Question separation method and device, equipment and storage medium

Abstract

The invention relates to the field of financial science and technology, in particular to the field of artificial intelligence, and particularly relates to a question separation method, device and equipment and a storage medium. The method comprises the steps that target phrases are traversed and matched in a domain knowledge tree to determine knowledge leaf node weights corresponding to the target phrases; in response to the fact that the number of the knowledge leaf node weights is at least two and the ratio of the maximum knowledge leaf node weight to the minimum knowledge leaf node weight is in a preset ratio range, matching the target word group in the syntactic tree to determine a syntactic leaf node weight corresponding to the target word group; responding to the condition that the number of the syntax leaf node weights is at least two and the ratio of the maximum syntax leaf node weight to the minimum syntax leaf node weight is in a preset ratio range, determining that at least two single questions exist in the customer questions; syntactic matching nodes obtained by matching the target phrases in the syntactic tree are determined, and client questions are separated based on the syntactic matching nodes to obtain single questions. The method and the device are convenient for improving the answering accuracy of the customer questions.

Description

Problem separation methods, apparatus, equipment and storage media Technical Field

[0001] This application relates to the field of financial technology, specifically to the field of artificial intelligence, and in particular to a problem separation method, apparatus, device, and storage medium. Background Technology

[0002] Within the banking industry, question-answering paradigms based on knowledge retrieval enhancement models have provided excellent assistance in resolving various inquiries from frontline customers, effectively reducing the costs associated with manual consultation. However, when constructing a knowledge base for a knowledge retrieval enhancement model, whether using question-answer pairs or breaking documents down into text blocks, the corresponding answers always focus on a single, specific question.

[0003] However, in real-world use cases, user questions are diverse. A user may ask multiple questions in a single conversation. When the knowledge retrieval enhancement model vectorizes these multiple questions and matches them with data in the knowledge base, it significantly reduces the accuracy of its answers to customer questions. Summary of the Invention

[0004] To improve the accuracy of answers to customer questions, this application provides a question separation method, apparatus, device, and storage medium.

[0005] Firstly, this application provides a problem separation method, including:

[0006] The target phrases in the customer's question are traversed and matched in the domain knowledge tree to determine the weight of the knowledge leaf node corresponding to the target phrases.

[0007] In response to the fact that the number of knowledge leaf node weights is at least two, and the ratio of the largest knowledge leaf node weight to the smallest knowledge leaf node weight is within a preset ratio range, the target word group is traversed and matched in the syntax tree to determine the syntax leaf node weight corresponding to the target word group.

[0008] In response to the fact that the number of the syntactic leaf node weights is at least two, and the ratio of the largest syntactic leaf node weight to the smallest syntactic leaf node weight is within the preset ratio range, it is determined that there are at least two individual problems in the customer problem.

[0009] The target phrase is identified by traversing and matching it in the syntax tree to obtain the syntax matching node. Based on the syntax matching node, the customer question is separated to obtain the individual question.

[0010] Through the above implementation, if the ratio of the largest knowledge leaf node weight to the smallest knowledge leaf node weight is within a preset ratio range, it indicates that the customer question may contain multiple individual questions. Furthermore, if the ratio of the largest syntactic leaf node weight to the smallest syntactic leaf node weight is also within the preset ratio range, it can be determined that the customer question contains multiple questions. Then, the customer question can be separated to obtain multiple individual questions. This makes it easier to answer each individual question one by one, thereby improving the accuracy of answering the customer question.

[0011] Preferably, before traversing and matching the target phrase in the customer question within the domain knowledge tree to determine the weight of the knowledge leaf node corresponding to the target phrase, the method further includes:

[0012] Based on at least one main business and a set of sub-businesses under the main business, a domain knowledge tree is constructed, the leaf node weights of the domain knowledge tree are determined, and the knowledge leaf node weights are obtained.

[0013] Based on syntactic components and industry terminology, a syntactic tree is constructed, and the leaf node weights of the syntactic tree are determined to obtain the syntactic leaf node weights; the syntactic components include at least subject, predicate, and object.

[0014] Through the above implementation, a domain knowledge tree and a syntax tree are created respectively, which facilitates the matching of phrases in customer questions, thereby making it easier to determine the number of individual questions in the customer questions based on the matching results.

[0015] Preferably, determining the leaf node weights of the domain knowledge tree to obtain the knowledge leaf node weights includes:

[0016] Determine the node order of each node in the domain knowledge tree, and the order weight corresponding to the node order;

[0017] Based on the order and weight of each node, the knowledge leaf node weight of each leaf node in the domain knowledge tree is calculated.

[0018] Through the above implementation, the knowledge leaf node weights of each leaf node in the domain knowledge tree are calculated by the order of each node and the corresponding order weights, which facilitates the preparation of data for the ratio between the weights of subsequent leaf nodes.

[0019] Preferably, the step of traversing and matching the target phrase in the customer question within the domain knowledge tree to determine the weight of the knowledge leaf node corresponding to the target phrase includes:

[0020] The customer's question is segmented into words to obtain the target phrase;

[0021] The target phrase is traversed and matched in the domain knowledge tree to obtain knowledge matching nodes;

[0022] The weight of the knowledge leaf node corresponding to the knowledge matching node is determined as the weight of the knowledge leaf node corresponding to the target word group.

