An intelligent outbound call method, device, equipment and storage medium

Abstract

Embodiments of the present application provide a kind of intelligent outbound method, device, equipment and storage medium.In the embodiments of the present application, after receiving the first user question issued by target user, the target question intent corresponding to the first user question is identified, if the reply answer adapted to the target question intent cannot be retrieved from the pre-set question and answer knowledge base, the first user question is marked as unreplied question and stored in question database;Based on the plurality of unreplied questions stored in question database, the question and answer knowledge base is expanded, so that the reply answer adapted to the target question intent is contained in the question and answer knowledge base after expansion.Based on this, when receiving the second user question corresponding to the target question intent, the reply answer adapted to the target question intent can be obtained from the question and answer knowledge base after expansion, to reply the second user question, effectively reduce the occurrence of the situation that user question cannot be replied, so as to improve the user experience in intelligent question and answer process.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to an intelligent outbound calling method, apparatus, device and storage medium. Background Technology

[0002] Currently, with the development of natural language processing technology, intelligent outbound calling robots have been widely used in industries such as finance, telecommunications, government, education, healthcare, and logistics. Technical personnel typically pre-configure user questions and corresponding responses for these robots based on their experience, forming a set of reply scripts. After automatically initiating outbound call tasks according to dialing rules and business scenarios, the intelligent outbound calling robot can respond to user-inputted questions based on the pre-set reply scripts.

[0003] However, the user questions and corresponding answers contained in the response scripts have certain limitations. If the intelligent outbound call robot cannot find a suitable response from the preset response scripts for the user's input question, it will be unable to respond to the user, resulting in a poor user experience. Summary of the Invention

[0004] This application provides a method, apparatus, device, and storage medium for intelligent outbound calling, which improves the user experience during intelligent question answering.

[0005] This application provides an intelligent outbound calling method, including:

[0006] In response to an outbound call instruction, an outbound call is initiated to the target user to conduct intelligent question and answer with the target user;

[0007] Receive the first user question issued by the target user during the intelligent question-answering process;

[0008] Perform a question intent recognition operation on the first user's question to obtain the target question intent corresponding to the first user's question;

[0009] Retrieve response answers that match the intent of the target question from a pre-set question-and-answer knowledge base, wherein the question-and-answer knowledge base contains a mapping relationship between question intent and response answers;

[0010] If no response matching the intent of the target question is found, the first user question is marked as an unanswered question and stored in the question database.

[0011] The question-and-answer knowledge base is expanded based on multiple unanswered questions in the question database to add response answers that are associated with the intent of the target question.

[0012] After the expansion is completed, if a second user question corresponding to the intent of the target question is received, a response answer that matches the intent of the target question is retrieved from the expanded question-and-answer knowledge base to reply to the second user question.

[0013] This application embodiment also provides an intelligent outbound calling device, including an outbound calling module, a question-and-answer module, and an expansion module;

[0014] The outbound call module is used to respond to an outbound call instruction, initiate an outbound call to the target user, and conduct intelligent question and answer with the target user; receive a first user question issued by the target user during the intelligent question and answer process; after expansion, if a second user question corresponding to the intent of the target question is received, the module uses the reply answer sent by the question and answer module that is adapted to the intent of the target question to reply to the second user question.

[0015] The question-and-answer module is used to perform a question intent recognition operation on the first user question to obtain the target question intent corresponding to the first user question; retrieve a reply answer that matches the target question intent from a preset question-and-answer knowledge base, wherein the question-and-answer knowledge base contains a mapping relationship between question intents and reply answers; for a second user question that corresponds to the target question intent, obtain a reply answer that matches the target question intent from an expanded question-and-answer knowledge base, and send the obtained reply answer that matches the target question intent to the outbound call module;

[0016] The expansion module is used to mark the first user question as an unanswered question and store the unanswered question in the question database when no reply answer matching the intent of the target question is found; and to expand the question-and-answer knowledge base based on multiple unanswered questions in the question database to add reply answers associated with the intent of the target question in the question-and-answer knowledge base.

[0017] This application also provides a computing device, including: a memory, a processor, and a communication component;

[0018] In response to an outbound call instruction, an outbound call is initiated to the target user to conduct intelligent question and answer with the target user;

[0019] The communication component receives the first user question raised by the target user during the intelligent question-answering process.

[0020] Perform a question intent recognition operation on the first user's question to obtain the target question intent corresponding to the first user's question;

[0021] Retrieve response answers that match the intent of the target question from a pre-set question-and-answer knowledge base, wherein the question-and-answer knowledge base contains a mapping relationship between question intent and response answers;

[0022] If no response matching the intent of the target question is found, the first user question is marked as an unanswered question and stored in the question database.

[0023] The question-and-answer knowledge base is expanded based on multiple unanswered questions in the question database to add response answers that are associated with the intent of the target question.

[0024] After the expansion is completed, if a second user question corresponding to the intent of the target question is received, a response answer that matches the intent of the target question is retrieved from the expanded question-and-answer knowledge base to reply to the second user question.

[0025] This application also provides a computer-readable storage medium for storing computer instructions, which, when executed by one or more processors, cause the one or more processors to perform the aforementioned intelligent outbound calling method.

[0026] This application also provides a computer program product, including a computer program / instructions; when the computer program / instructions are executed by a processor, the processor performs the aforementioned intelligent outbound calling method.

[0027] In this embodiment of the application, during the intelligent outbound call process, after receiving a first user question from a target user, the intent of the target question corresponding to the first user question can be identified. If a reply answer that matches the intent of the target question cannot be retrieved from the preset question-and-answer knowledge base, the first user question is marked as an unanswered question and stored in the question database. Based on the multiple unanswered questions stored in the question database, a reply answer associated with the intent of the target question is added to the question-and-answer knowledge base. So that when a second user question corresponding to the intent of the target question is received, a reply answer that matches the intent of the target question can be obtained from the expanded question-and-answer knowledge base to reply to the second user question.

