CN111382239A - Method and device for optimizing interaction flow - Google Patents
Method and device for optimizing interaction flow Download PDFInfo
- Publication number
- CN111382239A CN111382239A CN201811613932.8A CN201811613932A CN111382239A CN 111382239 A CN111382239 A CN 111382239A CN 201811613932 A CN201811613932 A CN 201811613932A CN 111382239 A CN111382239 A CN 111382239A
- Authority
- CN
- China
- Prior art keywords
- node
- feedback information
- flow
- ending
- interactive
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Abstract
The invention provides an optimization method of an interactive process, wherein the interactive process comprises a plurality of process nodes, and the optimization method comprises the following steps: acquiring end process nodes in all interaction logs related to the interaction process and end feedback information of the end process nodes; and recommending an optimization scheme for optimizing the interactive process based on the end feedback information of the end process node.
Description
Technical Field
The invention relates to the field of intelligent question answering, in particular to an interactive process optimization method and device.
Background
In the field of intelligent question answering, it is necessary to first determine questions that a user may input, and set preset answers based on the questions. When the questions input in the actual interaction process of the user can be matched with the questions stored in the question-answer database, answers corresponding to the matched questions in the question-answer database can be output.
In the interactive process of the interactive process with a plurality of process nodes, the intelligent question-answering system also obtains the output corresponding to the input of the user based on the special database of the interactive process, and the user continues to generate the next input based on the output, thereby completing the interaction of one process node. However, the interactive processes are also interactive in a preset manner, that is, for a process node, the output of the process node is preset, and the preset input of the user is set based on the output, and when the input of the user can be matched with the preset input, the interactive process flows to the next process node. It is conceivable that there may be various vulnerabilities in the preset interaction flow, and the input of the user is likely not within the preset input range, so that the interaction flow may be ended in the circulation process of any flow node of the interaction flow. However, the terminations may not be the user's intent, but rather because the intelligent question-and-answer system is unable to recognize the user's input. There is therefore a need for a method that can optimize the interaction flow based on the user's input.
Disclosure of Invention
In order to overcome the above-mentioned drawbacks, the present invention aims to provide a method capable of optimizing an interaction flow based on user input.
According to an aspect of the present invention, there is provided an optimization method for an interactive process, the interactive process including a plurality of process nodes, the optimization method including: acquiring end process nodes in all interaction logs related to the interaction process and end feedback information of the end process nodes; and recommending an optimization scheme for optimizing the interactive process based on the end feedback information of the end process node.
Still further, the optimization method further comprises: acquiring all interaction logs related to the interaction flow; acquiring the last piece of feedback information generated by a user in each interactive log and a process node corresponding to the last piece of feedback information; and setting the process node corresponding to the last piece of feedback information as an end process node, wherein the last piece of feedback information is the end feedback information of the end process node.
Further, the recommending and optimizing scheme for optimizing the interactive process based on the feedback information of the end process node includes: and in response to the fact that the frequency of any flow node of the interactive flow as the flow node ending is greater than a preset threshold value, recommending an optimization scheme for optimizing the interactive flow by using the feedback information of the flow node ending.
Further, the recommending and optimizing scheme for optimizing the interactive process by using the feedback information of the process node further includes: classifying or clustering all the ending feedback information of the utilized process nodes based on the node knowledge of the utilized process nodes to obtain a plurality of corpus sets; and recommending an optimization scheme for optimizing the interaction flow based on one or more corpus sets in response to the number of the ending feedback information included in any one or more corpus sets being greater than a preset threshold value.
Further, the optimizing the interaction process based on one or more corpus sets comprises: recommending and optimizing node knowledge corresponding to the classified and generated corpus set based on the ending feedback information in the classified and generated corpus set to serve as the optimization scheme; and recommending and optimizing the dialogs of the previous process nodes of the utilized process nodes or adding the next process nodes of the utilized process nodes as the optimization scheme based on the ending feedback information in the corpus set generated by clustering.
Still further, the obtaining the plurality of corpus sets further comprises: matching all the ending feedback information of the utilized process nodes with all the node knowledge of the utilized process nodes; in response to the fact that any one piece of end feedback information is successfully matched with any one piece of node knowledge of the utilized process node, classifying the end feedback information into a corpus set corresponding to the node knowledge successfully matched with the end feedback information; and in response to failure of matching of any ending feedback information with all node knowledge of the utilized process nodes, clustering the ending feedback information to a corpus set similar to the semantics of the ending feedback information.
Still further, the optimization method further comprises: outputting the optimization scheme for manual validation and executing the manually validated optimization scheme.
According to an aspect of the present invention, there is provided an optimization apparatus for an interactive process, the interactive process including a plurality of process nodes, the optimization apparatus including: an obtaining module, configured to obtain an end flow node in all interaction logs related to the interaction flow and end feedback information of the end flow node; and the recommending module is coupled with the acquiring module and receives the ending process node and the ending feedback information thereof acquired by the acquiring module, and the recommending module is further used for recommending an optimizing scheme for optimizing the interactive process based on the ending feedback information of the ending process node.
According to an aspect of the invention, there is provided a computer apparatus comprising a memory, a processor and a computer program stored on the memory, characterized in that the processor is adapted to carry out the steps of the optimization method according to any of the above when the computer program stored on the memory is executed.
According to an aspect of the present invention, there is provided a computer storage medium having a computer program stored thereon, wherein the computer program when executed implements the steps of the optimization method as described in any one of the above.
Drawings
The above features and advantages of the present disclosure will be better understood upon reading the detailed description of embodiments of the disclosure in conjunction with the following drawings.