[0023] Through the above implementation, the knowledge matching node in the domain knowledge tree that matches the target phrase is determined, which makes it easier to determine the weight of the knowledge leaf node corresponding to the target phrase through the knowledge matching node.

[0024] Preferably, the step of traversing and matching the target word group in the syntax tree to determine the weight of the corresponding syntactic leaf node includes:

[0025] The target phrase is traversed and matched in the predicate branch of the syntax tree to obtain the predicate matching node;

[0026] In response to the fact that the number of predicate matching nodes is at least two, the preceding and following word groups of the target word group are determined;

[0027] The preceding phrase is traversed and matched in the subject branch of the syntax tree to obtain the subject matching node;

[0028] The following phrase is traversed and matched in the object branch of the syntax tree to obtain the object matching node;

[0029] Determine the corresponding syntactic leaf nodes of the subject matching node, the predicate matching node, and the object matching node in the syntax tree;

[0030] The syntactic leaf node weights of the syntactic leaf nodes are determined to be the syntactic leaf node weights corresponding to the target word group.

[0031] Through the above implementation, the target phrase is traversed and matched in the syntax tree to obtain the corresponding subject matching node, predicate matching node and object matching node. It is convenient to determine the syntactic leaf node weight of the target phrase through the leaf nodes corresponding to the subject matching node, predicate matching node and object matching node respectively.

[0032] Preferably, after separating the customer question based on the syntactic matching node to obtain the individual question, the method further includes:

[0033] Determine the question order of the individual questions; perform vector transformation on the individual questions to obtain question vectors;

[0034] Determine the knowledge content of the question vector in the knowledge base, and generate a single question answer based on the knowledge content;

[0035] Based on the order of the questions, the answers to each individual question are assembled to obtain the target answer set.

[0036] By implementing the above method, the questions corresponding to each individual question are assembled in sequence. This makes it easier to answer each individual question in the customer's question according to the order of the individual questions in the answer, and improves the logic of the answer.

[0037] Preferably, the number of responses to the syntactic leaf node weights is at least two, and the ratio of the largest syntactic leaf node weight to the smallest syntactic leaf node weight is within the preset ratio range, determining that there are at least two individual problems in the customer problem includes:

[0038] In response to the fact that the number of the syntactic leaf node weights is at least two, the smallest syntactic leaf node weight is determined to obtain the denominator syntactic leaf node weights;

[0039] Calculate the ratio of the weights of the syntactic leaf nodes (excluding the weights of the denominator syntactic leaf nodes) to the weights of the denominator syntactic leaf nodes to obtain the syntactic target ratio set.

[0040] Determine the largest ratio in the set of syntactic target ratios to obtain the maximum ratio;

[0041] In response to the maximum ratio being within the preset ratio range, it is determined that there are at least two individual problems among the customer problems.

[0042] By implementing the above methods, the number of individual problems in a customer's problem can be determined by comparing the number of syntactic leaf node weights and the ratio of the maximum to the minimum syntactic leaf node weights with a preset ratio range; this facilitates improving the accuracy of determining the number of individual problems in a customer's problem.

[0043] Secondly, this application provides a problem separation device, comprising:

[0044] The knowledge weight determination module is used to traverse and match the target phrases in the customer's question in the domain knowledge tree to determine the weight of the knowledge leaf node corresponding to the target phrases.

[0045] The syntactic weight determination module is used to determine the syntactic weight of the target word group by traversing and matching the target word group in the syntactic tree in response to the fact that the number of knowledge leaf node weights is at least two and the ratio of the largest knowledge leaf node weight to the smallest knowledge leaf node weight is within a preset ratio range.

[0046] The question quantity determination module is used to determine that there are at least two individual questions among the customer questions in response to the fact that the number of the syntactic leaf node weights is at least two, and the ratio of the largest syntactic leaf node weight to the smallest syntactic leaf node weight is within the preset ratio range.

[0047] The customer question separation module is used to determine the syntactic matching node obtained by traversing and matching the target phrase in the syntactic tree, and to separate the customer question based on the syntactic matching node to obtain the individual question.

[0048] Through the above implementation, if the ratio of the largest knowledge leaf node weight to the smallest knowledge leaf node weight is within a preset ratio range, it indicates that the customer question may contain multiple individual questions. Furthermore, if the ratio of the largest syntactic leaf node weight to the smallest syntactic leaf node weight is also within the preset ratio range, it can be determined that the customer question contains multiple questions. Then, the customer question can be separated to obtain multiple individual questions. This makes it easier to answer each individual question one by one, thereby improving the accuracy of answering the customer question.

[0049] Thirdly, this application provides a computer device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the method described above.

[0050] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the above-described method.

[0051] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the steps in any of the above method embodiments.

