Intelligent question answering method and device, equipment, medium and vehicle

Abstract

The present disclosure relates to an intelligent question answering method, device, equipment, medium and vehicle. The intelligent question answering method comprises: obtaining a question sentence; performing intent recognition on the question sentence to obtain a real question intent of the question sentence; determining a target answer template corresponding to the real question intent from a plurality of preset answer templates; and combining the target answer template and target style information corresponding to a target commodity style to generate a standard answer corresponding to the question sentence, the target commodity style being a commodity style corresponding to a commodity involved in the question sentence. According to the embodiment of the present disclosure, the real question intent can be accurately determined, so that the standard answer of the question sentence can be accurately generated, and the user experience is improved.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence technology, and in particular to an intelligent question-answering method, apparatus, device, medium, and vehicle. Background Technology

[0002] In the era of big data, intelligent question-answering technology has been applied to many industries and fields, especially in some sales industries, where users can obtain answers to questions about products they are interested in through intelligent question-answering devices.

[0003] However, existing intelligent question-answering devices rely entirely on existing answer databases to retrieve a pre-set, fixed answer after obtaining the question's intent. However, the answers retrieved in this way are often inaccurate, which in turn affects the user experience. Summary of the Invention

[0004] To address the aforementioned technical problems, this disclosure provides an intelligent question-answering method, apparatus, device, medium, and vehicle.

[0005] Firstly, this disclosure provides an intelligent question-answering method, including:

[0006] Get the question statement;

[0007] Perform intent recognition on the question statement to obtain the true intent of the question statement;

[0008] Among multiple preset answer templates, determine the target answer template that corresponds to the true intent of the question;

[0009] The target answer template and the target product style information are combined to generate the standard answer to the question statement. The target product style is the product style corresponding to the product mentioned in the question statement.

[0010] Secondly, this disclosure provides an intelligent question-answering device, including:

[0011] The problem retrieval module is used to retrieve problem statements;

[0012] The intent recognition module is used to identify the intent of the question statement and obtain the true intent of the question statement;

[0013] The template determination module is used to determine the target answer template that corresponds to the true question intent from multiple preset answer templates;

[0014] The answer generation module is used to combine the target answer template and the target product style information to generate the standard answer to the question statement. The target product style is the product style corresponding to the product mentioned in the question statement.

[0015] Thirdly, this disclosure provides an intelligent question-answering device, including:

[0016] processor;

[0017] Memory, used to store executable instructions;

[0018] The processor is used to read executable instructions from memory and execute the executable instructions to implement the intelligent question-answering method of the first aspect.

[0019] Fourthly, this disclosure provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to implement the intelligent question-answering method of the first aspect.

[0020] Fifthly, this disclosure provides a vehicle including the intelligent question-and-answer device described above.

[0021] The technical solution provided in this disclosure has the following advantages compared with the prior art:

[0022] The intelligent question-answering method, apparatus, device, medium, and vehicle disclosed in this embodiment can acquire question statements, perform intent recognition on the question statements to obtain the true question intent, then determine the target answer template corresponding to the true question intent from multiple preset answer templates, and finally combine the target answer template with the target product style information to generate the standard answer corresponding to the question statement. The target product style is the product style corresponding to the product mentioned in the question statement. Therefore, the true question intent of the question statement can be accurately identified and the corresponding target answer template can be determined. Finally, the standard answer corresponding to the question statement is generated based on the target answer template and the target style information. This allows for the accurate generation of a standard answer matching the question statement by combining the true question intent and the target style information of the question statement, improving the accuracy of the generated answer and enhancing the user experience. Attached Figure Description

[0023] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0024] Figure 1 A flowchart illustrating an intelligent question-answering method provided in an embodiment of this disclosure;

[0025] Figure 2 A schematic diagram of an intent recognition model provided in an embodiment of this disclosure;

[0026] Figure 3A flowchart illustrating an intent determination method provided in an embodiment of this disclosure;

[0027] Figure 4 This is a schematic diagram of the structure of an intelligent question-answering device provided in an embodiment of the present disclosure;

[0028] Figure 5 This is a schematic diagram of the structure of an intelligent question-and-answer device provided in an embodiment of this disclosure. Detailed Implementation

[0029] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0030] It should be understood that the steps described in the method embodiments of this disclosure may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of this disclosure is not limited in this respect.

[0031] The term "comprising" and its variations as used herein are open-ended inclusions, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the description below.

[0032] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0033] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0034] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0035] This disclosure provides an intelligent question-answering method, apparatus, device, medium, and vehicle. The following will first combine... Figures 1 to 3The intelligent question-answering method provided in the embodiments of this disclosure will be described in detail.

[0036] Figure 1 A flowchart illustrating an intelligent question-answering method provided in an embodiment of this disclosure is shown.

[0037] In this embodiment of the disclosure, the intelligent question-answering method can be executed by an electronic device. The electronic device can be a device with intelligent question-answering functionality. Specifically, the electronic device can be, but is not limited to, mobile terminals such as mobile phones, in-vehicle devices, vehicle controllers, tablet computers, wearable devices, and smart home devices.

[0038] like Figure 1 As shown, the intelligent question-answering method may include the following steps.

[0039] S110, Obtain the problem statement.

