Answer generation method and device, readable medium, electronic equipment and product

Abstract

The present disclosure relates to an answer generation method, device, readable medium, electronic equipment and product. The method comprises: in response to receiving a target question, determining an intent recognition result corresponding to the target question; determining a first answer from a first type of question and answer group or a second type of question and answer group stored in a target answer library according to the intent recognition result, the first type of question and answer group being used to indicate a corresponding relationship between a question and an answer, and the second type of question and answer group being used to indicate a corresponding relationship between an intent and an answer; and generating a target answer corresponding to the target question according to at least the first answer. Thus, the maintenance and update requirements of the knowledge base are greatly reduced, and even if the knowledge base is no longer frequently maintained and updated, the quality of answer generation can still be ensured, and the answer generation can be applied to more diverse question and answer scenarios.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and more specifically, to an answer generation method, apparatus, readable medium, electronic device, and product. Background Technology

[0002] Currently, in intelligent question-answering scenarios, knowledge bases are typically used as the corpus for question answering. Natural language processing is generally used to parse the question, and then the answer is determined based on relevant information from various domains and answers to common questions contained in the knowledge base. It is evident that the problem-solving rate in intelligent question-answering scenarios largely depends on the richness of the knowledge base; therefore, the maintenance and updating of the knowledge base plays a crucial role. Related technologies generally require targeted maintenance and updates of the knowledge base based on the specific needs of the actual scenario, which requires additional effort and incurs significant maintenance costs. Summary of the Invention

[0003] This summary section is provided to briefly introduce the concepts, which will be described in detail in the detailed description section below. This summary section is not intended to identify key or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.

[0004] Firstly, this disclosure provides a method for generating answers, the method comprising:

[0005] In response to receiving a target question, determine the intent recognition result corresponding to the target question;

[0006] Based on the intent recognition result, a first answer is determined from a first type of question-and-answer group or a second type of question-and-answer group stored in the target answer database. The first type of question-and-answer group is used to indicate the correspondence between questions and answers, and the second type of question-and-answer group is used to indicate the correspondence between diagrams and answers.

[0007] Based at least on the first answer, generate a target answer corresponding to the target question.

[0008] Secondly, this disclosure provides an answer generation apparatus, the apparatus comprising:

[0009] The first determining module is configured to determine the intent recognition result corresponding to the target question in response to receiving the target question;

[0010] The second determining module is used to determine a first answer from a first type of question-and-answer group or a second type of question-and-answer group stored in the target answer database based on the intent recognition result. The first type of question-and-answer group is used to indicate the correspondence between questions and answers, and the second type of question-and-answer group is used to indicate the correspondence between diagrams and answers.

[0011] The first generation module is used to generate a target answer corresponding to the target question, based at least on the first answer.

[0012] Thirdly, this disclosure provides a computer-readable medium having a computer program stored thereon, which, when executed by a processing device, implements the steps of the method described in the first aspect of this disclosure.

[0013] Fourthly, this disclosure provides an electronic device, comprising:

[0014] A storage device having at least one computer program stored thereon;

[0015] At least one processing means is configured to execute the at least one computer program in the storage device to implement the steps of the method described in the first aspect of this disclosure.

[0016] Fifthly, this disclosure provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in the first aspect of this disclosure.

[0017] The above technical solution provides a target answer database storing a first type of question-and-answer group and a second type of question-and-answer group. The first type of question-and-answer group indicates the correspondence between questions and answers, while the second type of question-and-answer group indicates the correspondence between diagrams and answers. Based on this, upon receiving a target question, intent recognition is first performed on the target question to determine the intent recognition result. Then, based on the intent recognition result, a first answer is determined from either the first type of question-and-answer group or the second type of question-and-answer group in the target answer database. At least based on the first answer, a target answer corresponding to the target question is generated. By providing a target answer database, answer generation no longer relies entirely on the knowledge base, greatly reducing the need for maintenance and updates to the knowledge base. Even without frequent effort to maintain and update the knowledge base, the quality of answer generation can still be guaranteed based on the target answer database. Furthermore, the target answer database not only stores a second type of question-and-answer group for indicating the correspondence between diagrams and answers, but also stores a first type of question-and-answer group for indicating the correspondence between questions and answers. This makes the question-and-answer corpus in the target answer database richer. Moreover, when the received target question corresponds to an intent, the answer can be obtained directly from the second type of question-and-answer group. When the target question does not correspond to an intent, the answer can also be obtained from the first type of question-and-answer group. Therefore, it can be applied to a wider range of question-and-answer scenarios.

[0018] Other features and advantages of this disclosure will be described in detail in the following detailed description section. Attached Figure Description

[0019] 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. In the drawings:

[0020] Figure 1 This is a flowchart of an answer generation method provided according to one embodiment of the present disclosure;

[0021] Figure 2 This is a flowchart of an answer generation method provided according to another embodiment of this disclosure;

[0022] Figure 3 This is a block diagram of an answer generation apparatus provided according to one embodiment of the present disclosure;

[0023] Figure 4 A schematic diagram of the structure of an electronic device suitable for implementing embodiments of the present disclosure is shown. Detailed Implementation

[0024] 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.