[0052] The aforementioned problem separation method, apparatus, device, and storage medium determine the weight of the knowledge leaf node corresponding to the target word group by traversing and matching the target word group in the domain knowledge tree; in response to the fact that the number of knowledge leaf node weights is at least two, and the ratio of the largest knowledge leaf node weight to the smallest knowledge leaf node weight is within a preset ratio range, the target word group is traversed and matched in the syntax tree to determine the weight of the syntax leaf node corresponding to the target word group; in response to the fact that the number of syntax leaf node weights is at least two, and the ratio of the largest syntax leaf node weight to the smallest syntax leaf node weight is within the preset ratio range, it is determined that there are at least two individual problems in the customer problem; the syntax matching node obtained by traversing and matching the target word group in the syntax tree is determined, and the customer problem is separated based on the syntax matching node to obtain the individual problem. Through the above implementation, if the ratio of the largest knowledge leaf node weight to the smallest knowledge leaf node weight is within a preset ratio range, it indicates that the customer question may contain multiple individual questions. Furthermore, if the ratio of the largest syntactic leaf node weight to the smallest syntactic leaf node weight is also within the preset ratio range, it can be determined that the customer question contains multiple questions. Then, the customer question can be separated to obtain multiple individual questions. This makes it easier to answer each individual question one by one, thereby improving the accuracy of answering the customer question.

[0053] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0054] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 is a flowchart of a problem separation method provided in an embodiment of this application;

[0056] Figure 2 is a schematic diagram of a problem separation device provided in an embodiment of this application;

[0057] Figure 3 is a schematic diagram of the structure of a computer device provided in an embodiment of this application;

[0058] Figure 4 is an internal structure diagram of a computer-readable storage medium provided in an embodiment of this application. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of this disclosure clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this disclosure.

[0060] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings herein are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0061] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0062] Example 1

[0063] Figure 1 is a flowchart of a problem separation method provided in Embodiment 1 of this application. Referring to Figure 1, the method can be executed by a device that performs the method. The device can be implemented by software and / or hardware. The method includes:

[0064] S110. Traverse and match the target phrases in the customer's question in the domain knowledge tree to determine the weight of the knowledge leaf node corresponding to the target phrases.

[0065] In order to improve the efficiency of answering customer questions, financial institutions such as banks have set up automatic question-and-answer systems. These systems retrieve and analyze customer questions to generate answers that correspond to the questions and provide feedback to the customers. The questions raised by customers are recorded as customer questions. Customer questions consist of multiple phrases, and these phrases are recorded as target phrases.

[0066] To improve the accuracy of generated customer questions, this embodiment pre-defines a domain knowledge tree. This domain knowledge tree is built upon the various businesses undertaken by the financial institution. It reflects these businesses and allows for convenient matching of target terms to the corresponding businesses. It should be noted that each node in the domain knowledge tree corresponds to a specific business, and the leaf nodes correspond to the most granular business. Each leaf node has its own weight, which is recorded as the knowledge leaf node weight.

[0067] Specifically, taking a target phrase in a customer problem as an example, the target phrase is traversed and matched in the domain knowledge tree. The corresponding business node can be matched in the domain knowledge tree. This node may be a leaf node or have leaf nodes below it. The weight corresponding to the leaf node is then used as the weight of the knowledge leaf node.

[0068] S120. In response to the fact that the number of knowledge leaf node weights is at least two, and the ratio of the largest knowledge leaf node weight to the smallest knowledge leaf node weight is within a preset ratio range, the target word group is traversed and matched in the syntax tree to determine the syntax leaf node weight corresponding to the target word group.

[0069] Taking a target phrase as an example, the target phrase may correspond to one or more knowledge leaf node weights. This embodiment aims to determine the number of individual questions in a customer's problem by the number of knowledge leaf node weights corresponding to the target phrase. A customer's problem may contain two or more individual questions. If the number of knowledge leaf node weights corresponding to a target phrase is 1, it indicates that the customer's problem contains only one individual question. If the number of knowledge leaf node weights corresponding to the target phrase is at least 2, it indicates that the customer's problem may contain multiple individual questions. To further determine whether multiple questions are indeed present, further judgment is needed. The further judgment method provided in this embodiment is as follows: obtain all knowledge leaf node weights corresponding to the target phrase, then determine the largest and smallest knowledge leaf node weights, and then use the smallest knowledge leaf node weight as the denominator to calculate the largest knowledge leaf node weight. The ratio of the leaf node weight to the smallest knowledge leaf node weight is used to determine whether the ratio is within a preset threshold range. This preset threshold range is set based on historical experience data, and the maximum value of the preset threshold range is recorded as the upper limit threshold, and the minimum value is recorded as the lower limit threshold. For example, the preset threshold range is [2, 10]. If the ratio is greater than the upper limit threshold, it indicates that the customer problem does indeed contain multiple individual problems. If the ratio is within the threshold range, it indicates that the customer problem may contain multiple individual problems, and further steps are needed to determine whether the customer problem does indeed contain multiple problems.

[0070] It should be noted that, in order to further determine whether the customer's problem actually contains multiple problems, this embodiment also presets a syntax tree. The syntax tree is constructed based on the syntactic components of Chinese (such as subject, predicate, and object) and industry terminology. The syntax tree is used for the target phrase to traverse and match its nodes. The matched node itself may be a leaf node of the syntax tree or have leaf nodes below it. Each leaf node of the syntax tree is preset with a corresponding weight, and this weight is recorded as the syntax leaf node weight.

[0071] S130. In response to the fact that the number of the syntactic leaf node weights is at least two, and the ratio of the largest syntactic leaf node weight to the smallest syntactic leaf node weight is within the preset ratio range, it is determined that there are at least two individual problems in the customer problem.

