CN111930917A - Conversation process mining method and device, computer equipment and storage medium - Google Patents
Conversation process mining method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN111930917A CN111930917A CN202011007891.5A CN202011007891A CN111930917A CN 111930917 A CN111930917 A CN 111930917A CN 202011007891 A CN202011007891 A CN 202011007891A CN 111930917 A CN111930917 A CN 111930917A
- Authority
- CN
- China
- Prior art keywords
- intention
- task
- conversation
- under
- log
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
- G06F16/353—Clustering; Classification into predefined classes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2216/00—Indexing scheme relating to additional aspects of information retrieval not explicitly covered by G06F16/00 and subgroups
- G06F2216/03—Data mining
Abstract
The application provides a conversation process mining method, a conversation process mining device, computer equipment and a storage medium, wherein the conversation process mining method comprises the following steps: performing task clustering on the set of the dialog text logs; performing intention recognition on each pair of spoken sentences in a dialog text log under the same task to obtain an intention set under the same task; for each task, mining the basic relationship among the intentions under the task according to the time sequence of each intention in the corresponding session in the intention set under the task; constructing a control flow structure according to the basic relation; and generating a conversation process corresponding to the task based on the control flow structure. The scheme can improve the construction efficiency of the conversation process.
Description
Technical Field
The present application relates to the field of computer technologies and artificial intelligence technologies, and in particular, to a method and an apparatus for mining a conversation process, a computer device, and a storage medium.
Background
With the development of artificial intelligence technology, intelligent dialogue robots appear in many application scenarios to automatically complete the whole task processing. For example, in a task-based customer service scene, the intelligent conversation robot can replace manual customer service to perform a conversation with the user to complete the whole task processing. The intelligent dialogue robot guides the user to provide word slot information according to a preset task flow and a word slot collection state in a conversation, so that the whole task processing is completed. Therefore, it is the key for the intelligent dialogue robot to complete the task processing through the intelligent dialogue to construct the task flow in advance.
In the traditional method, one or more designated manual customer services summarize the wiring experience of the customer service, so that a task flow is constructed. Thus, the manual customer service can summarize the experience of the customer service itself, which results in low efficiency of constructing the task flow. Moreover, the number of designated manual customer services is limited, so that the experience is limited, a relatively perfect conversation process is difficult to summarize at one time, and the optimization needs to be repeated for many times. Therefore, the traditional method has the problem of low construction efficiency of the task flow.
Disclosure of Invention
In view of the above, it is necessary to provide a method, an apparatus, a computer device and a storage medium for mining a conversation process, which can improve efficiency.
A conversation process mining method, comprising:
performing task clustering on the set of the dialog text logs;
performing intention recognition on each pair of spoken sentences in a dialog text log under the same task to obtain an intention set under the same task;
for each task, mining the basic relationship among the intentions under the task according to the time sequence of each intention in the corresponding session in the intention set under the task;
constructing a control flow structure according to the basic relation;
and generating a conversation process corresponding to the task based on the control flow structure.
In one embodiment, the method further comprises:
obtaining a customer service log;
performing voice recognition on the voice logs in the customer service logs to generate text logs;
and determining a dialog text log according to the text log.
In one embodiment, the customer service log further comprises a text chat log;
said determining a dialog text log according to the text log comprises:
and generating a dialog text log according to the text chatting record and the text log.
In one embodiment, the performing intent recognition on each pair of utterances in the dialog text log under the same task to obtain an intent set under the same task includes:
determining the conversation to which the dialog text logs under the same task belong; each conversation comprises at least one dialogue statement;
performing intention recognition on the conversation sentences in each conversation to obtain at least one intention in each conversation;
and determining an intention set under the same task according to the intention in each session.
In one embodiment, for each task, mining a basic relationship between intentions under the task according to a time sequence of each intention in a corresponding session in the intention set under the task includes:
numbering each intention in an intention set under each task to obtain an intention identifier of each intention under each task;
determining a timestamp corresponding to an intention in each session under the task;
generating a conversation log corresponding to the task according to the conversation identification of each conversation under the task, the intention identification of the intention in the conversation and the timestamp corresponding to the intention in each conversation;
and mining the basic relationship between the intentions under the task from the conversation log.
In one embodiment, the mining, from the conversation log, the basic relationship between the intentions under the task includes:
for any two intentions in the set of intentions under the task, when one of the two intentions occurs before the other intention in one or more sessions in the session log, determining that there is an accompanying relationship in which a later intention of the two intentions is accompanied by a previous intention;
determining that a causal relationship exists between the two intentions when one of the two intentions occurs before the other intent in all sessions of the session log;
and when the two intentions are not before in all the conversations of the conversation log, judging that an unrelated relation exists between the two intentions.
In one embodiment, the constructing a control flow structure according to the basic relationship includes:
and aiming at any two intentions in the intention set under the task, when a causal relationship exists between the two intentions, constructing a reason node and an effect node corresponding to the two intentions respectively, and generating a sequential structure according to the reason node and the effect node.
In one embodiment, the constructing a control flow structure according to the basic relationship includes:
for any first intention, second intention and third intention in an intention set under a task, when the second intention and the third intention are both accompanied with the first intention and an irrelevant relation exists between the second intention and the third intention, constructing a bifurcation node corresponding to the first intention and constructing branch nodes corresponding to the second intention and the third intention respectively, and generating a selection bifurcation structure according to the bifurcation node and the branch nodes.
In one embodiment, the constructing a control flow structure further includes, according to the basic relationship:
when the third intention is respectively accompanied with the first intention and the second intention and an irrelevant relation exists between the first intention and the second intention, respectively constructing nodes to be merged corresponding to the first intention and the second intention and constructing merged nodes corresponding to the third intention, and generating a selection merged structure according to the nodes to be merged and the merged nodes.