FIG. 1 is a flow diagram illustrating a method of creating an interaction flow according to one aspect of the present invention;
FIG. 2 is a flow diagram illustrating a method for optimizing interaction flow according to one embodiment of the present invention;
FIG. 3 is a flow diagram illustrating a classification and clustering process of an optimization method according to one embodiment of the present invention;
fig. 4 is a hardware schematic block diagram of an optimization apparatus according to an embodiment shown in another aspect of the present invention.
Detailed Description
The invention is described in detail below with reference to the figures and specific embodiments. It is noted that the aspects described below in connection with the figures and the specific embodiments are only exemplary and should not be construed as imposing any limitation on the scope of the present invention.
The interactive flow of multiple flow nodes is generally applicable to some specific session scenarios, such as credit card activation interactive flow, and the final purpose of the interactive flow may be to teach the user how to activate or assist the user in performing an activation action. In the interactive process, information such as a credit card number of a user, an activation mode desired by the user, a signature confirmation input by the user, a password setting and the like may need to be acquired, and the information to be acquired constitutes a process element required by the interactive process. The interaction of each flow node is actually the process of asking the user for the specific element instance corresponding to the flow element by the intelligent question-answering system.
To facilitate an understanding of the specific inventive concepts of the present invention, a method 100 for creating a multi-element interaction flow is briefly described. It is understood that the optimization method of the interactive process of the present invention can be the optimization of the interactive process created by the creating method 100 or by other creating methods. The method 100 is created merely to facilitate an understanding of the present invention by those skilled in the art.
As shown in fig. 1, the creation method 100 includes:
s110: determining flow elements required by the interactive flow based on a session scene applicable to the interactive flow; in some specific session scenarios, such as credit card activation, in the interactive process, the smart question-answering system may need to acquire information such as a credit card number of the user, an activation manner desired by the user, signature confirmation input by the user, and password setting, and the information that needs to be acquired constitutes a process element required by the interactive process. The interaction of each flow node is actually the process of asking the user for the specific element instance corresponding to the flow element by the intelligent question-answering system. The process elements can be determined by enterprises based on the possible dialog contents generated by the session scene of the intelligent question-answering system application prepared by the enterprises based on the interactive process, or can be determined statistically through manual interaction records in the session scene.
S120: establishing a first process node related to a first process element, wherein the first process node is provided with a node knowledge base, and the node knowledge base stores element inquiry and at least one node knowledge of the process element corresponding to the first process node;
s130: and establishing a second process node related to the second process element and a circulation path for circulating the node knowledge to the second process node based on each node knowledge stored in the node knowledge base of the first process node.
The element query is a query statement output to a user when the interactive flow flows to a flow node, and is used for expressing specific information required to be provided by the user of the flow node to the user. For example, in a conference room reservation flow, if the number of flow elements corresponding to a certain flow node is the number of people, the corresponding element query may be "how many people participate in the request? "; for example, in a credit card activation process, if a process element corresponding to a certain process node is in an activation mode, the corresponding element query may be "ask you to activate by which mode: short message service or fixed telephone. ". It will be appreciated that in different interaction flows, the manner in which the element query is expressed may be based on the user-specific setting to which it is directed.
The node knowledge is similar to a preset answer set for the element query, and the node knowledge of each process node may be one or more. The content of the feedback generated by the user based on the element query should be within the one or more preset answers. For example, in a conference room reservation flow, node knowledge corresponding to a flow node whose flow element is the number of people should be related to the number of people, and assuming that a conference room is divided into a small size, a medium size, and a large size and that there are 5, 20, and 100 seats corresponding to the flow node, the node knowledge corresponding to the flow node may be 3, and may be 5 or less, 20 or less, and 100 or less, respectively. The feedback information corresponds to less than 5 persons of node knowledge when the feedback information of the user is 1-5 persons, corresponds to less than 20 persons of node knowledge when the feedback information of the user is 6-20 persons, and corresponds to less than 100 persons of node knowledge when the feedback information of the user is 21-100 persons. In the credit card activation process, the node knowledge corresponding to the process node whose process element is the activation mode may be 2, which are respectively a mobile phone short message and a fixed phone. And when the feedback information of the user is the mobile phone short message, the feedback information corresponds to the node knowledge mobile phone short message.
When each process node is created, at least a node knowledge base of the process node is created. The node knowledge base is used for storing element queries of the process nodes and all node knowledge, and each node knowledge comprises a running path. And when the interactive process flow is transferred to the process node, acquiring the element query from the node knowledge base of the process node and outputting the element query to the user. The user generates corresponding feedback information based on the element inquiry, the feedback information is preferentially matched with the node knowledge stored in the node knowledge base of the process node, and when the feedback information is successfully matched with the node knowledge, the interactive process flows to a second process node along the flowing path of the successfully matched node knowledge; when the feedback information cannot be matched with any node knowledge stored in the node knowledge base of the process node, it indicates that the user may generate feedback information that is not preset in the current process node, and the current interactive process cannot determine the next circulation path, and the current interactive process is ended.
In the actual circulation process of the interactive process, the situation that the feedback information of the user cannot be matched with the node knowledge of the circulating process node often occurs, and the interactive process generally needs to be completely circulated to the preset last process node to calculate the successful circulation, so that the aim of setting the intelligent question-answering system by an enterprise is fulfilled. Therefore, how to optimize the interactive flow based on the actual circulation situation of the interactive flow is a problem to be solved urgently.
In order to solve the above problem, according to an aspect of the present invention, an interactive process optimization method is provided, which is suitable for an interactive process including a plurality of process nodes.
In one embodiment, as shown in fig. 2, the optimization method 200 includes:
s210: the method comprises the steps of obtaining end process nodes in all interaction logs related to an interaction process and end feedback information of each end process node.