[0025] 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.

[0026] 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.

[0027] 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.

[0028] 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".

[0029] 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.

[0030] It is understood that before using the technical solutions disclosed in the various embodiments of this disclosure, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this disclosure in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0031] For example, upon receiving a user's active request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose whether to provide personal information to the software or hardware, such as the electronic device, application, server, or storage medium performing the operations of this disclosed technical solution, based on the prompt message.

[0032] As an optional but non-limiting implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0033] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of this disclosure. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this disclosure.

[0034] Meanwhile, it is understood that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) shall comply with the requirements of relevant laws, regulations and related provisions.

[0035] As described in the background section, in intelligent question-answering scenarios (such as intelligent customer service and intelligent search), the generation of answers by intelligent question-answering robots heavily relies on the intents and corresponding answers stored in the knowledge base. Therefore, if the knowledge base cannot be updated and maintained in a timely manner, it is highly likely that questions will not be answered successfully (e.g., the answer does not exist or is incorrect), affecting the questioner's experience in the intelligent question-answering scenario. Maintaining and updating the knowledge base requires significant costs. For example, in the e-commerce sector, intelligent chatbots are commonly used as intelligent customer service. For merchants (especially small and medium-sized enterprises), it is difficult to have sufficient resources or capital to maintain and update the knowledge base, resulting in a low problem-solving rate for intelligent customer service and a poor customer experience during conversations with intelligent customer service.

[0036] To address the aforementioned problems, this disclosure provides an answer generation method, apparatus, readable medium, electronic device, and product that reduces the dependence of answer generation on a knowledge base, lowers the maintenance and update requirements of the knowledge base, ensures the quality of answer generation, and is applicable to various question-and-answer scenarios.

[0037] Figure 1 This is a flowchart of an answer generation method provided according to one embodiment of this disclosure. For example... Figure 1 As shown, the method may include steps 11 to 13.

[0038] In step 11, in response to receiving the target question, the intent recognition result corresponding to the target question is determined.

[0039] The intent recognition result can be achieved by performing intent recognition on the target question.

[0040] If the intent can be identified from the target question through intent recognition, the intent recognition result can be used to indicate the intent identified from the target question, as the target intent. If the intent cannot be identified from the target question through intent recognition, the intent recognition result can be used to indicate the target question itself.

[0041] Optionally, a pre-generated intent recognition model can be used to identify the intent of the target question, thereby obtaining the intent recognition result. For example, a dataset containing multiple sets of text and their intent labels (e.g., labels representing no intent, labels representing at least one intent, etc.) can be collected. The text in the dataset can be used as model input, and the corresponding intent labels of the text can be used as the model's target output to train a neural network (e.g., convolutional neural network, recurrent neural network, attention mechanism neural network, or fine-tuning of a pre-trained language model) to generate the aforementioned intent recognition model. In this way, based on the intent recognition model, it can be determined whether the target question corresponds to an intent, and if so, what kind of intent it corresponds to.

[0042] In step 12, based on the intent recognition result, the first answer is determined from either the first type of question-and-answer group or the second type of question-and-answer group stored in the target answer database.

[0043] The first type of question-and-answer group is used to indicate the correspondence between questions and answers, while the second type of question-and-answer group is used to indicate the correspondence between diagrams and answers.

[0044] In this disclosure, the target answer database is a question-and-answer corpus independent of the knowledge base mentioned above. It stores both the correspondence between intent and answer, as well as the correspondence between question and answer, and is therefore applicable to a wide range of question-and-answer scenarios.

[0045] The generation method of the target answer database will be explained in detail below. Optionally, the target answer database can be generated in the following ways:

[0046] Identify multiple dialogue samples, each dialogue sample consisting of a question text and an answer text corresponding to that question text;

[0047] Intent recognition is performed on the question text in the dialogue samples to identify the first type of dialogue samples from which the intent cannot be identified and the second type of dialogue samples from which the intent can be identified.

[0048] Generate a first-type question-and-answer group based on the question and answer texts in the first-type dialogue sample;

[0049] Based on the intent corresponding to the second type of dialogue sample and the answer text in the second type of dialogue sample, generate a second type of question-answer group.

[0050] Generally, multiple sets of dialogue information can be collected, and at least a portion of the collected sets of dialogue information can be selected as dialogue samples. Each set of dialogue information may include an original question and an original answer.

[0051] In one possible implementation, since the collected multiple sets of dialogue information may contain useless data that is not helpful in generating the target answer database or even adversely affects the quality of the answers, the collected multiple sets of dialogue information can be filtered to select the aforementioned multiple dialogue samples. In this implementation, the multiple dialogue samples can be determined in the following way:

[0052] Obtain multiple sets of dialogue information;

[0053] The dialogue information is filtered to obtain multiple dialogue samples that meet preset conditions.

[0054] The preset conditions may include at least one of the following: no noise data, semantic similarity between question text and answer text, and no preset content.

