A Jumpable Multi-Round Conversation Method, Device and Electronic Device

Abstract

The present invention provides a multi-round conversation method, apparatus and electronic device capable of jumping. Among them, the method includes: generating a plurality of knowledge matrices each containing a plurality of items, and when jumping is possible, jumping to a second knowledge matrix or a calibrated item in the second knowledge matrix, and then outputting a corresponding content result. Through the multi-round conversation method, apparatus and electronic device provided by the embodiments of the present invention, when generating a knowledge matrix, it is not necessary to accurately determine entities, and it is even less necessary to pay attention to the relationships between entities. The generation process is simple. When extracting the knowledge points of items and determining the knowledge types of knowledge points, even if there are deviations, it does not affect the finally determined content required by the user; multi-round conversations can still be better realized in complex application scenarios. By associating a plurality of knowledge matrices or items in the knowledge matrices in a jumping manner, a more complete content can be constructed through a plurality of small knowledge matrices, and the construction efficiency of the knowledge matrix can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-round conversation, and in particular, to a multi-round conversation method, device, electronic device and computer-readable storage medium that can jump. Background Technique

[0002] In the field of multi-round conversation, the most common application currently is still task-driven multi-round dialogue, where users have clear purposes, such as ordering food, booking tickets, etc. Existing multi-round conversations almost completely set each step of the multi-round conversation manually, such as: setting keywords (or instructions), establishing slots, setting whether the slots are optional, selecting the corresponding dictionary, setting clarification words, setting context change strategies, etc. If complex logic is involved, elaborate planning of logical conditions such as AND, OR, and NOT is required. If there are multiple similar tasks, the debugging cost is extremely high.

[0003] In order to apply complex logic, multi-round conversation can also be currently implemented based on a knowledge graph. However, the knowledge graph essentially represents the relationships between entities. Therefore, the multi-round conversation scheme based on the knowledge graph needs to extract entities and determine the relationships between entities. For simple scenarios such as ordering food and booking tickets, the entities are keywords or slots, and it is easy to define the relationships between entities. In complex scenarios, it is difficult to determine the relationships between entities, and it is even difficult to extract entities.

[0004] In some fields with extensive and professional knowledge, such as the government affairs field, there are many and complex matters. Since users do not understand what effective information needs to be provided, the questions raised by users are somewhat blind, and users are likely to hit the wrong entities, making it impossible for users to locate the required answers. Summary of the Invention

[0005] To solve the existing technical problems, embodiments of the present invention provide a multi-round conversation method, device, electronic device and computer-readable storage medium that can jump.

[0006] In a first aspect, embodiments of the present invention provide a multi-round conversation method that can jump, including:

[0007] Generating a plurality of knowledge matrices each including a plurality of matters, and setting an association relationship between a jump matter in a first knowledge matrix and a second knowledge matrix or a calibrated matter in the second knowledge matrix; wherein, each of the matters includes one or more knowledge points divided according to knowledge types, and the elements of the knowledge matrix represent the knowledge points corresponding to the corresponding knowledge types in the corresponding matters;

[0008] Determining a corresponding current knowledge matrix according to the information input by the user, and positioning to a first target matter in the current knowledge matrix according to the information input by the user;

[0009] If the first target item is a jump item in the first knowledge matrix, determine the second knowledge matrix corresponding to the first target item or the calibration item in the second knowledge matrix according to the association relationship;

[0010] Output a corresponding content result according to the second knowledge matrix or the calibration item in the second knowledge matrix.

[0011] In a second aspect, an embodiment of the present invention further provides a multi-round session device with jump function, including:

[0012] A generation module, configured to generate a plurality of knowledge matrices each including a plurality of items, and set an association relationship between a jump item in the first knowledge matrix and a second knowledge matrix or a calibration item in the second knowledge matrix; wherein, each of the items includes one or more knowledge points divided according to knowledge types, and the elements of the knowledge matrix represent the knowledge points corresponding to the corresponding knowledge types in the corresponding items;

[0013] A positioning module, configured to determine a corresponding current knowledge matrix according to the information input by the user, and locate a first target item in the current knowledge matrix according to the information input by the user;

[0014] A jump module, configured to, if the first target item is a jump item in the first knowledge matrix, determine the second knowledge matrix corresponding to the first target item or the calibration item in the second knowledge matrix according to the association relationship;

[0015] A content output module, configured to output a corresponding content result according to the second knowledge matrix or the calibration item in the second knowledge matrix.

[0016] In a third aspect, an embodiment of the present invention provides an electronic device, including a bus, a transceiver, a memory, a processor, and a computer program stored on the memory and executable on the processor. The transceiver, the memory, and the processor are connected through the bus. When the computer program is executed by the processor, the steps in the multi-round session method with jump function described in any one of the above are implemented.

[0017] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps in the multi-round session method with jump function described in any one of the above are implemented.

[0018] The multi-round conversation method, device, electronic device, and computer-readable storage medium with jump function provided by the embodiments of the present invention record the knowledge points of matters according to matters and knowledge types, so that a knowledge matrix with a simple structure can be generated; when generating the knowledge matrix, it is not necessary to accurately determine entities, let alone pay attention to the relationships between entities, and the generation process is simple. Moreover, when extracting the knowledge points of matters and determining the knowledge types of the knowledge points, even if there are deviations, it does not affect the content required by the user finally determined; therefore, even if a complex application scenario contains a large number of matters and it is difficult to extract the knowledge points of matters according to a unified standard, and it is also difficult to accurately classify the knowledge points to determine the knowledge types, the multi-round conversation can still be better implemented based on this knowledge matrix. By associating multiple knowledge matrices or matters in the knowledge matrix in a jump manner, more complete content can be constructed through multiple small knowledge matrices, which can improve the construction efficiency of the knowledge matrix; and through the jump method, the matter required by the user can also be finally located, providing a more accurate content result for the user. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the background technology, the drawings required to be used in the embodiments of the present invention or the background technology will be described below.

[0020] Figure 1 The flowchart of a multi-round conversation method with jump function provided by the embodiments of the present invention is shown;

[0021] Figure 2a A schematic structural diagram of a knowledge matrix in the multi-round conversation method provided by the embodiments of the present invention is shown;

[0022] Figure 2b Another schematic structural diagram of a knowledge matrix in the multi-round conversation method provided by the embodiments of the present invention is shown;

[0023] Figure 3 A schematic diagram showing the realization of jump of the knowledge matrix provided by the embodiments of the present invention is shown;

[0024] Figure 4 Another schematic diagram showing the realization of jump of the knowledge matrix provided by the embodiments of the present invention is shown;

[0025] Figure 5 The schematic structural diagram of a multi-round conversation device with jump function provided by the embodiments of the present invention is shown;

[0026] Figure 6 The schematic structural diagram of an electronic device for executing the multi-round conversation method with jump function provided by the embodiments of the present invention is shown. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] The embodiments of the present invention will be described below with reference to the accompanying drawings in the embodiments of the present invention.

