of each discriminative keyword in each prevention and control problem, and using the repetition rate as the selection coefficient two of the keyword; selecting supplementary inquiry keywords from the discriminative keywords based on the selection coefficient one and the selection coefficient two; obtaining the inquiry text corresponding to the supplementary inquiry keywords, and using the inquiry text as the supplementary inquiry text; the inquiry text is a text for asking questions about the supplementary inquiry keywords.

[0032] Preferably, selecting the supplementary inquiry keywords from the discriminative keywords based on the selection coefficient one and the selection coefficient two includes:

[0033] Inputting the selection coefficient one and the selection coefficient two into a set comprehensive evaluation function to obtain a comprehensive score corresponding to the discriminative keyword; using the discriminative keyword with the maximum comprehensive score as the inquiry keyword.

[0034] Preferably, the discriminative keywords are also used to construct a problem text library for the next query keywords, including:

[0035] Obtaining the discriminative keywords of each problem text; using the discriminative keywords as the characteristic keywords of the corresponding prevention and control problems; integrating each prevention and control problem and its corresponding characteristic keywords into a problem text library, and using the problem text library to update the current problem text library.

[0036] Another aspect of the present application provides an intelligent voice interaction method for pest control consultation, including the following steps:

[0037] Step 1: Obtaining the user's question voice;

[0038] Step 2: Inputting the question voice into a voice conversion and recognition model to obtain the dialect attribute, several keywords and their corresponding confidence levels;

[0039] Step 3: Obtaining the recognition state of the current voice based on the keywords and the confidence levels; determining whether the recognition state is a normal state; if so, proceeding to Step 4; if not, querying the abnormal answer text based on the recognition state and proceeding to Step 6;

[0040] Step 4: Using a lookup table to convert the keywords to obtain corresponding query keywords;

[0041] Step 5: Query in the current question text library based on the query keywords and generate a normal answer text;

[0042] Step 6: Convert the answer text into a dialect answer text corresponding to the dialect attribute; the answer text includes an abnormal answer text and a normal answer text;

[0043] Step 7: Reply according to the answer voice corresponding to the dialect answer text.

[0044] Compared with the prior art, the beneficial effects of the present application are:

[0045] 1. In the present application, the question voice of the user is obtained; the question voice is input into the voice conversion and recognition model to obtain the dialect attribute, several keywords and their corresponding confidence levels; when the recognition status is the normal status, the lookup table is used to convert the keywords to obtain the corresponding query keywords; query in the current question text library based on the query keywords and generate a normal answer text; otherwise, query the abnormal answer text based on the recognition status; convert the answer text into a dialect answer text corresponding to the dialect attribute; the answer text includes an abnormal answer text and a normal answer text; reply according to the answer voice corresponding to the dialect answer text; by setting multiple answer texts according to the actual inquiry, the interaction process between the system and the user is strengthened; through multiple interactions, the accuracy of the answer is improved.

[0046] 2. In each inquiry of one round of inquiry in this embodiment, the question text library will be updated step by step to narrow down the question text library; which is convenient for reducing the calculation amount. Description of the Drawings

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to these drawings.

[0048] Figure 1 It is a schematic diagram of the module connection principle of the intelligent voice interaction system in the present application;

[0049] Figure 2 It is a schematic flowchart of the intelligent voice interaction method in the present application. Detailed Embodiments

[0050] The technical solution of the present application will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts belong to the scope of protection of the present application.

[0051] Please refer to Figure 1 , a first aspect of the present application provides an intelligent voice interaction system for pest control consultation, including: a voice acquisition module, a voice recognition module, a knowledge query module, an answer module, and a database;

[0052] Voice acquisition module: Obtain the user's question voice through the user's Q&A device; the question voice is the voice of the user's inquiry input through the voice acquisition module of the system; the Q&A device includes a microphone and a speaker;

[0053] Voice recognition module: Obtain the question voice, input the question voice into the voice conversion recognition model to obtain the dialect attribute, several keywords, and their corresponding confidence levels; the voice attribute is the dialect type of the recognized question voice; the keyword is the recognized keyword of the question voice;

[0054] The knowledge query module includes a status recognition unit, an input conversion unit, a text query unit, and an answer conversion unit;

[0055] Status recognition unit: Obtain the recognition status of this voice based on the keyword and the confidence level; the recognition status represents the credibility of this voice recognition; includes a normal state, a high-noise state, and a low-accuracy state; the high-noise state means that the recognition credibility of the question voice is low due to large noise; the low-accuracy state means that the recognition credibility of some keywords of the question voice is low due to reasons such as noise; when the recognition status is the normal state, transmit the keyword to the input conversion unit; otherwise, query the abnormal answer text based on the recognition status and input it into the answer conversion unit; the abnormal answer text includes the set high-noise answer text and low-accuracy answer text; the high-noise answer text corresponds to the high-noise state; the low-accuracy answer text corresponds to the low-accuracy state; in this embodiment, the content of the high-noise answer text is "Can you repeat the question just now loudly?" and the content of the low-accuracy answer text is "Can you repeat the question just now?"