In one embodiment, the method further comprises:
acquiring a sentence input by a user;
after the intention of the statement is identified, locating a node corresponding to the identified intention in the task flow;
and determining a next node pointed by the node in the task flow, and outputting a corresponding answer statement based on the corresponding intention of the next node.
A conversation process mining apparatus, the apparatus comprising:
the task dividing module is used for clustering tasks of the set of the dialog text logs;
the intention identification module is used for carrying out intention identification on each pair of spoken sentences in the dialog text log under the same task to obtain an intention set under the same task;
the intention relation mining module is used for mining the basic relation among the intentions under the tasks according to the time sequence of each intention in the corresponding session in the intention set under the tasks;
the conversation process generating module is used for constructing a control flow structure according to the basic relationship; and generating a conversation process corresponding to the task based on the control flow structure.
A computer device comprising a memory and a processor, the memory storing a computer program, the processor implementing the steps of the present application when executing the computer program:
performing task clustering on the set of the dialog text logs;
performing intention recognition on each pair of spoken sentences in a dialog text log under the same task to obtain an intention set under the same task;
for each task, mining the basic relationship among the intentions under the task according to the time sequence of each intention in the corresponding session in the intention set under the task;
constructing a control flow structure according to the basic relation;
and generating a conversation process corresponding to the task based on the control flow structure.
A computer-readable storage medium, on which a computer program is stored which, when executed by a processor, carries out the steps of:
performing task clustering on the set of the dialog text logs;
performing intention recognition on each pair of spoken sentences in a dialog text log under the same task to obtain an intention set under the same task;
for each task, mining the basic relationship among the intentions under the task according to the time sequence of each intention in the corresponding session in the intention set under the task;
constructing a control flow structure according to the basic relation;
and generating a conversation process corresponding to the task based on the control flow structure.
According to the conversation process mining method, the conversation process mining device, the computer equipment and the storage medium, task clustering is firstly carried out on a set of conversation text logs; and then performing intention recognition on each pair of spoken sentences in the dialog text log under the same task to obtain an intention set under the same task. Since the intentions can represent the purpose of the conversation, the basic relationship among the intentions under each task is mined according to the time sequence of each intention in the corresponding conversation in the intention set under each task, and then a control flow structure is constructed. Based on the control flow structure, the conversation process corresponding to each task can be automatically generated. The method and the device realize automatic excavation of the conversation process from the conversation log, and improve efficiency compared with the traditional manual construction of the conversation process.
Drawings
FIG. 1 is a diagram of an application environment for a conversation process mining method in one embodiment;
FIG. 2 is a flow diagram that illustrates a method for conversation process mining, according to one embodiment;
FIG. 3 is a diagram of a dialog text log in one embodiment;
FIG. 4 is a simplified flow diagram of a method for conversation process mining in one embodiment;
FIG. 5 is a block diagram showing the construction of a conversation process mining apparatus according to one embodiment;
FIG. 6 is a block diagram showing the construction of a conversation process mining apparatus according to another embodiment;
FIG. 7 is a diagram illustrating an internal structure of a computer device according to an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
The conversation process mining method provided by the application can be applied to the application environment shown in fig. 1. Wherein the server 102 communicates with the terminal 104 via a network. The terminal 104 may be, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices, and the server 102 may be implemented by an independent server or a server cluster composed of a plurality of servers.
Server 102 may perform task clustering on the set of dialog text logs; performing intention recognition on each pair of spoken sentences in a dialog text log under the same task to obtain an intention set under the same task; for each task, mining the basic relationship between intentions under the task according to the time sequence of each intention in the corresponding session in an intention set under the task; constructing a control flow structure according to the basic relation; and generating a conversation flow corresponding to the task based on the control flow structure. Subsequently, when the user initiates a dialog to the server 102 using the terminal 104, the intelligent dialog robot may execute the generated dialog flow under each task to perform an intelligent dialog with the user. It is understood that the intelligent conversation robot is a program for executing a conversation process to realize intelligent conversation.
In one embodiment, as shown in fig. 2, a conversation process mining method is provided, which may be applied to a terminal or a server, and may also be applied to a system including the terminal and the server, and is implemented through interaction between the terminal and the server. The embodiment of the present application is exemplified by being applied to a server, and includes the following steps:
step 202, task clustering is carried out on the set of the dialog text logs.
The dialog text log is a dialog log in a text format. Task clustering refers to a process of clustering dialog text logs under the same task into one class. The dialog text log in the same task refers to a log in a text format for processing a dialog in the same task.
In one embodiment, the server may obtain the dialog text log directly from the database, or may collect the dialog text log in the voice format first and then convert the dialog text log in the voice format into the dialog text log in the text format.
It can be understood that the dialog is generated in the process of processing the task, so the set of dialog text logs can be dialog text logs under a plurality of tasks, the server can perform task clustering on the set of dialog text logs, the dialog text logs under the same task are clustered into one type, and the dialog text logs for processing different tasks are distributed in different types.
In one embodiment, the dialog text log may include a service log in text format-i.e., a text chat record between the service and the user. It can be understood that a customer service session, used to execute a task, records an actual execution of a business process. The business of a company is usually diversified, each customer service person can interface with hundreds of customers every month, and the service log records the processing of repeated tasks for many times. Therefore, the server can obtain a large number of text chatting records of different tasks, and the text chatting records are clustered through a text clustering technology of natural language processing. Taking a bank customer service scene as an example, a text chatting record set under the task of applying for a credit card, a text chatting record set under the task of inquiring balance of a bank card, a text chatting record set under the task of reporting loss of the credit card and the like under a plurality of different tasks can be obtained through clustering.
It is to be understood that the dialog text log is not limited to the service log in text format, and may be a dialog text log in other scenarios, for example, a text chat record in a social scenario.