[0055] Optionally, the filtering process may include methods for filtering noisy data. Noisy data in the dialogue information is data irrelevant to the generation of the target answer database and can be filtered out. What constitutes noisy data can be pre-defined according to actual needs. For example, greetings and small talk can be classified as noisy data.

[0056] Based on this, noise identification can be performed on the dialogue information to identify noisy data within it. Then, by filtering out dialogue information containing noisy data, and / or by filtering out strings corresponding to noisy data in the original question and answer, the dialogue samples can be made free of noisy data. Filtering out dialogue information containing noisy data is equivalent to directly filtering out the entire dialogue information if it contains noisy data. Filtering out strings corresponding to noisy data in the original question and answer is equivalent to using the new text generated after filtering out the noise from the original question as the question sample in the dialogue sample, and using the text generated after filtering out the noise from the original answer as the answer sample in the dialogue sample. For example, noise data can be identified through keyword matching.

[0057] Optionally, the filtering process may include methods for filtering semantically dissimilar dialogue information. In a question-and-answer scenario, if the question and answer are too semantically unrelated, it may lead to an incorrect answer. Therefore, for each dialogue information, the semantic similarity between the original question and the original answer in that dialogue information can be determined, and only dialogue information with semantically similar questions and answers can be retained as dialogue samples.

[0058] For example, semantic similarity between question text and answer text can be determined in the following ways:

[0059] Input the question text and answer text into a pre-generated sentence embedding representation model to obtain the first sentence representation vector corresponding to the question text and the second sentence representation vector corresponding to the answer text;

[0060] Determine the similarity between the representation vectors of the first sentence and the representation vectors of the second sentence;

[0061] If the similarity score is greater than the similarity threshold, the question text and the answer text are considered semantically similar. Otherwise, if the similarity score is not greater than the similarity threshold, the question text and the answer text are considered semantically dissimilar and will be filtered out.

[0062] This involves collecting a dataset containing questions, answers, and semantic similarity labels (e.g., a first label representing semantic similarity and a second label representing semantic dissimilarity). This dataset is then used to fine-tune a pre-trained model, minimizing the distance between the sentence representation vectors of the questions and answers corresponding to the first label and maximizing the distance between the two sentence representation vectors of the questions and answers corresponding to the second label, thus obtaining the aforementioned sentence embedding representation model. For example, the pre-trained model can be a BERT model; correspondingly, the resulting sentence embedding representation model is the Sentence-BERT model.

[0063] After inputting the question text and answer text into the sentence embedding representation model, the model outputs a first sentence representation vector corresponding to the question text and a second sentence representation vector corresponding to the answer text. Then, the semantic similarity between the two can be determined based on the distance between them. If the distance is small, it indicates that the two are close in semantic space and have similar semantics.

[0064] The distance between two sentence representation vectors can be reflected by calculating their similarity. Therefore, the similarity between the first and second sentence representation vectors can be determined, and by comparing it to a pre-set similarity threshold, it can be determined whether the question and answer are semantically similar. For example, the similarity between the first and second sentence representation vectors can be calculated using cosine similarity.

[0065] Based on this, semantically dissimilar dialogue information between the original question and the original answer can be filtered out, and the remaining dialogue information can be used as dialogue samples. The original question in the remaining dialogue information can be used as question text and the original answer can be used as answer text, so that the question samples and answer samples in the dialogue samples are semantically similar.

[0066] Optionally, the filtering process may include methods for filtering preset content. The preset content can be pre-set according to actual needs. For example, content that is not allowed to appear in the answer can be set as preset content.

[0067] Based on this, content recognition can be performed on dialogue information to identify preset content within the dialogue. Furthermore, by filtering out dialogue information containing preset content, and / or by filtering out strings corresponding to preset content in the original question and answer, the dialogue sample can be made free of preset content. Filtering out dialogue information containing preset content is equivalent to directly filtering out the entire dialogue information group if it contains preset content. Filtering out strings corresponding to preset content in the original question and answer is equivalent to using the new text generated after filtering out the preset content from the original question as the question sample in the dialogue sample, and using the new text generated after filtering out the preset content from the original answer as the answer sample in the dialogue sample. For example, preset content can be identified through keyword matching. Alternatively, preset content can be identified through a pre-generated recognition model that outputs labels for each word segment in the text; these labels characterize whether the word segment is preset content.

[0068] Based on the above method, multiple dialogue samples can be identified for generating the target answer library.

[0069] Based on the aforementioned dialogue samples, intent recognition can be performed on the question texts within the dialogue samples to determine which question texts lack intent and classify the dialogue samples containing those question texts as the first type of dialogue sample. Conversely, it can be determined which question texts correspond to intent and classify the dialogue samples containing those question texts as the second type of dialogue sample. In this way, dialogue samples can be divided into two main categories based on whether the question texts correspond to intent. For example, the above intent recognition can be implemented using an intent recognition model, and the relevant implementation methods have been described above and will not be repeated here.

[0070] For the first type of dialogue sample, since the questions in the first type of dialogue sample do not have corresponding intentions, the first type of question-answer group can be generated directly based on the question text and answer text in the first type of dialogue sample.