[0056] Input conversion unit: Use a lookup table to convert the keyword to obtain the corresponding query keyword; the lookup table is a table established in advance consisting of various scientific names related to pest control and their common names in various dialects; the query keyword is the scientific name corresponding to the keyword in the lookup table;

[0057] Text query unit: Query in the current question text library based on the query keywords and generate a normal answer text; the normal answer text includes a question answer text and a supplementary inquiry text; the question answer text is the accurate answer already found for the above-mentioned question voice; the supplementary inquiry text is a supplementary inquiry to the user generated because the above-mentioned question voice cannot find an accurate answer.

[0058] Answer conversion unit: Obtain the dialect attribute and the answer text, and convert the answer text into a dialect answer text corresponding to the dialect attribute; specifically, a text conversion model can be trained to convert the answer text; the answer text includes an abnormal answer text and a normal answer text.

[0059] Answer module: Reply according to the answer voice corresponding to the dialect answer text.

[0060] In this embodiment, by obtaining the user's question voice; inputting the question voice into the voice conversion and recognition model to obtain the dialect attribute, several keywords and their corresponding confidence levels; when the recognition state is the normal state, use the lookup table to convert the keywords to obtain the corresponding query keywords; query in the current question text library based on the query keywords and generate a normal answer text; otherwise, query the abnormal answer text based on the recognition state; convert the answer text into a dialect answer text corresponding to the dialect attribute; the answer text includes an abnormal answer text and a normal answer text; reply according to the answer voice corresponding to the dialect answer text; by setting multiple answer texts according to the actual inquiry, the interaction process between the system and the user is strengthened; through multiple interactions, the accuracy of the answer is improved.

[0061] The voice conversion and recognition model is trained through an artificial intelligence model, including: obtaining several voice data from the database, as well as their corresponding dialect attributes and several keywords; the several voice data include voices recorded in various dialects, and the dialect attribute is the dialect type of the voice data; the keywords are the respective keywords after the voice data corresponding to the keywords is translated into text; integrating the voice data, the voice data and several keywords into training data and test data.

[0062] Use the training data to train the artificial intelligence model; use the test data to test the trained artificial intelligence model to obtain a voice conversion and recognition model with the input being voice data and the output being its corresponding dialect attribute, several keywords and the confidence levels corresponding to the keywords; among them, the artificial intelligence model includes a BP neural network model and an RBF neural network model; the specific training process is similar to the training of the existing artificial intelligence model and will not be elaborated here.

[0063] Obtain the recognition status of the current speech based on keywords and confidence levels, including: obtaining each keyword and its corresponding confidence level; sequentially determining whether the confidence level corresponding to each keyword is less than a set confidence threshold; if so, mark the keyword as an abnormal keyword; otherwise, mark the keyword as a normal keyword; a confidence threshold of a keyword less than the set confidence threshold indicates that the accuracy of the keyword recognition is relatively low; it may be caused by noise or other reasons; the speech needs to be re-recorded.

[0064] Obtain the proportion of abnormal keywords; determine whether the proportion is greater than a set high abnormal proportion threshold; if so, set the recognition status to a high-noise state; otherwise, when the proportion is greater than a set low abnormal proportion threshold; set the recognition status to a low-accuracy state; otherwise, set the recognition status to a normal state.

[0065] In this embodiment, by setting a high-noise state and a low-accuracy state, the accuracy of the text recognized from the speech is ensured as much as possible during the speech recognition process; thereby ensuring the accuracy of subsequent question queries.