[0028] Figure 1 The flowchart of a multi-round session method with jump function provided by the embodiments of the present invention is shown. As Figure 1 shown, the method includes:

[0029] Step 101: Generate a plurality of knowledge matrices each containing a plurality of items, and set the association relationship between the jump items in the first knowledge matrix and the second knowledge matrix or the calibrated items in the second knowledge matrix; wherein, each item contains one or more knowledge points divided according to knowledge types, and the elements of the knowledge matrix represent the knowledge points corresponding to the corresponding knowledge types in the corresponding items.

[0030] In the embodiments of the present invention, the keywords involved in the multi-round session are represented in the form of a matrix. Specifically, in complex scenarios such as government affairs, users essentially need to query the content they need, and this content generally has a corresponding title or abstract. In this embodiment, the title or abstract of the content is called an "item", that is, each item corresponds to a piece of content. When it can be determined which item the user is concerned about based on the information input by the user, the content corresponding to the item, that is, the content result, can be output. In this embodiment, each item contains one or more key information, that is, knowledge points. The knowledge points of each item are represented in an array structure, so that all the knowledge points of all items can be represented by a matrix formed by a plurality of array structures; at the same time, each knowledge point has an attribute of knowledge type, and the knowledge points with the same knowledge type are used as a row or a column of the matrix, so that a knowledge matrix can be formed. A form of this knowledge matrix can be seen in Figure 2a shown.

[0031] Figure 2a In, the items are used as rows and the knowledge types are used as columns to form a knowledge matrix. This knowledge matrix contains 8 items. Each row corresponds to the knowledge points of an item, and the knowledge points of all items are divided into five categories: A, B, C, D, and E, forming an 8×5 knowledge matrix. The elements of this knowledge matrix represent the knowledge points corresponding to the corresponding knowledge types in the corresponding items; for example, Figure 2a the element a1 in the first row and the first column in represents the knowledge point corresponding to knowledge type A in item 1. If an element is empty, it means that there is no knowledge point belonging to the corresponding knowledge type in this item; for example, the element in the first row and the second column is empty, which means that there is no knowledge point belonging to knowledge type B in item 1. In addition, under the same knowledge type, the knowledge points of different items can be the same. For example, a1, a2, and a3 can be the same.

[0032] In this embodiment, multi-round conversations are implemented through information in matrix form (i.e., the knowledge matrix). This knowledge matrix records corresponding knowledge points according to matters and knowledge types, with a simple structure. Moreover, in the process of generating the knowledge matrix, knowledge points of each matter can be determined through methods such as word segmentation processing and semantic recognition. Furthermore, the knowledge types of different knowledge points can be summarized, and the process of generating the knowledge matrix is also simple. Different from a knowledge graph, when generating a knowledge matrix, it is not necessary to accurately determine entities, let alone pay attention to the relationships between entities. In addition, when extracting knowledge points of a matter, even if the extraction methods are different and different knowledge points are extracted, it does not affect the subsequent unique positioning of the corresponding matter based on the knowledge points. Two similar knowledge points of different matters, even if they are not classified into the same knowledge type, do not affect the result of multi-round conversations. For example, as shown in Figure 2a If the difference between matter 3 and matter 4 lies in that a3 and a4 are different, that is, b3 = b4 and c3 = c4. At this time, even if d3 and e4 are two similar knowledge points, although they are classified into knowledge type D and knowledge type E respectively, based on the information input by the user, it is still possible to determine whether the knowledge point the user is concerned about is a3 or a4, that is, whether the user is concerned about matter 3 or matter 4. Therefore, when extracting knowledge points of a matter and determining the knowledge type of the knowledge points, even if there are deviations, it does not affect the content required by the user finally determined, so it has high reliability. In complex application scenarios, because the number of matters included is large, it is difficult to extract knowledge points of matters according to a unified standard, and it is also difficult to accurately classify knowledge points to determine the knowledge type. However, based on this knowledge matrix, multi-round conversations can still be better implemented.

[0033] In addition, since fields such as government affairs contain different types of matters, and the number of corresponding matters of each type is huge. For example, in the field of government affairs, there are topics such as provident fund, marriage registration, and social assistance. Each topic corresponds to a category, and each topic can contain multiple matters. If all topics are comprehensively generated into a single knowledge matrix, it will result in an overly large knowledge matrix, which not only occupies a large amount of storage space but also makes it difficult to achieve positioning when using the knowledge matrix to implement multi-round conversations, and the efficiency of multi-round conversations is low. In the embodiment of the present invention, multiple knowledge matrices are generated to reduce the scale of each knowledge matrix; and if a certain matter is associated with other knowledge matrices or matters in other knowledge matrices, corresponding association relationships are established to achieve the association between knowledge matrices.

[0034] Specifically, the first knowledge matrix and the second knowledge matrix are two knowledge matrices. If a certain item in the first knowledge matrix can be associated with the second knowledge matrix, then this item is called a jump item, and the corresponding association relationship is established. In addition, this association relationship is divided into two cases: First, there is an association relationship between the jump item and the second knowledge matrix; Second, there is an association relationship between the jump item and a certain item in the second knowledge matrix, and this item in the second knowledge matrix is called a calibration item. In the embodiments of the present invention, both the "jump item" and the "calibration item" are a type of item, but there can be an association relationship between the two, and it is possible to transfer from the jump item to the calibration item.

[0035] See Figure 3 As shown, if there are currently two knowledge matrices, namely the "provident fund" knowledge matrix and the "provident fund query" knowledge matrix, and there is an association relationship between the item "how to query the deposit and withdrawal of personal housing provident fund" in the "provident fund" knowledge matrix and the "provident fund query" knowledge matrix, that is, the item "how to query the deposit and withdrawal of personal housing provident fund" is a jump item, and the "provident fund query" knowledge matrix is the second knowledge matrix associated with this jump item.

[0036] In addition, optionally, the jump matrix can establish an association relationship with one or more second knowledge matrices, or establish an association relationship with one or more calibration items, which can be specifically determined based on the actual situation.

[0037] Step 102: Determine the corresponding current knowledge matrix according to the information input by the user, and locate the first target item in the current knowledge matrix according to the information input by the user.

[0038] In the embodiments of the present invention, during the multi-round conversation process, the user will input corresponding information. According to the information input by the user, it can be determined which preset knowledge matrices correspond to the user's information, and the corresponding knowledge matrix is used as the current knowledge matrix. Then, through the multi-round conversation method, the item required by the user in the current knowledge matrix, that is, the first target item, is determined.

[0039] Optionally, the above step 102 "locate the first target item in the current knowledge matrix according to the information input by the user" includes:

[0040] Step A1: Obtain the current-round information input by the user in the current round, extract the target knowledge point in the current-round information, and determine the relevant items in the current knowledge matrix based on the target knowledge points of all current rounds; when the first target item required in the current knowledge matrix cannot be located based on the target knowledge points of all current rounds, generate the current-round reply according to other knowledge types except the knowledge type to which the target knowledge point belongs.

[0041] Step A2: Obtain the next-round information input by the user according to the current-round reply, extract the target knowledge points in the next-round information, and determine the relevant matters in the current knowledge matrix again until the first target matter is located based on the target knowledge points of all current rounds.

[0042] In the embodiments of the present invention, the user can obtain the content he needs based on the multi-round conversation mode. In each round, the user inputs corresponding information, such as questions, etc. Specifically, in the current round, the user inputs the current-round information. At this time, the knowledge points included in the current-round information, that is, the target knowledge points, can be extracted based on methods such as word segmentation processing and semantic recognition, and the knowledge matrix is matched based on all the determined target knowledge points at present. If the target matter required in the knowledge matrix, that is, the first target matter, can be located, it means that the content required can be located based on all the target knowledge points input by the user. At this time, the user does not need to submit information anymore, that is, there is no need to continue the multi-round conversation. On the contrary, if the first target matter in the knowledge matrix cannot be located based on all the target knowledge points obtained so far, it means that the information provided by the user is not enough. In the current round, a current-round reply can be generated to instruct and guide the user to input information based on this current-round reply, that is, the next-round information. Similar to the current-round information, the next-round information can also include target knowledge points, and it is determined again whether the first target matter can be located at this time. Repeat the above multi-round conversation process until the first target matter in the knowledge matrix is located.

[0043] In the embodiments of the present invention, the target matter is the matter currently required by the user; the target matter is not a pre-determined matter, and in different multi-round conversations, the target matter may also be different. Specifically, it can be determined whether the required target matter is located based on the number of matters that can be located currently. In this embodiment, when the number of matters located based on the target knowledge points of all rounds exceeds the preset threshold, it is considered that the required target matter in the knowledge matrix cannot be located currently. For example, the preset threshold is set to 2. If the user can still determine more than 2 matters after inputting the current-round information, it means that the information input by the user at this time is not enough, and the next-round conversation is needed to guide the user to input more information; if the user can locate less than two matters, or even can locate a unique matter after inputting the current-round information, it can be considered that the target matter is located at this time. If the preset threshold is 1, the multi-round conversation needs to end when a unique matter in the knowledge matrix is located.

[0044] Alternatively, among the matters located based on the target knowledge points of all rounds, if there is an overlap among the matters corresponding to the knowledge points of at least two knowledge types (or all knowledge types) other than the knowledge type to which the target knowledge point belongs, it is also considered that the target matter (such as the first target matter) required in the knowledge matrix cannot be located based on the target knowledge points of all rounds; otherwise, if there is no overlap, that is, the knowledge points of at least two knowledge types (or all knowledge types) among other knowledge types all correspond to different matters, it means that generating the current round of response based on one knowledge type is not sufficient to cover all possibilities. At this time, it is not suitable to continue the multi-round conversation, so it is considered that the target matter has been located. As Figure 2a shown, if the knowledge points corresponding to matters 3, 4, and 7 of knowledge types A, B, and C are respectively the same, that is, a3, a4, and a7 are the same, b3, b4, and b7 are the same, c3, c4, and c7 are the same, and currently matters 3, 4, and 7 can be located based on knowledge types A, B, and C, and at this time, the current round of response has not been generated based on knowledge types D and E; among matters 3, 4, and 7, the knowledge points d3 and d7 of knowledge type D correspond to matters 3 and 7, and the knowledge point e4 of knowledge type E corresponds to matter 4, that is, the knowledge points of the two knowledge types D and E both correspond to different matters. At this time, the conversation ends, and matters 3, 4, and 7 are the target matters. At the same time, when "generating the current round of response according to other knowledge types other than the knowledge type to which the target knowledge point belongs", it is necessary to generate the current round of response according to the knowledge types among other knowledge types other than the knowledge type whose knowledge points correspond to different matters from those of another other knowledge type. In another example, if the knowledge points a3, a4, and a7 are the same, the knowledge points b3, b4, and b7 are the same, and currently matters 3, 4, and 7 are located based on knowledge types A and B, at this time, since the knowledge points of knowledge types D and E correspond to different matters, the current round of response can only be generated based on knowledge type C.

[0045] In addition, when generating the current round of response, it is necessary to be based on other knowledge types other than the knowledge type to which the target knowledge point belongs. Specifically, when determining the knowledge type, a name with semantic meaning can be defined for this knowledge type. At this time, the current round of response can be generated based on the knowledge type with semantic meaning. For example, the knowledge points belonging to a certain knowledge type are "adult", "minor", etc., which are related to age. At this time, the knowledge type can be defined as "age", and when generating the current round of response based on this knowledge type, the current round of response can be "What is your age?". Or, the current round of response can also be generated based on the knowledge points corresponding to other knowledge types. At this time, it is allowed that the knowledge type only has an ID representation. Still taking the knowledge points "adult" and "minor" as an example, at this time its knowledge type can be "Knowledge Type A", and the current round of response can be a selection form of response, such as "Are you an adult or a minor?".

[0046] Optionally, the above step A1 "generating a current round reply according to other knowledge types except the knowledge type to which the target knowledge point belongs" includes:

[0047] Step A11: other knowledge types except the knowledge type to which the target knowledge point belongs are regarded as pending knowledge types.

[0048] Step A12: Determine the priority of each pending knowledge type, take the pending knowledge type with the highest priority as the target knowledge type, and generate the current round response according to the target knowledge type.

[0049] In an embodiment of the present invention, the priority of each knowledge type can be predetermined. For example, the fewer the text types of knowledge points corresponding to a knowledge type, the higher the probability that the user will exclude more knowledge points that the user does not pay attention to when entering the knowledge points under the knowledge type, and the fewer the text types, the more convenient it is for the user to select, so the priority of the knowledge type can be higher. Among them, the text type of the knowledge point refers to the category determined by classifying the text similarities and differences of the knowledge point. For example, in knowledge type 3 in Table 2, the knowledge points of item 1 and item 8 are both "minimum living security objects", and the texts of the two are the same, so the two belong to the same text category; accordingly, knowledge type 3 can be divided into multiple text categories such as "minimum living security", "low-income families", "extremely poor people", and "temporary relief objects". Alternatively, the more easy-to-understand the knowledge points corresponding to the knowledge type are, the easier it is for the user to accurately enter the knowledge points under the knowledge type, and the higher the priority of the knowledge type. Other methods can also be used to determine the priority of the knowledge type, which is not limited here. In this embodiment, the pending knowledge type with the highest priority is used as the target knowledge type to generate the current round of replies, making it easier for users to enter key knowledge points in the next round and facilitating the location of the first target item as soon as possible, which is conducive to simplifying the number and process of multi-round conversations.

[0050] In addition, those skilled in the art can understand that when determining the pending knowledge types, it is necessary to exclude all knowledge types to which the target knowledge points of all current rounds belong. Figure 2a As shown, in a multi-round conversation process, if knowledge type A is determined in the first round, and knowledge type B is used as the target knowledge type in the second round, in the third round, all knowledge types except knowledge types A and B (i.e., knowledge types C, D, and E) can be used as pending knowledge types.

[0051] by Figure 2b For example, Figure 2bDifferent letter symbols represent different knowledge points. In the first round, if the user input is knowledge point a1, at this time, a response for the current round can be generated in knowledge type C, and [c1, c2] can be output for the user to choose; if the user inputs knowledge point a2, then [b4, b6, b7] can be output; if the user inputs a2 and b4 at one time, then item 5 can be uniquely located and the conversation ends. If the content input by the user does not contain knowledge points, then [a1, a2, a3] can be output for the user to choose. Taking the complete conversation process as an example, in the first round, the user input is knowledge point a1 and the output is [c1, c2]; in the second round, the knowledge point input by the user is c2 and the output is [b3, b4]; in the third round, the knowledge point input by the user is b3, and at this time item 3 can be located, and this item 3 is the first target item required by the user.

[0052] Step 103: If the first target item is a jump item in the first knowledge matrix, determine the corresponding second knowledge matrix or the calibrated item in the second knowledge matrix according to the association relationship.

[0053] In the embodiment of the present invention, after determining the first target item in the current knowledge matrix, if this first target item can jump to other knowledge matrices, that is, this first target item is a jump item in the first knowledge matrix (the current knowledge matrix is the first knowledge matrix), at this time, the corresponding second knowledge matrix or the calibrated item in the second knowledge matrix can be determined according to the pre-set association relationship. As Figure 3 shown, if the first target item is "how to query the deposit and withdrawal situation of personal housing provident fund", since this item is a jump item, so at this time, according to the determined first target item, it can be further jumped to the corresponding second knowledge matrix, that is, the "provident fund query" knowledge matrix.

[0054] Step 104: Output the corresponding content result according to the second knowledge matrix or the calibrated item in the second knowledge matrix.

[0055] In the embodiment of the present invention, after multiple rounds of conversations, that is, after the user inputs multiple rounds of information, if the second knowledge matrix or the calibrated item in the second knowledge matrix is located, then multiple rounds of conversations can continue based on the second knowledge matrix, and finally the item required by the user can be determined and the corresponding content result can be output; if the calibrated item in the second knowledge matrix is directly located, then the content result related to this calibrated item can be output for the user to view.

[0056] A method for multi-round conversations provided by an embodiment of the present invention records the knowledge points of matters according to matters and knowledge types, so that a knowledge matrix with a simple structure can be generated; when generating the knowledge matrix, it is not necessary to accurately determine entities, let alone pay attention to the relationships between entities, and the generation process is simple. Moreover, when extracting the knowledge points of matters and determining the knowledge types of the knowledge points, even if there are deviations, it does not affect the content finally determined by the user; therefore, even if the complex application scenario contains a large number of matters and it is difficult to extract the knowledge points of matters according to a unified standard, and it is also difficult to accurately classify the knowledge points to determine the knowledge types, the multi-round conversation can still be better implemented based on this knowledge matrix. By associating multiple knowledge matrices or matters in the knowledge matrix in a jump manner, more complete content can be constructed through multiple small knowledge matrices, which can improve the construction efficiency of the knowledge matrix; and by the jump method, the matter required by the user can also be finally located, providing a more accurate content result for the user.

[0057] Based on the above embodiment, step 104 "output the corresponding content result according to the second knowledge matrix or the calibrated matter in the second knowledge matrix" includes:

[0058] Step B1: If there is an association relationship between the jump matter and the second knowledge matrix, then jump to the second knowledge matrix, locate the second target matter according to the second knowledge matrix and the information input by the user, and output the content result corresponding to the second target matter.

[0059] Step B2: If there is an association relationship between the jump matter and the calibrated matter in the second knowledge matrix, then output the content result corresponding to the calibrated matter.

[0060] In the embodiment of the present invention, as described above, the association relationship is divided into two cases. If there is an association relationship between the jump matter and the second knowledge matrix, then based on the first target matter, it is only possible to jump to the second knowledge matrix. At this time, it is still necessary to determine the required matter from the second knowledge matrix, that is, the second target matter, and then output the content result corresponding to the second target matter. Among them, the process principle of "locating the second target matter according to the second knowledge matrix and the information input by the user" is the same as that of step 102 for determining the first target matter above, and will not be elaborated here.

[0061] Alternatively, if there is an association relationship between the jump item and the calibrated item in the second knowledge matrix, it indicates that the corresponding calibrated item can be directly located based on the jump item. In this case, the content result corresponding to the calibrated item can be directly output. Those skilled in the art can understand that the items in the second knowledge matrix can also continue to jump to other knowledge matrices. Even the calibrated items can jump to other knowledge matrices, that is, an item in the second knowledge matrix can be both a calibrated item and a jump object of other knowledge matrices (including the first knowledge matrix), which can be determined based on the actual situation.

[0062] For example, referring to Figure 4 as shown, Figure 4 shows two knowledge matrices related to marriage registration, and item 3 (i.e., the jump item) of one knowledge matrix can jump to item 1 (i.e., the calibrated item) of the other knowledge matrix. That is, when the user needs to query the establishment basis of the marriage registration method (i.e., item 3), they can directly jump to item 1. At this time, the content result corresponding to item 1 is provided to the user. For example, it can be the full text of the marriage registration method. Through the association between the jump item and the calibrated item, items that need to display the same content result can be associated, and there is no need to repeatedly record the same content result for multiple items. In addition, the two knowledge matrices can also be combined into a larger knowledge matrix, that is, jumps can also be realized within the knowledge matrix. As Figure 4 shown, the two knowledge matrices can be combined into a knowledge matrix containing 7 items.

[0063] Optionally, the above step B1 "locate the second target item according to the second knowledge matrix and the information input by the user" includes:

[0064] Step B11: Determine the relevant valid items in the second knowledge matrix according to the information previously input by the user.

[0065] Step B12: Locate the required second target item from the valid items according to the information input by the user later.

[0066] In the embodiment of the present invention, during the multi-round conversation process, the user will input information multiple times. In this embodiment, taking the jump to the second knowledge matrix as the demarcation point, the information input by the user is divided into "information previously input by the user" and "information input by the user later", that is, "information previously input by the user" is the information input by the user before jumping to the second knowledge matrix, and "information input by the user later" is the information input by the user after jumping to the second knowledge matrix. Since the user has input some information before jumping to the second knowledge matrix, based on this part of the information, the second knowledge matrix may also be preliminarily screened to filter out valid items, and then the second target item is located from the valid items. For example, referring to Figure 3As shown, if the information previously input by the user is "housing provident fund payment and withdrawal", at this time, after positioning to the jump item, it jumps to the "housing provident fund query" knowledge matrix; at the same time, according to the previous information, it can be determined that the matters required by the user are those related to "payment and withdrawal", rather than those related to "loan". Therefore, the matters related to "payment and withdrawal" in the "housing provident fund query" knowledge matrix can be used as valid matters, and then the second target matter can be determined through multiple rounds of conversations.

[0067] Based on the above embodiments, the above step 101 "generate multiple knowledge matrices each containing multiple matters" includes:

[0068] Step C1: Obtain multiple matters, perform clustering processing on the matters, and determine the theme of each category of matters.

[0069] In the embodiments of the present invention, multiple matters are obtained, and the knowledge points included in each matter are extracted based on keyword extraction, semantic recognition, etc. Among them, in order to simplify the structure of the knowledge matrix, the knowledge matrix of matters under the same theme can be determined, that is, one subject corresponds to one knowledge matrix. Specifically, all matters can be clustered, so that the matters can be classified, each category of objects corresponds to a theme, and then the knowledge matrix of the theme can be generated according to all the matters corresponding to each category of themes.

[0070] For example, after clustering processing, it is determined that the current theme includes social assistance, and there are 17 matters involved in the social assistance theme. Specifically, see Table 1 below. Taking the first matter as an example, the knowledge points in it can include "recognition", "minimum living security object", "city; rural area"; in addition, the knowledge points can also be expanded synonymously. For example, "confirmation" is a synonym of "recognition", and at this time, the knowledge point "recognition" can be expanded to "recognition / confirmation".

[0071] Table 1

[0072]

[0073]

[0074] Step C2: Extract one or more knowledge points included in each matter of the theme, classify all the knowledge points, and determine the knowledge type corresponding to each type of knowledge point.

[0075] In the embodiments of the present invention, taking the theme as a unit, the knowledge matrix corresponding to each theme is determined respectively. Specifically, for the current theme, the knowledge types can be preset in advance, and the knowledge points are classified based on these knowledge types; alternatively, it is difficult to define the knowledge types in complex scenarios, and in this case, the knowledge points can also be directly classified, for example, by using the clustering method; after the classification is completed, the corresponding knowledge types can be determined.

[0076] Step C3: Generate an M×N or N×M knowledge matrix corresponding to the theme, where M is the number of matters in the theme and N is the number of knowledge types in the theme.

[0077] In this embodiment, after determining the knowledge points of all matters corresponding to the theme and the knowledge types to which the knowledge points belong, the knowledge matrix can be generated. In this embodiment, an M×N knowledge matrix can be generated with matters as rows and knowledge types as columns, or an N×M knowledge matrix can be generated with knowledge types as rows and matters as columns. A form of expression of a knowledge matrix corresponding to Table 1 above can be seen in Table 2. The rows of this knowledge matrix represent matters and the columns represent knowledge types. In addition, in this embodiment, the knowledge types of the knowledge matrix are only used to distinguish different categories of knowledge points, such as Knowledge Type 1, Knowledge Type 2, etc.; this knowledge type can also have semantic meanings. For example, Knowledge Type 3 in Table 2 can be "social security recipients". In addition, for the convenience of description in Table 2, some knowledge points of different types are grouped into the same knowledge type. For example, "city; rural area" and "subsidy" are both grouped into Knowledge Type 4. Generally, only knowledge points with lower priority or importance are likely to be grouped into the same knowledge type. The knowledge matrices corresponding to other themes can also be determined in the same way, which will not be elaborated here.

[0078] Table 2

[0079]

[0080]

[0081] Based on the above embodiments, after step 101 "generate multiple knowledge matrices each including multiple matters", the method further includes:

[0082] Step D1: Determine the matrix entry of each knowledge matrix, and the matrix entry includes multiple keywords extracted from the knowledge matrix.

[0083] In addition, after step A1 "extract the target knowledge points in the current round of information", the method further includes:

[0084] Step A3: When the information in the current round is the first-round information, match the target knowledge point of the first-round information with each matrix entry, and use the knowledge matrix corresponding to the matched matrix entry as the current knowledge matrix.

[0085] In the embodiments of the present invention, when there are multiple knowledge matrices, a simplified matrix entry is set for each knowledge matrix, and based on this matrix entry, it is determined which knowledge matrix the user input information best matches, facilitating quickly locating the required knowledge matrix from multiple knowledge matrices. Specifically, in the first round (i.e., the first round) of a multi-round conversation, the user inputs valid information for the first time, that is, the first-round information. At this time, the corresponding knowledge matrix can be located based on this first-round information. If the information input by the user is useless information, such as "Hello", etc., this process is not regarded as part of the multi-round conversation in this embodiment, that is, the invalid information input by the user at this time is not the first-round information.

[0086] A method for multi-round conversation provided by the embodiments of the present invention records the knowledge points of matters according to matters and knowledge types, so that a knowledge matrix with a simple structure can be generated; when generating the knowledge matrix, it is not necessary to accurately determine entities, and it is even less necessary to pay attention to the relationships between entities, and the generation process is simple. And when extracting the knowledge points of matters and determining the knowledge types of the knowledge points, even if there are deviations, it does not affect the content finally required by the user; therefore, even if a complex application scenario contains a large number of matters and it is difficult to extract the knowledge points of matters according to a unified standard and it is also difficult to accurately classify the knowledge points to determine the knowledge types, the multi-round conversation can still be better implemented based on this knowledge matrix. By associating multiple knowledge matrices or matters in the knowledge matrix in a jump manner, more complete content can be constructed through multiple small knowledge matrices, which can improve the construction efficiency of the knowledge matrix; and through the jump method, the matter required by the user can also be finally located, providing a more accurate content result for the user. Taking the undetermined knowledge type with the highest priority as the target knowledge type to generate the reply in the current round makes it easier for the user to input key knowledge points in the next round, facilitating quickly locating the first target matter, which is beneficial to simplifying the number of rounds and processes of the multi-round conversation.

[0087] The above has described in detail the jumpable multi-round conversation method provided by the embodiments of the present invention. This method can also be implemented by a corresponding device. The following details the jumpable multi-round conversation device provided by the embodiments of the present invention.

[0088] Figure 5 The structural schematic diagram of a jumpable multi-round conversation device provided by the embodiments of the present invention is shown. As Figure 5 shown, the jumpable multi-round conversation device includes:

[0089] A generation module 51, configured to generate a plurality of knowledge matrices each including a plurality of matters, and set an association relationship between a jump matter in a first knowledge matrix and a second knowledge matrix or a calibrated matter in the second knowledge matrix; wherein, each of the matters includes one or more knowledge points divided according to knowledge types, and elements of the knowledge matrix represent knowledge points corresponding to the corresponding knowledge types in the corresponding matters.

[0090] A positioning module 52, configured to determine a corresponding current knowledge matrix according to information input by a user, and position to a first target matter in the current knowledge matrix according to the information input by the user.

[0091] A jump module 53, configured to, if the first target matter is a jump matter in the first knowledge matrix, determine a corresponding second knowledge matrix or a calibrated matter in the second knowledge matrix according to the association relationship.

[0092] A content output module 54, configured to output a corresponding content result according to the second knowledge matrix or the calibrated matter in the second knowledge matrix.

[0093] Based on the above embodiments, the content output module 54 outputs a corresponding content result according to the second knowledge matrix or the calibrated matter in the second knowledge matrix, including:

[0094] If there is an association relationship between the jump matter and the second knowledge matrix, jump to the second knowledge matrix, position to a second target matter according to the second knowledge matrix and the information input by the user, and output a content result corresponding to the second target matter.

[0095] If there is an association relationship between the jump matter and the calibrated matter in the second knowledge matrix, output a content result corresponding to the calibrated matter.

[0096] Based on the above embodiments, the content output module 54 positions to a second target matter according to the second knowledge matrix and the information input by the user, including:

[0097] Determine relevant valid matters in the second knowledge matrix according to information previously input by the user.

[0098] Locate a required second target matter from the valid matters according to information subsequently input by the user.

[0099] Based on the above embodiments, the positioning module 52 positions to a first target matter in the current knowledge matrix according to the information input by the user, including:

[0100] Obtain the current-round information input by the user in the current round, extract the target knowledge points in the current-round information, and determine the relevant matters in the current knowledge matrix based on the target knowledge points in all current rounds; when the first target matter required in the current knowledge matrix cannot be located based on the target knowledge points in all current rounds, generate a current-round response according to other knowledge types except the knowledge type to which the target knowledge points belong.

[0101] Obtain the next-round information input by the user according to the current-round response, extract the target knowledge points in the next-round information, and determine the relevant matters in the current knowledge matrix again until the first target matter is located based on the target knowledge points in all current rounds.

[0102] Based on the above embodiments, the positioning module 52 generates a current-round response according to other knowledge types except the knowledge type to which the target knowledge points belong, including:

[0103] Regard other knowledge types except the knowledge type to which the target knowledge points belong as undetermined knowledge types;

[0104] Determine the priority of each undetermined knowledge type, regard the undetermined knowledge type with the highest priority as the target knowledge type, and generate a current-round response according to the target knowledge type.

[0105] Based on the above embodiments, the device further includes: an entrance setting module;

[0106] After the generation module 51 generates multiple knowledge matrices each including multiple matters, the entrance setting module is used for:

[0107] Determine the matrix entrance of each knowledge matrix, and the matrix entrance includes multiple keywords extracted from the knowledge matrix;

[0108] After the positioning module 52 extracts the target knowledge points in the current-round information, the positioning module 52 is further used for:

[0109] When the current-round information is the first-round information, perform a matching process on the target knowledge points of the first-round information and each matrix entrance, and regard the knowledge matrix corresponding to the matched matrix entrance as the current knowledge matrix.

[0110] Based on the above embodiments, the generation module 51 generates multiple knowledge matrices each including multiple matters, including:

[0111] Obtain multiple matters, perform a clustering process on the matters, and determine the theme of each category of matters;

[0112] Extract one or more knowledge points included in each matter of the said theme, classify all the said knowledge points, and determine the knowledge types corresponding to each type of knowledge points;

[0113] Generate a knowledge matrix of M×N or N×M corresponding to the said theme, where M is the number of matters in the said theme and N is the number of knowledge types in the said theme.

[0114] In addition, an embodiment of the present invention further provides an electronic device, including a bus, a transceiver, a memory, a processor, and a computer program stored on the memory and executable on the processor. The transceiver, the memory, and the processor are respectively connected through the bus. When the computer program is executed by the processor, it implements each process of the above-mentioned multi-round session method embodiment that can jump, and can achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0115] Specifically, referring to Figure 6 As shown, an embodiment of the present invention further provides an electronic device, which includes a bus 1110, a processor 1120, a transceiver 1130, a bus interface 1140, a memory 1150, and a user interface 1160.

[0116] In an embodiment of the present invention, the electronic device further includes: a computer program stored on the memory 1150 and executable on the processor 1120. When the computer program is executed by the processor 1120, it implements each process of the above-mentioned multi-round session method embodiment that can jump.

[0117] The transceiver 1130 is configured to receive and send data under the control of the processor 1120.

[0118] In an embodiment of the present invention, the bus architecture (represented by the bus 1110), the bus 1110 may include any number of interconnected buses and bridges. The bus 1110 connects various circuits including one or more processors represented by the processor 1120 and the memory represented by the memory 1150 together.

[0119] Bus 1110 represents one or more of any of several types of bus structures, including a memory bus and memory controller, a peripheral bus, an Accelerated Graphics Port (AGP), a processor, or a local bus using any of a variety of bus architectures. By way of example and not limitation, such architectures include: Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MCA) bus, Enhanced ISA (EISA) bus, Video Electronics Standards Association (VESA), Peripheral Component Interconnect (PCI) bus.

[0120] Processor 1120 may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method embodiments may be completed by the integrated logic circuit in the processor or instructions in software form. The above processors include: general-purpose processors, central processing units (CPUs), network processors (NPs), digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), complex programmable logic devices (CPLDs), programmable logic arrays (PLAs), microcontroller units (MCUs), or other programmable logic devices, discrete gates, transistor logic devices, discrete hardware components. The various methods, steps, and logic block diagrams disclosed in the embodiments of the present invention may be implemented or executed. For example, the processor may be a single-core processor or a multi-core processor, and the processor may be integrated on a single chip or located on multiple different chips.

[0121] The processor 1120 may be a microprocessor or any conventional processor. The method steps disclosed in connection with the embodiments of the present invention may be directly executed by a hardware decoding processor or executed by a combination of hardware and software modules in the decoding processor. The software module may be located in a readable storage medium well known in the art such as a Random Access Memory (RAM), a Flash Memory, a Read-Only Memory (ROM), a Programmable ROM (PROM), an Erasable PROM (EPROM), a register, etc. The readable storage medium is located in the memory, and the processor reads the information in the memory and combines it with its hardware to complete the steps of the above method.

[0122] The bus 1110 may also connect together various other circuits such as, for example, peripheral devices, voltage regulators, or power management circuits. The bus interface 1140 provides an interface between the bus 1110 and the transceiver 1130, which are all well known in the art. Therefore, the embodiments of the present invention will not be further described herein.

[0123] The transceiver 1130 may be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. For example, the transceiver 1130 receives external data from other devices, and the transceiver 1130 is used to send the data processed by the processor 1120 to other devices. Depending on the nature of the computer system, a user interface 1160 may also be provided, such as a touch screen, a physical keyboard, a display, a mouse, a speaker, a microphone, a trackball, a joystick, a stylus.

[0124] It should be understood that in the embodiments of the present invention, the memory 1150 may further include a memory remotely disposed relative to the processor 1120, and these remotely disposed memories can be connected to the server through a network. One or more parts of the above networks can be an ad hoc network, an intranet, an extranet, a virtual private network (VPN), a local area network (LAN), a wireless local area network (WLAN), a wide area network (WAN), a wireless wide area network (WWAN), a metropolitan area network (MAN), the Internet, a public switched telephone network (PSTN), a plain old telephone service network (POTS), a cellular telephone network, a wireless network, a wireless fidelity (Wi-Fi) network, and a combination of two or more of the above networks. For example, the cellular telephone network and the wireless network can be a Global System for Mobile Communications (GSM) system, a Code Division Multiple Access (CDMA) system, a Worldwide Interoperability for Microwave Access (WiMAX) system, a General Packet Radio Service (GPRS) system, a Wideband Code Division Multiple Access (WCDMA) system, a Long Term Evolution (LTE) system, an LTE Frequency Division Duplexing (FDD) system, an LTE Time Division Duplexing (TDD) system, an Advanced Long Term Evolution (LTE-A) system, a Universal Mobile Telecommunications System (UMTS) system, an Enhance Mobile Broadband (eMBB) system, a massive Machine Type of Communication (mMTC) system, an UltraReliable Low Latency Communications (uRLLC) system, etc.

[0125] It should be understood that the memory 1150 in the embodiments of the present invention can be a volatile memory or a non-volatile memory, or can include both a volatile memory and a non-volatile memory. Among them, the non-volatile memory includes: Read-Only Memory (ROM), Programmable ROM (PROM), Erasable PROM (EPROM), Electrically Erasable PROM (EEPROM), or Flash Memory.

[0126] The volatile memory includes: a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as: static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchlink dynamic random access memory (SLDRAM), and direct rambus random access memory (DRRAM). The memory 1150 of the electronic device described in the embodiments of the present invention includes, but is not limited to, the above and any other suitable types of memory.

[0127] In the embodiments of the present invention, the memory 1150 stores the following elements of the operating system 1151 and the application program 1152: executable modules, data structures, or subsets or extended sets thereof.

[0128] Specifically, the operating system 1151 includes various system programs, such as: framework layer, core library layer, driver layer, etc., for implementing various basic services and processing hardware-based tasks. The application program 1152 includes various application programs, such as: media player, browser, for implementing various application services. The program for implementing the method of the embodiments of the present invention may be included in the application program 1152. The application program 1152 includes: applets, objects, components, logics, data structures, and other computer system executable instructions for performing specific tasks or implementing specific abstract data types.

[0129] In addition, the embodiments of the present invention also provide a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements each process of the above-described embodiments of the multi-round session method that can be jumped, and can achieve the same technical effect. To avoid repetition, it will not be described in detail here.

[0130] A computer-readable storage medium includes: permanent and non-permanent, removable and non-removable media, which are tangible devices that can retain and store instructions for use by an instruction execution device. A computer-readable storage medium includes: electronic storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, and any suitable combination of the foregoing. A computer-readable storage medium includes: 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), non-volatile random access memory (NVRAM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tape storage, magnetic tape disk storage or other magnetic storage devices, memory sticks, mechanical encoding devices (such as punched cards or raised structures in grooves on which instructions are recorded), or any other non-transmission medium that can be used to store information accessible by a computing device. As defined in the embodiments of the present invention, a computer-readable storage medium does not include transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagated through a waveguide or other transmission medium (such as light pulses passing through an optical fiber cable), or electrical signals transmitted through a wire.

[0131] In several embodiments provided in the present application, it should be understood that the disclosed devices, electronic devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules or units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed couplings or direct couplings or communication connections to each other can be indirect couplings or communication connections through some interfaces, devices, or units, and can also be in the form of electrical, mechanical, or other connections.

[0132] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units. They can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to solve the problems to be solved by the solution of the embodiments of the present invention.

[0133] In addition, in each embodiment of the present invention, the functional units can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0134] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the embodiments of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (including: personal computer, server, data center or other network devices) to execute all or part of the steps of the methods described in the embodiments of the present invention. And the above-mentioned storage medium includes various media that can store program codes as listed above.

[0135] In the description of the embodiments of the present invention, those skilled in the art should know that the embodiments of the present invention can be implemented as methods, devices, electronic devices and computer-readable storage media. Therefore, the embodiments of the present invention can be specifically implemented in the following forms: complete hardware, complete software (including firmware, resident software, microcode, etc.), and a combination of hardware and software. In addition, in some embodiments, the embodiments of the present invention can also be implemented in the form of a computer program product in one or more computer-readable storage media, and the computer-readable storage medium contains computer program code.