[0072] In this process, the number of syntactic leaf node weights obtained by traversing and matching the target phrase in the syntactic tree is generally at least two. To determine whether there are indeed at least two individual problems in the customer's problem, it is also necessary to determine the largest and smallest syntactic leaf node weights from the syntactic leaf node weights determined by the matching, and then calculate the ratio of the largest to the smallest syntactic leaf node weights. If the ratio is within the preset ratio range, it indicates that there are indeed at least two individual problems in the customer's problem.

[0073] S140. Determine the syntactic matching node obtained by traversing and matching the target phrase in the syntactic tree, and separate the customer question based on the syntactic matching node to obtain the single question.

[0074] In step S120, after identifying at least two individual problems within the customer problem, the next step is to separate each individual problem from the customer problem. During this step, the target phrase is traversed and matched in the syntax tree to find the corresponding node, which is then recorded as a syntax matching node. Each syntax matching node corresponds to an individual problem, and the customer problem can be separated through each syntax matching node, thus obtaining the individual problems corresponding to each syntax matching node.

[0075] It should be noted that in this embodiment, the target word group in the customer question is traversed and matched in the domain knowledge tree to determine the weight of the knowledge leaf node corresponding to the target word group; in response to the fact that the number of knowledge leaf node weights is at least two, and the ratio of the largest knowledge leaf node weight to the smallest knowledge leaf node weight is within a preset ratio range, the target word group is traversed and matched in the syntax tree to determine the weight of the syntax leaf node corresponding to the target word group; in response to the fact that the number of syntax leaf node weights is at least two, and the ratio of the largest syntax leaf node weight to the smallest syntax leaf node weight is within the preset ratio range, it is determined that there are at least two individual questions in the customer question; the syntax matching node obtained by traversing and matching the target word group in the syntax tree is determined, and the customer question is separated based on the syntax matching node to obtain the individual questions. Through the above implementation, if the ratio of the largest knowledge leaf node weight to the smallest knowledge leaf node weight is within a preset ratio range, it indicates that the customer question may contain multiple individual questions. Furthermore, if the ratio of the largest syntactic leaf node weight to the smallest syntactic leaf node weight is also within the preset ratio range, it can be determined that the customer question contains multiple questions. Then, the customer question can be separated to obtain multiple individual questions. This makes it easier to answer each individual question one by one, thereby improving the accuracy of answering the customer question.

[0076] Example 2

[0077] This application provides a problem separation method in Embodiment 2, which supplements the method described in Embodiment 1. It should be noted that for parts not detailed in this embodiment, please refer to the descriptions in other embodiments. The method includes:

[0078] S210. Based on at least one main business and a set of sub-businesses under the main business, construct a domain knowledge tree, determine the leaf node weights of the domain knowledge tree, and obtain the knowledge leaf node weights.

[0079] Financial institutions typically have multiple main businesses, such as corporate banking, lending, and risk control. Each main business contains secondary businesses, such as corporate deposits, loans, and intermediary services. Secondary businesses may also contain tertiary businesses, such as demand deposits, time deposits, and margin deposits. And so on. Taking a main business as an example, the set of all levels of businesses under that main business is also the set of sub-businesses under that main business. The main business and its corresponding set of sub-businesses are used to construct a domain knowledge tree, where the main business is the root node, the businesses between the secondary and penultimate levels in the sub-business set are internal nodes, and the last level of business in the sub-business set is a leaf node.

[0080] It should be noted that, based on the preset leaf node weight calculation method, the weight of each leaf node in the knowledge tree of this domain can be determined, and this weight is recorded as the knowledge leaf node weight.

[0081] S220. Based on syntactic components and industry terminology, construct a syntactic tree, determine the leaf node weights of the syntactic tree, and obtain the syntactic leaf node weights; the syntactic components include at least subject, predicate, and object.

[0082] In this embodiment, the syntactic components include at least a subject, a predicate, and an object. Taking one of the syntactic components as an example, the syntactic component serves as the root node of the syntactic tree. Each syntactic component has a corresponding set of industry terms. For example, one industry term under the predicate is "execution". In this embodiment, each industry term under the syntactic component is used as a leaf node of the syntactic tree.

[0083] It should be noted that, based on the preset leaf node weight calculation method, the weight of each leaf node of the syntax tree can be determined, and this weight is recorded as the syntax leaf node weight.

[0084] S230. Traverse and match the target phrases in the customer's question in the domain knowledge tree to determine the weight of the knowledge leaf node corresponding to the target phrases.

[0085] S240. In response to the fact that the number of knowledge leaf node weights is at least two, and the ratio of the largest knowledge leaf node weight to the smallest knowledge leaf node weight is within a preset ratio range, the target word group is traversed and matched in the syntax tree to determine the syntax leaf node weight corresponding to the target word group.

[0086] S250. In response to the fact that the number of the syntactic leaf node weights is at least two, and the ratio of the largest syntactic leaf node weight to the smallest syntactic leaf node weight is within the preset ratio range, it is determined that there are at least two individual problems in the customer problem.

[0087] S260. Determine the syntactic matching node obtained by traversing and matching the target phrase in the syntactic tree, and separate the customer question based on the syntactic matching node to obtain the single question.