[0028] In this way, when the first user's question cannot be answered, multiple unanswered questions from the historical outbound call records can be used to expand the question-and-answer knowledge base. This expands the knowledge base to include answers that match the target question intent of the first user's question. Based on this, the second user's question corresponding to the target question intent can be answered according to the expanded knowledge base, reducing the occurrence of situations where user questions cannot be answered and thus effectively improving the user experience in the intelligent question-and-answer process. Attached Figure Description

[0029] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0030] Figure 1 A flowchart illustrating an exemplary embodiment of this application for a smart outbound calling method;

[0031] Figure 2 A flowchart illustrating an optional implementation of an expanded question-answering knowledge base, provided as another exemplary embodiment of this application;

[0032] Figure 3 A schematic diagram illustrating the distribution of at least one cluster type, provided as another exemplary embodiment of this application;

[0033] Figure 4 A logical diagram of an intelligent outbound calling scenario is provided as another exemplary embodiment of this application;

[0034] Figure 5 A schematic diagram of the structure of an intelligent outbound calling device provided as another exemplary embodiment of this application;

[0035] Figure 6 This is a schematic diagram of the structure of a computing device provided as another exemplary embodiment of this application. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0037] Currently, when intelligent robots cannot find a suitable response from preset reply scripts, they cannot respond to users, leading to interruptions in outbound calls and a poor user experience. To improve this technical problem, this application provides a solution. The basic idea is as follows: After receiving a first user question from a target user, the intent of the first user question is identified. If a suitable response cannot be found from a preset question-and-answer knowledge base, the first user question is marked as an unanswered question and stored in a question database. Based on the multiple unanswered questions stored in the question database, a response associated with the intent of the target question is added to the question-and-answer knowledge base. This allows the robot to retrieve a suitable response from the expanded question-and-answer knowledge base when a second user question corresponding to the intent of the target question is received, thus responding to the second user question. In this way, when the first user's question cannot be answered, multiple unanswered questions from the historical outbound call records can be used to expand the question-and-answer knowledge base. This expands the knowledge base to include answers that match the target question intent of the first user's question. Based on this, the second user's question corresponding to the target question intent can be answered according to the expanded knowledge base, reducing the occurrence of situations where user questions cannot be answered and thus effectively improving the user experience in the intelligent question-and-answer process.

[0038] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0039] Figure 1 This is a flowchart illustrating an intelligent outbound calling method provided as an exemplary embodiment of this application. The method can be executed by a data processing device, which can be implemented as a combination of software and / or hardware, and can be integrated into a computing device. (Reference) Figure 1 The method includes:

[0040] Step 100: In response to the outbound call instruction, initiate an outbound call to the target user to conduct intelligent question and answer with the target user;

[0041] Step 101: Receive the first user question raised by the target user during the intelligent question-answering process;

[0042] Step 102: Perform a question intent recognition operation on the first user's question to obtain the target question intent corresponding to the first user's question;

[0043] Step 103: Retrieve response answers that match the target question intent from a preset question-and-answer knowledge base. The question-and-answer knowledge base contains a mapping relationship between question intent and response answers.

[0044] Step 104: If no response matching the target question intent is found, the first user question is marked as an unanswered question and stored in the question database;

[0045] Step 105: Expand the question-and-answer knowledge base based on multiple unanswered questions in the question database to add response answers that are related to the intent of the target question in the question-and-answer knowledge base;

[0046] Step 106: After the expansion is completed, if a second user question corresponding to the target question intent is received, a reply answer that matches the target question intent is obtained from the expanded question-answer knowledge base to reply to the second user question.

[0047] The intelligent outbound calling method provided in this embodiment can be executed by an intelligent outbound calling platform. An intelligent outbound calling platform is a dialogue initiation platform based on artificial intelligence technology, which can initiate dialogues via voice or text. The intelligent outbound calling method provided in this embodiment is applicable to various scenarios requiring intelligent outbound calling platforms to execute intelligent outbound calls. For example, in a furniture repair scenario, intelligent outbound calls can be used to confirm on-site repair times with customers; in a smart express delivery scenario, intelligent outbound calls can be used to notify customers to pick up their packages; of course, it can also be applied to scenarios such as product promotion, event notifications, and after-sales service, which are not limited in this embodiment.

[0048] In step 100, in response to an outbound call instruction, an outbound call can be initiated to the target user for intelligent question-and-answer interaction. Technical personnel can configure corresponding outbound call rules for business scenarios on the intelligent outbound call platform to generate outbound call instructions based on these rules. For example, for the business scenario of express delivery, the outbound call rule can be configured as follows: when the express delivery is in the delivery status, a phone call can be made to the recipient to inquire about their preferred delivery confirmation method. When the logistics system detects that the express delivery is in the delivery status, it sends the express delivery information to the intelligent outbound call platform. The intelligent outbound call platform can generate an outbound call instruction based on the delivery status information contained in the express delivery information. The outbound call instruction can include the contact information of the target recipient (i.e., the target user) of the express delivery, and an outbound call can be initiated to the target recipient in response to the outbound call instruction.

[0049] After initiating an outbound call to the target user, the platform can engage in a Q&A session with the user according to the Q&A rules corresponding to the business scenario. These rules can contain multiple questions requiring the target user's response, and the intelligent outbound calling platform can send these questions to the user based on these rules. Furthermore, the Q&A rules are linked to a Q&A knowledge base, which contains appropriate responses tailored to various question intents. The intelligent outbound calling platform can then respond to the target user's questions based on this knowledge base.

[0050] In the process of responding to questions raised by target users, the system can receive the first user question raised by the target user during the intelligent question-and-answer process. The first user question can be in the form of voice or text. A question intent recognition operation is performed on the first user question to obtain the target question intent corresponding to the first user question. For example, an intent recognition model can be used to identify the target question intent of the first user question, or the semantic features of the first user question can be extracted first, and then the target question intent can be combined based on the extracted semantic features. This embodiment does not limit this approach.