[0040] In this embodiment of the disclosure, when a user wants to get the answer to a question, they can input the question into the electronic device, and the electronic device can obtain the corresponding question statement based on the user's input.

[0041] Optionally, the question statement can be a statement text that includes the question entered by the user.

[0042] Specifically, when a user wants to get the answer to a question, they can input the corresponding question into the electronic device by inputting voice, image, or text. The electronic device can receive the voice, image, and text information input by the user and obtain the corresponding question text through voice recognition, image recognition, and text recognition.

[0043] S120. Perform intent recognition on the question statement to obtain the true question intent of the question statement.

[0044] In this embodiment of the disclosure, after acquiring the question statement, the electronic device can perform intent recognition on the question statement to obtain the true question intent of the question statement.

[0045] Alternatively, intent recognition can be used to identify the intent in the question statement.

[0046] Alternatively, the true question intent can be the purpose that the question statement most intends to ask.

[0047] Specifically, after acquiring the question statement, the electronic device can identify the true question intent of the question statement, that is, perform intent recognition, thereby obtaining the true question intent of the question statement.

[0048] For example, taking the question "What is the price of phone A?" as an example, the electronic device can perform intent recognition on the question statement to obtain the corresponding true question intent, which can be asking about "price", that is, the question statement is asking about the price of "phone A"; taking the question "What is the fuel consumption of car B?" as an example, the electronic device can perform intent recognition on the question statement to obtain the corresponding true question intent, which can be asking about "fuel consumption", that is, the question statement is asking about the fuel consumption of "car B".

[0049] S130. Among multiple preset answer templates, determine the target answer template that corresponds to the true intent of the question.

[0050] In this embodiment of the disclosure, after obtaining the true question intent, the electronic device can search for the target answer template corresponding to the true question intent among multiple preset answer templates.

[0051] Optionally, the preset answer template can be a pre-defined answer template. For example, the preset answer template can be "Hello, the price of (product model) is xx yuan" or "Hello, the fuel consumption of (product model) is xx liters", etc., without limitation.

[0052] Optionally, the target answer template can be an answer template from the preset answer templates that corresponds to the actual question intent. For example, the actual question intent corresponding to the preset answer template "Hello, the price of (product style) is xx yuan" can be the intent of "selling price", "down payment", etc., which is not limited here.

[0053] Specifically, after obtaining the true question intent, the electronic device can search for the answer template corresponding to the true question intent from among a number of pre-set answer templates, i.e., the target answer template.

[0054] S140. Combine the target answer template and the target product style information to generate the standard answer corresponding to the question statement. The target product style is the product style corresponding to the product involved in the question statement.

[0055] In this embodiment of the disclosure, after determining the target answer template corresponding to the true question intent, the electronic device can combine the target answer template and the target style information corresponding to the target product style to generate the standard answer corresponding to the question statement.

[0056] Optionally, the target product style can be the product style corresponding to the product mentioned in the question statement.

[0057] Optionally, the goods mentioned in the question can be vehicles, mobile phones, computers, etc., without limitation.

[0058] Optionally, the target product style can be the product's design. For example, if the product is a vehicle, the target product style can be the vehicle's model number, year of manufacture, etc.; if the product is a mobile phone, the target product style can be the mobile phone's model number, year of manufacture, etc.

[0059] Optionally, the target style information can be the style information of the target product. For example, when the target product style is the model of a vehicle, the target style information can be the model information of the vehicle; when the target product style is the year of a vehicle, the target style information can be the year of the vehicle.

[0060] Optionally, the products involved in the question statement can be obtained by performing named entity recognition on the question statement.

[0061] For example, electronic devices can input a question statement into a pre-trained named entity recognition model, which then performs named entity recognition on the question statement and outputs the target entity, i.e., the corresponding product.

[0062] Specifically, after determining the target answer template and the target product style, the electronic device can combine the target answer template and the target product style information corresponding to the target style to generate the standard answer corresponding to the question statement.

[0063] For example, taking the question "What is the fuel consumption of car B?" as an example, the electronic device can first determine that the true intention of the question is to ask about "fuel consumption". Then, it can determine that the corresponding target answer template is "Hello, the fuel consumption of (product model) is xx liters". The electronic device can also determine the product model corresponding to the product mentioned in the question, such as "car B". At this time, the electronic device can combine the target answer template and the target answer template to generate the standard answer corresponding to the question "Hello, the fuel consumption of car B is 3 liters".

[0064] Therefore, in this embodiment, a question statement can be acquired and its intent identified to obtain the true intent of the question statement. Then, among multiple preset answer templates, a target answer template corresponding to the true intent of the question statement is determined. Finally, the target answer template and the target product style information corresponding to the target product style are combined to generate a standard answer corresponding to the question statement. Here, the target product style is the product style corresponding to the product involved in the question statement. Thus, the true intent of the question statement can be accurately identified and the corresponding target answer template can be determined. Finally, a standard answer corresponding to the question statement is generated based on the target answer template and the target product style information. This allows for the accurate generation of a standard answer that matches the question statement by combining the true intent of the question statement and the target product style information of the question statement, thereby improving the accuracy of the generated answer and enhancing the user experience.

[0065] Optionally, S120 may specifically include: performing intent recognition on the question statement to obtain the probability that the question statement belongs to each intent; and performing probability judgment based on the probability of each intent to determine the true question intent of the question statement.