[0136] As described above, only the specific implementation manners of the embodiments of the present invention are concerned, but the protection scope of the embodiments of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the embodiments of the present invention, and all of them should be covered by the protection scope of the embodiments of the present invention. Therefore, the protection scope of the embodiments of the present invention shall be subject to the protection scope of the claims.

Claims

1. A multi-round conversation method with jump function, characterized in that, Including: Generating a plurality of knowledge matrices each including a plurality of matters, and setting an association relationship between a jump matter in a first knowledge matrix and a second knowledge matrix or a calibrated matter in the second knowledge matrix; wherein, each of the matters includes one or more knowledge points divided according to knowledge types, and elements of the knowledge matrix represent knowledge points corresponding to the corresponding knowledge types in the corresponding matters; rows of the knowledge matrix are matters; columns of the knowledge matrix are knowledge types; the knowledge matrix includes multiple knowledge types; wherein, the process of generating the knowledge matrix includes determining knowledge points of each matter through word segmentation processing and semantic recognition to summarize knowledge types of different knowledge points; Determining a corresponding current knowledge matrix according to information input by a user, and positioning to a first target matter in the current knowledge matrix according to the information input by the user; wherein, positioning to the first target matter in the current knowledge matrix according to the information input by the user includes: obtaining current round information input by the user in the current round, extracting target knowledge points in the current round information, and determining relevant matters in the current knowledge matrix based on the target knowledge points of all current rounds; when the first target matter required in the current knowledge matrix cannot be positioned based on the target knowledge points of all current rounds, generating a current round reply according to other knowledge types except the knowledge type to which the target knowledge points belong; obtaining next round information input by the user according to the current round reply, extracting target knowledge points in the next round information, and determining relevant matters in the current knowledge matrix again until the first target matter is positioned based on the target knowledge points of all current rounds; generating the current round reply according to other knowledge types except the knowledge type to which the target knowledge points belong includes: generating the current round reply according to knowledge types among other knowledge types except the knowledge type to which the target knowledge points belong and except knowledge types corresponding to matters that are different from knowledge points of another other knowledge type; If the first target matter is a jump matter in the first knowledge matrix, determining the second knowledge matrix or the calibrated matter in the second knowledge matrix corresponding to the first target matter according to the association relationship; Outputting a corresponding content result according to the second knowledge matrix or the calibrated matter in the second knowledge matrix.

2. The method according to claim 1, characterized in that, The outputting a corresponding content result according to the second knowledge matrix or the calibrated matter in the second knowledge matrix includes: If there is an association relationship between the jump matter and the second knowledge matrix, jumping to the second knowledge matrix, positioning to a second target matter according to the second knowledge matrix and information input by the user, and outputting a content result corresponding to the second target matter; If there is an association relationship between the jump matter and the calibrated matter in the second knowledge matrix, outputting a content result corresponding to the calibrated matter.

3. The method according to claim 2, characterized in that, The positioning to a second target matter according to the second knowledge matrix and information input by the user includes: Determining relevant valid matters in the second knowledge matrix according to information input by the user before; Locate the required second target matter from the valid matters according to the information input by the user later.

4. The method according to claim 1, wherein Generating the current round of response according to other knowledge types except the knowledge type to which the target knowledge point belongs includes: Regarding other knowledge types except the knowledge type to which the target knowledge point belongs as the pending knowledge types; Determine the priority of each of the pending knowledge types, take the pending knowledge type with the highest priority as the target knowledge type, and generate the current round of response according to the target knowledge type.

5. The method according to claim 1, wherein After generating multiple knowledge matrices each containing multiple matters, it further includes: Determine the matrix entry of each of the knowledge matrices, and the matrix entry includes multiple keywords extracted from the knowledge matrix; After extracting the target knowledge point in the current round of information, it further includes: When the current round of information is the first-round information, perform a matching process on the target knowledge point of the first-round information and each of the matrix entries, and use the knowledge matrix corresponding to the matching matrix entry as the current knowledge matrix.

6. The method according to any one of claims 1 to 3, characterized in that, Generating multiple knowledge matrices each containing multiple matters includes: Obtain multiple matters, perform clustering processing on the matters, and determine the theme of each category of the matters; Extract one or more knowledge points included in each matter of the theme, classify all the knowledge points, and determine the knowledge type corresponding to each category of knowledge points; Generate a knowledge matrix of M×N or N×M corresponding to the theme, where M is the number of matters in the theme and N is the number of knowledge types in the theme.

7. A multi-round conversation device capable of jumping, characterized in that, It includes: A generating module for generating multiple knowledge matrices each containing multiple matters, and setting an association relationship between the jump matter in the first knowledge matrix and the second knowledge matrix or the calibrated matter in the second knowledge matrix; wherein, each of the matters contains one or more knowledge points divided according to knowledge types, and the elements of the knowledge matrix represent the knowledge points corresponding to the corresponding knowledge types in the corresponding matters; the rows of the knowledge matrix are matters; the columns of the knowledge matrix are knowledge types; the knowledge matrix includes multiple knowledge types; wherein, the process of generating the knowledge matrix includes determining the knowledge points of each matter through word segmentation processing and semantic recognition to summarize the knowledge types of different knowledge points. A positioning module, configured to determine a corresponding current knowledge matrix according to the information input by the user, and locate a first target item in the current knowledge matrix according to the information input by the user; wherein, the positioning module is specifically configured to: obtain the current round information input by the user in the current round, extract the target knowledge points in the current round information, and determine the relevant items in the current knowledge matrix based on the target knowledge points of all current rounds; when the first target item required in the current knowledge matrix cannot be located based on the target knowledge points of all current rounds, generate a current round response according to other knowledge types except the knowledge type to which the target knowledge points belong; obtain the next round information input by the user according to the current round response, extract the target knowledge points in the next round information, and determine the relevant items in the current knowledge matrix again until the first target item is located based on the target knowledge points of all current rounds; the generating the current round response according to other knowledge types except the knowledge type to which the target knowledge points belong includes: generating the current round response according to the knowledge types except the knowledge type to which the target knowledge points belong and except the knowledge types corresponding to items that are different from each other with the knowledge points of another other knowledge type; A jump module, configured to, if the first target item is a jump item in the first knowledge matrix, determine a corresponding second knowledge matrix or a calibrated item in the second knowledge matrix according to the association relationship; A content output module, configured to output a corresponding content result according to the second knowledge matrix or the calibrated item in the second knowledge matrix.

8. An electronic device, comprising a bus, a transceiver, a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the transceiver, the memory, and the processor are connected via the bus, and is characterized in that, When the computer program is executed by the processor, it implements the steps in the multi-round session method with jump function as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps in the multi-round session method with jump function as described in any one of claims 1 to 6.

Images