[0088] Example 3

[0089] This application provides a problem separation method in Embodiment 3, which optimizes the "determining the leaf node weights of the domain knowledge tree and obtaining the knowledge leaf node weights" step in Embodiment 2. It should be noted that for parts not detailed in this embodiment, please refer to the descriptions in other embodiments. The method includes:

[0090] S311. Construct a domain knowledge tree based on at least one main business and a set of sub-businesses under the main business.

[0091] S312. Determine the node order of each node in the domain knowledge tree, and the order weight corresponding to the node order.

[0092] In this domain knowledge tree, each node has a one-to-one corresponding node order. For example, the node order of each root node in the domain knowledge tree is denoted as 1, 2, 3, ... Taking the root node with node order 1 as an example, there are two levels of nodes under this root node, denoted as intermediate nodes and leaf nodes, with the leaf nodes being the last level of nodes. The node order of each intermediate node is denoted as 1.1, 1.2, 1.3, ... Taking the intermediate node with node order 1.1 as another example, the node order of each leaf node under this intermediate node is denoted as 1.1.1, 1.1.2, 1.1.3, ... and so on, without further elaboration.

[0093] Taking the leaf node with node order 1.12.13 as an example, the order weight corresponding to the node order of this leaf node is (10, 10). -1 10 -2 ), where 10 corresponds to 1, 10 -1 Corresponding to 12, 10 -2 Corresponds to 13.

[0094] S313. Based on the order of each node and the weight of each order, calculate the knowledge leaf node weight of each leaf node in the domain knowledge tree.

[0095] Taking the leaf node with node order 1.12.13 as an example, the knowledge leaf node weight C of this leaf node is: C = 1 * 10 + 12 * 10 -1+13*10 -2 .

[0096] S320. Based on syntactic components and industry terminology, construct a syntactic tree, determine the leaf node weights of the syntactic tree, and obtain the syntactic leaf node weights; the syntactic components include at least subject, predicate, and object.

[0097] S330. Traverse and match the target phrases in the customer's question in the domain knowledge tree to determine the weight of the knowledge leaf node corresponding to the target phrases.

[0098] S340. In response to the fact that the number of knowledge leaf node weights is at least two, and the ratio of the largest knowledge leaf node weight to the smallest knowledge leaf node weight is within a preset ratio range, the target word group is traversed and matched in the syntax tree to determine the syntax leaf node weight corresponding to the target word group.

[0099] S350. In response to the fact that the number of the syntactic leaf node weights is at least two, and the ratio of the largest syntactic leaf node weight to the smallest syntactic leaf node weight is within the preset ratio range, it is determined that there are at least two individual problems in the customer problem.

[0100] S360. Determine the syntactic matching node obtained by traversing and matching the target phrase in the syntactic tree, and separate the customer question based on the syntactic matching node to obtain the single question.

[0101] Example 4

[0102] This application provides a problem separation method in Embodiment 4, which optimizes the step in Embodiment 1 of "traversing and matching target terms in customer questions in a domain knowledge tree to determine the weight of the knowledge leaf nodes corresponding to the target terms". It should be noted that for parts not detailed in this embodiment, please refer to the descriptions in other embodiments. This method includes:

[0103] S411. Segment the customer's question into words to obtain the target phrase.

[0104] In this embodiment, a word segmentation algorithm is preset. This word segmentation algorithm is used to segment the customer's question into multiple word groups, and the word group is recorded as the target word group.

[0105] S412. The target phrase is traversed and matched in the domain knowledge tree to obtain knowledge matching nodes.

[0106] For example, if a target phrase is "deposit", the target phrase is traversed and matched in the domain knowledge tree. That is, nodes that are semantically similar to the target phrase are determined from the domain knowledge tree, and the nodes that are easily matched are recorded as knowledge matching nodes. There may be multiple knowledge matching nodes corresponding to a target phrase. For example, when the target phrase is "deposit", the knowledge matching nodes that can be obtained by traversing and matching in the domain knowledge tree are: current deposit business, time deposit business and margin deposit business.

[0107] S413. The weight of the knowledge leaf node corresponding to the knowledge matching node is determined as the weight of the knowledge leaf node corresponding to the target word group.

[0108] The weight of the knowledge leaf node corresponding to the target phrase is also the weight of the knowledge matching node corresponding to the target phrase.

[0109] S420. In response to the fact that the number of knowledge leaf node weights is at least two, and the ratio of the largest knowledge leaf node weight to the smallest knowledge leaf node weight is within a preset ratio range, the target word group is traversed and matched in the syntax tree to determine the syntax leaf node weight corresponding to the target word group.

[0110] S430. In response to the fact that the number of the syntactic leaf node weights is at least two, and the ratio of the largest syntactic leaf node weight to the smallest syntactic leaf node weight is within the preset ratio range, it is determined that there are at least two individual problems in the customer problem.

[0111] S440. Determine the syntactic matching node obtained by traversing and matching the target phrase in the syntactic tree, and separate the customer question based on the syntactic matching node to obtain the single question.