When a user interacts with the intelligent question-answering system in an interaction process, the intelligent question-answering system records all inputs of the user and all outputs of the intelligent question-answering system based on each interaction, and all conversation contents between the start of the interaction process and the end of the interaction process of one user are called an interaction log. Since the flow of the interactive process is actually the interaction of a plurality of process nodes, the interaction of each process node comprises an output and an input. When a certain process node obtains the input of the user in the circulation process but does not generate the output of the next process node, the process node is called an end process node, and the feedback information of the process node is the end feedback information of the end process node.
S220: and recommending and optimizing an optimization scheme of the interactive process based on the end feedback information of the end process node of the interactive process.
It can be understood that the reason for ending the interactive process can be obtained by performing statistical analysis on the ending feedback information of the ending process node, so as to obtain an optimization scheme for optimizing the interactive process.
Further, step S210 includes:
s211: acquiring all interaction logs related to the interaction flow;
s212: acquiring the last piece of feedback information generated by a user in each interactive log and a process node corresponding to the last piece of feedback information;
s213: and setting the process node corresponding to the last piece of feedback information as an end process node, and setting the last piece of feedback information as end feedback information of the end process node.
It can be understood that the abnormal ending of the interactive process is generally because the feedback information of the user cannot be matched with the starting knowledge of the corresponding process node, and the interactive process cannot start another process node based on any starting knowledge. Therefore, the last record in each interaction log must be the ending feedback information input by the user, and the last record of the ending feedback information is the element query corresponding to the ending feedback information, and the flow node corresponding to the element query is the ending flow node.
Specifically, each element query in the interaction log may have information indicating the corresponding process node, and then the process node corresponding to the information indicating the last process node in the interaction log may be directly obtained as the end process node. Or, the end flow node is obtained based on the last output record in the interaction log, that is, the last element query, for example, the element query is matched with the element queries of all flow nodes in the interaction flow associated with the interaction log, and the flow node successfully matched is set as the end flow node.
Furthermore, it can be understood that statistics is performed on the end flow nodes in all the interaction logs of an interaction flow, and when the number of times that a certain flow node is used as an end flow node is large, it indicates that the flow node may have an optimized space. Therefore, step S220 further includes:
s221: and in response to the fact that the frequency of any flow node of an interactive flow as a flow node ending is greater than a preset threshold value, recommending an optimization scheme for optimizing the interactive flow by using the feedback information of the flow node ending.
Specifically, step S221 includes:
s2211: classifying or clustering all the ending feedback information of the utilized process nodes based on the node knowledge of the utilized process nodes to obtain a plurality of corpus sets; assuming that the number of times that the ith process node of an interactive process is taken as the process ending node is greater than a preset threshold value, taking the ith process node as the process ending node, and taking the feedback information when the ith process node is taken as the process ending node as the feedback ending information. It can be understood that, in some interaction logs, the ith process node has a normal flow, that is, a user can determine a process node of a next flow based on feedback information input by a factor query of the ith process node, in the interaction logs, the feedback information of the ith process node is meaningless for optimizing an interaction process, and only the feedback information when the ith process node is used as an end process node may be used for analyzing an end reason of the interaction process.
S2212: and recommending an optimization scheme for optimizing the interactive process based on the one or more corpus sets in response to the number of the ending feedback information included in any one or more corpus sets being larger than a preset threshold.
It can be understood that, when the amount of the end feedback information included in a certain corpus set is large, it can be laterally proved that the probability that the user generates the end feedback information in the corpus set based on the element query of the ith process node is large, and then the optimization effect of the optimization scheme recommended based on the end feedback information in the corpus set is larger. When the amount of the ending feedback information included in a certain corpus set is small, it can be laterally proved that the probability that the user generates the ending feedback information in the corpus set based on the element query of the ith process node is small, that is, the ending feedback information in the corpus combination is not representative, and may even be the wrong input of the user, which is not suitable for optimizing the interactive process. Therefore, the purpose of steps S2211 to S2212 is actually to hope to generate an optimization scheme for optimizing the interactive flow based on representative end feedback information, and to complete the screening of the end feedback information, so as to reduce the number of recommended optimization schemes and improve the quality of the optimization schemes.
Specifically, the classification or clustering method in step S2211 can be completed by the following steps, as shown in fig. 3:
s310: matching all the ending feedback information of the utilized process nodes with all the node knowledge of the utilized process nodes;
it will be appreciated that the method by which the end flow node matches the feedback information to all of its node knowledge during the flow process should be different from the method by which the end flow node matches the feedback information to its node knowledge during the flow process.
Specifically, whether the node knowledge successfully matched with the end feedback information exists can be judged by calculating the similarity between each end feedback information and all the node knowledge of the utilized process nodes and according to the size of the similarity value. For example, the node knowledge with the maximum similarity to the end feedback information is selected, the maximum similarity is compared with a preset threshold, when the maximum similarity is greater than the preset threshold, it can be determined that the matching is successful, otherwise, the matching is failed.
The specific similarity calculation method may adopt a combination of one or more of the following modes: a calculation method based on a Space Vector Space Model (VSM), a calculation method based on an invisible semantic indexing Model (LSI), a semantic similarity calculation method based on an attribute theory, or a semantic similarity calculation method based on a hamming distance. Those skilled in the art will appreciate that the similarity calculation method may also be or be combined with other semantic similarity calculation methods.
S320: in response to the fact that any one piece of end feedback information is successfully matched with any one piece of node knowledge of the utilized process node, classifying the end feedback information into a corpus set corresponding to the node knowledge successfully matched with the end feedback information;
a corpus set can be set based on knowledge of each node of the node for ending the process, and ending feedback information successfully matched with the knowledge of the node is classified into the corpus set corresponding to the knowledge of the node.