[0071] For the second type of dialogue sample, since the questions in the second type of dialogue sample have corresponding intentions, in order to be compatible with more question formats corresponding to the same intention, a second type of question-answer group can be generated based on the intentions corresponding to the second type of dialogue sample and the answer text in the second type of dialogue sample.

[0072] It should be noted that when generating the first or second type of question-and-answer groups, answers can be clustered based on the question or the intent, depending on the actual needs. For example, multiple answers corresponding to the same question can be clustered into multiple answers for that question, forming a one-to-many relationship between questions and answers. Similarly, multiple answers corresponding to the same intent can be clustered into multiple answers for that intent, forming a one-to-many relationship between intent and answers.

[0073] Optionally, for the data in the target answer database (including the first type of question and answer group and the second type of question and answer group), considering that the data volume is large in actual application scenarios and the real-time requirements are not high, it can be written into the Hive (a data warehouse tool) table of the offline task, and then the data in the table can be imported into ES (Elasticsearch, a distributed search engine) offline for subsequent answer generation.

[0074] In another possible implementation, all dialogue information can be used as dialogue samples, and the first type of dialogue samples and the second type of dialogue samples can be filtered as described above (e.g., filtering noisy data, filtering semantically dissimilar dialogue samples, filtering preset content).

[0075] In another possible implementation, the order of filtering can be reasonably set according to actual needs. For example, by filtering the dialogue information for noise data and semantic dissimilarity, dialogue samples are obtained. After obtaining the first type of dialogue samples and the second type of dialogue samples based on the dialogue samples, the first type of dialogue samples and the second type of dialogue samples are then filtered for preset content respectively.

[0076] The filtering methods described above have been described in detail above and will not be repeated here.

[0077] The above method can be used to construct a target answer library with rich first-type and second-type question-answer groups, which is conducive to compatibility with a variety of different question-answering scenarios.

[0078] Returning to step 12, based on the intent recognition results, determine the first answer from either the first or second type of question-and-answer group stored in the target answer database.

[0079] In one possible implementation, step 12 may include the following steps:

[0080] If the intent recognition result is the target question, then based on the target question, at least one first alternative answer that matches the target question is determined from the first type of question-answer group, and one of the first alternative answers is determined as the first answer;

[0081] If the intent recognition result is the target intent, then based on the target intent, at least one second alternative answer that matches the target intent is determined from the second type of question and answer group, and one of the second alternative answers is determined as the first answer.

[0082] If the intent recognition result is the target question, it means there is no intent corresponding to the target question, and therefore the answer cannot be determined based on the intent. In this case, at least one first alternative answer that matches the target question can be determined from the first question-answer group. That is, based on the first question-answer group, questions that match the target question in the first question-answer group are identified, and the answer to that matching question is taken as the first alternative answer. Furthermore, one of the first alternative answers is determined as the first answer.

[0083] Optionally, the matching degree between each question in the first question-and-answer group and the target question can be calculated. If the matching degree is higher than a specified threshold, the question is considered to match the target question. For example, the answer with the highest matching degree among the first candidate answers can be determined as the first answer.

[0084] Optionally, as mentioned above, the data in the target answer database can be stored in Elasticsearch (ES). Since ES has query and scoring functions, the target question can be used as a query condition in ES to obtain the answers and their scores returned by ES. For example, the answer with the highest average score among the first candidate answers can be determined as the first answer.

[0085] If the intent recognition result is the target intent, it means that there exists an intent corresponding to the target question, and the answer can be determined directly through the intent. In this case, based on the target intent, at least one second alternative answer that matches the target intent can be determined from the second question-answer group. That is, based on the second question-answer group, the intent in the second question-answer group that can match the target intent is determined, and the answer of the matching intent is taken as the second alternative answer. Furthermore, one of the second alternative answers is determined as the first answer.

[0086] Optionally, the matching degree between each intent in the second question-and-answer group and the target intent can be calculated. If the matching degree is higher than a specified threshold, the intent is considered to match the target intent. For example, the one with the highest matching degree among the second alternative answers can be determined as the first answer.

[0087] Optionally, as mentioned above, the data in the target answer database can be stored in Elasticsearch (ES). Since ES has query and scoring functions, the target intent can be used as the query condition to obtain the answers returned by ES and their scores. For example, the answer with the highest average score among the second alternative answers can be determined as the first answer.

[0088] In step 13, at least based on the first answer, a target answer corresponding to the target question is generated.

[0089] In one possible implementation, after determining the first answer, it can be designated as the target answer corresponding to the target question. In this implementation, the target answer database can independently generate answers to the target question without relying on a knowledge base.

[0090] In another possible implementation, the method provided in this disclosure may further include the following steps:

[0091] If the intent recognition result is the target intent, determine the second answer from the preset knowledge base.

[0092] The preset knowledge base refers to the correspondence between the diagram and the answer. Optionally, the preset knowledge base can be the knowledge base mentioned above that needs to be maintained and updated.

[0093] For example, based on the target intent, the text similarity between each intent in the preset knowledge base and the target intent can be determined, and the answer corresponding to the intent with the highest text similarity can be used as the second answer. Furthermore, the method for determining the second answer can also refer to the method for determining the first answer described above, and will not be repeated here.