[0066] Use a lookup table to convert the corresponding keyword to obtain a corresponding query keyword, including: obtaining the scientific name corresponding to the keyword in the lookup table; marking the scientific name as the query keyword corresponding to the keyword; the lookup table is constructed from several scientific names and their corresponding common names; specifically, obtain several scientific names and their corresponding common names, and the common names include the various dialectal names of the physical objects corresponding to the scientific names; integrate a scientific name and its corresponding several common names into a group of retrieval keyword groups, and use the scientific name as the query keyword; sequentially obtain the retrieval keyword groups and query keywords corresponding to each scientific name; integrate the retrieval keyword groups and query keywords corresponding to each scientific name into a lookup table; when looking up, search according to the keyword in the retrieval keyword group, and if there is the same keyword in a certain retrieval keyword group; then use the query keyword corresponding to the retrieval keyword group as the query keyword corresponding to the keyword.

[0067] Query based on the query keyword in the current question text library and generate a normal answer text, including: obtaining several feature keywords of each prevention and control question in the current question text library; recording the prevention and control question containing all query keywords as a matching prevention and control question; the prevention and control question is a question related to pests and diseases, and one type of pest and disease corresponds to one prevention and control question, and the feature keywords are the keywords related to the prevention and control question and the pests, including the size, color, spots of the pests, and the problem phenomena caused to plants or livestock; such as the shape of the eaten leaves; the size and color of the spots on the bitten parts of livestock, etc.; the matching prevention and control question is the prevention and control question screened to contain all the query keywords recognized from the speech.

[0068] Determine whether the matching prevention and control problem is unique; if so, obtain the solution corresponding to the matching prevention and control problem, and generate question answering text according to the solution; if not, obtain several characteristic keywords of each prevention and control problem; remove the characteristic keywords that appear in all prevention and control problems, and mark the remaining characteristic keywords as the distinguishing keywords corresponding to the corresponding prevention and control problems; generate supplementary inquiry text according to the distinguishing keywords; the normal answering text includes the question answering text and the supplementary inquiry text;

[0069] When the current inquiry can obtain a unique prevention and control problem, generate the solution corresponding to the corresponding prevention and control problem as the question answering text; when the current inquiry cannot obtain a unique prevention and control problem, it is necessary to generate supplementary inquiry text according to the differences of the remaining matching prevention and control problems; the question answering text is the text used to answer the user's inquiry question; the supplementary inquiry text is the text used to inquire the user.

[0070] In this embodiment, by removing the already matched query keywords, the data volume during subsequent keyword matching is reduced.

[0071] Generating supplementary inquiry text according to the distinguishing keywords includes: obtaining the distinguishing keywords of each prevention and control problem and their corresponding keyword attributes; the keyword attribute is the common attribute of a certain type of keyword; for example, the keyword attribute is the spot color; its corresponding distinguishing keywords include that the spot color is brown, white, black, yellow, etc.

[0072] Obtain the number of distinguishing keywords in the keyword attributes corresponding to each distinguishing keyword; perform normalization processing on the number of each distinguishing keyword to obtain the selection coefficient one corresponding to each keyword; the larger the selection coefficient one, the more screening the user's answer will be when asking questions about this keyword.

[0073] Obtain the repetition rate of each distinguishing keyword in each prevention and control problem, and use the repetition rate as the selection coefficient two of the keyword; the larger the selection coefficient two, the more serious the repetition of this keyword in each prevention and control problem; it is almost a common feature of each prevention and control problem, and asking questions about the common feature will not have a high screening rate for the answers; select supplementary inquiry keywords from the distinguishing keywords based on the selection coefficient one and the selection coefficient two; obtain the inquiry text corresponding to the supplementary inquiry keywords, and use the inquiry text as the supplementary inquiry text; the inquiry text is the text preset for asking questions about the supplementary inquiry keywords; such as the size of the pest, whether there are spots on the pest, the color of the spots on the pest, the shape of the spots on the pest, and so on.

[0074] Selecting the supplementary inquiry keywords from the distinguishing keywords based on the selection coefficient one and the selection coefficient two includes:

[0075] Input the selection coefficient one and the selection coefficient two into the set comprehensive evaluation function to obtain the comprehensive score corresponding to the distinguishing keyword; use the distinguishing keyword with the maximum comprehensive score as the query keyword.

[0076] The comprehensive evaluation function in this embodiment is: ZP = α1×XQ1 - α2×XQ2; where ZP is the comprehensive score; α1 and α2 are the set proportionality coefficients; XQ1 is the selection coefficient one; XQ2 is the selection coefficient two; it can be understood that the selection coefficient one and the selection coefficient two are on the same scale; the larger the comprehensive score, the stronger the screening ability for the distinguishing keyword obtained by querying.