S330: and in response to failure of matching of any ending feedback information with all node knowledge of the utilized process nodes, clustering the ending feedback information to a corpus set similar to the semantics of the ending feedback information.
For the finishing feedback information which cannot be successfully matched with any node knowledge, two finishing feedback information with the similarity value larger than a preset threshold value can be clustered into a corpus set by calculating the similarity between the two finishing feedback information with failed matching. And then, continuing to calculate the similarity between the other ending feedback information and the ending feedback information in the corpus set, and when the similarity between the other ending feedback information and any ending feedback information in the corpus set is greater than a preset threshold, judging that the semantics of the other ending feedback information is similar to that of the ending feedback information in the corpus set, and clustering to the corpus set. Otherwise, a corpus set may be generated based on the other ending feedback information, or the similarity may be continuously calculated with the ending feedback information in the other corpus set until the similarity with the other ending feedback information is clustered to a preset threshold corpus set with semantic similarity.
In other embodiments, the classification or clustering method in step S2211 can also be performed by a deep learning method, and one or more corpus sets can also be obtained.
Further, the recommending an optimization scheme for optimizing the interaction flow based on the one or more corpus sets in step S2212 may specifically include:
s2213: and recommending and optimizing the node knowledge corresponding to the ending feedback information in the corpus set generated by classification to serve as an optimization scheme for optimizing the interactive flow. It can be understood that the interaction process cannot continue to flow unless the end feedback information is successfully matched with any node knowledge in the normal flow process of the end feedback node, and the matching of the end feedback information and a node knowledge in step S310 is successful through another matching method, which indicates that the node knowledge of the end process node may need to be optimized.
Specifically, all the finishing feedback information in the corpus collection generated by each classification can be added to the generalization knowledge of the node knowledge successfully matched with the finishing feedback information, and the generalization knowledge and the corresponding node knowledge share the same circulation path and have the same function.
In the process of matching the feedback information, the node knowledge with the generalized knowledge is added, and as long as the feedback information is successfully matched with the node knowledge or the generalized knowledge of the node knowledge, the interaction process can continue to flow through the flow path of the node knowledge, so that the probability of successful matching of the feedback information in the interaction process is greatly increased, and the flow success rate of the interaction process is improved.
S2214: and recommending and optimizing the dialogs of the previous process nodes of the utilized process nodes or adding the next process nodes of the utilized process nodes as the optimization scheme for the ending feedback information in the corpus set generated by clustering.
All the ending feedback information in the corpus set generated by clustering is ending feedback information which cannot be successfully matched with any node knowledge of the ending process node, and the ending feedback information may express other intentions of the user. Step S2214 is presented below distinctively based on two hypothetical scenarios.
In the circulation process of a flow node corresponding to a flow element of an interactive flow for purchasing a financial product, the intelligent question and answer system recommends names of multiple financial products for a user, the user is required to input the name of one financial product to continue circulation of the next flow node, but when the user is not satisfied with the recommended multiple financial products or cannot immediately decide which money to purchase, the user may input 'do other products exist' or 'I consider again'. It will be appreciated that this situation is particularly likely to occur during the interactive process of product purchase, and that the feedback information does not match the knowledge of any node of the process node, and thus the feedback information is identified as the end feedback information. However, the end feedback information indicates that the user actually has a desire to purchase, but the recommended product does not meet the user's desire, and for the feedback information, the optimal optimization scheme is to add a next process node of the end process node, and the end feedback information can be used as node knowledge for starting the next process node. In this scenario, the added subordinate process node may be to recommend another different batch of financial products.
In the interactive process of a flow node corresponding to a flow element 'financial product type' of an interactive flow for purchasing a financial product, after a user inputs the name of the financial product, the interactive flow is streamed to the next confirmed purchase flow node along the streaming path of the node knowledge corresponding to the name of the financial product. In the circulation process of confirming the purchase process node, the intelligent question-answering system needs to purchase the financing product input by the user of the previous process node through the account of the user, in the process, the intelligent question-answering system may find that the account balance of the user is not enough to purchase the financing product, the intelligent question-answering system may automatically finish circulation or output a prompt message of ' purchase failure ', and the user may continue to generate ' how much money there is or ' what product i can purchase ' according to the prompt message. However, for the interactive process, in the process of confirming the purchase process node, the interactive process is ended due to insufficient balance, the feedback information subsequently input by the user cannot be matched with the next flow process node for the interactive process, and therefore when the user outputs the feedback information of "how much money is left by i" or "what product can be purchased by i", the interactive process does not start the interaction of the next flow process node. However, in the above scenario, the user is actually having a desire to purchase, simply because the selected financial product amount does not match its account balance, resulting in a failed purchase. Then the words of the flow node immediately preceding the end flow node may be modified based on the end feedback information "how much i have money there" or "what product i can buy", for example, during the interaction of the flow node corresponding to the flow element "kind of financing product", the basic query is modified to "your account balance is, you want to buy which financing product: A. b, C'.
Actually, according to the optimization scheme in the previous scenario, for the ending feedback information "how much money i have" or "what product i can buy" a process node may be added to the interactive process based on the ending feedback information to output the account balance of the user. However, it is conceivable that the preset of the interactive process is set based on information required by the interactive process, and when unnecessary process nodes are added to the interactive process to optimize the interactive process, the number of the process nodes of the interactive process will increase, and the database for storing the related information of the interactive process will also increase. Moreover, due to the specificity of the user, various requirements may be generated, and if the process nodes of the interactive process are continuously added based on the requirements of the user, the final appearance of the interactive process may be completely different from the initial setting, even deviating from the center of gravity. The process of optimizing the previous process node does not increase the process node of the interactive process, but achieves the purpose of optimization, so that the process of optimizing the previous process node of the end process node based on the end feedback information can be selected when the newly added process node based on the end feedback information is probably unrelated to the process element of the interactive process.