[0112] Example 5

[0113] This application provides a problem separation method in Embodiment 5, which optimizes the step of "traversing and matching the target word group in the syntax tree to determine the weight of the corresponding syntactic leaf node" in Embodiment 1. It should be noted that for parts not detailed in this embodiment, please refer to the descriptions in other embodiments. This method includes:

[0114] S510. Traverse and match the target phrases in the customer's question in the domain knowledge tree to determine the weight of the knowledge leaf node corresponding to the target phrases.

[0115] S521. In response to the fact that the number of knowledge leaf node weights is at least two, and the ratio of the largest knowledge leaf node weight to the smallest knowledge leaf node weight is within a preset ratio range, the target phrase is traversed and matched in the predicate branch of the syntax tree to obtain the predicate matching node.

[0116] In this embodiment, each root node in the syntax tree is a syntactic component, such as one of the subject, predicate, and object; each root node has at least one leaf node attached to it, and the root node and each leaf node attached to it are branches in the syntax tree, and the branch corresponding to the predicate of the root node is recorded as the predicate branch; the predicate branch can also be traversed and matched by the target phrase; and the nodes matched by the traversal are recorded as predicate matching nodes.

[0117] S522. In response to the number of predicate matching nodes being at least two, determine the preceding and following word groups of the target word group.

[0118] There may be multiple predicate matching nodes. When there are multiple predicate matching nodes, it is necessary to further traverse and match the preceding and following word groups of the target word group corresponding to the predicate matching node in the syntax tree. The preceding word group is the word group in the client question that is before the target word group, which is usually the subject. The following word group is the word group in the client question that is after the target word group, which is usually the object.

[0119] S523. The preceding phrase is traversed and matched in the subject branch of the syntax tree to obtain the subject matching node.

[0120] The syntax tree contains not only the predicate branch mentioned above, but also the subject branch; the subject branch is used for traversal matching of the preceding phrase; and the nodes obtained by traversing and matching the preceding phrase in the subject branch are recorded as subject matching nodes.

[0121] S524. The phrase following the grammar tree is traversed and matched in the object branch to obtain the object matching node.

[0122] The syntax tree contains not only the predicate branch and subject branch mentioned above, but also the object branch; the object branch is used for traversal matching of subsequent phrases; and the nodes obtained by traversing and matching subsequent phrases in the object branch are recorded as object matching nodes.

[0123] S525. Determine the corresponding syntactic leaf nodes of the subject matching node, the predicate matching node and the object matching node in the syntax tree.

[0124] Among them, the subject matching node, the predicate matching node, and the object matching node may each be a leaf node of the syntax tree, or the nodes attached to them may be leaf nodes of the syntax tree. That is, the subject matching node, the predicate matching node, and the object matching node each have their own corresponding leaf node, and the leaf node is recorded as the syntax leaf node.

[0125] S526. The syntactic leaf node weights of the syntactic leaf nodes are determined as the syntactic leaf node weights corresponding to the target word group.

[0126] It should be noted that, similar to the method for calculating the weight of knowledge leaf nodes in the above embodiments, the weight of syntactic leaf nodes can be calculated by the node order and order weight of the syntactic leaf nodes, which will not be elaborated here.

[0127] S530. In response to the fact that the number of the syntactic leaf node weights is at least two, and the ratio of the largest syntactic leaf node weight to the smallest syntactic leaf node weight is within the preset ratio range, it is determined that there are at least two individual problems in the customer problem.

[0128] S540. Determine the syntactic matching node obtained by traversing and matching the target phrase in the syntactic tree, and separate the customer question based on the syntactic matching node to obtain the single question.

[0129] Example 6

[0130] This application provides a problem separation method in Embodiment Six, which supplements the method described in Embodiment One. It should be noted that for parts not detailed in this embodiment, please refer to the descriptions in other embodiments. The method includes:

[0131] S610. Traverse and match the target phrases in the customer's question in the domain knowledge tree to determine the weight of the knowledge leaf node corresponding to the target phrases.

[0132] S620. In response to the fact that the number of knowledge leaf node weights is at least two, and the ratio of the largest knowledge leaf node weight to the smallest knowledge leaf node weight is within a preset ratio range, the target word group is traversed and matched in the syntax tree to determine the syntax leaf node weight corresponding to the target word group.

[0133] S630. In response to the fact that the number of the syntactic leaf node weights is at least two, and the ratio of the largest syntactic leaf node weight to the smallest syntactic leaf node weight is within the preset ratio range, it is determined that there are at least two individual problems in the customer problem.

[0134] S640. Determine the syntactic matching node obtained by traversing and matching the target phrase in the syntactic tree, and separate the customer question based on the syntactic matching node to obtain the single question.

[0135] S650. Determine the question order of the individual questions; perform vector transformation on the individual questions to obtain a question vector.

[0136] In this embodiment, each individual question corresponds to a target phrase in the customer question. The customer question contains multiple target phrases, and each target phrase has a one-to-one corresponding order among the multiple target phrases. This order is used as the question order of the corresponding individual question. This embodiment provides a vector conversion algorithm, which is used to convert an individual question into a corresponding question vector.

[0137] S660. Determine the knowledge content of the question vector in the knowledge base, and generate a single question answer based on the knowledge content.

[0138] The question vector is used to traverse and match the corresponding knowledge content in a preset knowledge base. This knowledge content is then used to generate the answer to a single question corresponding to the question vector.