[0051] Based on this, responses matching the intent of the target question can be retrieved from a pre-defined question-and-answer knowledge base. During this process, the following two scenarios may occur:

[0052] In one scenario, if a response that matches the intent of the target question can be retrieved from the question-and-answer knowledge base, the retrieved response will be provided directly to the target user to answer the first user question raised by the target user.

[0053] In another scenario, if a response matching the target question's intent cannot be retrieved from the question-and-answer knowledge base, the first user's question can be marked as an unanswered question, and the marked first user's question can be stored in the question database. The question database contains unanswered questions from multiple business scenarios, with multiple unanswered questions in each scenario. These unanswered questions can originate from the same user or different users. Alternatively, a separate question database can be established for each business scenario, and the first user's question can be marked as an unanswered question in the history of intelligent outbound calls, without storing the marked first user's question in the question database. In this case, in step 105, the question-and-answer knowledge base can be expanded based on the multiple unanswered questions in the question database to add response answers associated with the target question's intent.

[0054] Optionally, the number of unanswered questions stored in the question database can be monitored. When the number of unanswered questions detected is greater than or equal to a preset expansion threshold, the question-and-answer knowledge base is expanded based on multiple unanswered questions in the question data. Alternatively, when the number of unanswered questions detected is less than the preset expansion threshold, and a second user question corresponding to the target question intent is received, the question-and-answer knowledge base is expanded based on multiple unanswered questions in the question data.

[0055] By monitoring the number of unanswered questions in the question database in real time, the question-and-answer knowledge base can be periodically expanded. This allows for the addition of new question intents and associated answers, thus resolving the issue of not being able to retrieve answers matching the target question intent from the knowledge base. Furthermore, if a second user question corresponding to the target question intent is received, the knowledge base will be expanded even if the preset expansion time has not been reached. This serves as a fallback strategy, effectively preventing multiple instances of unanswerable questions for the same intent, thereby improving the user experience during the intelligent question-and-answer process.

[0056] Based on this, the expanded question-and-answer knowledge base includes new responses associated with the intent of the target question. After the expansion is completed, if a second user question corresponding to the intent of the target question is received, a response matching the intent of the target question is retrieved from the expanded question-and-answer knowledge base, and the retrieved response is used to answer the second user question.

[0057] In this embodiment, the first user question received during the intelligent outbound call process can be subjected to question intent recognition to obtain the target question intent corresponding to the first user question. A response matching the target question intent is then retrieved from a preset question-and-answer knowledge base. If no response matching the target question intent is found, the first user question is marked as an unanswered question and stored in a question database. The question-and-answer knowledge base is then expanded based on the multiple unanswered questions stored in the question database. The expanded question-and-answer knowledge base includes response answers associated with the target question intent. This allows for the retrieval of a response matching the target question intent from the expanded question-and-answer knowledge base when a second user question corresponding to the target question intent is received, thus significantly reducing the occurrence of situations where user questions cannot be answered. This effectively lowers the rejection rate of intelligent outbound calls, thereby further improving the user experience during the intelligent question-and-answer process.

[0058] In the above or below embodiments, various implementation methods can be used to expand the question-answering knowledge base based on multiple unanswered questions in the question database. Figure 2 A flowchart illustrating an optional implementation of an expanded question-answering knowledge base, provided as another exemplary embodiment of this application, is shown below. Figure 2 The steps in this optional implementation may include:

[0059] Step 200: Cluster multiple unanswered questions into at least one cluster.

[0060] Step 201: Determine the problem intent corresponding to at least one cluster;

[0061] Step 202: For any cluster in at least one cluster, obtain the response answer configured for the question intent corresponding to the cluster;

[0062] Step 203: Establish the mapping relationship between the intent of the question and the answer;

[0063] Step 204: Add the mapping relationship to the question-and-answer knowledge base to expand the question-and-answer knowledge base; wherein, the mapping relationship added to the question-and-answer database includes the mapping relationship between the target question intent and the response answer.

[0064] In step 201, multiple unanswered questions can first be classified to obtain the question types corresponding to each unanswered question. Then, according to the types corresponding to each unanswered question, the multiple unanswered questions can be clustered into at least one cluster. Alternatively, multiple unanswered questions can be clustered into at least one cluster based on the semantic features corresponding to each unanswered question. Of course, these are just examples, and this embodiment is not limited to them. Other methods can also be used to cluster multiple unanswered questions into at least one cluster.

[0065] If multiple unanswered questions are clustered into at least one cluster based on their respective semantic features, then each unanswered question can be input into a semantic feature extraction model. This model can then perform semantic feature extraction on the unanswered questions to obtain their respective semantic features. For example, the M3E (Multi-Modal Metric-Embedding Model) can be used as the semantic feature extraction model. The M3E model supports multimodal input, can process and fuse information of different data types, and can simultaneously handle unanswered questions in audio, text, image, or other formats without requiring different semantic feature extraction models for different data types. This unifies the standard for semantic feature extraction and improves its accuracy. Of course, models such as Word2Vec, GloVe (Global Vectors for Word Representation), and CLIP (Contrastive Language-Image Pre-training) can also be used as semantic feature extraction models; this embodiment does not limit the specific model used.

[0066] Based on this, multiple unanswered questions can be clustered according to their respective semantic features to obtain at least one cluster and at least one semantic feature group corresponding to each cluster. During this process, the similarity between any two semantic features can be calculated. If the similarity between two semantic features is higher than a preset similarity threshold, these two semantic features are assigned to the same semantic feature group. Furthermore, the unanswered questions corresponding to the semantic features in the same semantic feature group are assigned to the same cluster.

[0067] In this embodiment, the algorithm used to cluster multiple unanswered questions can be k-means clustering, density-based spatial clustering of applications with noise (DBSCAN), or the balanced iterative reducing and clustering using hierarchies (BIRCH) algorithm. Among these, density-based spatial clustering (DBSCAN) is preferred to cluster multiple unanswered questions into at least one cluster. This is because DBSCAN does not require specifying the number of clusters; it can determine the optimal number of clusters based on the actual situation without being limited by the number of clusters. Furthermore, it effectively handles noise in the clustering data, eliminating the influence of noise on the clustering results, thereby significantly improving the accuracy of the clustering results.