The dialog text log may be a question and answer log or a chat log of a non-question and answer type.
In one embodiment, the server may pre-process the set of dialog text logs and perform task clustering on the pre-processed set of dialog text logs.
In one embodiment, the server may perform validity detection on the dialog text logs, remove invalid dialog text logs, and perform task clustering processing based on a set of dialog text logs from which invalid logs are removed.
In one embodiment, the server may determine the dialog text log including only the sentence output by one party as an invalid dialog text log and eliminate the dialog text log. It will be appreciated that if, for example, only the customer service outputs a statement on the one hand, and never gets a reply from the user, the statement is an invalid dialog text log.
And step 204, performing intention recognition on each pair of uttered sentences in the dialog text log under the same task to obtain an intention set under the same task.
The intention is a dialogue intention for indicating the purpose to be achieved by the dialogue. An intent set comprising at least one intent. The dialog text log includes dialog statements. A conversational sentence is a sentence that is generated during a conversation. A conversational sentence may be a single sentence or a set of sentences of a set of conversations.
Specifically, each dialog sentence has a corresponding intention, and the server can perform intention recognition on each pair of dialog sentences in the dialog text log under the same task. The server may use natural semantic analysis techniques to identify the intent in each conversational sentence.
The server can use a text clustering method to perform text clustering on all conversation sentences under the same task, the conversation sentences with similar intentions are clustered into one class, the conversation sentences with dissimilar intentions are dispersed in different classes, and each cluster represents one class of intentions. Since there are multiple dialog statements under the same task, there will be at least one (i.e. one or more) intention under the same task, and thus a set of intentions under the same task can be obtained. The same intent may occur in one or more tasks.
And step 206, mining the basic relationship among the intentions under the tasks according to the time sequence of each intention in the corresponding session in the intention set under the tasks.
It is to be understood that a session may correspond to at least a portion (i.e., part or all) of the dialog statements under a task. Then, the corresponding session of the intention refers to the session corresponding to the dialog sentence representing the intention. The time sequence of the intention is the time sequence of the dialog sentence representing the intention in the corresponding conversation.
In one embodiment, the primary relationships between the intents may include at least one of an accompanying relationship, a causal relationship, an unrelated relationship, and the like.
In one embodiment, for each task, the server may determine a dialog statement corresponding to each intention in the corresponding session in the intention set under the task, determine a time sequence of each dialog statement in the session, obtain a time sequence between each intention in the intention set under the task, and mine a basic relationship between the intentions under the task according to the time sequence.
In one embodiment, the server may also use a timestamp of a corresponding dialog statement of each intention in the session corresponding to the task as a timestamp corresponding to each intention. And then mining the basic relationship between the intentions under the task according to the time sequence between the intentions represented by the time stamps among the intentions.
It will be appreciated that the server can extract commonalities between chronologically adjacent intents in each session under the same task to determine the underlying relationship between the intents. The adjacent time-series intentions refer to adjacent intentions arranged in time series in the same session.
At step 208, a control flow structure is constructed based on the base relationships.
The control flow structure controls the conversation process of the intelligent conversation robot and comprises conversation nodes and a process direction.
Specifically, the server may construct corresponding nodes and determine corresponding flow directions for each basic relationship, construct sub-control flow structures according to the constructed nodes and the corresponding flow directions, and then combine the sub-control flow structures to obtain a final control flow structure.
Specifically, the server may generate a conversation process corresponding to the task based on the control flow structure. Subsequently, when the user initiates a conversation for the task, the intelligent conversation robot can intelligently converse with the user according to the conversation process.
In one embodiment, the method further comprises: acquiring a sentence input by a user; after the intention of the sentence is identified, positioning a node corresponding to the identified intention in the task flow; and determining the next node pointed by the node in the task flow, and outputting a corresponding answer sentence based on the corresponding intention of the next node.
Specifically, the user may input a sentence through voice or text, the intelligent dialogue robot program in the server may perform intent recognition on the sentence, locate a node corresponding to the recognized intent in the generated task flow, and then determine, according to a structure in the task flow, a next node pointed to by the located node in the task flow. The intelligent dialogue robot program can obtain the corresponding intention of the determined next node and output a corresponding answer sentence based on the intention.
In one embodiment, the intelligent dialogue robot may implement intelligent dialogue by clarifying a conversational output answer sentence that guides a user to provide word slot information based on the determined intent and the collection status of word slots in the conversation.
The conversation process mining method comprises the steps of firstly clustering tasks of a set of conversation text logs; and then performing intention recognition on each pair of spoken sentences in the dialog text log under the same task to obtain an intention set under the same task. Since the intentions can represent the purpose of the conversation, the basic relationship among the intentions under each task is mined according to the time sequence of each intention in the corresponding conversation in the intention set under each task, and then a control flow structure is constructed. Based on the control flow structure, the conversation process corresponding to each task can be automatically generated. The method and the device realize automatic excavation of the conversation process from the conversation log, and improve efficiency compared with the traditional manual construction of the conversation process.
In addition, labor cost and subsequent process optimization cost are saved. Finally, the limitation caused by only depending on manual construction of the conversation process is avoided, and the operation modes and means of the intelligent conversation robot are enriched.
In one embodiment, the method further comprises: obtaining a customer service log; performing voice recognition on the voice logs in the customer service logs to generate text logs; and determining a dialog text log according to the text log.
The customer service log is a log of a conversation between a manual customer service person and a user. The customer service log may include a voice log. The voice log is a customer service log in a voice format.
In one embodiment, the voice log, may include a telephone chat log. In addition, the voice log may also include a log of voice formats generated by routes other than telephony.
Specifically, the server may perform Automatic Speech Recognition (ASR) on the voice log in the customer service log, and convert it into a text log. The server may determine a dialog text log from the text log. It is understood that when only the voice log is included in the customer service log, the text log generated by conversion can be directly used as the dialog text log.