[0139] S670. Assemble the individual answers corresponding to each individual question based on the question order to obtain the target answer set.

[0140] The target answer set contains multiple answers to individual questions, and the order of each answer in the target answer set is the question order of its corresponding individual question.

[0141] Example 7

[0142] This application provides a problem separation method in Embodiment Seven, which optimizes the method in Embodiment One, which states that "in response to the fact that the number of syntactic leaf node weights is at least two, and the ratio of the largest syntactic leaf node weight to the smallest syntactic leaf node weight is within the preset ratio range, it is determined that there are at least two individual problems among the customer problems." It should be noted that for parts not detailed in this embodiment, please refer to the descriptions in other embodiments. This method includes:

[0143] S710. Traverse and match the target phrases in the customer's question in the domain knowledge tree to determine the weight of the knowledge leaf node corresponding to the target phrases.

[0144] S720. In response to the fact that the number of knowledge leaf node weights is at least two, and the ratio of the largest knowledge leaf node weight to the smallest knowledge leaf node weight is within a preset ratio range, the target word group is traversed and matched in the syntax tree to determine the syntax leaf node weight corresponding to the target word group.

[0145] S731. In response to the fact that the number of the syntactic leaf node weights is at least two, determine the smallest syntactic leaf node weight to obtain the denominator syntactic leaf node weight.

[0146] The weights of each syntactic leaf node are different, and the smallest syntactic leaf node weight is recorded as the denominator syntactic leaf node weight. This denominator syntactic leaf node weight is used as the denominator when dividing other syntactic leaf node weights.

[0147] S732. Calculate the ratio of the weights of the syntactic leaf nodes other than the weights of the denominator syntactic leaf nodes to the weights of the denominator syntactic leaf nodes to obtain the syntactic target ratio set.

[0148] Taking the weight of the syntactic leaf node other than the weight of the denominator syntactic leaf node as an example, the weight of the syntactic leaf node is used as the numerator, and the weight of the denominator syntactic leaf node is used as the denominator for division to obtain the ratio corresponding to the weight of the syntactic leaf node; the syntactic target ratio group is the set of ratios of each syntactic leaf node weight to the weight of the denominator syntactic leaf node.

[0149] S733. Determine the largest ratio in the set of syntactic target ratios to obtain the maximum ratio.

[0150] The maximum ratio is the maximum value in the syntactic target ratio group.

[0151] S734. In response to the maximum ratio being within the preset ratio range, it is determined that there are at least two individual problems among the customer problems.

[0152] If the maximum ratio is determined to be within the preset ratio range, it indicates that there are at least two individual problems among the customer's issues.

[0153] S740. Determine the syntactic matching node obtained by traversing and matching the target phrase in the syntactic tree, and separate the customer question based on the syntactic matching node to obtain the single question.

[0154] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0155] Example 8

[0156] Based on the same inventive concept, this embodiment also provides a problem separation device for implementing the problem separation method described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations in one or more problem separation device embodiments provided below can be found in the limitations of the problem separation method above, and will not be repeated here.

[0157] In this embodiment, as shown in FIG2, a problem separation device is provided, including:

[0158] The knowledge weight determination module is used to traverse and match the target phrases in the customer's question in the domain knowledge tree to determine the weight of the knowledge leaf node corresponding to the target phrases.

[0159] The syntactic weight determination module is used to determine the syntactic weight of the target word group by traversing and matching the target word group in the syntactic tree in response to the fact that the number of knowledge leaf node weights is at least two and the ratio of the largest knowledge leaf node weight to the smallest knowledge leaf node weight is within a preset ratio range.

[0160] The question quantity determination module is used to determine that there are at least two individual questions among the customer questions in response to the fact that the number of the syntactic leaf node weights is at least two, and the ratio of the largest syntactic leaf node weight to the smallest syntactic leaf node weight is within the preset ratio range.

[0161] The customer question separation module is used to determine the syntactic matching node obtained by traversing and matching the target phrase in the syntactic tree, and to separate the customer question based on the syntactic matching node to obtain the individual question.

[0162] Each module in the aforementioned problem separation device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0163] It should be noted that in this embodiment, the target word group in the customer question is traversed and matched in the domain knowledge tree to determine the weight of the knowledge leaf node corresponding to the target word group; in response to the fact that the number of knowledge leaf node weights is at least two, and the ratio of the largest knowledge leaf node weight to the smallest knowledge leaf node weight is within a preset ratio range, the target word group is traversed and matched in the syntax tree to determine the weight of the syntax leaf node corresponding to the target word group; in response to the fact that the number of syntax leaf node weights is at least two, and the ratio of the largest syntax leaf node weight to the smallest syntax leaf node weight is within the preset ratio range, it is determined that there are at least two individual questions in the customer question; the syntax matching node obtained by traversing and matching the target word group in the syntax tree is determined, and the customer question is separated based on the syntax matching node to obtain the individual questions. Through the above implementation, if the ratio of the largest knowledge leaf node weight to the smallest knowledge leaf node weight is within a preset ratio range, it indicates that the customer question may contain multiple individual questions. Furthermore, if the ratio of the largest syntactic leaf node weight to the smallest syntactic leaf node weight is also within the preset ratio range, it can be determined that the customer question contains multiple questions. Then, the customer question can be separated to obtain multiple individual questions. This makes it easier to answer each individual question one by one, thereby improving the accuracy of answering the customer question.