[0094] In this embodiment, step 13 may include the following steps:

[0095] Choose either the first or second answer as the target answer.

[0096] In this implementation, it is equivalent to combining the target answer library and the preset knowledge base to generate an answer to the target question, thereby obtaining a better target answer.

[0097] Optionally, if the answer cannot be determined through the preset knowledge base, that is, if the second answer is empty, the first answer can be directly determined as the target answer.

[0098] Optionally, if the answer cannot be determined through the target answer database, i.e., the first answer is empty, the second answer can be directly determined as the target answer.

[0099] Optionally, the priority of the answers can be preset, and when neither the first nor the second answer is empty, the one with the higher priority is determined as the target answer.

[0100] For example, the priority of answers in the target answer base can be set to be higher than that of answers in the preset knowledge base, so that when neither the first answer nor the second answer is empty, the first answer is taken as the target answer.

[0101] For example, answer priorities can be set based on the update status of answers in the preset knowledge base. If a user updates an answer while maintaining the preset knowledge base, the current answer better reflects the user's expectations. Based on this, the highest priority can be given to answers updated by the user in the preset knowledge base, followed by answers in the target answer database, and the lowest priority can be given to answers in the preset knowledge base that have not been updated by the user, so that the answers better meet the actual needs of the questioner.

[0102] Accordingly, determining the first or second answer as the target answer may include the following steps:

[0103] Determine if there is a record of answer modifications corresponding to the second answer;

[0104] If there is a history of answer modifications, the second answer will be selected as the target answer.

[0105] If there is no record of answer modification, the first answer will be set as the target answer.

[0106] The answer modification record reflects the user's updates to the answers in the preset knowledge base. If a modification record exists corresponding to the second answer, it means the user updated the second answer while maintaining the preset knowledge base; therefore, the current second answer better reflects the user's expectations. Based on this, if neither the first nor the second answer is empty, and a modification record exists corresponding to the second answer, the second answer is determined as the target answer; otherwise, the first answer is determined as the target answer.

[0107] For example, Figure 2 This is an exemplary flowchart for generating a target answer by combining a pre-defined knowledge base and a target answer database. It mainly includes the following steps:

[0108] Step 101: In response to receiving the target question, determine the intent recognition result corresponding to the target question;

[0109] If the intent recognition result is the target intent, proceed to steps 103, 104 and 107; if the intent recognition result is the target question, proceed to steps 105, 106 and 109.

[0110] Step 103: Based on the target intent, determine at least one second alternative answer that matches the target intent from the second type of question and answer group, and determine one of the second alternative answers as the first answer;

[0111] Step 104: Determine the second answer from the preset knowledge base;

[0112] Step 105: Based on the target question, determine at least one first alternative answer that matches the target question from the first type of question-answer group, and determine one of the first alternative answers as the first answer;

[0113] Step 106: Determine from the preset knowledge that the second answer is empty;

[0114] Step 107: Determine if there is an answer modification record corresponding to the second answer; if there is, proceed to step 108; if not, proceed to step 109.

[0115] Step 108: Determine the second answer as the target answer;

[0116] Step 109: Determine the first answer as the target answer.

[0117] based on Figure 2 The provided flowchart can yield a better target answer and optimize the answer.

[0118] The above technical solution provides a target answer database storing a first type of question-and-answer group and a second type of question-and-answer group. The first type of question-and-answer group indicates the correspondence between questions and answers, while the second type of question-and-answer group indicates the correspondence between diagrams and answers. Based on this, upon receiving a target question, intent recognition is first performed on the target question to determine the intent recognition result. Then, based on the intent recognition result, a first answer is determined from either the first type of question-and-answer group or the second type of question-and-answer group in the target answer database. At least based on the first answer, a target answer corresponding to the target question is generated. By providing a target answer database, answer generation no longer relies entirely on the knowledge base, greatly reducing the need for maintenance and updates to the knowledge base. Even without frequent effort to maintain and update the knowledge base, the quality of answer generation can still be guaranteed based on the target answer database. Furthermore, the target answer database not only stores a second type of question-and-answer group for indicating the correspondence between diagrams and answers, but also stores a first type of question-and-answer group for indicating the correspondence between questions and answers. This makes the question-and-answer corpus in the target answer database richer. Moreover, when the received target question corresponds to an intent, the answer can be obtained directly from the second type of question-and-answer group. When the target question does not correspond to an intent, the answer can also be obtained from the first type of question-and-answer group. Therefore, it can be applied to a wider range of question-and-answer scenarios.

[0119] Figure 3 This is a block diagram of an answer generation apparatus provided according to one embodiment of the present disclosure. Figure 3 As shown, the device 30 includes:

[0120] The first determining module 31 is configured to determine the intent recognition result corresponding to the target question in response to receiving the target question;

[0121] The second determining module 32 is used to determine a first answer from a first type of question-and-answer group or a second type of question-and-answer group stored in the target answer database based on the intent recognition result. The first type of question-and-answer group is used to indicate the correspondence between questions and answers, and the second type of question-and-answer group is used to indicate the correspondence between diagrams and answers.