Still further, the optimization method 200 further comprises:
s230: outputting the optimization scheme for manual confirmation and executing the manually confirmed optimization scheme.
It should be noted that the preset thresholds related to different attributes in different steps of the optimization method 200 are different, for example, the preset threshold of the similarity is different from the preset threshold of the number of times that the process node is used as the end process node or the preset threshold of the number of end feedback information in the corpus problem set.
While, for purposes of simplicity of explanation, the methodologies are shown and described as a series of acts, it is to be understood and appreciated that the methodologies are not limited by the order of acts, as some acts may, in accordance with one or more embodiments, occur in different orders and/or concurrently with other acts from that shown and described herein or not shown and described herein, as would be understood by one skilled in the art.
According to another aspect of the present invention, an apparatus for optimizing an interaction flow is provided.
In one embodiment, as shown in fig. 4, the optimization apparatus 400 includes an obtaining module 410 and a recommending module 420.
The obtaining module 410 is used for obtaining the end process nodes in all the interaction logs related to an interaction process and the end feedback information of each end process node.
When a user interacts with the intelligent question-answering system in an interaction process, the intelligent question-answering system records all inputs of the user and all outputs of the intelligent question-answering system based on each interaction, and all conversation contents between the start of the interaction process and the end of the interaction process of one user are called an interaction log. Since the flow of the interactive process is actually the interaction of a plurality of process nodes, the interaction of each process node comprises an output and an input. When a certain process node obtains the input of the user in the circulation process but does not generate the output of the next process node, the process node is called an end process node, and the feedback information of the process node is the end feedback information of the end process node.
The recommending module 420 is coupled to the obtaining module 410, and is configured to recommend an optimization scheme for optimizing an interactive process based on the ending feedback information of the ending process node of the interactive process obtained by the obtaining module 410.
It can be understood that the reason for ending the interactive process can be obtained by performing statistical analysis on the ending feedback information of the ending process node, so as to obtain an optimization scheme for optimizing the interactive process.
Specifically, the obtaining module 410 first obtains all interaction logs related to the interaction process, obtains the last feedback information generated by the user in each interaction log and the process node corresponding to the last feedback information, sets the process node corresponding to the last feedback information as the end process node, and sets the last feedback information as the end feedback information of the end process node.
It can be understood that the abnormal ending of the interactive process is generally because the feedback information of the user cannot be matched with the starting knowledge of the corresponding process node, and the interactive process cannot start another process node based on any starting knowledge. Therefore, the last record in each interaction log must be the ending feedback information input by the user, and the last record of the ending feedback information is the element query corresponding to the ending feedback information, and the flow node corresponding to the element query is the ending flow node.
Specifically, each element query in the interaction log may have information indicating the corresponding process node, and then the process node corresponding to the information indicating the last process node in the interaction log may be directly obtained as the end process node. Or, the end flow node is obtained based on the last output record in the interaction log, that is, the last element query, for example, the element query is matched with the element queries of all flow nodes in the interaction flow associated with the interaction log, and the flow node successfully matched is set as the end flow node.
Further, the optimization apparatus 400 further includes a statistics module 430, and the recommendation module 420 is coupled to the obtaining module 410 through the statistics module 430, and is configured to count the number of times that any flow node of an interactive flow is taken as a flow node end, where it can be understood that when a certain flow node is taken as a flow node end more times, it indicates that the flow node may have an optimization space. Therefore, the recommending module 420 recommends the optimization scheme of the interactive process based on the end feedback information of the process node with a high number of times of ending the process node counted by the counting module 430.
Further, the optimization apparatus 400 further includes a corpus set generating module 440, the recommending module 420 is coupled to the counting module 430 through the corpus set generating module 440, and the corpus set generating module 440 is configured to classify or cluster the ending feedback information of the ending process node, which is counted by the counting module 430 and has the number of times of ending the process node larger than a preset threshold, so as to obtain a plurality of corpus sets. The recommending module 420 recommends an optimization scheme for optimizing the interaction process based on the corpus set of which the number of the ending feedback information counted by the counting module 430 is greater than a preset threshold.
Further, the corpus generating set module 440 further includes a matching unit 441, a classifying unit 442 and a clustering unit 443.
The matching unit 441 is configured to match all the end feedback information in the corpus set, of which the number of the end feedback information counted by the counting module 430 is greater than a preset threshold, with all the node knowledge of the corresponding process node. It will be appreciated that the method by which the end flow node matches the feedback information to all of its node knowledge during the flow process should be different from the method by which the end flow node matches the feedback information to its node knowledge during the flow process.
Specifically, whether the node knowledge successfully matched with the end feedback information exists can be judged by calculating the similarity between each end feedback information and all the node knowledge of the utilized process nodes and according to the size of the similarity value. For example, the node knowledge with the maximum similarity to the end feedback information is selected, the maximum similarity is compared with a preset threshold, when the maximum similarity is greater than the preset threshold, it can be determined that the matching is successful, otherwise, the matching is failed.
The specific similarity calculation method may adopt a combination of one or more of the following modes: a calculation method based on a Space Vector Space Model (VSM), a calculation method based on an invisible semantic indexing Model (LSI), a semantic similarity calculation method based on an attribute theory, or a semantic similarity calculation method based on a hamming distance. Those skilled in the art will appreciate that the similarity calculation method may also be or be combined with other semantic similarity calculation methods.
The classifying unit 442 is coupled to the matching unit 441, and in response to the matching unit 441 determining that any ending feedback information is successfully matched with any node knowledge of the utilized process node, the classifying unit 442 classifies the ending feedback information into the corpus set corresponding to the node knowledge successfully matched therewith.