[0066] In this embodiment of the disclosure, after acquiring the question statement, the electronic device can input the question statement into a pre-trained intent recognition model, so that the pre-trained intent recognition model can perform intent recognition on the question statement and obtain the real question intent output by the pre-trained intent recognition model.

[0067] Optionally, the pre-trained intent recognition model can be a pre-trained model used to identify the true question intent in a question statement.

[0068] Optionally, each intent can be one of several predefined intents.

[0069] Specifically, after acquiring the question statement, the electronic device can input the question statement into a pre-trained intent recognition model. The intent recognition model can receive the question statement and perform intent recognition processing, identify the probability of the question statement for each intent, and output the probability of each intent to which the question statement belongs to to the electronic device. The electronic device can receive and acquire the probability of each intent to which the question statement belongs.

[0070] Furthermore, after acquiring the probability of each intent, the electronic device can make a probability judgment based on the probability of each intent, thereby determining the true intent of the question statement.

[0071] Specifically, after acquiring the probability of each intent, the electronic device can perform a probability judgment, that is, determine the true intent of the question statement based on the probability of each intent. The specific implementation method is described in detail below.

[0072] Therefore, in this embodiment of the disclosure, the true intent of the question statement can be determined by a pre-trained intent recognition model, thereby accurately generating the standard answer to the question statement and improving the user experience.

[0073] The following is combined with Figure 2 A detailed explanation of the intent recognition model is provided.

[0074] Optionally, the intent recognition model may include a word vector sub-model, an encoding sub-model, and a multilayer perceptron model.

[0075] Figure 2 A schematic diagram of an intent recognition model provided by an embodiment of this disclosure is shown.

[0076] like Figure 2As shown, the intent recognition model may include a word vector sub-model 201, an encoding sub-model 202, and a multilayer perceptron model 203.

[0077] Optionally, the question statement is input into a pre-trained intent recognition model so that the intent recognition model can identify the intent of the question statement and obtain the probability that the question statement output by the intent recognition model belongs to each intent. This can specifically include: inputting the question statement into a word vector sub-model so that the word vector sub-model can perform vector mapping on the question statement to obtain word vectors output by the word vector sub-model; inputting the word vectors into an encoding sub-model so that the encoding sub-model can perform vector encoding on the word vectors to obtain fixed-length vectors output by the encoding sub-model; and inputting the fixed-length vectors into a multilayer perceptron model so that the multilayer perceptron model can perform probability calculation on the fixed-length vectors to obtain the probability that the question statement output by the multilayer perceptron model belongs to each intent.

[0078] In this embodiment of the disclosure, the electronic device can input a question statement into a word vector sub-model, thereby enabling the word vector sub-model to perform vector mapping on the question statement, thus obtaining the word vectors output by the word vector sub-model.

[0079] Optionally, the word vector sub-model can be a model that transforms text into corresponding word vectors.

[0080] Specifically, the electronic device can input the question statement into the word vector sub-model. The word vector sub-model can receive the text of the question statement, perform vector mapping on the question statement, map the question statement into the corresponding word vector, and output the word vector.

[0081] Continue to refer to Figure 2 Electronic devices can input the question statement "How much fuel does it consume", i.e. "W1, W2, W3, W4", into the word vector sub-model 201. The word vector sub-model can perform vector mapping (embedding) on ​​the question statement to obtain the corresponding word vectors.

[0082] Furthermore, after the word vector sub-model outputs word vectors, the word vectors can be input into the encoding sub-model, so that the encoding sub-model can perform vector encoding on the word vectors to obtain the fixed-length vectors output by the encoding sub-model.

[0083] Alternatively, the encoding sub-model can be a model that performs vector encoding on word vectors.

[0084] Optionally, the fixed-length vector can be a vector of fixed length.

[0085] Specifically, after the word vector sub-model outputs word vectors, the word vectors can be input into the encoding sub-model. The encoding sub-model can receive the word vectors and perform vector encoding on them, so that the length of the word vectors becomes a fixed length, that is, output a fixed-length vector.

[0086] Continue to refer to Figure 2 After the word vector sub-model 201 outputs the word vector, it is input into the encoding sub-model 202. The encoding sub-model 202 can perform vector encoding on the word vector, thereby outputting a fixed-length vector.

[0087] Furthermore, after the encoding sub-model outputs the fixed-length vector, the fixed-length vector can be input into the multilayer perceptron model, thereby enabling the multilayer perceptron model to perform probability calculations on the fixed-length vector and obtain the probability that the question statement output by the multilayer perceptron model belongs to each intent.

[0088] Alternatively, the multilayer perceptron model can be a model for calculating the probability of intent in a fixed vector.

[0089] Specifically, after the encoding sub-model outputs the fixed-length vector, the fixed-length vector can be input into the multilayer perceptron model. The multilayer perceptron model can receive the fixed-length vector and perform probability calculations on it, calculating the probability of each intent corresponding to the fixed-length vector, and thus outputting the probability of each intent.

[0090] Continue to refer to Figure 2 After the encoding sub-model 202 outputs the fixed vector, the fixed vector can be input into the multilayer perceptron model 203. The multilayer perceptron model 203 can perform probability calculation on the fixed vector, thereby outputting the probability (prob) of each intention.