[0068] Before clustering multiple unanswered questions, the semantic features corresponding to each unanswered question can be dimensionality-reduced to obtain dimensionality-reduced semantic features. Based on these dimensionality-reduced semantic features, the multiple unanswered questions are then clustered into at least one cluster. In this embodiment, the algorithms used for dimensionality reduction include, but are not limited to, Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), and t-distributed Stochastic Neighbor Embedding (t-SNE). No specific limitations are imposed on the algorithms used for dimensionality reduction. By dimensionality-reducing the semantic features, high-dimensional data is transformed into a low-dimensional space for easier analysis and visualization, while preserving as much important information as possible from the original data. This effectively reduces the computational resources required for clustering, thereby improving clustering efficiency. Furthermore, it more clearly displays the relationships between different clusters, resulting in better visualization of the clustering results.

[0069] Figure 3 This is a schematic diagram illustrating the distribution of at least one cluster type, provided as an exemplary embodiment of this application. (Reference) Figure 3 Based on the semantic features after dimensionality reduction using the t-SNE algorithm, multiple unanswered questions can be clustered into several clear clusters. Each cluster is represented by a different shape or color to clearly show the distribution of different clusters.

[0070] Continue to refer to Figure 2 In step 202, for any cluster within at least one cluster, at least one common feature can be extracted from the target semantic feature set corresponding to that cluster. The target semantic feature set contains semantic features corresponding to at least one unanswered question, and each unanswered question may correspond to one or more semantic features. When extracting the common feature, the intersection of the semantic features corresponding to any two unanswered questions can be taken, and the semantic features located in the intersection can be used as the common feature. For example, the semantic features corresponding to unanswered question A include a, b, and c, and the semantic features corresponding to unanswered question B include a, b, and d. Taking the intersection of these two semantic features yields the common features a and b.

[0071] However, determining common features by taking the intersection might miss some not-quite-identical similar semantic features. To improve this, common features can be determined by calculating similarity: calculate the semantic similarity between the semantic features corresponding to any two unanswered questions, and identify the semantic features with a similarity greater than a specified threshold as common features. For example, the semantic features corresponding to unanswered question A include a1, b1, and c1, and the semantic features corresponding to unanswered question B include a1, b2, and d1. Calculating the similarity between any two semantic features, only the similarity between a1 and a1, and between b1 and b2, is greater than the specified threshold. Therefore, the common features can be determined as a1, b1, and b2.

[0072] For any given cluster, after extracting at least one common feature corresponding to that cluster, these features can be combined to obtain the question intent corresponding to that cluster. In one optional implementation, the weight value of each common feature in the target semantic feature group can be calculated. Common features with weight values ​​greater than a preset threshold are combined to obtain the question intent corresponding to the cluster. In this way, only the common features that appear most frequently in the target feature group are used as components of the question intent, while less frequent common features are discarded. This eliminates some common features determined by chance, thereby effectively improving the accuracy of the question intent.

[0073] In one optional implementation for calculating weights for common features, for any common feature in a cluster, the number N of unanswered questions in that cluster can be counted, and the frequency n of the common feature appearing in the semantic features corresponding to each of the N unanswered questions can be calculated. The ratio n / N between the frequency of the common feature and the number of unanswered questions is used as the weight of the common feature in the target semantic feature group. It should be noted that each unanswered question can correspond to one or more unique semantic features. Of course, other implementations can also be used to calculate weights for common features, which will not be detailed here.

[0074] For example, a cluster contains 10 unanswered questions. At least one common feature extracted from the target semantic feature group corresponding to this cluster is "venue", "time", "time", "open", and "go". Calculating the proportion of the common feature in the target semantic feature group is essentially calculating which of the 10 unanswered questions contains the common feature in its semantic feature. We can obtain the weight value of "venue" as 9 / 10, "time" as 8 / 10, "time" as 2 / 10, "open" as 9 / 10, and "go" as 2 / 10. Based on this, we can filter out the common features with weight values ​​greater than a preset threshold as "venue", "time", and "open". By combining the filtered common features, we can infer that the question intent corresponding to this cluster is to inquire about the opening time of the venue.

[0075] After obtaining the question intent corresponding to each cluster, the question intent can be displayed to technical personnel, who can then perform answer configuration operations for the question intent corresponding to the cluster. In response to the answer configuration operations performed by the technical personnel, a corresponding response strategy can be obtained. The data format of the response strategy may include, but is not limited to, regular expressions, configuration text, flowcharts, and data tables, etc., which are not limited in this embodiment. The obtained response strategy is parsed to obtain at least one candidate response answer associated with the question intent corresponding to the cluster, and this candidate response answer is used as the response answer adapted to the question intent corresponding to the cluster. In this way, when configuring response answers for question intents, the various question types covered by the question intents are considered, and candidate response answers are configured separately for each question type, enabling more accurate responses to question intents.

[0076] For example, if the question is about how much a used car can be sold for, the response might vary depending on the brand and condition of the car. Therefore, the response strategy for configuring the answer to this question could include multiple candidate answers, such as the price range for a specific brand of used car, the price range for a 90% new used car, and the price range for a 50% new used car. These candidate answers can then be used as the response configured for this question.

[0077] In step 204, a mapping relationship between question intents and response answers can be established and added to the question-and-answer knowledge base to expand it. The question-and-answer knowledge base can store question intents and response answers in list form, and the mapping relationship between question intents obtained from processing unanswered questions and their response answers can be added to the list in the question-and-answer knowledge base. In addition, intent categories and associated corpora can be configured for question intents in the question-and-answer knowledge base; these corpora are used to assist in retrieving question intents that match the user's question.