Specifically, a corpus set may be set based on knowledge of each node of the end flow node, and the end feedback information that is successfully matched with the knowledge of the node is classified into the corpus set corresponding to the knowledge of the node.
The clustering unit 443 is coupled to the matching unit 441, and in response to the matching unit 441 determining that any ending feedback information fails to match with all node knowledge of the utilized process nodes, the clustering unit 443 clusters the ending feedback information into a corpus set with similar semantics.
Specifically, for the end feedback information that cannot be successfully matched with any node knowledge, the clustering unit 443 may cluster two end feedback information with similarity values greater than a preset threshold into a corpus set by calculating similarity between the end feedback information that fails to be matched with other end feedback information. And then, continuing to calculate the similarity between the other ending feedback information and the ending feedback information in the corpus set, and when the similarity between the other ending feedback information and any ending feedback information in the corpus set is greater than a preset threshold, judging that the semantics of the other ending feedback information is similar to that of the ending feedback information in the corpus set, and clustering to the corpus set. Otherwise, a corpus set may be generated based on the other ending feedback information, or the similarity may be continuously calculated with the ending feedback information in the other corpus set until the similarity with the other ending feedback information is clustered to a preset threshold corpus set with semantic similarity.
In other embodiments, the classification or clustering method of the classification unit 442 and the clustering unit 443 can be further performed by a deep learning method, and a corpus can also be obtained.
Further, the recommending module 420 recommends, as an optimization scheme for optimizing the interactive process, node knowledge corresponding to the end feedback information in the corpus set generated by classification. It can be understood that the end feedback information is not successfully matched with any node knowledge in the normal streaming process of the end feedback node, the interactive flow cannot continue to stream, and the matching unit 441 successfully matches the end feedback information with a node knowledge through another matching method, which indicates that the node knowledge of the end flow node may need to be optimized.
Specifically, all the finishing feedback information in the corpus collection generated by each classification can be added to the generalization knowledge of the node knowledge successfully matched with the finishing feedback information, and the generalization knowledge and the corresponding node knowledge share the same circulation path and have the same function.
In the process of matching the feedback information, the node knowledge with the generalized knowledge is added, and as long as the feedback information is successfully matched with the node knowledge or the generalized knowledge of the node knowledge, the interaction process can continue to flow through the flow path of the node knowledge, so that the probability of successful matching of the feedback information in the interaction process is greatly increased, and the flow success rate of the interaction process is improved.
The recommending module 420 recommends, as the optimizing scheme, a word for optimizing a previous flow node of the utilized flow nodes or adding a next flow node of the utilized flow nodes to the end feedback information in the corpus set generated by clustering.
All the ending feedback information in the corpus set generated by clustering is ending feedback information which cannot be successfully matched with any node knowledge of the ending process node, and the ending feedback information may express other intentions of the user. The following procedure for recommending an optimization based on a corpus set generated by clustering by the recommendation module 420 is presented distinctively based on two hypothetical scenarios.
In the circulation process of a flow node corresponding to a flow element of an interactive flow for purchasing a financial product, the intelligent question and answer system recommends names of multiple financial products for a user, the user is required to input the name of one financial product to continue circulation of the next flow node, but when the user is not satisfied with the recommended multiple financial products or cannot immediately decide which money to purchase, the user may input 'do other products exist' or 'I consider again'. It will be appreciated that this situation is particularly likely to occur during the interactive process of product purchase, and that the feedback information does not match the knowledge of any node of the process node, and thus the feedback information is identified as the end feedback information. However, the end feedback information indicates that the user actually has a desire to purchase, but the recommended product does not meet the user's desire, and for the feedback information, the optimal optimization scheme is to add a next process node of the end process node, and the end feedback information can be used as node knowledge for starting the next process node. In this scenario, the added subordinate process node may be to recommend another different batch of financial products.
In the interactive process of a flow node corresponding to a flow element 'financial product type' of an interactive flow for purchasing a financial product, after a user inputs the name of the financial product, the interactive flow is streamed to the next confirmed purchase flow node along the streaming path of the node knowledge corresponding to the name of the financial product. In the circulation process of confirming the purchase process node, the intelligent question-answering system needs to purchase the financing product input by the user of the previous process node through the account of the user, in the process, the intelligent question-answering system may find that the account balance of the user is not enough to purchase the financing product, the intelligent question-answering system may automatically finish circulation or output a prompt message of ' purchase failure ', and the user may continue to generate ' how much money there is or ' what product i can purchase ' according to the prompt message. However, for the interactive process, in the process of confirming the purchase process node, the interactive process is ended due to insufficient balance, the feedback information subsequently input by the user cannot be matched with the next flow process node for the interactive process, and therefore when the user outputs the feedback information of "how much money is left by i" or "what product can be purchased by i", the interactive process does not start the interaction of the next flow process node. However, in the above scenario, the user is actually having a desire to purchase, simply because the selected financial product amount does not match its account balance, resulting in a failed purchase. Then the words of the flow node immediately preceding the end flow node may be modified based on the end feedback information "how much i have money there" or "what product i can buy", for example, during the interaction of the flow node corresponding to the flow element "kind of financing product", the basic query is modified to "your account balance is, you want to buy which financing product: A. b, C'.