[0091] Therefore, in this embodiment of the disclosure, the probability of each intent can be accurately calculated through word vector sub-model, encoding sub-model and multilayer perceptron model, thereby accurately determining the true question intent and accurately generating the standard answer to the question statement, thus improving the user experience.

[0092] The coding sub-model will be explained in detail below.

[0093] Optionally, the encoding sub-model may further include a bidirectional encoding layer, an attention layer, and a pooling layer.

[0094] Optionally, inputting word vectors into an encoding sub-model to perform vector encoding on the word vectors and obtain a fixed-length vector output by the encoding sub-model can specifically include: inputting word vectors into a bidirectional encoding layer to perform bidirectional vector encoding on the word vectors and obtain a bidirectional vector output by the bidirectional encoding layer; inputting the bidirectional vector into an attention layer to assign weights to the bidirectional vector and obtain a weight vector output by the attention layer; and inputting the weight vector into a pooling layer to perform fixed-length sampling on the weight vector and obtain a fixed-length vector output by the pooling layer.

[0095] In this embodiment of the disclosure, after the word vector sub-model outputs word vectors, the word vectors can be input into the bidirectional encoding layer in the encoding sub-model, so that the bidirectional encoding layer performs bidirectional vector encoding on the word vectors to obtain the bidirectional vectors output by the bidirectional encoding layer.

[0096] Alternatively, the bidirectional encoding layer can be a convolutional layer used for bidirectional vector encoding of word vectors.

[0097] Specifically, after the word vector sub-model outputs word vectors, it can input the word vectors into the bidirectional encoding layer in the encoding sub-model. The bidirectional encoding layer can encode the word vectors bidirectionally to obtain the bidirectional vectors corresponding to the word vectors, and then output the bidirectional vectors.

[0098] See also Figure 2 After outputting word vectors, the word vector sub-model 201 inputs the word vectors into the bidirectional encoding layer (bilstm) in the encoding sub-model 202. The bidirectional encoding layer can encode the word vectors bidirectionally to obtain the bidirectional vectors corresponding to the word vectors, such as "h1, h2, h3, h4".

[0099] Furthermore, after the bidirectional encoding layer outputs the bidirectional vector, it can input the bidirectional vector into the attention layer, so that the attention layer can assign weights to the bidirectional vector and obtain the weight vector output by the attention layer.

[0100] Alternatively, the attention layer can be a convolutional layer used to assign weights to bidirectional vectors.

[0101] Specifically, after the bidirectional encoding layer outputs the bidirectional vector, it can input the bidirectional vector into the attention layer. The attention layer can assign weights to the bidirectional vector, that is, assign different weights to each vector in the bidirectional vector, thereby obtaining the corresponding weight vector.

[0102] See also Figure 2 After the bidirectional encoding layer outputs bidirectional vectors, it can input the bidirectional vectors into the attention layer (self-attention). The attention layer can assign weights to the bidirectional vectors "h1, h2, h3, h4" to obtain the corresponding weight vectors, such as "m1, m2, m3, m4".

[0103] Furthermore, after the attention layer outputs the weight vector, it can input the weight vector into the pooling layer, so that the pooling layer performs fixed-length sampling on the weight vector to obtain the fixed-length vector output by the pooling layer.

[0104] Alternatively, the pooling layer can be a convolutional layer used for fixed-length sampling of the weight vector.

[0105] Specifically, after the attention layer outputs the weight vector, it can input the weight vector into the pooling layer. The pooling layer can receive the weight vector and perform fixed-length sampling on the weight vector to obtain a vector of fixed length, i.e., a fixed-length vector.

[0106] See also Figure 2 After the attention layer outputs the weight vector, it can input the weight vector into the pooling layer. The pooling layer can perform fixed-length sampling on the weight vector to obtain a fixed-length vector.

[0107] Therefore, in this embodiment of the disclosure, a fixed-length vector can be accurately calculated through a bidirectional encoding layer, an attention layer, and a pooling layer, thereby accurately determining the true intent of the question and generating the standard answer to the question statement, thus improving the user experience.

[0108] In some embodiments of this disclosure, after obtaining the probabilities of each intent, the electronic device can also determine the true intent of the question statement by the magnitude of the probabilities of each probability, which will be described in detail below.

[0109] Optionally, determining the true question intent of a question statement based on the probability of each intent may specifically include: sorting the probabilities of each intent in descending order; determining whether the probability of the first intent, which is ranked first, is greater than a first preset threshold; if the probability of the first intent is less than or equal to the first preset threshold, determining whether the sum of the probabilities of the first intent and the second intent, which is ranked second, is greater than a second preset threshold; if the sum of the probabilities of the first intent and the second intent is greater than the second preset threshold, then the first intent and the second intent are taken as the true question intent.

[0110] In this embodiment of the disclosure, after obtaining the probability of each intention, the electronic device can first sort the probabilities of each intention in descending order.

[0111] Alternatively, descending sort can be sorted from largest to smallest probability.

[0112] Specifically, after acquiring the probability of each intention, the electronic device can sort the intentions in descending order of probability.

[0113] Furthermore, after sorting the probabilities of each intent in descending order, the electronic device can determine whether the probability of the first intent, which is ranked first, is greater than a first preset threshold.

[0114] Optionally, the first preset threshold can be a pre-set probability value. For example, 0.80, 0.81, etc., are not limited here.