[0078] Accordingly, in this embodiment, cluster analysis is performed on the unanswered data to determine the question intent corresponding to the unanswered question, and the question intent and corresponding target answer are added to the question-and-answer knowledge base to expand the question-and-answer knowledge base. Based on the expanded question-and-answer knowledge base, the occurrence of the situation where the question intent and / or answer that matches the user's question cannot be retrieved can be reduced, thereby effectively reducing the rejection rate in the intelligent question-and-answer process and greatly improving the user experience.

[0079] In the above or following embodiments, unanswered questions stored in the question database can be categorized. When selecting multiple unanswered questions to expand the question-and-answer knowledge base, only unanswered questions included in one of the categorization results can be selected to expand the knowledge base. For example, the unanswered questions stored in the question database can be categorized according to the question-and-answer stage in the intelligent question-and-answer process. In this case, when selecting multiple unanswered questions to expand the knowledge base, the target process node to which the first user's question belongs can be determined first. The target process node can be used to characterize the question-and-answer stage in which the first user's question is in the intelligent question-and-answer process. The target process node may include, but is not limited to, opening node, identity confirmation node, guiding action node, and closing node. Then, multiple unanswered questions associated with the target process node are queried from the question database, and the queried multiple unanswered questions associated with the target process node are used as multiple unanswered questions to expand the question-and-answer knowledge base. Of course, the unanswered questions stored in the question database can also be categorized according to the time when they were stored in the question database. This embodiment does not limit this.

[0080] In this way, unanswered questions stored in the question database are classified into unanswered questions located in different question-and-answer stages. Unanswered data generated in each question-and-answer stage can be clustered separately. The semantic similarity within the same question-and-answer stage is higher, allowing for more granular clustering. This eliminates interference from other question-and-answer stages in the intelligent question-and-answer process, improving the accuracy of clustering and, consequently, the precision of the question intent determined for the clusters.

[0081] Furthermore, multiple mapping tables corresponding to different process nodes can be configured in the question-and-answer knowledge base. These tables store the mapping relationships between question intents and response answers applicable to different question-and-answer stages. Based on this, when expanding the question-and-answer knowledge base based on multiple unanswered questions under a target process node, the target mapping table corresponding to the target process node can be searched within the knowledge base, and the response answers associated with the target question intent can be added to the target mapping table, thus expanding the knowledge base. Storing response answers associated with question intents according to process nodes in the knowledge base allows for the categorization and arrangement of mapping relationships. When a response answer matching the target question intent needs to be retrieved later, the required response answer can be quickly retrieved from the knowledge base, effectively improving the efficiency of intelligent outbound calling.

[0082] Figure 4 This is a logical diagram illustrating an intelligent outbound calling scenario, provided as an exemplary embodiment of this application. (Reference) Figure 4 The intelligent outbound calling scenario includes an intelligent outbound calling platform, a user client, and a question-and-answer database. The intelligent outbound calling platform communicates with both the user client and the question-and-answer database. The intelligent outbound calling platform has a question database, and the question-and-answer knowledge base has at least one pre-set mapping relationship between "question intent and response answer." The user client can contain multiple users.

[0083] The intelligent outbound calling platform can trigger outbound calling instructions according to preset outbound calling rules and respond to these instructions to initiate outbound calls to target users on the user end for intelligent question-and-answer communication. During the intelligent question-and-answer process, the platform can receive the first user question posed by the target user and perform intent recognition on it to identify the target intent. Based on this, the platform can request a response matching the target intent from the question-and-answer knowledge base. If it receives a search result from the knowledge base containing "not found," it marks the first user question as an unanswered question and stores the marked question in the question database.

[0084] Based on this, the intelligent outbound calling platform can cluster multiple unanswered questions in the question database into at least one cluster, determine the question intent corresponding to each cluster, obtain the response answer configured for the question intent corresponding to each cluster, and establish a mapping relationship between the question intent and the response answer. This mapping relationship is then sent to a question-and-answer knowledge base. The question-and-answer knowledge base can save the mapping relationship sent by the intelligent outbound calling platform, which includes the mapping relationship between the target question intent and the response answer. When the intelligent outbound calling platform receives a second user question corresponding to the target question intent, it can request the question-and-answer knowledge base to retrieve a response answer that matches the target question intent, and send the response answer returned by the question-and-answer knowledge base to the user to respond to the second user question.

[0085] Accordingly, in this embodiment, by performing cluster analysis on the unanswered data, the question intent corresponding to each cluster is determined, and the question intent and corresponding target answer are added to the question-and-answer knowledge base to expand the question-and-answer knowledge base. Based on the expanded question-and-answer knowledge base, the occurrence of situations where no answer matching the user's question intent can be found can be reduced, thereby effectively reducing the rejection rate in the intelligent question-and-answer process and further improving the user experience.

[0086] It should be noted that the execution subject of each step of the method provided in the above embodiments can be the same device, or the method can be executed by different devices. For example, the execution subject of steps 100 to 103 can be device A; or the execution subject of steps 101 and 102 can be device A, and the execution subject of step 103 can be device B; and so on.

[0087] Furthermore, some processes described in the above embodiments and accompanying drawings include multiple operations appearing in a specific order. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or they may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not represent any execution order. Additionally, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first" and "second" in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to different types.

[0088] Figure 5 This is a schematic diagram of the structure of an intelligent outbound calling device provided as another exemplary embodiment of this application. (See diagram below.) Figure 5 As shown, the intelligent outbound calling device 50 may include an outbound calling module 51, a question-and-answer module 52, and an expansion module 53. Wherein:

[0089] Outbound call module 51 is used to respond to outbound call instructions, initiate outbound calls to target users, and conduct intelligent question and answer with target users; receive the first user question issued by the target user during the intelligent question and answer process; after expansion, if a second user question corresponding to the intent of the target question is received, the second user question is answered using the reply answer sent by question and answer module 52 that is adapted to the intent of the target question.

[0090] The question-and-answer module 52 is used to perform question intent recognition on the first user's question to obtain the target question intent corresponding to the first user's question; retrieve the corresponding answer from the preset question-and-answer knowledge base, which contains the mapping relationship between question intent and answer; for the second user's question corresponding to the target question intent, obtain the corresponding answer from the expanded question-and-answer knowledge base, and send the obtained corresponding answer to the outbound call module 51.