In one embodiment, the customer service log further comprises a text chat record ((i.e., a log in text format). in this embodiment, determining a conversation text log from the text log comprises generating a conversation text log from the text chat record and the text log.
The text chatting record is a record of the dialog between the customer service personnel and the user in a text mode.
Specifically, the server may use the text chat records included in the customer service log, together with the text log generated by converting the voice log in the customer service log, as the dialog text log.
In the embodiment, the voice recognition processing is automatically performed on the voice logs in the customer service logs, and massive analysis samples can be rapidly provided, so that the accuracy and the efficiency of the conversation process mining are improved.
In one embodiment, the step 204 of performing intent recognition on each pair of utterances in the dialog text log under the same task to obtain an intent set under the same task includes: determining the conversation to which the dialog text logs under the same task belong; each conversation comprises at least one dialogue statement; performing intention recognition on the conversation sentences in each conversation to obtain at least one intention in each conversation; and determining an intention set under the same task according to the intention in each session.
It is understood that the dialog text log under the same task is equivalent to a session log for executing a task flow once. One or more sessions may be included in a session log, and a session may be represented by a sequence of dialog statements in a dialog text log. Each session includes at least one dialog statement.
Specifically, the server may determine sessions to which the dialog text logs under the same task belong, perform intent recognition on dialog sentences in each session, recognize at least one intent in each session, and determine an intent set under the same task according to the intent in each session.
In one embodiment, the server may determine a set of sub-intents corresponding to each session according to the intention corresponding to each session. Then, the server may compose an intention set under the same task according to a sub-intention set corresponding to each session under the same task. That is, the final intent set under the same task includes the sub-intent set of each session.
For ease of understanding, reference is now made to fig. 3 for illustration. Fig. 3 takes a dialog text log as an example of a customer service log in a customer service scene, where the customer service log includes question and answer sentences, a session ID is 1, each row represents a question and answer sentence, various information related to the question and answer is recorded, and 8 question and answer sentences all belong to the same session. The question-answer sentences are natural languages and are presentation forms of intentions, and each question-answer sentence has a corresponding intention, so that the question-answer sentences can be normalized and recognized as the intentions. For example, the question-answer sentence "good," asking if your bank card is "normalized to" ask whether your bank card is "intention," yes "is normalized to" positive "intention," no "is" normalized to "negative" intention, "please input your inquiry password, for example, 123456, confirm" normalization to "inquiry password" intention by well number after input is completed, and the like. Further, the server may perform cluster division according to the intentions to obtain a sub-intention set corresponding to the session. When a plurality of sessions exist in the same task, the intention set in the same task can be determined according to the sub-intention sets corresponding to the sessions.
In other embodiments, the server may also aggregate the intentions of all sessions under the same task to obtain an intention set as the intention set under the task.
In the embodiment, the intention recognition is performed on the spoken sentence on the conversation level, and the intention set under the same task is determined according to the intention in each conversation, so that the accuracy is improved compared with the overall general summary intention.
In one embodiment, for each task, mining the basic relationship between the intentions under the task according to the time sequence of the intentions in the intent set under the task in the corresponding session includes: numbering each intention in an intention set under each task to obtain an intention identifier of each intention under each task; determining a timestamp corresponding to an intention in each session under the task; generating a conversation log corresponding to the task according to the conversation identification of each conversation under the task, the intention identification of the intention in the conversation and the timestamp corresponding to the intention in each conversation; and mining the basic relationship between the intentions under the task from the conversation log.
One session log may include one or more sessions, and one session may correspond to one or more intention identifications. An intention identification may occur in different sessions. That is, intent identification is used to uniquely identify an intent, and the same intent may occur in different sessions. It will be appreciated that a timestamp may be used to indicate the timing of the intent in the session.
Specifically, the server may also number the intentions in the same task to obtain the intention identifier. The server may use a timestamp of a dialog statement corresponding to an intention in each session under the task as a timestamp corresponding to each intention. The server can generate a session log corresponding to the task according to the session identifier of each session under the task, the intention identifier of the intention in the session and the timestamp corresponding to the intention in each session; and mining the basic relationship between the intentions under the task from the conversation log.
As can be seen from fig. 3, the timestamp "2020/6/1812: 00: 01" corresponding to the dialog statement "i want to check the primary balance" is the timestamp corresponding to the intention of "inquiring the balance" corresponding to "i want to check the primary balance".
In one embodiment, the server may arrange the intention identifications of the intentions in the session represented by each session identification according to corresponding timestamps, and form a session log according to a set of the intention identifications arranged correspondingly to each session. The server can mine the basic relationship among the intentions under the task according to the sequence of the timestamps of the intention identifications in the session log corresponding to the task.
For ease of understanding, the session log is now illustrated with reference to an example. Let T be the set of intent identifiers corresponding to the same session,
wherein, a, b, c, d belong to 4 disagreeable graph identifiers, and the set of intention identifiers of the same conversation is arranged according to the time stamp sequence and has time sequence. It is assumed that,
，
it is a session log containing 6 sessions. Wherein, the upper corner labels "3" and "2" respectively represent the number of sessions corresponding to the intention set, that is, L indicates that one task correspondingly generates a session log, the session log of the task includes 6 sessions, wherein the set of intention labels corresponding to 3 sessions is
The set of 2 session intentions is
The set of 1 session's intention identifiers is
. As can be appreciated, the same intent, or the same set of intents, can be included in different sessions.
It is understood that the server can extract the commonalities between the intentions characterized by the adjacent intention identifications corresponding to each session in the session log, and determine the basic relationship between the intentions.