[0115] Specifically, after sorting the probabilities of each intent in descending order, the electronic device can obtain the probability of the first intent ranked first, that is, obtain the probability of the intent with the highest probability, and determine whether the probability of the first intent is greater than a first preset threshold.

[0116] Furthermore, if the probability of the first intention is less than or equal to the first preset threshold, it is further determined whether the sum of the probability of the first intention and the probability of the second intention ranked second is greater than the second preset threshold.

[0117] Optionally, the second preset threshold can be a pre-set probability value. For example, 0.90, 0.91, etc., are not limited here.

[0118] Specifically, after determining whether the probability of the first intention is greater than the first preset threshold, if the electronic device determines that the probability of the first intention is less than or equal to the first preset threshold, it continues to obtain the probability of the second intention ranked second, and sums the probability of the first intention and the probability of the second intention, and determines whether the sum of probabilities is greater than the second preset threshold.

[0119] Furthermore, after the electronic device judges the probability of the first intention and the sum of the probabilities of the second intention, if the sum of the probabilities of the first intention and the second intention is greater than a second preset threshold, then the first intention and the second intention are regarded as the real question intentions.

[0120] Specifically, if an electronic device determines that the sum of the probabilities of the first intent and the second intent is greater than a second preset threshold, then the first intent and the second intent can be used together as the true question intent of the question statement.

[0121] Therefore, in this embodiment of the disclosure, the electronic device can determine the true question intent of the question statement based on the probability of each intent, thereby accurately generating the standard answer to the question statement and improving the user experience.

[0122] In other embodiments of this disclosure, after the electronic device determines whether the probability of the first intent ranked first is greater than a first preset threshold, the intelligent question answering method may further include: if the probability of the first intent is greater than the first preset threshold, the first intent is taken as the real question intent.

[0123] In this embodiment of the disclosure, after the electronic device judges the probability of the first intent ranked first, if the probability of the first intent is greater than a first preset threshold, the first intent can be regarded as the real question intent.

[0124] Therefore, in this embodiment of the disclosure, the electronic device can determine the true question intent of the question statement based on the probability of each intent, thereby accurately generating the standard answer to the question statement and improving the user experience.

[0125] In some further embodiments of this disclosure, after the electronic device determines whether the sum of the probability of the first intention and the probability of the second intention (ranked second) is greater than a second preset threshold, the intelligent question-answering method may further include: generating a general answer corresponding to the target product style if the sum of the probabilities of the first intention and the second intention is less than or equal to the second preset threshold.

[0126] In this embodiment of the disclosure, after the electronic device judges the probability of the first intention and the sum of the probabilities of the second intention, if the probability of the first intention and the sum of the probabilities of the second intention are less than or equal to a second preset threshold, the electronic device can generate a general answer corresponding to the target product style.

[0127] Optionally, the general answer can be a pre-set answer corresponding to the target product style. For example, if the target product style is "B car", the general answer could be "Excuse me, what question do you have about B car?"; if the target product style is "A mobile phone", the general answer could be "The relevant features of A mobile phone are as follows, please select", and so on. There are no restrictions here.

[0128] Therefore, in this embodiment of the disclosure, the electronic device can determine the true question intent of the question statement based on the probability of each intent, thereby accurately generating the standard answer to the question statement and improving the user experience.

[0129] Optionally, after S130, the intelligent question answering method may further include: identifying the target entity of the question statement to obtain the target entity corresponding to the question statement; and determining the target product style corresponding to the product to which the target entity belongs.

[0130] In this embodiment of the disclosure, after the electronic device obtains the question statement, it can perform named entity recognition on the question statement to obtain the target entity contained in the question statement.

[0131] Optionally, named entity recognition can be used to identify entities with specific meanings in the question statement.

[0132] Optionally, the target entity can be a product entity. For example, the target entity can be a vehicle, mobile phone, computer, etc., without limitation.

[0133] Specifically, after acquiring the question statement, the electronic device can identify the product entity in the question statement, that is, perform named entity recognition, thereby obtaining the product entity contained in the question statement, that is, the target entity.

[0134] For example, if the question is "What is the price of a 2010 model A car?", the electronic device can perform named entity recognition on the product entity "A car" in the question to obtain the target entity "A car". Similarly, if the question is "What is the price of a B mobile phone?", the electronic device can perform named entity recognition on the product entity "B mobile phone" in the question to obtain the target entity "B mobile phone".

[0135] Furthermore, after obtaining the target entity contained in the question statement, the electronic device can determine the target product style corresponding to the product to which the target entity belongs.

[0136] Optionally, the target product style can be the style of the physical product. For example, when the physical product is a vehicle, the target product style can be the vehicle model, the year of manufacture, etc.; when the physical product is a mobile phone, the target product style can be the mobile phone model, the year of manufacture, etc.

[0137] Specifically, after obtaining the target entity contained in the question statement, the electronic device can determine the product to which it belongs and the corresponding target product style based on the target entity.

[0138] For example, if the question is "What is the price of a 2010 model A car?", the electronic device can determine that the target entity is "A car", and can first determine that the product to which the target entity belongs is a vehicle. Then, it can determine that the target product model of the vehicle is "vehicle model A". Similarly, if the question is "What is the price of a B mobile phone?", the electronic device can determine that the target entity is "B mobile phone", and can first determine that the product to which the target entity belongs is a mobile phone. Then, it can determine that the target product model of the mobile phone is "mobile phone model B".