[0122] The first generation module 33 is used to generate a target answer corresponding to the target question, based at least on the first answer.

[0123] Optionally, the target answer database is generated through the following modules:

[0124] The third determining module is used to determine multiple dialogue samples, wherein each dialogue sample includes a question text and an answer text corresponding to the question text;

[0125] The fourth determining module is used to perform intent recognition on the question text in the dialogue samples to determine a first type of dialogue samples from which intent cannot be identified and a second type of dialogue samples from which intent can be identified.

[0126] The second generation module is used to generate the first type of question-answer group based on the question text and answer text in the first type of dialogue sample;

[0127] The third generation module is used to generate the second type of question-and-answer group based on the intent corresponding to the second type of dialogue sample and the answer text in the second type of dialogue sample.

[0128] Optionally, the third determining module includes:

[0129] The acquisition submodule is used to acquire multiple sets of dialogue information, where each set of dialogue information includes an original question and an original answer.

[0130] The filtering submodule is used to filter the dialogue information to obtain multiple dialogue samples that meet preset conditions.

[0131] The preset conditions include at least one of the following: no noise data, semantic similarity between the question text and the answer text, and no preset content.

[0132] Optionally, semantic similarity between question and answer texts can be determined using the following modules:

[0133] The processing module is used to input the question text and the answer text into a pre-generated sentence embedding representation model to obtain a first sentence representation vector corresponding to the question text and a second sentence representation vector corresponding to the answer text;

[0134] The fifth determining module is used to determine the similarity between the first sentence representation vector and the second sentence representation vector;

[0135] The sixth determining module is used to determine that the question text and the answer text are semantically similar if the similarity is greater than a similarity threshold.

[0136] Optionally, the second determining module 32 includes:

[0137] The first determining submodule is configured to, if the intent recognition result is the target question, determine at least one first alternative answer matching the target question from the first type of question-answer group, and determine one of the first alternative answers as the first answer;

[0138] The second determining submodule is used to determine at least one second alternative answer that matches the target intent from the second type of question-and-answer group if the intent recognition result is a target intent, and to determine one of the second alternative answers as the first answer.

[0139] Optionally, the device 30 further includes:

[0140] The seventh determining module is used to determine a second answer from a preset knowledge base if the intent recognition result is the target intent. The preset knowledge base is used to refer to the correspondence between the schematic diagram and the answer.

[0141] The first generation module 33 includes:

[0142] The third determining submodule is used to determine the first answer or the second answer as the target answer.

[0143] Optionally, the third determining submodule includes:

[0144] The fourth determination submodule is used to determine whether there is an answer modification record corresponding to the second answer;

[0145] The fifth determination submodule is used to determine the second answer as the target answer if the answer modification record exists;

[0146] The sixth determination submodule is used to determine the first answer as the target answer if no answer modification record exists.

[0147] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0148] Based on the same inventive concept, this disclosure also provides a computer-readable medium having a computer program stored thereon, which, when executed by a processing device, implements the steps of the answer generation method described in any embodiment of this disclosure.

[0149] Based on the same inventive concept, this disclosure also provides an electronic device, comprising:

[0150] A storage device having at least one computer program stored thereon;

[0151] At least one processing device is configured to execute the at least one computer program in the storage device to implement the steps of the answer generation method described in any embodiment of the present disclosure.

[0152] Based on the same inventive concept, this disclosure also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the answer generation method described in any embodiment of this disclosure.

[0153] The following is for reference. Figure 4 This diagram illustrates a structural schematic of an electronic device 600 suitable for implementing embodiments of the present disclosure. The terminal devices in the embodiments of the present disclosure may include, but are not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 4 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0154] like Figure 4 As shown, electronic device 600 may include a processing device (e.g., a central processing unit, a graphics processor, etc.) 601, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 602 or a program loaded from storage device 608 into random access memory (RAM) 603. RAM 603 also stores various programs and data required for the operation of electronic device 600. Processing device 601, ROM 602, and RAM 603 are interconnected via bus 604. Input / output (I / O) interface 605 is also connected to bus 604.

[0155] Typically, the following devices can be connected to I / O interface 605: input devices 606 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 607 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 608 including, for example, magnetic tapes, hard disks, etc.; and communication devices 609. Communication device 609 allows electronic device 600 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 4 An electronic device 600 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively.

[0156] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 609, or installed from a storage device 608, or installed from a ROM 602. When the computer program is executed by the processing device 601, it performs the functions defined in the methods of embodiments of this disclosure.

[0157] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0158] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0159] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0160] The aforementioned computer-readable medium carries one or more programs that, when executed by the electronic device, cause the electronic device to: in response to receiving a target question, determine an intent recognition result corresponding to the target question; determine a first answer from a first type of question-and-answer group or a second type of question-and-answer group stored in a target answer database, wherein the first type of question-and-answer group indicates the correspondence between questions and answers, and the second type of question-and-answer group indicates the correspondence between diagrams and answers; and generate a target answer corresponding to the target question, at least based on the first answer.