In the above embodiment, the session log corresponding to the task is generated according to the session identifier of each session, the intention identifier of the intention in the session, and the timestamp corresponding to the intention in each session under the same task, and further, the basic relationship between the intentions under the task is mined from the session log, so that the timing sequence between the intentions in each session can be considered, and the basic relationship between the intentions can be mined more accurately. Further, the conversation process can be mined more accurately.
In one embodiment, said mining, from said conversation log, the fundamental relationship between intentions under said task, comprises: for any two intentions in the set of intentions under the task, when one of the two intentions occurs before the other intention in one or more sessions in the session log, determining that there is an accompanying relationship in which a later intention of the two intentions is accompanied by a previous intention.
It is understood that one task corresponds to one session log, and for any two intents in the intent set under the task, when one of the two intents occurs before the other one in one or more sessions in the session log (i.e., one or more sessions in all sessions under one task), it is determined that there is an accompanying relationship between the two intents, and the accompanying relationship is that a following intent is accompanied by a preceding intent. That is, a prior intent may occur before a subsequent intent, and a subsequent intent may occur after the prior intent occurs. For example, if there are one or more sessions in the session log and intention a occurs before intention b, it indicates that there is an accompanying relationship of a > b, i.e., there is an accompanying relationship of b accompanying a.
In one embodiment, said mining, from said conversation log, the basic relationship between the intentions under said task further comprises: when one of the two intentions occurs before the other in all sessions of the session log, it is determined that a causal relationship exists between the two intentions.
Specifically, for any two intentions in the set of intentions under the task, when one of the two intentions occurs before the other in all sessions of the session log (i.e., in all sessions under one task), it is determined that a causal relationship exists between the two intentions, that is, a causal relationship exists between the former intent and the latter intent. For example, in one or more sessions in the session log, if an intention a occurs before an intention b, and if no intention b occurs before the intention a, it indicates that a- > b causal relationship exists, i.e., there is a causal relationship that causes the intention b by the intention a.
In one embodiment, when the two intentions are not before each other in all sessions of the session log, it is determined that an unrelated relationship exists between the two intentions. For example, for any two intents in the set of intents under the task, when there is no intention a occurring before intention b or no intention b occurring before intention a in all sessions of the session log (i.e., in all sessions under one task), it is determined that there is no relation between intention a and intention b, i.e., there is an irrelevant relation of a # b.
It should be noted that, in the embodiment of the present application, only three basic relationships, namely, an accompanying relationship, a causal relationship, and an unrelated relationship, are not limited to mining, and other basic relationships may be included.
In the embodiment, the time sequence relationship among the intentions in each session of the session log is integrated, so that the basic relationships such as the accompanying relationship, the causal relationship and the irrelevant relationship among the intentions can be accurately mined, and the accuracy of the subsequent mining conversation process is improved.
In one embodiment, said constructing a control flow structure from said fundamental relationships comprises: aiming at any two intentions in an intention set under the same task, when a causal relationship exists between the two intentions, a reason node and an effect node are respectively constructed corresponding to the two intentions, and a sequence structure is generated according to the reason node and the effect node.
Wherein the reason node is a previous node pointing to the result node. The sequential structure is a structure in which the cause node points to the result node.
For example, the causal relationship of the intention a — > intention b can be used to construct a cause node corresponding to the intention a and an effect node corresponding to the intention b, and then generate a sequential structure in which the cause node corresponding to the intention a points to the effect node corresponding to the intention b.
It can be understood that, after the intent corresponding to the reason node is generated, the intent corresponding to the result node pointed to by the reason node in the corresponding sequential structure is triggered.
For example, if the intention corresponding to the reason node is an intention of "want to inquire balance", the intention corresponding to the result node pointed by the reason node in the sequence result may be "inquire whether the result node is the bank card of the person", so that the intelligent dialogue robot may be triggered to output an inquiry statement corresponding to the intention.
In the embodiment, the sequence structure is automatically constructed according to the causal relationship among intentions, which is equivalent to automatically realizing the construction steps of the conversation process, so that the efficiency is improved, and the cost is saved.
In one embodiment, said constructing a control flow structure from said fundamental relationships comprises: for any first intention, second intention and third intention in an intention set under the same task, when the second intention and the third intention are both accompanied with the first intention and an irrelevant relation exists between the second intention and the third intention, constructing a bifurcation node corresponding to the first intention and constructing branch nodes corresponding to the second intention and the third intention respectively, and generating a selection bifurcation structure according to the bifurcation node and the branch nodes.
The first intention, the second intention and the third intention are used for representing any three different intentions in an intention set under the same task, and do not have other limitations in terms of dependency, size and the like. A bifurcation node refers to a bifurcation point pointing to a different node. A branch node is a node pointed to by a bifurcation node.
Specifically, when the second intention and the third intention are both accompanied by the first intention and there is an irrelevant relationship between the second intention and the third intention, it means that the second intention and the third intention both occur after the first intention and there is no directional relationship between the second intention and the third intention. Then, the server may construct a bifurcation node corresponding to the first intention, construct branch nodes corresponding to the second intention and the third intention, and point to two parallel branch nodes by the bifurcation node, respectively, to generate a selective bifurcation structure.
For example, the accompanying relationship of intention a > intention b, the accompanying relationship of intention a > intention c, and the unrelated relationship of b # c may be used to construct the bifurcation node corresponding to intention a, and the branch nodes corresponding to intents b and c, respectively, and then generate the selective bifurcation structure in which the bifurcation node corresponding to intention a points to the branch nodes corresponding to intents b and c, respectively.
In the embodiment, the selection bifurcation structure is automatically constructed according to the accompanying relation and the irrelevant relation among intentions, which is equivalent to automatically realizing the construction step of the conversation process, so that the efficiency is improved, the cost is saved, and the accuracy is ensured.