[0139] Therefore, in this embodiment of the disclosure, the target style information corresponding to the target entity in the question statement can be determined, thereby combining the real question intent and the target style information of the question statement to accurately generate a standard answer that matches the question statement, improving the accuracy of the generated answer and enhancing the user experience.

[0140] Figure 3 A flowchart illustrating an intent determination method provided in an embodiment of this disclosure is shown.

[0141] like Figure 3 As shown, the method for determining intent may include:

[0142] S310. Sort the probabilities of each intention in descending order.

[0143] In this embodiment of the disclosure, after obtaining the probability of each intention, the electronic device can first sort the probabilities of each intention in descending order.

[0144] Specifically, after acquiring the probability of each intention, the electronic device can sort the intentions in descending order of probability.

[0145] S320. Determine whether the probability of the first intention ranked first is greater than the first preset threshold.

[0146] In this embodiment of the disclosure, after sorting the probabilities of each intent in descending order, the electronic device can determine whether the probability of the first intent, which is ranked first, is greater than a first preset threshold. If the probability of the first intent is greater than the first preset threshold, step S330 is executed; if the probability of the first intent is less than or equal to the first preset threshold, step S340 is executed.

[0147] S330. Treat the primary intent as the true problem intent.

[0148] In this embodiment of the disclosure, if the sum of the probabilities of the first intention and the second intention is greater than a second preset threshold, then the first intention and the second intention are regarded as the real problem intentions.

[0149] Specifically, if an electronic device determines that the sum of the probabilities of the first intent and the second intent is greater than a second preset threshold, then the first intent and the second intent can be used together as the true question intent of the question statement.

[0150] S340. Determine whether the sum of the probability of the first intention and the probability of the second intention ranked second is greater than a second preset threshold.

[0151] In this embodiment of the disclosure, if the probability of the first intention is less than or equal to the first preset threshold, it is further determined whether the sum of the probability of the first intention and the probability of the second intention ranked second is greater than the second preset threshold.

[0152] Specifically, after determining whether the probability of the first intention is greater than a first preset threshold, if the probability of the first intention is less than or equal to the first preset threshold, the electronic device continues to obtain the probability of the second intention, which is ranked second, and sums the probabilities of the first intention and the second intention, and determines whether the sum of the probabilities is greater than the second preset threshold. If the sum of the probabilities of the first intention and the second intention is greater than the second preset threshold, S360 is executed; if the sum of the probabilities of the first intention and the second intention is less than or equal to the second preset threshold, S350 is executed.

[0153] S350, Generate a general answer corresponding to the target product style.

[0154] In this embodiment of the disclosure, after the electronic device judges the probability of the first intention and the sum of the probabilities of the second intention, if the probability of the first intention and the sum of the probabilities of the second intention are less than or equal to a second preset threshold, the electronic device can generate a general answer corresponding to the target product style.

[0155] Optionally, the general answer can be a pre-set answer corresponding to the target product style. For example, if the target product style is "B car", the general answer could be "Excuse me, what question do you have about B car?"; if the target product style is "A mobile phone", the general answer could be "The relevant features of A mobile phone are as follows, please select", and so on. There are no restrictions here.

[0156] S360. Treat the first and second intentions as the real problem intentions.

[0157] In this embodiment of the disclosure, after the electronic device judges the probability of the first intention and the sum of the probabilities of the second intention, if the sum of the probabilities of the first intention and the second intention is greater than a second preset threshold, then the first intention and the second intention are regarded as the real question intention.

[0158] Specifically, if an electronic device determines that the sum of the probabilities of the first intent and the second intent is greater than a second preset threshold, then the first intent and the second intent can be used together as the true question intent of the question statement.

[0159] Therefore, in this embodiment of the disclosure, the electronic device can determine the true question intent of the question statement based on the probability of each intent, thereby accurately generating the standard answer to the question statement and improving the user experience.

[0160] Figure 4 A schematic diagram of the structure of an intelligent question-answering device provided in an embodiment of this disclosure is shown.

[0161] In some embodiments of this disclosure, Figure 4 The intelligent question-and-answer device shown can be installed within an electronic device. The electronic device can be a device with intelligent question-and-answer functionality. Specifically, the electronic device can include, but is not limited to, mobile terminals such as mobile phones, in-vehicle devices, vehicle controllers, tablets, wearable devices, and smart home devices.

[0162] like Figure 4 As shown, the intelligent question-answering device 400 may include a question acquisition module 410, an intent recognition module 420, a template determination module 430, and an answer generation module 440.

[0163] The problem retrieval module 410 can be used to retrieve problem statements.

[0164] The intent recognition module 420 can be used to recognize the intent of a question statement and obtain the true intent of the question statement.

[0165] The template determination module 430 can be used to determine the target answer template corresponding to the true question intent among multiple preset answer templates.

[0166] The answer generation module 440 can be used to combine the target answer template and the target product style information to generate the standard answer corresponding to the question statement. The target product style is the product style corresponding to the product involved in the question statement.