[0161] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including but not limited to object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0162] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0163] The modules described in the embodiments of this disclosure can be implemented in software or in hardware. The names of the modules are not necessarily limiting in certain circumstances; for example, the first determining module can also be described as "a module for determining the intent recognition result corresponding to the target problem".

[0164] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: Field Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), Application Standard Products (ASSPs), System-on-Chip (SoCs), Complex Programmable Logic Devices (CPLDs), and so on.

[0165] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0166] According to one or more embodiments of this disclosure, an answer generation method is provided, the method comprising:

[0167] In response to receiving a target question, determine the intent recognition result corresponding to the target question;

[0168] Based on the intent recognition result, a first answer is determined from a first type of question-and-answer group or a second type of question-and-answer group stored in the target answer database. The first type of question-and-answer group is used to indicate the correspondence between questions and answers, and the second type of question-and-answer group is used to indicate the correspondence between diagrams and answers.

[0169] Based at least on the first answer, generate a target answer corresponding to the target question.

[0170] According to one or more embodiments of this disclosure, an answer generation method is provided, wherein the target answer database is generated in the following manner:

[0171] Determine multiple dialogue samples, each dialogue sample comprising a question text and an answer text corresponding to the question text;

[0172] Intent recognition is performed on the question text in the dialogue samples to determine a first type of dialogue samples from which intent cannot be identified and a second type of dialogue samples from which intent can be identified.

[0173] Generate the first type of question-answer group based on the question text and answer text in the first type of dialogue sample;

[0174] Based on the intent corresponding to the second type of dialogue sample and the answer text in the second type of dialogue sample, generate the second type of question-answer group.

[0175] According to one or more embodiments of this disclosure, an answer generation method is provided, wherein determining multiple dialogue samples includes:

[0176] Acquire multiple sets of dialogue information, where each set of dialogue information includes an original question and an original answer;

[0177] The dialogue information is filtered to obtain multiple dialogue samples that meet preset conditions;

[0178] The preset conditions include at least one of the following: no noise data, semantic similarity between the question text and the answer text, and no preset content.

[0179] According to one or more embodiments of this disclosure, an answer generation method is provided that determines semantic similarity between question text and answer text by:

[0180] The question text and the answer text are input into a pre-generated sentence embedding representation model to obtain a first sentence representation vector corresponding to the question text and a second sentence representation vector corresponding to the answer text;

[0181] Determine the similarity between the first sentence representation vector and the second sentence representation vector;

[0182] If the similarity is greater than the similarity threshold, the question text and the answer text are determined to be semantically similar.

[0183] According to one or more embodiments of this disclosure, an answer generation method is provided, wherein determining a first answer from a first type of question-and-answer group or a second type of question-and-answer group stored in a target answer database based on the intent recognition result includes:

[0184] If the intent recognition result is the target question, then, based on the target question, at least one first alternative answer matching the target question is determined from the first type of question-answer group, and one of the first alternative answers is determined as the first answer;

[0185] If the intent recognition result is a target intent, then based on the target intent, at least one second alternative answer that matches the target intent is determined from the second type of question and answer group, and one of the second alternative answers is determined as the first answer.

[0186] According to one or more embodiments of this disclosure, an answer generation method is provided, the method further comprising:

[0187] If the intent recognition result is the target intent, a second answer is determined from a preset knowledge base, whereby the preset knowledge base is used to indicate the correspondence between the diagram and the answer;

[0188] The step of generating a target answer corresponding to the target question based at least on the first answer includes:

[0189] The first answer or the second answer is determined as the target answer.

[0190] According to one or more embodiments of this disclosure, an answer generation method is provided, wherein determining the first answer or the second answer as the target answer includes:

[0191] Determine if there is an answer modification record corresponding to the second answer;

[0192] If the aforementioned answer modification record exists, the second answer will be determined as the target answer;

[0193] If no such answer modification record exists, the first answer will be designated as the target answer.

[0194] According to one or more embodiments of this disclosure, an answer generation apparatus is provided, the apparatus comprising:

[0195] The first determining module is configured to determine the intent recognition result corresponding to the target question in response to receiving the target question;

[0196] The second determining module is used to determine a first answer from a first type of question-and-answer group or a second type of question-and-answer group stored in the target answer database based on the intent recognition result. The first type of question-and-answer group is used to indicate the correspondence between questions and answers, and the second type of question-and-answer group is used to indicate the correspondence between diagrams and answers.

[0197] The first generation module is used to generate a target answer corresponding to the target question, based at least on the first answer.

[0198] According to one or more embodiments of the present disclosure, a computer-readable medium is provided having a computer program stored thereon that, when executed by a processing device, implements the steps of the answer generation method described in any embodiment of the present disclosure.

[0199] According to one or more embodiments of this disclosure, an electronic device is provided, comprising:

[0200] A storage device having at least one computer program stored thereon;

[0201] At least one processing device is configured to execute the at least one computer program in the storage device to implement the steps of the answer generation method described in any embodiment of the present disclosure.

[0202] According to one or more embodiments of the present disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the answer generation method described in any embodiment of the present disclosure.

[0203] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features disclosed in this disclosure that have similar functions.