[0164] Example 9

[0165] In this embodiment, a computer device is provided, which may be a server, and its internal structure diagram is shown in Figure 3. The computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores data. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements a problem separation method.

[0166] Those skilled in the art will understand that the structure shown in Figure 3 is merely a block diagram of a portion of the structure related to the present disclosure and does not constitute a limitation on the computer device to which the present disclosure is applied. A specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0167] Example 10

[0168] In this embodiment, a computer-readable storage medium is provided, as shown in FIG4, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps in the above-described method embodiments.

[0169] Example 11

[0170] In this embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0171] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties.

[0172] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this disclosure can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this disclosure may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this disclosure may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0173] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0174] The embodiments described above are merely illustrative of several implementations of this disclosure, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent disclosure. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this disclosure, and these all fall within the protection scope of this disclosure. Therefore, the protection scope of this disclosure should be determined by the appended claims.

Claims

1. A problem separation method, characterized in that, include: The target phrases in the customer's question are traversed and matched in the domain knowledge tree to determine the weight of the knowledge leaf node corresponding to the target phrases. In response to the fact that the number of knowledge leaf node weights is at least two, and the ratio of the largest knowledge leaf node weight to the smallest knowledge leaf node weight is within a preset ratio range, the target word group is traversed and matched in the syntax tree to determine the syntax leaf node weight corresponding to the target word group; in response to the fact that the number of syntax leaf node weights is at least two, and the ratio of the largest syntax leaf node weight to the smallest syntax leaf node weight is within the preset ratio range, it is determined that there are at least two individual questions in the customer question; The target phrase is identified by traversing and matching it in the syntax tree to obtain the syntax matching node. Based on the syntax matching node, the customer question is separated to obtain the individual question.

2. The method according to claim 1, characterized in that, Before traversing and matching the target phrases in the customer's question within the domain knowledge tree to determine the weights of the corresponding knowledge leaf nodes, the method further includes: constructing a domain knowledge tree based on at least one main business and a set of sub-businesses under the main business, determining the weights of the leaf nodes of the domain knowledge tree, and obtaining the knowledge leaf node weights; constructing a syntax tree based on syntactic components and industry terminology, determining the weights of the leaf nodes of the syntax tree, and obtaining the syntactic leaf node weights; the syntactic components include at least a subject, a predicate, and an object.

3. The method according to claim 2, characterized in that, Determining the leaf node weights of the domain knowledge tree to obtain the knowledge leaf node weights includes: determining the node order of each node in the domain knowledge tree and the order weight corresponding to the node order; and calculating the knowledge leaf node weights of each leaf node in the domain knowledge tree based on each node order and each order weight.

4. The method according to claim 1, characterized in that, The step of traversing and matching the target phrase in the customer question within the domain knowledge tree to determine the weight of the knowledge leaf node corresponding to the target phrase includes: segmenting the customer question into words to obtain the target phrase; traversing and matching the target phrase within the domain knowledge tree to obtain the knowledge matching node; and determining the weight of the knowledge leaf node corresponding to the knowledge matching node as the weight of the knowledge leaf node corresponding to the target phrase.

5. The method according to claim 1, characterized in that, The step of traversing and matching the target phrase in the syntax tree to determine the weight of the corresponding syntactic leaf node includes: traversing and matching the target phrase in the predicate branch of the syntax tree to obtain a predicate matching node; determining the preceding and following phrases of the target phrase in response to the number of predicate matching nodes being at least two; traversing and matching the preceding phrase in the subject branch of the syntax tree to obtain a subject matching node; traversing and matching the following phrase in the object branch of the syntax tree to obtain an object matching node; determining the corresponding syntactic leaf nodes of the subject matching node, the predicate matching node, and the object matching node in the syntax tree; and determining the weight of the syntactic leaf node as the weight of the syntactic leaf node corresponding to the target phrase.

6. The method according to claim 1, characterized in that, After separating the customer question based on the syntactic matching node to obtain the individual question, the method further includes: determining the question order of the individual questions; performing vector transformation on the individual questions to obtain question vectors; determining the knowledge content of the question vectors in the knowledge base, and generating individual question answers based on the knowledge content; and assembling the individual question answers corresponding to each individual question based on the question order to obtain a target answer set.

7. The method according to claim 1, characterized in that, The determination that at least two individual problems exist in the customer problem is made in response to the fact that the number of syntactic leaf node weights is at least two, and the ratio of the largest syntactic leaf node weight to the smallest syntactic leaf node weight is within the preset ratio range. This includes: determining the smallest syntactic leaf node weight to obtain the denominator syntactic leaf node weight in response to the fact that the number of syntactic leaf node weights is at least two; calculating the ratios of the other syntactic leaf node weights (excluding the denominator syntactic leaf node weight) to the denominator syntactic leaf node weight to obtain a syntactic target ratio set; determining the largest ratio in the syntactic target ratio set to obtain the maximum ratio; and determining that at least two individual problems exist in the customer problem in response to the fact that the maximum ratio is within the preset ratio range.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, 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 to 7.

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 to 7.

Images