[0167] In this embodiment, a question statement can be acquired and its intent identified to obtain the true intent of the question statement. Then, among multiple preset answer templates, a target answer template corresponding to the true intent of the question statement is determined. Finally, the target answer template and the target product style information corresponding to the target product style are combined to generate a standard answer corresponding to the question statement. Here, the target product style is the product style corresponding to the product involved in the question statement. Thus, the true intent of the question statement can be accurately identified and the corresponding target answer template can be determined. Finally, the standard answer corresponding to the question statement is generated based on the target answer template and the target product style information. This allows for the accurate generation of a standard answer that matches the question statement by combining the true intent of the question statement and the target product style information of the question statement, thereby improving the accuracy of the generated answer and enhancing the user experience.

[0168] In some embodiments of this disclosure, the intent recognition module 420 may include a first training unit and an intent determination unit.

[0169] The first training unit can be used to input the question statement into the pre-trained intent recognition model, so that the intent recognition model can identify the intent of the question statement and obtain the probability that the question statement output by the intent recognition model belongs to each intent.

[0170] This intent determination unit can be used to determine the true question intent of a question statement based on the probability of each intent.

[0171] In some embodiments of this disclosure, the intent recognition model may include a word vector sub-model, an encoding sub-model, and a multilayer perceptron model.

[0172] In some embodiments of this disclosure, the first training unit may include a first training subunit, a second training subunit, and a third training subunit.

[0173] This first training subunit can be used to input the question statement into the word vector sub-model, so that the word vector sub-model can perform vector mapping on the question statement to obtain the word vectors output by the word vector sub-model.

[0174] This second training subunit can be used to input word vectors into the encoding submodel, so that the encoding submodel can perform vector encoding on the word vectors to obtain a fixed-length vector output by the encoding submodel.

[0175] This third training subunit can be used to input a fixed vector into a multilayer perceptron model, so that the multilayer perceptron model can perform probability calculations on the fixed vector, thereby obtaining the probability that the question statement output by the multilayer perceptron model belongs to each intent.

[0176] In some embodiments of this disclosure, the coding sub-model may include a bidirectional coding layer, an attention layer, and a pooling layer.

[0177] In some embodiments of this disclosure, the second training subunit may be specifically used to input word vectors into a bidirectional coding layer, so that the bidirectional coding layer performs bidirectional vector encoding on the word vectors to obtain bidirectional vectors output by the bidirectional coding layer.

[0178] In some embodiments of this disclosure, the second training subunit may also be specifically used to input a bidirectional vector into the attention layer, so that the attention layer assigns weights to the bidirectional vector to obtain a weight vector output by the attention layer.

[0179] In some embodiments of this disclosure, the second training subunit may also be specifically used to input the weight vector into the pooling layer, so that the pooling layer performs fixed-length sampling on the weight vector to obtain the fixed-length vector output by the pooling layer.

[0180] In some embodiments of this disclosure, the intent determination unit may include a descending sorting subunit, a first judgment subunit, a second judgment subunit, and a first determination subunit.

[0181] This descending sorting subunit can be used to sort the probabilities of various intentions in descending order.

[0182] The first judgment subunit can be used to determine whether the probability of the first intention in the first ranking is greater than the first preset threshold.

[0183] The second judgment subunit can be used to determine whether the sum of the probability of the first intention and the probability of the second intention (ranked second) is greater than the second preset threshold if the probability of the first intention is less than or equal to the first preset threshold.

[0184] The first determining subunit can be used to treat the first intention and the second intention as real problem intentions when the sum of the probabilities of the first intention and the second intention is greater than a second preset threshold.

[0185] In some embodiments of this disclosure, the intent-determining unit may further include a second determination subunit.

[0186] The second determining subunit can be used to determine whether the probability of the first intention in the first ranking is greater than the first preset threshold. If the probability of the first intention is greater than the first preset threshold, the first intention is taken as the real question intention.

[0187] In some embodiments of this disclosure, the intent determination unit may further include a third determination subunit.

[0188] The third determining subunit can be used to generate a general answer corresponding to the target product style after determining whether the sum of the probability of the first intention and the probability of the second intention in the second order is greater than a second preset threshold, and if the sum of the probability of the first intention and the probability of the second intention is less than or equal to the second preset threshold.

[0189] Figure 5 A schematic diagram of the structure of an intelligent question-answering device provided in an embodiment of this disclosure is shown.

[0190] In some embodiments of this disclosure, Figure 5 The intelligent question-and-answer device shown can be any electronic device that a user wants to use for intelligent question-and-answer. Specifically, the electronic device can be a device with intelligent question-and-answer functionality. This electronic device can include, but is not limited to, mobile terminals such as mobile phones, in-vehicle devices, vehicle controllers, tablets, wearable devices, and smart home devices.

[0191] like Figure 5 As shown, the intelligent question-answering device may include a processor 501 and a memory 502 storing computer program instructions.

[0192] Specifically, the processor 501 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0193] Memory 502 may include a large-capacity storage for information or instructions. For example, and not limitingly, memory 502 may include a hard disk drive (HDD), a floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 502 may include removable or non-removable (or fixed) media. Where appropriate, memory 502 may be internal or external to the integrated gateway device. In a particular embodiment, memory 502 is a non-volatile solid-state memory. In a particular embodiment, memory 502 includes read-only memory (ROM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (Electrically Programmable ROM, EPROM), an electrically erasable programmable PROM (EEPROM), an electrically alterable ROM (EAROM), or flash memory, or a combination of two or more of these.