[0204] Furthermore, while the operations are described in a specific order, this should not be construed as requiring these operations to be performed in the specific order shown or in a sequential order. In certain environments, multitasking and parallel processing may be advantageous. Similarly, while several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of individual embodiments may also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented individually or in any suitable sub-combination in multiple embodiments.

[0205] Although the subject matter has been described using language specific to structural features and / or methodological logic, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are merely illustrative forms of implementing the claims. Regarding the apparatus in the above embodiments, the specific manner in which the various modules perform their operations has been described in detail in the embodiments relating to the method, and will not be elaborated upon here.

Claims

1. An answer generation method, characterized by, The method includes: In response to receiving a target question, determine the intent recognition result corresponding to the target question; Based on the intent recognition result, a first answer is determined from a first type of question-and-answer group or a second type of question-and-answer group stored in the target answer database. The first type of question-and-answer group is used to indicate the correspondence between questions and answers, and the second type of question-and-answer group is used to indicate the correspondence between diagrams and answers. At least based on the first answer, generate a target answer corresponding to the target question; The determination of the intent recognition result corresponding to the target question includes: If an intent can be identified from the target question, the intent identification result is used to indicate the intent identified from the target question as the target intent; If the intent cannot be identified from the target question, the intent identification result is used to indicate the target question; The step of determining the first answer from the first or second type of question-answer group stored in the target answer database based on the intent recognition result includes: If the intent recognition result is the target question, then, based on the target question, at least one first alternative answer matching the target question is determined from the first type of question-answer group, and one of the first alternative answers is determined as the first answer; If the intent recognition result is a target intent, then based on the target intent, at least one second alternative answer that matches the target intent is determined from the second type of question and answer group, and one of the second alternative answers is determined as the first answer.

2. The method of claim 1, wherein, The target answer database is generated in the following way: Determine multiple dialogue samples, each dialogue sample comprising a question text and an answer text corresponding to the question text; Intent recognition is performed on the question text in the dialogue samples to determine a first type of dialogue samples from which intent cannot be identified and a second type of dialogue samples from which intent can be identified. Generate the first type of question-answer group based on the question text and answer text in the first type of dialogue sample; Based on the intent corresponding to the second type of dialogue sample and the answer text in the second type of dialogue sample, generate the second type of question-answer group.

3. The method of claim 2, wherein, The determination of multiple dialogue samples includes: Acquire multiple sets of dialogue information, where each set of dialogue information includes an original question and an original answer; The dialogue information is filtered to obtain multiple dialogue samples that meet preset conditions; The preset conditions include at least one of the following: no noise data, semantic similarity between the question text and the answer text, and no preset content.

4. The method according to claim 3, characterized in that, Determine semantic similarity between question and answer texts using the following methods: The question text and the answer text are input into a pre-generated sentence embedding representation model to obtain a first sentence representation vector corresponding to the question text and a second sentence representation vector corresponding to the answer text; Determine the similarity between the first sentence representation vector and the second sentence representation vector; If the similarity is greater than the similarity threshold, the question text and the answer text are determined to be semantically similar.

5. The method of claim 1, wherein, The method further includes: If the intent recognition result is the target intent, a second answer is determined from a preset knowledge base, whereby the preset knowledge base is used to indicate the correspondence between the diagram and the answer; The step of generating a target answer corresponding to the target question based at least on the first answer includes: The first answer or the second answer is determined as the target answer.

6. The method of claim 5, wherein, The step of determining the first answer or the second answer as the target answer includes: Determine if there is an answer modification record corresponding to the second answer; If the aforementioned answer modification record exists, the second answer will be determined as the target answer; If no such answer modification record exists, the first answer will be designated as the target answer.

7. An answer generation apparatus characterized by comprising: The device includes: The first determining module is configured to determine the intent recognition result corresponding to the target question in response to receiving the target question; The second determining module is used to determine a first answer from a first type of question-and-answer group or a second type of question-and-answer group stored in the target answer database based on the intent recognition result. The first type of question-and-answer group is used to indicate the correspondence between questions and answers, and the second type of question-and-answer group is used to indicate the correspondence between diagrams and answers. The first generation module is configured to generate a target answer corresponding to the target question, based at least on the first answer. The first determining module is further configured to: If an intent can be identified from the target question, the intent identification result is used to indicate the intent identified from the target question as the target intent; If the intent cannot be identified from the target question, the intent identification result is used to indicate the target question; The second determining module includes: The first determining submodule is configured to, if the intent recognition result is the target question, determine at least one first alternative answer matching the target question from the first type of question-answer group, and determine one of the first alternative answers as the first answer; The second determining submodule is used to determine at least one second alternative answer that matches the target intent from the second type of question-and-answer group if the intent recognition result is a target intent, and to determine one of the second alternative answers as the first answer.

8. A computer readable medium having stored thereon a computer program, characterized in that When executed by the processing device, the program implements the steps of the method according to any one of claims 1-6.

9. An electronic device, comprising: include: A storage device having at least one computer program stored thereon; At least one processing device is configured to execute the at least one computer program in the storage device to implement the steps of the method according to any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-6.

Images