[0077] This embodiment sets reasonable query texts according to the properties of the remaining distinguishing keywords; while ensuring accuracy, minimize the number of queries as much as possible.

[0078] The distinguishing keywords are also used to construct the question text library for the next query keyword, including:

[0079] Obtain the distinguishing keywords of each question text; use the distinguishing keywords as the characteristic keywords corresponding to the prevention and control questions; integrate each prevention and control question and its corresponding characteristic keywords into the question text library, and update the current question text library with the question text library.

[0080] In each query during one round of query in this embodiment, the question text library will be updated step by step to narrow down the question text library; this is convenient for reducing the calculation amount; it should be noted that when there is only one prevention and control question in the question text library, use the solution corresponding to the prevention and control question as the question answer text.

[0081] Please refer to Figure 2 , another aspect of this application provides an intelligent voice interaction method for pest control consultation, including the following steps:

[0082] Step 1: Obtain the user's question voice;

[0083] Step 2: Input the question voice into the voice conversion and recognition model to obtain the dialect attribute, several keywords and their corresponding confidence levels;

[0084] Step 3: Obtain the recognition status of this voice based on the keywords and confidence levels; determine whether the recognition status is a normal status; if yes, go to Step 4; if not, query the abnormal answer text based on the recognition status and enter Step 6;

[0085] Step 4: Use the lookup table to convert the keywords to obtain the corresponding query keywords;

[0086] Step 5: Query based on the query keywords in the current question text library and generate a normal answer text;

[0087] Step 6: Convert the response text into a dialect response text corresponding to the dialect attribute; the response text includes an abnormal response text and a normal response text;

[0088] Step 7: Reply according to the response voice corresponding to the dialect response text.

[0089] Some data in the above formula is calculated by removing the dimension and taking its numerical value. The formula is obtained by software simulation of a large amount of collected data to get a formula closest to the actual situation; the preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

[0090] The working principle of this application:

[0091] This application obtains the user's question voice; inputs the question voice into a voice conversion and recognition model to obtain the dialect attribute, several keywords and their corresponding confidence levels; when the recognition status is normal, uses a lookup table to convert the keywords to obtain the corresponding query keywords; queries in the current question text library based on the query keywords and generates a normal response text; otherwise, queries the abnormal response text based on the recognition status; converts the response text into a dialect response text corresponding to the dialect attribute; the response text includes an abnormal response text and a normal response text; replies according to the response voice corresponding to the dialect response text; by setting multiple response texts according to the actual inquiry, the interaction process between the system and the user is strengthened; through multiple interactions, the accuracy of the answer is improved.

[0092] The above embodiments are only used to illustrate the technical method of this application and not to limit it. Although this application has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of this application can be modified or equivalently replaced without departing from the spirit and scope of the technical method of this application.

Claims

1. An intelligent voice interaction system for pest control consultation, characterized in that Speech recognition module: Input the question voice into the speech conversion and recognition model to obtain the dialect attribute, several keywords and their corresponding confidence levels; Status recognition unit: Obtain the recognition status of this speech based on the keywords and confidence levels; when the recognition status is the normal state, transmit the keywords to the input conversion unit; otherwise, query the abnormal answer text based on the recognition status and input it into the answer conversion unit; Input conversion unit: Use the lookup table to convert the keywords to obtain the corresponding query keywords; Text query unit: Query based on the query keywords in the current question text library and generate the normal answer text; the normal answer text includes supplementary inquiry text and question answer text; Answer conversion unit: Convert the answer text into the dialect answer text corresponding to the dialect attribute; the answer text includes abnormal answer text and normal answer text; Answer module: Reply according to the answer voice corresponding to the dialect answer text.

2. The intelligent voice interaction system for pest control consultation according to claim 1, wherein The speech conversion and recognition model is obtained through training of an artificial intelligence model, including: Obtain several speech data from the database, as well as their corresponding dialect attributes and several keywords; integrate the speech data, speech data and several keywords into training data and test data; Use the training data to train the artificial intelligence model; use the test data to test the trained artificial intelligence model to obtain a speech conversion and recognition model with speech data as input and its corresponding dialect attribute, several keywords and the confidence levels corresponding to the keywords as output; among them, the artificial intelligence model includes a BP neural network model and an RBF neural network model.

3. An intelligent voice interaction system for pest control consultation according to claim 1, characterized in that, The obtaining the recognition status of this speech based on the keywords and confidence levels includes: Obtain each keyword and its corresponding confidence level; sequentially determine whether the confidence level corresponding to each keyword is less than the set confidence level threshold; if so, mark the keyword as an abnormal keyword; otherwise, mark the keyword as a normal keyword; Obtain the proportion of abnormal keywords; determine whether the proportion is greater than the set high abnormal proportion threshold; if so, set the recognition status to the high noise state; otherwise, when the proportion is greater than the set low abnormal proportion threshold; set the recognition status to the low accuracy state; otherwise, set the recognition status to the normal state.

4. The intelligent voice interaction system for pest control consultation according to claim 3, wherein, The abnormal answer text includes the set high noise answer text and low accuracy answer text; the high noise answer text corresponds to the high noise state; the low accuracy answer text corresponds to the low accuracy state.

5. An intelligent voice interaction system for pest control consultation according to claim 1, characterized in that, The using the lookup table to convert the corresponding keywords to obtain the corresponding query keywords includes: Obtain the corresponding scientific name of the keyword in the lookup table; mark the scientific name as the query keyword corresponding to the keyword; the lookup table is constructed from several scientific names and their corresponding common names.

6. The intelligent voice interaction system for pest control consultation according to claim 1, characterized in that, Query based on the query keywords in the current question text library and generate the normal answer text, including: Obtain several feature keywords of each pest control question in the current question text library; record the pest control question containing all query keywords as the matching pest control question; Determine whether the matching prevention and control problem is unique; if yes, obtain the solution corresponding to the matching prevention and control problem, and generate question answering text according to the solution; if not, obtain several characteristic keywords of each prevention and control problem; remove the characteristic keywords that appear in all prevention and control problems, and mark the remaining characteristic keywords as the distinguishing keywords corresponding to the corresponding prevention and control problems; generate supplementary inquiry text according to the distinguishing keywords; the normal answering text includes the question answering text and the supplementary inquiry text.

7. An intelligent voice interaction system for pest control consultation according to claim 6, characterized in that, Generating the supplementary inquiry text according to the distinguishing keywords includes: Obtain the distinguishing keywords of each prevention and control problem and their corresponding keyword attributes; Obtain the number of distinguishing keywords in the keyword attributes corresponding to each distinguishing keyword; perform normalization processing on the number of each distinguishing keyword to obtain the selection coefficient one corresponding to each keyword; Obtain the repetition rate of each distinguishing keyword in each prevention and control problem, and use the repetition rate as the selection coefficient two of the keyword; select supplementary inquiry keywords from the distinguishing keywords based on the selection coefficient one and the selection coefficient two; obtain the inquiry text corresponding to the supplementary inquiry keywords, and use the inquiry text as the supplementary inquiry text; the inquiry text is the text for asking questions about the supplementary inquiry keywords.

8. An intelligent voice interaction system for pest control consultation according to claim 7, characterized in that, Selecting the supplementary inquiry keywords from the distinguishing keywords based on the selection coefficient one and the selection coefficient two includes: Input the selection coefficient one and the selection coefficient two into a set comprehensive evaluation function to obtain the comprehensive score corresponding to the distinguishing keyword; use the distinguishing keyword with the largest comprehensive score as the inquiry keyword.

9. An intelligent voice interaction system for pest control consultation according to claim 6, characterized in that, The distinguishing keywords are also used to construct a question text library for the next query keyword, including: Obtain the distinguishing keywords of each question text; use the distinguishing keywords as the characteristic keywords of the corresponding prevention and control problem; integrate each prevention and control problem and its corresponding characteristic keywords into a question text library, and use the question text library to update the current question text library.

10. An intelligent voice interaction method for pest control consultation, which is an application of intelligent voice interaction for pest control consultation according to any one of claims 1 to 9; characterized in that, Including the following steps: Step 1: Obtain the user's question voice; Step 2: Input the question voice into a voice conversion and recognition model to obtain the dialect attribute, several keywords and their corresponding confidence levels; Step 3: Obtain the recognition status of this voice based on the keywords and the confidence level; determine whether the recognition status is a normal status; If yes, go to Step 4; if not, query the abnormal answering text based on the recognition status and go to Step 6; Step 4: Use a lookup table to convert the keywords to obtain the corresponding query keywords; Step 5: Query in the current question text library based on the query keywords and generate normal answering text; Step 6: Convert the answering text into a dialect answering text corresponding to the dialect attribute; the answering text includes the abnormal answering text and the normal answering text; Step 7: Reply according to the answering voice corresponding to the dialect answering text.

Images