[0194] The processor 501 reads and executes computer program instructions stored in the memory 502 to perform the steps of the intelligent question-answering method provided in this embodiment of the disclosure.

[0195] In one example, the intelligent question-answering device may also include a transceiver 503 and a bus 504. Wherein, as... Figure 5 As shown, the processor 501, memory 502 and transceiver 503 are connected via bus 504 and communicate with each other.

[0196] Bus 504 may include hardware, software, or both. For example, and not limitingly, a bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industrial Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a MicroChannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 504 may include one or more buses. Although specific buses are described and illustrated in the embodiments of this application, this application considers any suitable bus or interconnection.

[0197] This disclosure also provides a computer-readable storage medium that can store a computer program that, when executed by a processor, enables the processor to implement the intelligent question-answering method provided in this disclosure.

[0198] The aforementioned storage medium may, for example, include a memory 502 containing computer program instructions, which can be executed by the processor 501 of the intelligent question-answering device to complete the intelligent question-answering method provided in this embodiment. Optionally, the storage medium may be a non-transitory computer-readable storage medium, such as a ROM, random access memory (RAM), compact disc ROM (CD-ROM), magnetic tape, floppy disk, and optical data storage device.

[0199] This disclosure also provides a vehicle including the intelligent question-answering device described above. It is understood that the vehicle may also include a processor, a memory, and a computer program. The computer program is stored in the memory and configured to be executed by the processor to implement the intelligent question-answering method provided in this disclosure. The processor and memory are already... Figure 5 The parts of the illustrated embodiments will not be repeated here.

[0200] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the term "comprising" is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus.

[0201] The above description is merely a specific embodiment of this disclosure, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent question-answering method, characterized in that, include: Get the question statement; The intent of the question statement is identified to obtain the true intent of the question statement. Among multiple preset answer templates, determine the target answer template corresponding to the true intent of the question; The target answer template and the target product style information corresponding to the target product style are combined to generate the standard answer corresponding to the question statement, wherein the target product style is the product style corresponding to the product involved in the question statement; The step of performing intent recognition on the question statement to obtain the true question intent includes: The intent of the question statement is identified to obtain the probability that the question statement belongs to each intent. Based on the probability of each of the aforementioned intentions, a probability judgment is made to determine the true intention of the question statement. The step of determining the true question intent of the question statement based on the probability of each intent includes: Sort the probabilities of each intention in descending order; Determine whether the probability of the first intention ranked first is greater than a first preset threshold; If the probability of the first intention is less than or equal to the first preset threshold, determine whether the sum of the probability of the first intention and the probability of the second intention ranked second is greater than the second preset threshold. If the sum of the probabilities of the first intention and the second intention is greater than a second preset threshold, the first intention and the second intention are taken as the true problem intention.

2. The method according to claim 1, characterized in that, After determining whether the probability of the first intention ranked first is greater than a first preset threshold, the method further includes: If the probability of the first intention is greater than the first preset threshold, the first intention is taken as the true question intention.

3. The method according to claim 1, characterized in that, After determining whether the sum of the probability of the first intent and the probability of the second intent (ranked second) is greater than a second preset threshold, the method further includes: If the sum of the probabilities of the first intention and the second intention is less than or equal to a second preset threshold, a general answer corresponding to the target product style is generated.

4. The method according to claim 1, characterized in that, After determining the target answer template corresponding to the true question intent from multiple preset answer templates, the method further includes: The target entity is identified by performing target entity recognition on the question statement to obtain the target entity corresponding to the question statement; Determine the target product style corresponding to the product to which the target entity belongs.

5. An intelligent question-and-answer device, characterized in that, include: The problem retrieval module is used to retrieve problem statements; The intent recognition module is used to recognize the intent of the question statement and obtain the true question intent of the question statement. The template determination module is used to determine the target answer template corresponding to the true question intent from multiple preset answer templates; The answer generation module is used to combine the target answer template and the target product style information to generate the standard answer to the question statement, wherein the target product style is the product style corresponding to the product involved in the question statement; The intent recognition module includes: The first training unit is used to perform intent recognition on the question statement and obtain the probability that the question statement belongs to each intent. The intent determination unit is used to make a probability judgment based on the probability of each intent to determine the true intent of the question statement; The intent determination unit includes: The descending sorting subunit is used to sort the probabilities of the various intentions in descending order. The first judgment subunit is used to determine whether the probability of the first intention ranked first is greater than the first preset threshold. The second judgment subunit is used to determine whether the sum of the probability of the first intention and the probability of the second intention ranked second is greater than the second preset threshold if the probability of the first intention is less than or equal to the first preset threshold. The first determining subunit is configured to determine the first intention and the second intention as the real question intention when the sum of the probabilities of the first intention and the second intention is greater than a second preset threshold.

6. An electronic device, characterized in that, include: processor; Memory, used to store executable instructions; The processor is configured to read the executable instructions from the memory and execute the executable instructions to implement the intelligent question-answering method according to any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, The storage medium stores a computer program that, when executed by a processor, causes the processor to implement the intelligent question-answering method according to any one of claims 1-4.

8. A vehicle, characterized in that, Including the intelligent question-and-answer device as described in claim 5.

Images