In one embodiment, said constructing a control flow structure according to said basic relationship further comprises: when the third intention is respectively accompanied with the first intention and the second intention and an irrelevant relation exists between the first intention and the second intention, respectively constructing nodes to be merged corresponding to the first intention and the second intention and constructing merged nodes corresponding to the third intention, and generating a selection merged structure according to the nodes to be merged and the merged nodes.
The node to be merged refers to a node pointing to the same merged node. The merging node is a node to which different nodes point.
Specifically, when the third intention is accompanied by the first intention and the second intention respectively, and there is an irrelevant relationship between the first intention and the second intention, it means that both the first intention and the second intention occur before the third intention, and there is no directional relationship between the second intention and the third intention. Then, the server may respectively construct nodes to be merged corresponding to the first intention and the second intention, respectively, construct a merged node corresponding to the third intention, and point to the merged node from the two nodes to be merged, so as to generate a selective merging structure.
For example, the accompanying relationship of intention a > intention c, the accompanying relationship of intention b > intention c, and the irrelevant relationship of a # b may be used to construct nodes to be merged corresponding to the intents a and b, respectively, and merge nodes corresponding to the intention c, and then generate a node to be merged corresponding to the intention a and a node to be merged corresponding to the intention b, which all point to the merge node corresponding to the intention c.
In the embodiment, the selection merging structure is automatically constructed according to the accompanying relation and the irrelevant relation among the intentions, which is equivalent to automatically realizing the construction step of the conversation process, so that the efficiency is improved, the cost is saved, and the accuracy is ensured.
FIG. 4 is a simplified flow diagram of a method for conversation process mining in one embodiment. Fig. 4 illustrates a customer service scenario, which may be implemented by first collecting a manual customer service log, performing ASR automatic speech recognition processing on a telephone log (i.e., a log in a speech format) therein, and converting the telephone log into a text log. And then, dividing the text logs according to the tasks by adopting a text clustering method. The text log includes dialogue sentences, so a text clustering method can be adopted again to divide the dialogue sentences according to intentions, and then the conversation log composed of a conversation ID (namely a conversation identifier), an intention ID (namely an intention identifier) and timestamps corresponding to the intentions is obtained. Then, a flow mining algorithm is adopted to mine the control flow structure of the task based on the conversation log. A conversation process may thus be generated based on the control flow structure.
It should be understood that, although the steps in the flowcharts of the embodiments of the present application are shown in sequence as indicated by the arrows, the steps are not necessarily performed in sequence as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a part of the steps in the embodiments of the present application may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of performing the steps or stages is not necessarily sequential, but may be performed alternately or alternately with other steps or at least a part of the steps or stages in other steps.
In one embodiment, as shown in fig. 5, there is provided a conversation process mining apparatus including: a task partitioning module 502, an intent recognition module 504, an intent relationship mining module 506, and a conversation process generation module 508, wherein:
and the task dividing module 502 is configured to perform task clustering on the set of dialog text logs.
And the intention recognition module 504 is configured to perform intention recognition on each pair of utterances in the dialog text log under the same task to obtain an intention set under the same task.
And the intention relation mining module 506 is used for mining the basic relation among the intentions under the tasks according to the time sequence of the intentions in the corresponding session in the intention set under the tasks.
A conversation process generating module 508, configured to construct a control flow structure according to the basic relationship; and generating a conversation process corresponding to the task based on the control flow structure.
In one embodiment, the apparatus further comprises:
a voice recognition module 501, configured to obtain a customer service log; performing voice recognition on the voice logs in the customer service logs to generate text logs; and determining a dialog text log according to the text log.
In one embodiment, the customer service log further comprises a text chat record; the task dividing module 502 is further configured to generate a dialog text log according to the text chat record and the text log.
In one embodiment, the intent recognition module 504 is further configured to determine a session to which the dialog text log belongs under the same task; each conversation comprises at least one dialogue statement; performing intention recognition on the conversation sentences in each conversation to obtain at least one intention in each conversation; and determining an intention set under the same task according to the intention in each session.
In one embodiment, the intention relationship mining module 506 is further configured to, for each task, number each intention in the intention set under the task to obtain an intention identifier of each intention under the task; determining a timestamp corresponding to an intention in each session under the task; generating a conversation log corresponding to the task according to the conversation identification of each conversation under the task, the intention identification of the intention in the conversation and the timestamp corresponding to the intention in each conversation; and mining the basic relationship between the intentions under the task from the conversation log.
In one embodiment, the intent relationship mining module 506 is further configured to determine, for any two intents in the set of intents under the task, that there is a companion relationship where a later one of the two intents is accompanied by a previous one when one of the two intents occurs before the other intent in one or more sessions in the session log; determining that a causal relationship exists between the two intentions when one of the two intentions occurs before the other intent in all sessions of the session log; and when the two intentions are not before in all the conversations of the conversation log, judging that an unrelated relation exists between the two intentions.
In one embodiment, the dialog flow generation module 508 is further configured to, for any two intentions in the set of intentions under the task, construct a reason node and an effect node corresponding to the two intentions, respectively, when a causal relationship exists between the two intentions, and generate a sequential structure according to the reason node and the effect node.
In one embodiment, the dialog flow generation module 508 is further configured to, for any first intention, second intention, and third intention in the set of intentions under the task, when the second intention and the third intention are both accompanied by the first intention and there is an unrelated relationship between the second intention and the third intention, construct a bifurcation node corresponding to the first intention and construct a branch node corresponding to the second intention and the third intention, respectively, and generate a selection bifurcation structure according to the bifurcation node and the branch node.
In one embodiment, the dialog flow generation module 508 is further configured to, when the third intention is respectively accompanied by the first intention and the second intention and an unrelated relationship exists between the first intention and the second intention, respectively construct a node to be merged corresponding to the first intention and the second intention and construct a merged node corresponding to the third intention, and generate a selection merged structure according to the node to be merged and the merged node.