Actually, according to the optimization scheme in the previous scenario, for the ending feedback information "how much money i have" or "what product i can buy" a process node may be added to the interactive process based on the ending feedback information to output the account balance of the user. However, it is conceivable that the preset of the interactive process is set based on information required by the interactive process, and when unnecessary process nodes are added to the interactive process to optimize the interactive process, the number of the process nodes of the interactive process will increase, and the database for storing the related information of the interactive process will also increase. Moreover, due to the specificity of the user, various requirements may be generated, and if the process nodes of the interactive process are continuously added based on the requirements of the user, the final appearance of the interactive process may be completely different from the initial setting, even deviating from the center of gravity. The process of optimizing the previous process node does not increase the process node of the interactive process, but achieves the purpose of optimization, so that the process of optimizing the previous process node of the end process node based on the end feedback information can be selected when the newly added process node based on the end feedback information is probably unrelated to the process element of the interactive process.
Further, the optimization apparatus 400 further includes an execution module 450 coupled to the recommendation module 420, wherein the recommendation module 420 outputs the recommended optimization scheme for confirmation by the user, and the execution module 450 executes the optimization scheme in response to the user confirming execution of the optimization scheme.
It should be noted that the preset thresholds related to different attributes in the above description of the optimization apparatus 400 are different, for example, the preset threshold of the similarity is different from the preset threshold of the number of times that the process node is taken as the end process node or the preset threshold of the number of end feedback information in the corpus problem set.
According to a further aspect of the present invention, there is provided an electronic device comprising a memory, a processor and a computer program stored on the memory, wherein the processor is adapted to carry out the steps of the optimization method 200 set forth in any of the embodiments described above when executing the computer program stored on the memory.
According to a further aspect of the present invention, a computer storage medium is provided, having a computer program stored thereon, wherein the computer program, when executed, implements the steps of the optimization method 200 set forth in any of the embodiments above.
Those of skill in the art would understand that information, signals, and data may be represented using any of a variety of different technologies and techniques. For example, data, instructions, commands, information, signals, bits (bits), symbols, and chips that may be referenced throughout the above description may be represented by voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields or particles, or any combination thereof.
Those of skill would further appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
The various illustrative logical modules, and circuits described in connection with the embodiments disclosed herein may be implemented or performed with a general purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing devices, e.g., a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module may reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor such the processor can read information from, and write information to, the storage medium. In the alternative, the storage medium may be integral to the processor. The processor and the storage medium may reside in an ASIC. The ASIC may reside in a user terminal. In the alternative, the processor and the storage medium may reside as discrete components in a user terminal.
In one or more exemplary embodiments, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software as a computer program product, the functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium. Computer-readable media includes both computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. A storage media may be any available media that can be accessed by a computer. By way of example, and not limitation, such computer-readable media can comprise RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. Any connection is properly termed a computer-readable medium. For example, if the software is transmitted from a web site, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair, Digital Subscriber Line (DSL), or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of medium. Disk (disk) and disc (disc), as used herein, includes Compact Disc (CD), laser disc, optical disc, Digital Versatile Disc (DVD), floppy disk and blu-ray disc where disks (disks) usually reproduce data magnetically, while discs (discs) reproduce data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.
The previous description is provided to enable any person skilled in the art to practice the various aspects described herein. It is to be understood that the scope of the invention is to be defined by the appended claims and not by the specific constructions and components of the embodiments illustrated above. Those skilled in the art can make various changes and modifications to the embodiments within the spirit and scope of the present invention, and these changes and modifications also fall within the scope of the present invention.
Claims (10)
1. A method for optimizing an interactive process, the interactive process comprising a plurality of process nodes, the method comprising:
acquiring end process nodes in all interaction logs related to the interaction process and end feedback information of the end process nodes; and
and recommending an optimization scheme for optimizing the interactive process based on the end feedback information of the end process node.
2. The optimization method of claim 1, further comprising:
acquiring all interaction logs related to the interaction flow;
acquiring the last piece of feedback information generated by a user in each interactive log and a process node corresponding to the last piece of feedback information; and
and setting the process node corresponding to the last piece of feedback information as an end process node, wherein the last piece of feedback information is the end feedback information of the end process node.
3. The optimization method of claim 1, wherein recommending an optimization scheme for optimizing an interactive flow based on the feedback information of the end-flow node comprises:
and in response to the fact that the frequency of any flow node of the interactive flow as the flow node ending is greater than a preset threshold value, recommending an optimization scheme for optimizing the interactive flow by using the feedback information of the flow node ending.
4. The optimization method of claim 3, wherein each process node includes node knowledge for starting a next process node, and the recommending an optimization scheme for optimizing an interactive process using feedback information of the process node further comprises:
classifying or clustering all the ending feedback information of the utilized process nodes based on the node knowledge of the utilized process nodes to obtain a plurality of corpus sets; and
and recommending an optimization scheme for optimizing the interaction flow based on the one or more corpus sets in response to the fact that the number of the ending feedback information included in any one or more corpus sets is larger than a preset threshold value.
5. The optimization method of claim 4, wherein optimizing the interaction flow based on the one or more corpus sets comprises:
recommending and optimizing node knowledge corresponding to the classified and generated corpus set based on the ending feedback information in the classified and generated corpus set to serve as the optimization scheme; and
recommending and optimizing the dialogs of the previous process nodes of the utilized process nodes or adding the next process nodes of the utilized process nodes as the optimization scheme based on the ending feedback information in the corpus set generated by clustering.
6. The optimization method of claim 5, wherein said obtaining a plurality of corpus sets further comprises:
matching all the ending feedback information of the utilized process nodes with all the node knowledge of the utilized process nodes;
in response to the fact that any one piece of end feedback information is successfully matched with any one piece of node knowledge of the utilized process node, classifying the end feedback information into a corpus set corresponding to the node knowledge successfully matched with the end feedback information; and
and in response to failure of matching of any ending feedback information with all node knowledge of the utilized process nodes, clustering the ending feedback information to a corpus set similar to the semantics of the ending feedback information.
7. The optimization method of claim 1, further comprising:
outputting the optimization scheme for manual validation and executing the manually validated optimization scheme.