[0091] The expansion module 53 is used to mark the first user question as an unanswered question and store the unanswered question in the question database when no reply answer matching the target question intent is found; and to expand the question-and-answer knowledge base based on multiple unanswered questions in the question database to add reply answers related to the target question intent in the question-and-answer knowledge base.

[0092] It is worth noting that the technical details of the above embodiments of the intelligent outbound calling device can be found in the descriptions of the operation of the data processing device or the intelligent outbound calling platform in the aforementioned embodiments of the intelligent outbound calling method. To save space, these details will not be repeated here, but this should not cause any loss to the scope of protection of this application.

[0093] This application also provides a computer program product, including a computer program. When the computer program is executed by a processor, it can implement the steps in the above-described intelligent outbound calling method. It should be noted that the technical solution of this computer program and the technical solution of the above-described intelligent outbound calling method belong to the same concept. Details not described in detail in the technical solution of the computer program can be found in the description of the technical solution of the above-described intelligent outbound calling method.

[0094] Figure 6 This is a schematic diagram of the structure of a computing device provided as another exemplary embodiment of this application. For example... Figure 6 As shown, the computing device includes: a memory 60, a processor 61, and a communication component 62.

[0095] Processor 61, coupled to memory 60, is used to execute computer programs in memory 60 for:

[0096] In response to an outbound call instruction, initiate an outbound call to the target user to conduct intelligent question and answer with the target user;

[0097] The communication component 62 receives the first user question posed by the target user during the intelligent question-answering process.

[0098] Perform a question intent recognition operation on the first user's question to obtain the target question intent corresponding to the first user's question;

[0099] Retrieve responses that match the intent of the target question from a pre-defined question-and-answer knowledge base. The question-and-answer knowledge base contains a mapping relationship between question intent and response responses.

[0100] If no response matching the intent of the target question is found, the first user's question is marked as an unanswered question and stored in the question database;

[0101] The question-and-answer knowledge base is expanded based on multiple unanswered questions in the question database to add response answers that are related to the intent of the target question.

[0102] After the expansion is completed, if a second user question corresponding to the intent of the target question is received, a response answer that matches the intent of the target question is retrieved from the expanded question-and-answer knowledge base to reply to the second user question.

[0103] In an optional embodiment, when expanding the question-and-answer knowledge base based on multiple unanswered questions in the question database, the processor 61 is further configured to:

[0104] Monitor the number of unanswered questions stored in the issue database;

[0105] When the number of unanswered questions detected is greater than or equal to a preset expansion threshold, the question-and-answer knowledge base is expanded based on multiple unanswered questions in the question data; or,

[0106] When the number of unanswered questions detected is less than the preset expansion threshold and a second user question corresponding to the target intent is received, the question-and-answer knowledge base is expanded based on the multiple unanswered questions in the question data.

[0107] In an optional embodiment, before expanding the question-and-answer knowledge base based on multiple unanswered questions in the question database, the processor 61 is further configured to:

[0108] Determine the target process node to which the first user question belongs. The target process node is used to represent the question-and-answer stage of the first user question in the intelligent question-and-answer process.

[0109] From the issue database, query multiple unanswered issues associated with the target process node;

[0110] Multiple unanswered questions associated with the target process node will be used as additional unanswered questions to expand the question-and-answer knowledge base.

[0111] In an optional embodiment, during the process of adding a response answer associated with the intent of the target question in the question-and-answer knowledge base, the processor 61 is further configured to:

[0112] In the question-and-answer knowledge base, find the target mapping relationship table corresponding to the target process node;

[0113] Add the responses that are associated with the target question's intent to the target mapping table.

[0114] In an optional embodiment, during the process of expanding the question-and-answer knowledge base based on multiple unanswered questions stored in the question database, to add response answers associated with the intent of the target question in the question-and-answer knowledge base, the processor 61 is further configured to:

[0115] Cluster multiple unanswered questions into at least one cluster;

[0116] Determine the problem intent corresponding to at least one cluster;

[0117] For any cluster in at least one cluster, obtain a response that matches the question intent corresponding to the cluster;

[0118] Establish a mapping relationship between the intent of a question and the answer to the question;

[0119] Add mapping relationships to the question-and-answer knowledge base to expand it. The mapping relationships added to the question-and-answer knowledge base include the mapping relationship between the target question meaning and the response answer.

[0120] In an optional embodiment, during the process of clustering multiple unanswered questions into at least one cluster, processor 61 is further configured to:

[0121] Multiple unanswered questions are input into the semantic feature extraction model;

[0122] Using a semantic feature extraction model, semantic feature extraction operations are performed on multiple unanswered questions to obtain the semantic features corresponding to each unanswered question.

[0123] Based on the semantic features corresponding to each unanswered question, multiple unanswered questions are clustered to obtain at least one cluster and at least one semantic feature group corresponding to each cluster.

[0124] In an alternative embodiment, during the process of determining the problem intent corresponding to at least one cluster, processor 61 is further configured to:

[0125] For any cluster in at least one cluster, extract at least one common feature from the target semantic feature group corresponding to the cluster;

[0126] Calculate the weight value of each common feature in the target semantic feature group;

[0127] Common features whose weight values ​​exceed a preset threshold are combined to obtain the problem intent corresponding to the cluster.

[0128] In an optional embodiment, during the process of obtaining a response answer that matches the question intent corresponding to the cluster, the processor 61 is further configured to:

[0129] In response to the answer configuration operation for the question intent corresponding to the cluster, obtain the response strategy corresponding to the answer configuration operation;

[0130] The response strategy is analyzed to obtain at least one candidate response answer that is associated with the question intent corresponding to the cluster cluster;

[0131] At least one candidate response is selected as the response that matches the question intent corresponding to the cluster.

[0132] Furthermore, such as Figure 6 As shown, the computing device also includes other components such as a display 63, a power supply component 64, and an audio component 65. Figure 6 The diagram only shows some components and does not mean that the computing device includes only these components. Figure 6 The components shown.

[0133] It is worth noting that the technical details of the above embodiments of the computing device can be found in the descriptions of the actions of the computing device in the aforementioned embodiments of the intelligent outbound calling method. To save space, these details will not be repeated here, but this should not cause any loss to the scope of protection of this application.

[0134] Accordingly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed, can perform the steps that can be executed by a computing device in the above method embodiments.

[0135] The above Figure 6The memory in a computer is used to store computer programs and can be configured to store various other data to support operation on a computing platform. Examples of this data include instructions for any application or method operating on the computing platform, contact data, phone book data, messages, pictures, videos, etc. The memory can be implemented from any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disks, or optical disks.

[0136] The above Figure 6 The communication component is configured to facilitate wired or wireless communication between the device containing the communication component and other devices. The device containing the communication component can access wireless networks based on communication standards, such as WiFi, 2G, 3G, 4G / LTE, 5G, or combinations thereof. In one exemplary embodiment, the communication component receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, the communication component further includes a Near Field Communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID), Infrared Data Association (IrDA) technology, Ultra-Wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0137] The above Figure 6 The display includes a screen, which may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen can be implemented as a touchscreen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors can sense not only the boundaries of the touch or swipe action, but also the duration and pressure associated with the touch or swipe operation.

[0138] The above Figure 6 The power supply component provides power to the various components of the device in which it resides. The power supply component may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device in which it resides.

[0139] The above Figure 6The audio component can be configured to output and / or input audio signals. For example, the audio component includes a microphone (MIC) configured to receive external audio signals when the device containing the audio component is in an operating mode, such as call mode, recording mode, or voice recognition mode. The received audio signals can be further stored in memory or transmitted via a communication component. In some embodiments, the audio component also includes a speaker for outputting audio signals.

[0140] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0141] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0142] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0143] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0144] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0145] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0146] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0147] It should also be noted that the terms "comprising," "including," or any other variations thereof are 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. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0148] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. An intelligent outbound calling method, characterized by, The method comprises the following steps: in response to an outbound call instruction, initiating an outbound call to a target user to conduct an intelligent question and answer with the target user; receiving a first user question issued by the target user in the intelligent question and answer process; performing a question intent recognition operation on the first user question to obtain a target question intent corresponding to the first user question; retrieving a reply answer adapted to the target question intent from a preset question and answer knowledge base, the question and answer knowledge base comprising a mapping relationship between question intents and reply answers; if a reply answer adapted to the target question intent is not retrieved, marking the first user question as an un-replied question and storing the un-replied question in a question database; determining a target process node to which the first user question belongs, the target process node being used to represent a question and answer stage in which the first user question is located in the intelligent question and answer process; querying a plurality of un-replied questions associated with the target process node from the question database; and taking the plurality of un-replied questions associated with the target process node as a plurality of un-replied questions used to expand the question and answer knowledge base; expanding the question and answer knowledge base based on the plurality of un-replied questions associated with the target process node in the question database to add a new question intent and a reply answer associated with the new question intent in the question and answer knowledge base; after the expansion is completed, if a second user question corresponding to the new question intent is received, obtaining a reply answer adapted to the new question intent from the expanded question and answer knowledge base to reply to the second user question; wherein the process of expanding the question and answer knowledge base comprises: obtaining semantic features corresponding to each of the un-replied questions associated with the target process node by using a semantic feature extraction model and performing dimension reduction processing to obtain dimension-reduced semantic features corresponding to each of the un-replied questions; dividing dimension-reduced semantic features with a similarity higher than a first threshold value between the dimension-reduced semantic features corresponding to each of the un-replied questions into a same semantic feature group, and dividing un-replied questions corresponding to the dimension-reduced semantic features in the same semantic feature group into a same clustering cluster to obtain at least one clustering cluster and a semantic feature group corresponding to each of the at least one clustering cluster; calculating a weight value of each common feature in a target semantic feature group corresponding to any clustering cluster, the common feature being a dimension-reduced semantic feature with a semantic similarity greater than a second threshold value between any two dimension-reduced semantic features in the target semantic feature group; combining common features with a weight value exceeding a third threshold value to obtain a question intent corresponding to the any clustering cluster; analyzing a reply strategy corresponding to the obtained question intent, taking at least one candidate reply answer obtained by the analysis as a reply answer adapted to the question intent corresponding to the any clustering cluster; and adding a mapping relationship between the question intent and the reply answer to the question and answer knowledge base to expand the question and answer knowledge base.

2. The method of claim 1, wherein, The expansion of the question and answer knowledge base based on the plurality of un-replied questions in the question database comprises: monitoring a number of un-replied questions stored in the question database; when the number of un-replied questions is greater than or equal to a preset expansion threshold, expanding the question and answer knowledge base based on a plurality of un-replied questions in the question database; or when the number of un-replied questions is less than the preset expansion threshold and a second user question corresponding to the target question intent is received, expanding the question and answer knowledge base based on a plurality of un-replied questions in the question database.

3. The method of claim 1, wherein, The question and answer knowledge base comprises a plurality of mapping relationship tables corresponding to respective process nodes, and the adding of the reply answer associated with the target question intent in the question and answer knowledge base comprises: in the question and answer knowledge base, finding a target mapping relationship table corresponding to the target process node; adding the reply answer associated with the target question intent to the target mapping relationship table.

4. The method of claim 1, wherein, adding the mapping relationship between the question intent and the reply answer to the question and answer knowledge base to expand the question and answer knowledge base, comprising: establishing a mapping relationship between the question intent and the reply answer; adding the mapping relationship to the question and answer knowledge base to expand the question and answer knowledge base, wherein the mapping relationship comprises a mapping relationship between a target question intent and a reply answer.

5. The method of claim 4, wherein, The obtaining of the reply answer adapted to the question intent corresponding to the clustering cluster comprises: in response to an answer configuration operation for the question intent corresponding to any clustering cluster, obtaining a reply strategy corresponding to the answer configuration operation; parsing the reply strategy to obtain at least one candidate reply answer associated with the question intent corresponding to any clustering cluster; the at least one candidate reply answer is used as a reply answer adapted to the question intent corresponding to any clustering cluster.

6. An intelligent outbound device, characterized by comprising an outbound call module, a question and answer module and an expansion module; The outbound call module is configured to initiate an outbound call to a target user in response to an outbound call instruction to perform intelligent question and answer with the target user, and receive a first user question issued by the target user in the intelligent question and answer process. After the expansion is completed, if a second user question corresponding to a target question intent is received, a reply answer adapted to the target question intent sent by the question and answer module is used to reply to the second user question; The question and answer module is configured to perform question intent identification on the first user question to obtain a target question intent corresponding to the first user question; retrieve a reply answer adapted to the target question intent from a preset question and answer knowledge base, wherein the question and answer knowledge base comprises a mapping relationship between a question intent and a reply answer; for a second user question corresponding to the target question intent, obtain a reply answer adapted to the target question intent from the expanded question and answer knowledge base, and send the obtained reply answer adapted to the target question intent to the outbound call module; The expansion module is configured to, in the case that no reply answer adapted to the target question intent is retrieved, mark the first user question as an un-replied question, and store the un-replied question in a question database. determine a target process node to which the first user question belongs, the target process node being used to represent a question and answer stage in which the first user question is located in an intelligent question and answer process; query a plurality of un-replied questions associated with the target process node from the question database; and take the plurality of un-replied questions associated with the target process node as a plurality of un-replied questions used to expand the question and answer knowledge base, the plurality of un-replied questions in the question and answer knowledge base being different questions; expand the question and answer knowledge base based on the plurality of un-replied questions associated with the target process node in the question database, so as to add a new question intent and a reply answer associated with the new question intent in the question and answer knowledge base; wherein the process of expanding the question and answer knowledge base comprises: obtaining semantic features corresponding to each of the un-replied questions associated with the target process node by using a semantic feature extraction model and performing dimension reduction processing, so as to obtain dimension-reduced semantic features corresponding to each of the un-replied questions; dividing dimension-reduced semantic features between which a similarity is higher than a first threshold to a same semantic feature group, and dividing un-replied questions corresponding to the dimension-reduced semantic features in the same semantic feature group to a same clustering cluster, so as to obtain at least one clustering cluster and a semantic feature group corresponding to each of the at least one clustering cluster; calculating a weight value of each common feature in a target semantic feature group corresponding to any clustering cluster, the common feature being a dimension-reduced semantic feature between which a semantic similarity in the target semantic feature group is greater than a second threshold; combining common features whose weight values exceed a third threshold, so as to obtain a question intent corresponding to the any clustering cluster; analyzing a reply strategy corresponding to the question intent obtained, taking at least one candidate reply answer obtained by the analysis as a reply answer adapted to the question intent corresponding to the any clustering cluster; and adding a mapping relationship between the question intent and the reply answer to the question and answer knowledge base, so as to expand the question and answer knowledge base.

7. A computing device, comprising: comprises a memory, a processor and a communication component; the memory is configured to store one or more computer instructions; the processor is coupled with the memory and the communication component, and is configured to execute the one or more computer instructions, so as to: initiate an outbound call to a target user in response to an outbound call instruction, so as to perform an intelligent question and answer with the target user; receive a first user question issued by the target user in the intelligent question and answer process through the communication component; perform a question intent recognition operation on the first user question, so as to obtain a target question intent corresponding to the first user question; retrieve a reply answer adapted to the target question intent from a preset question and answer knowledge base, the question and answer knowledge base comprising a mapping relationship between question intents and reply answers; if no reply answer adapted to the target question intent is retrieved, mark the first user question as an un-replied question, and store the un-replied question in a question database. determining a target process node to which the first user question belongs, the target process node being used to represent a question-answering stage in which the first user question is located in an intelligent question-answering process; querying, from the question database, a plurality of unreplied questions associated with the target process node, and taking the plurality of unreplied questions associated with the target process node as a plurality of unreplied questions used to expand the question-answering knowledge base; expanding the question-answering knowledge base based on the plurality of unreplied questions associated with the target process node in the question database, so as to add a new question intent and a reply answer associated with the new question intent in the question-answering knowledge base; after the expansion is completed, if a second user question corresponding to the target question intent is received, obtaining a reply answer adapted to the target question intent from the expanded question-answering knowledge base, so as to reply to the second user question; wherein the process of expanding the question-answering knowledge base comprises: obtaining semantic features corresponding to each of the unreplied questions associated with the target process node by using a semantic feature extraction model and performing dimension reduction processing, so as to obtain dimension-reduced semantic features corresponding to each of the unreplied questions; dividing dimension-reduced semantic features between which a similarity is higher than a first threshold to a same semantic feature group, and dividing unreplied questions corresponding to the dimension-reduced semantic features in the same semantic feature group to a same clustering cluster, so as to obtain at least one clustering cluster and a semantic feature group corresponding to each of the at least one clustering cluster; calculating a weight value of each common feature in a target semantic feature group corresponding to any clustering cluster, the common feature being a dimension-reduced semantic feature between which a semantic similarity is greater than a second threshold in the target semantic feature group; combining common features whose weight values exceed a third threshold, so as to obtain a question intent corresponding to the any clustering cluster; analyzing a reply strategy corresponding to the question intent obtained, taking at least one candidate reply answer obtained by the analysis as a reply answer adapted to the question intent corresponding to the any clustering cluster; and adding a mapping relationship between the question intent and the reply answer to the question-answering knowledge base, so as to expand the question-answering knowledge base.

8. A computer readable storage medium storing computer instructions, wherein, The computer instructions, when executed by one or more processors, cause the one or more processors to perform the intelligent outbound call method of any one of claims 1-5.

Images