As shown in fig. 6, in one embodiment, the apparatus further comprises: a speech recognition module 501 and a dialog module 510; wherein:
a dialogue module 510, configured to obtain a sentence input by a user; after the intention of the statement is identified, locating a node corresponding to the identified intention in the task flow; and determining a next node pointed by the node in the task flow, and outputting a corresponding answer statement based on the corresponding intention of the next node.
For the specific definition of the conversation process mining device, reference may be made to the above definition of the conversation process mining method, which is not described herein again. The modules in the conversation process mining device can be wholly or partially realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as shown in fig. 7. The computer device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used to store a collection of dialog text logs. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a conversation process mining method.
Those skilled in the art will appreciate that the architecture shown in fig. 7 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is provided, which includes a memory and a processor, the memory storing a computer program, and the processor implementing the steps of the conversation process mining method described in the embodiments of the present application when executing the computer program.
In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the conversation process mining method described in the embodiments of the present application.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database or other medium used in the embodiments provided herein can include at least one of non-volatile and volatile memory. Non-volatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical storage, or the like. Volatile Memory can include Random Access Memory (RAM) or external cache Memory. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others.
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (15)
1. A conversation process mining method, the method comprising:
performing task clustering on the set of the dialog text logs;
performing intention recognition on each pair of spoken sentences in a dialog text log under the same task to obtain an intention set under the same task;
for each task, mining the basic relationship among the intentions under the task according to the time sequence of each intention in the corresponding session in the intention set under the task;
constructing a control flow structure according to the basic relation;
and generating a conversation process corresponding to the task based on the control flow structure.
2. The method of claim 1, further comprising:
obtaining a customer service log;
performing voice recognition on the voice logs in the customer service logs to generate text logs;
and determining a dialog text log according to the text log.
3. The method of claim 2, wherein the customer service log further comprises a text chat log;
said determining a dialog text log according to the text log comprises:
and generating a dialog text log according to the text chatting record and the text log.
4. The method of claim 1, wherein the performing intent recognition on each pair of utterances in the dialog text log under the same task to obtain an intent set under the same task comprises:
determining the conversation to which the dialog text logs under the same task belong; each conversation comprises at least one dialogue statement;
performing intention recognition on the conversation sentences in each conversation to obtain at least one intention in each conversation;
and determining an intention set under the same task according to the intention in each session.
5. The method according to claim 1, wherein mining, for each task, the basic relationship between the intentions under the task according to the time sequence of the intentions in the set of intentions under the task in the corresponding session comprises:
numbering each intention in an intention set under each task to obtain an intention identifier of each intention under each task;
determining a timestamp corresponding to an intention in each session under the task;
generating a conversation log corresponding to the task according to the conversation identification of each conversation under the task, the intention identification of the intention in the conversation and the timestamp corresponding to the intention in each conversation;
and mining the basic relationship between the intentions under the task from the conversation log.
6. The method of claim 5, wherein mining the fundamental relationships between the intentions under the task from the conversation logs comprises:
for any two intentions in the set of intentions under the task, when one of the two intentions occurs before the other intention in one or more sessions in the session log, determining that there is an accompanying relationship in which a later intention of the two intentions is accompanied by a previous intention;
determining that a causal relationship exists between the two intentions when one of the two intentions occurs before the other intent in all sessions of the session log;
and when the two intentions are not before in all the conversations of the conversation log, judging that an unrelated relation exists between the two intentions.
7. The method of claim 6, wherein constructing a control flow structure based on the primitive relationships comprises:
and aiming at any two intentions in the intention set under the task, when a causal relationship exists between the two intentions, constructing a reason node and an effect node corresponding to the two intentions respectively, and generating a sequential structure according to the reason node and the effect node.
8. The method of claim 6, wherein constructing a control flow structure from the base relationships comprises:
for any first intention, second intention and third intention in an intention set under a task, when the second intention and the third intention are both accompanied with the first intention and an irrelevant relation exists between the second intention and the third intention, constructing a bifurcation node corresponding to the first intention and constructing branch nodes corresponding to the second intention and the third intention respectively, and generating a selection bifurcation structure according to the bifurcation node and the branch nodes.
9. The method of claim 8, wherein constructing a control flow structure according to the fundamental relationship further comprises:
when the third intention is respectively accompanied with the first intention and the second intention and an irrelevant relation exists between the first intention and the second intention, respectively constructing nodes to be merged corresponding to the first intention and the second intention and constructing merged nodes corresponding to the third intention, and generating a selection merged structure according to the nodes to be merged and the merged nodes.
10. The method according to any one of claims 1 to 9, further comprising:
acquiring a sentence input by a user;
after the intention of the statement is identified, locating a node corresponding to the identified intention in the task flow;
and determining a next node pointed by the node in the task flow, and outputting a corresponding answer statement based on the corresponding intention of the next node.
11. A conversation process mining apparatus, comprising:
the task dividing module is used for clustering tasks of the set of the dialog text logs;
the intention identification module is used for carrying out intention identification on each pair of spoken sentences in the dialog text log under the same task to obtain an intention set under the same task;
the intention relation mining module is used for mining the basic relation among the intentions under the tasks according to the time sequence of each intention in the corresponding session in the intention set under the tasks;
the conversation process generating module is used for constructing a control flow structure according to the basic relationship; and generating a conversation process corresponding to the task based on the control flow structure.
12. The apparatus of claim 11, further comprising:
the voice recognition module is used for acquiring a customer service log; performing voice recognition on the voice logs in the customer service logs to generate text logs; and determining a dialog text log according to the text log.
13. The apparatus according to claim 11, wherein the intention relationship mining module is further configured to, for each task, number each intention in the set of intentions under the task to obtain an intention identifier of each intention under the task; determining a timestamp corresponding to an intention in each session under the task; generating a conversation log corresponding to the task according to the conversation identification of each conversation under the task, the intention identification of the intention in the conversation and the timestamp corresponding to the intention in each conversation; and mining the basic relationship between the intentions under the task from the conversation log.
14. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor realizes the steps of the method of any one of claims 1 to 10 when executing the computer program.
15. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 10.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011007891.5A CN111930917B (en) | 2020-09-23 | 2020-09-23 | Conversation process mining method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011007891.5A CN111930917B (en) | 2020-09-23 | 2020-09-23 | Conversation process mining method and device, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111930917A true CN111930917A (en) | 2020-11-13 |
| CN111930917B CN111930917B (en) | 2021-02-05 |
Family
ID=73335165
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011007891.5A Active CN111930917B (en) | 2020-09-23 | 2020-09-23 | Conversation process mining method and device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111930917B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170337261A1 (en) * | 2014-04-06 | 2017-11-23 | James Qingdong Wang | Decision Making and Planning/Prediction System for Human Intention Resolution |
| CN110532361A (en) * | 2019-08-09 | 2019-12-03 | 深圳追一科技有限公司 | Recognition methods, device, computer equipment and the storage medium that user is intended to |
| CN110674287A (en) * | 2018-06-07 | 2020-01-10 | 阿里巴巴集团控股有限公司 | Method and device for establishing hierarchical intention system |
| CN110942769A (en) * | 2018-09-20 | 2020-03-31 | 九阳股份有限公司 | Multi-turn dialogue response system based on directed graph |
| CN111145745A (en) * | 2019-12-27 | 2020-05-12 | 苏州思必驰信息科技有限公司 | Conversation process customizing method and device |
| CN111368085A (en) * | 2020-03-05 | 2020-07-03 | 北京明略软件系统有限公司 | Recognition method and device of conversation intention, electronic equipment and storage medium |
| CN111581375A (en) * | 2020-04-01 | 2020-08-25 | 车智互联（北京）科技有限公司 | Dialog intention type identification method, multi-turn dialog method, device and computing equipment |
| CN111667833A (en) * | 2019-03-07 | 2020-09-15 | 国际商业机器公司 | Speech recognition based on conversation |
-
2020
- 2020-09-23 CN CN202011007891.5A patent/CN111930917B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170337261A1 (en) * | 2014-04-06 | 2017-11-23 | James Qingdong Wang | Decision Making and Planning/Prediction System for Human Intention Resolution |
| CN110674287A (en) * | 2018-06-07 | 2020-01-10 | 阿里巴巴集团控股有限公司 | Method and device for establishing hierarchical intention system |
| CN110942769A (en) * | 2018-09-20 | 2020-03-31 | 九阳股份有限公司 | Multi-turn dialogue response system based on directed graph |
| CN111667833A (en) * | 2019-03-07 | 2020-09-15 | 国际商业机器公司 | Speech recognition based on conversation |
| CN110532361A (en) * | 2019-08-09 | 2019-12-03 | 深圳追一科技有限公司 | Recognition methods, device, computer equipment and the storage medium that user is intended to |
| CN111145745A (en) * | 2019-12-27 | 2020-05-12 | 苏州思必驰信息科技有限公司 | Conversation process customizing method and device |
| CN111368085A (en) * | 2020-03-05 | 2020-07-03 | 北京明略软件系统有限公司 | Recognition method and device of conversation intention, electronic equipment and storage medium |
| CN111581375A (en) * | 2020-04-01 | 2020-08-25 | 车智互联（北京）科技有限公司 | Dialog intention type identification method, multi-turn dialog method, device and computing equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111930917B (en) | 2021-02-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11018885B2 (en) | Summarization system | |
| CN107038220B (en) | Method, intelligent robot and system for generating memorandum | |
| CN109002510B (en) | Dialogue processing method, device, equipment and medium | |
| WO2022160707A1 (en) | Human-machine interaction method and apparatus combined with rpa and ai, and storage medium and electronic device | |
| CN112235470B (en) | Incoming call client follow-up method, device and equipment based on voice recognition | |
| KR102312993B1 (en) | Method and apparatus for implementing interactive message using artificial neural network | |
| CN112035630A (en) | Dialogue interaction method, device, equipment and storage medium combining RPA and AI | |
| CN110275703B (en) | Method and device for assigning key value to data, computer equipment and storage medium | |
| CN109887483A (en) | Self-Service processing method, device, computer equipment and storage medium | |
| CN111930917B (en) | Conversation process mining method and device, computer equipment and storage medium | |
| CN111510566B (en) | Method and device for determining call label, computer equipment and storage medium | |
| CN110931002B (en) | Man-machine interaction method, device, computer equipment and storage medium | |
| CN115393077B (en) | Data processing method based on loan transaction man-machine conversation system and related device | |
| CN111241249A (en) | Man-machine conversation method, device, computer equipment and storage medium | |
| CN112929499B (en) | Dialogue interaction method and device, computer equipment and computer-readable storage medium | |
| CN111182117A (en) | Call processing method and device, computer equipment and computer readable storage medium | |
| CN112269473B (en) | Man-machine interaction method and system based on flexible scene definition | |
| CN111783415B (en) | Template configuration method and device | |
| US7908143B2 (en) | Dialog call-flow optimization | |
| CN111428018B (en) | Intelligent question-answering method and device | |
| CN113556430B (en) | Outbound system and outbound method | |
| US20230281396A1 (en) | Message mapping and combination for intent classification | |
| CN112037796A (en) | Data processing method, device, equipment and medium | |
| CN115567646A (en) | Intelligent outbound method, device, computer equipment and storage medium | |
| CN117251631A (en) | Information recommendation method, device, equipment and storage medium based on artificial intelligence |
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 |