8. An optimization apparatus for an interactive process, the interactive process comprising a plurality of process nodes, the optimization apparatus comprising:
an obtaining module, configured to obtain an end flow node in all interaction logs related to the interaction flow and end feedback information of the end flow node; and
and the recommending module is coupled with the acquiring module and receives the ending process node and the ending feedback information thereof acquired by the acquiring module, and is further used for recommending an optimizing scheme for optimizing the interactive process based on the ending feedback information of the ending process node.
9. A computer arrangement comprising a memory, a processor and a computer program stored on the memory, characterized in that the processor is adapted to carry out the steps of the optimization method according to any one of claims 1 to 7 when executing the computer program stored on the memory.
10. A computer storage medium having a computer program stored thereon, wherein the computer program, when executed, performs the steps of the optimization method according to any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811613932.8A CN111382239B (en) | 2018-12-27 | 2018-12-27 | Interaction flow optimization method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811613932.8A CN111382239B (en) | 2018-12-27 | 2018-12-27 | Interaction flow optimization method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111382239A true CN111382239A (en) | 2020-07-07 |
| CN111382239B CN111382239B (en) | 2023-06-23 |
Family
ID=71219865
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811613932.8A Active CN111382239B (en) | 2018-12-27 | 2018-12-27 | Interaction flow optimization method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111382239B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112015880A (en) * | 2020-10-23 | 2020-12-01 | 和美（深圳）信息技术股份有限公司 | Automatic transfer method and system for man-machine conversation process of intelligent robot |
| CN112015881A (en) * | 2020-10-23 | 2020-12-01 | 和美（深圳）信息技术股份有限公司 | Automatic switching method and system for man-machine conversation process of intelligent robot |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105488185A (en) * | 2015-12-01 | 2016-04-13 | 上海智臻智能网络科技股份有限公司 | Optimization method and apparatus for knowledge base |
| CN105912697A (en) * | 2016-04-25 | 2016-08-31 | 北京光年无限科技有限公司 | Optimization method and device of dialog system knowledge base |
| CN106295807A (en) * | 2016-08-23 | 2017-01-04 | 上海智臻智能网络科技股份有限公司 | A kind of method and device of information processing |
| CN106897067A (en) * | 2017-02-26 | 2017-06-27 | 广州衡昊数据科技有限公司 | A kind of method and expert system based on human-computer interaction technology modeling |
-
2018
- 2018-12-27 CN CN201811613932.8A patent/CN111382239B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105488185A (en) * | 2015-12-01 | 2016-04-13 | 上海智臻智能网络科技股份有限公司 | Optimization method and apparatus for knowledge base |
| CN105912697A (en) * | 2016-04-25 | 2016-08-31 | 北京光年无限科技有限公司 | Optimization method and device of dialog system knowledge base |
| CN106295807A (en) * | 2016-08-23 | 2017-01-04 | 上海智臻智能网络科技股份有限公司 | A kind of method and device of information processing |
| CN106897067A (en) * | 2017-02-26 | 2017-06-27 | 广州衡昊数据科技有限公司 | A kind of method and expert system based on human-computer interaction technology modeling |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112015880A (en) * | 2020-10-23 | 2020-12-01 | 和美（深圳）信息技术股份有限公司 | Automatic transfer method and system for man-machine conversation process of intelligent robot |
| CN112015881A (en) * | 2020-10-23 | 2020-12-01 | 和美（深圳）信息技术股份有限公司 | Automatic switching method and system for man-machine conversation process of intelligent robot |
| CN112015881B (en) * | 2020-10-23 | 2021-02-19 | 和美（深圳）信息技术股份有限公司 | Automatic switching method and system for man-machine conversation process of intelligent robot |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111382239B (en) | 2023-06-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| TWI700632B (en) | User intention recognition method and device | |
| CN107886949B (en) | Content recommendation method and device | |
| US10678894B2 (en) | Disambiguation and authentication of device users | |
| US11106983B2 (en) | Intelligent interaction method and intelligent interaction system | |
| CN107797984B (en) | Intelligent interaction method, equipment and storage medium | |
| JP2019159309A (en) | Method and apparatus for determining speech interaction satisfaction | |
| US10412225B2 (en) | Systems and methods of natural language processing to rank users of real time communications connections | |
| CN110399609B (en) | Intention recognition method, device, equipment and computer readable storage medium | |
| US20200143115A1 (en) | Systems and methods for improved automated conversations | |
| CN108364650B (en) | Device and method for adjusting voice recognition result | |
| CN110071915B (en) | Identity verification product pushing method, device, equipment and system architecture | |
| CN106486120B (en) | Interactive voice response method and answering system | |
| CN111382235A (en) | Question-answer knowledge base optimization method and device | |
| CN111177338A (en) | Context-based multi-turn dialogue method | |
| CN111382239B (en) | Interaction flow optimization method and device | |
| CN112561535A (en) | Transaction dispute data processing method, device, equipment and storage medium | |
| US20190286713A1 (en) | Systems and methods for enhanced natural language processing for machine learning conversations | |
| US11133022B2 (en) | Method and device for audio recognition using sample audio and a voting matrix | |
| US20220019178A1 (en) | Intelligent device and method for controlling the same | |
| CN116881429B (en) | Multi-tenant-based dialogue model interaction method, device and storage medium | |
| CN112182046B (en) | Information recommendation method, device, equipment and medium | |
| CN112735374B (en) | Automatic voice interaction method and device | |
| WO2020139865A1 (en) | Systems and methods for improved automated conversations | |
| CN111310453B (en) | User theme vectorization representation method and system based on deep learning | |
| CN114898755B (en) | Voice processing method and related device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |