CN112115249B - Statistical analysis and result display method and device for user intention - Google Patents
Statistical analysis and result display method and device for user intention Download PDFInfo
- Publication number
- CN112115249B CN112115249B CN202011034837.XA CN202011034837A CN112115249B CN 112115249 B CN112115249 B CN 112115249B CN 202011034837 A CN202011034837 A CN 202011034837A CN 112115249 B CN112115249 B CN 112115249B
- Authority
- CN
- China
- Prior art keywords
- user
- node
- statistical analysis
- round
- intention
- 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.)
- Active
Links
- 238000007619 statistical method Methods 0.000 title claims abstract description 163
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000013145 classification model Methods 0.000 claims abstract description 18
- 238000006243 chemical reaction Methods 0.000 claims description 11
- 238000012216 screening Methods 0.000 claims description 10
- 230000008859 change Effects 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 10
- 238000007689 inspection Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 230000006399 behavior Effects 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000007670 refining Methods 0.000 description 2
- 238000007792 addition Methods 0.000 description 1
- 230000001174 ascending effect Effects 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 230000003442 weekly effect Effects 0.000 description 1
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
- G06F16/31—Indexing; Data structures therefor; Storage structures
- G06F16/316—Indexing structures
- G06F16/322—Trees
-
- 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
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- 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
- G06F16/335—Filtering based on additional data, e.g. user or group profiles
-
- 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
- G06F16/338—Presentation of query results
-
- 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
Abstract
The embodiment of the specification provides a statistical analysis and result display method and device for user intention, wherein the method comprises the following steps: acquiring historical multi-round conversations of each user and customer service in a preset time period; for any historical multi-round dialogue, determining the user intention corresponding to each round of dialogue in the historical multi-round dialogue by using an intention classification model; carrying out statistical analysis on user intentions corresponding to the first-round conversations in each historical multi-round conversation respectively to obtain statistical analysis results of the first-round conversations, wherein the statistical analysis results at least comprise the user intentions of the first-round conversations; taking each user intention of the first round as a first-layer node in a tree hierarchy to be displayed, and establishing paths of the first-layer node and a root node of the tree hierarchy; and according to the turn sequence of each turn of dialogue in the historical multi-turn dialogue, carrying out statistical analysis for each turn, and establishing paths from the plurality of i-layer nodes to the node at the upper layer. The user intention can be accurately identified to provide effective information, and the working efficiency is improved.
Description
Technical Field
One or more embodiments of the present disclosure relate to the field of computers, and more particularly, to a method and apparatus for statistical analysis and result display of user intent.
Background
Currently, a machine is often used instead of manually talking to a user to achieve a predetermined business objective. In the process of interaction between the machine and the user for multiple rounds of dialogue, the user statement needs to be understood, the intention of the user is obtained, and then the strategy corresponding to the intention is selected to answer the user statement. Because the population of users is constantly changing and because of the complexity of population talk, statistical analysis of user intent in historical multi-turn conversations is required to more accurately identify user intent based on the results of the statistical analysis.
In the prior art, a statistical analysis and result display scheme of user intention is generally relatively simple, only the change condition of the user intention is counted and displayed, and for more accurately identifying that the user intention fails to provide effective information, further analysis is needed by manual intervention, and the working efficiency is low.
Therefore, an improved solution is desired, which can provide effective information for more accurately recognizing the intention of the user, and improve the working efficiency.
Disclosure of Invention
One or more embodiments of the present disclosure describe a method and apparatus for statistical analysis and result display of user intent, which can provide effective information for more accurately identifying user intent, and improve working efficiency.
In a first aspect, a statistical analysis and result display method of user intent expressed by user sentences in a historical multi-round dialog of user and customer service is provided, the method comprising:
acquiring historical multi-round conversations of each user and customer service in a preset time period;
determining user intentions corresponding to each dialog in each history multi-turn dialog by using an intention classification model according to any history multi-turn dialog in each history multi-turn dialog;
carrying out statistical analysis on user intentions corresponding to the first-round conversations in each historical multi-round conversation respectively to obtain statistical analysis results of the first-round conversations, wherein the statistical analysis results at least comprise the user intentions of the first-round conversations;
taking each user intention of a first round as a first-layer node in a tree hierarchy to be displayed, and establishing paths of the first-layer node and a root node of the tree hierarchy;
according to the turn sequence of each turn of dialogue in the historical multi-turn dialogue, carrying out statistical analysis on each turn, wherein the statistical analysis on any ith turn comprises, for any first node of the ith-1 layer, screening out current turn dialogue with user intention track conforming to the path from the root node to the first node from each historical multi-turn dialogue, carrying out statistical analysis on user intention corresponding to each current turn dialogue, and obtaining statistical analysis results of the current turn dialogue, wherein the statistical analysis results at least comprise a plurality of current turn user intentions;
And taking the intentions of the current round users as a plurality of i-layer nodes, and establishing paths from the i-layer nodes to the first node.
In a possible implementation manner, the statistical analysis result further comprises flow data and business index data corresponding to each user intention; the method further comprises the steps of:
and taking the flow data and the business index data corresponding to the intention of each user as node additional information which can be displayed of the corresponding node.
In one possible embodiment, the method further comprises: and sequentially displaying the nodes of each level in the tree hierarchy chart according to the gradual operation of the user.
Further, the flow data includes the number of user sentences corresponding to the user intention, and/or the proportion of the number of user sentences corresponding to the user intention in the total number of user sentences.
Further, the business index data includes conversion rate.
In one possible embodiment, before the statistical analysis, the method further comprises:
acquiring a confidence score based on which the intention classification model determines the intention of the user;
the statistical analysis includes:
determining a user statement in a current dialog in a preset number of historical multi-dialog rounds with the lowest confidence score as a low confidence sample corresponding to the user intention for any user intention in all user intentions included in the statistical analysis result; the statistical analysis result also comprises a low confidence coefficient sample corresponding to each user intention, and the low confidence coefficient sample is used as node additional information of a corresponding node for display.
In one possible embodiment, the statistical analysis comprises:
dividing each historical multi-round dialogue of the user and the manual customer service into a plurality of class clusters by using a clustering mode;
matching any user intention in the user intentions included in the statistical analysis result to a target class cluster in the plurality of class clusters by using a text matching mode;
extracting conversation sentences of high-conversion-rate manual customer service from a plurality of historical multi-round conversations included in the target class cluster, wherein the conversation sentences are used as recommendation strategies of conversation sentences of robot customer service in multi-round conversations of a user and the robot customer service; the statistical analysis result also comprises a recommendation strategy corresponding to each user intention, and the recommendation strategy is used as node additional information which can be displayed for the corresponding node.
Further, the tree hierarchy is a dynamic structure, and the method further comprises:
when the business index data is changed from the business index data of the first index to the business index data of the second index, the business index data included in the node additional information is changed; or when the business index data of the third index is newly added, inserting the business index data of the third index into the node additional information; or deleting the business index data of the fourth index in the node additional information when the business index data cancel the business index data of the fourth index.
Further, the step-by-step operation according to the user sequentially displays the nodes of each level in the tree hierarchy chart, including:
receiving a first operation of a user on any second node in the tree hierarchy, wherein the first operation is used for indicating to expand a next layer node of the second node;
a next level node connected to the second node through the path is shown.
Further, according to the step-by-step operation of the user, the method sequentially displays the nodes of each level in the tree hierarchy chart, and further includes:
receiving a second operation of a user on the second node, wherein the second operation is used for indicating to pack up the unfolded next-layer node of the second node;
the next level node that has been shown is collapsed.
Further, the step-by-step operation according to the user sequentially displays the nodes of each level in the tree hierarchy chart, including:
receiving third operation of a user on any second node in the tree hierarchy, wherein the third operation is used for indicating node additional information for expanding the second node;
node-attached information of the second node is displayed in a predetermined area associated with the second node.
Further, according to the step-by-step operation of the user, the method sequentially displays the nodes of each level in the tree hierarchy chart, and further includes:
receiving a fourth operation of a user on the second node, wherein the fourth operation is used for indicating unfolded node additional information of the second node to be folded;
folding the displayed node additional information.
In a second aspect, there is provided a statistical analysis and result presentation apparatus of user intent expressed by user sentences in a historical multi-round conversation of a user with customer service, the apparatus comprising:
the first acquisition unit is used for acquiring historical multi-round conversations between each user and customer service in a preset time period;
the intention determining unit is used for determining the user intention corresponding to each dialog in each history multi-turn dialog by using the intention classifying model according to any history multi-turn dialog in each history multi-turn dialog acquired by the first acquiring unit;
the first statistical unit is used for carrying out statistical analysis on the user intentions respectively corresponding to the first-round conversations in the historical multi-round conversations determined by the intent determining unit to obtain statistical analysis results of the first-round conversations, wherein the statistical analysis results at least comprise the user intentions of the first-round conversations;
The first path establishing unit is used for respectively taking each user intention of the first round obtained by the first statistical unit as a first layer node in the tree hierarchy to be displayed and establishing a path between the first layer node and a root node of the tree hierarchy;
the second statistical unit is used for carrying out statistical analysis on each round according to the round sequence of each round of dialogue in the historical multi-round dialogue, wherein the statistical analysis on any ith round comprises screening out current round dialogue with user intention track conforming to the path from the root node to the first node from the historical multi-round dialogue for any first node of the ith layer-1, and carrying out statistical analysis on the user intention corresponding to each current round dialogue respectively to obtain the statistical analysis result of the current round dialogue, wherein the statistical analysis result at least comprises a plurality of current round user intentions;
and the second path establishing unit is used for taking the current round user intentions obtained by the second statistical unit as a plurality of i-layer nodes and establishing paths from the i-layer nodes to the first node.
In a third aspect, there is provided a computer readable storage medium having stored thereon a computer program which, when executed in a computer, causes the computer to perform the method of the first aspect.
In a fourth aspect, there is provided a computing device comprising a memory having executable code stored therein and a processor which, when executing the executable code, implements the method of the first aspect.
Through the method and the device provided by the embodiment of the specification, firstly, each historical multi-round dialogue between each user and customer service in a preset time period is obtained; then, aiming at any one of the historical multi-round conversations, determining the user intention corresponding to each round of conversations in the historical multi-round conversations by using an intention classification model; then, carrying out statistical analysis on user intentions corresponding to the first-round conversations in the historical multi-round conversations respectively to obtain statistical analysis results of the first-round conversations, wherein the statistical analysis results at least comprise the user intentions of the first-round conversations; respectively taking each user intention of a first round as a first-layer node in a tree hierarchy to be displayed, and establishing paths of the first-layer node and a root node of the tree hierarchy; then, according to the turn sequence of each turn of dialogue in the historical multi-turn dialogue, carrying out statistical analysis on each turn, wherein the statistical analysis on any ith turn comprises, for any first node of the ith-1 layer, screening out the current turn dialogue with user intention track conforming to the path from the root node to the first node from each historical multi-turn dialogue, carrying out statistical analysis on the user intention corresponding to each current turn of dialogue, and obtaining the statistical analysis result of the current turn of dialogue, wherein the statistical analysis result at least comprises a plurality of current turn of user intention; and taking the intentions of the current round users as a plurality of i-layer nodes, and establishing paths from the i-layer nodes to any first node of the i-1 layer. From the above, in the embodiment of the present disclosure, by performing statistical analysis on the user intentions corresponding to each dialog in the historical multi-dialog, determining each level node of the tree-like hierarchical structure to be displayed, and constructing the path between each level node, the embodiment of the present disclosure can provide effective information for more accurately identifying the user intentions, thereby improving the working efficiency.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic illustration of an implementation scenario of an embodiment disclosed herein;
FIG. 2 illustrates a statistical analysis of user intent and a result presentation method flow diagram in accordance with one embodiment;
FIG. 3 illustrates a view frame construction schematic diagram in accordance with one embodiment;
FIG. 4 illustrates a view frame morphology diagram in accordance with one embodiment;
FIG. 5 shows a schematic block diagram of a statistical analysis of user intent and results presentation device, according to one embodiment.
Detailed Description
The following describes the scheme provided in the present specification with reference to the drawings.
Fig. 1 is a schematic diagram of an implementation scenario of an embodiment disclosed in the present specification. The implementation scene relates to statistical analysis and result display of user intention, wherein the user intention is the intention expressed by a user sentence in a plurality of historical conversations of a user and customer service, the user intention can be a standard question sentence or a knowledge point element which is arranged in advance, and the knowledge point element can be a business element or a requirement element. The customer service can be an artificial customer service or a robot customer service, that is, a plurality of historical conversations aimed by statistical analysis, a plurality of historical conversations of a user and the artificial customer service, and a plurality of historical conversations of the user and the robot customer service. In addition, the historical multi-round dialogue can be a dialogue actively initiated by a user or a dialogue actively initiated by customer service. Referring to fig. 1, taking a case that a history multi-turn dialog specifically includes 3-turn dialogues as an example, a customer service sentence 1 and a user sentence 1 constitute a first-turn dialog (i.e., a first-turn dialog), a user intention corresponding to the first-turn dialog is a user intention 1, a customer service sentence 2 and a user sentence 2 constitute a second-turn dialog, a user intention corresponding to the second-turn dialog is a user intention 2, a customer service sentence 3 and a user sentence 3 constitute a third-turn dialog, and a user intention corresponding to the third-turn dialog is a user intention 3. In the embodiment of the present specification, the customer service sentence may also be referred to as customer service utterances, and the user sentence may also be referred to as user utterances.
Generally, as the number of times of each dialog in the historical multi-turn dialog increases, the obtained information expressed by the user is more sufficient, so that the identified user intention can better represent the user's request, that is, the user intention corresponding to the last dialog can usually better represent the user's request. In the embodiment of the specification, when the user intention is statistically analyzed, instead of determining the user intention corresponding to the last dialog of each historical multi-turn dialog, the user intention corresponding to each dialog of each historical multi-turn dialog is determined, statistical analysis is performed on the user intention corresponding to each dialog of each turn, statistical analysis is performed on the user intention corresponding to each dialog of the historical multi-turn dialog, each layer node of the tree hierarchy to be displayed is determined, and paths among the layers nodes are constructed, so that effective information can be provided for more accurately identifying the user intention, and the working efficiency is improved.
FIG. 2 illustrates a flow diagram of a statistical analysis and results presentation method of user intent expressed by user statements in a historical multi-turn dialog of a user with customer service, which may be based on the implementation scenario illustrated in FIG. 1, according to one embodiment. As shown in fig. 2, the statistical analysis and result display method of the user intention in this embodiment includes the following steps: step 21, acquiring historical multi-round conversations of each user and customer service in a preset time period; step 22, determining user intentions respectively corresponding to each of the historical multi-round dialogues by using an intention classification model according to any of the historical multi-round dialogues; step 23, carrying out statistical analysis on user intentions corresponding to the first-round conversations in each historical multi-round conversation respectively to obtain statistical analysis results of the first-round conversations, wherein the statistical analysis results at least comprise the user intentions of the first-round conversations; step 24, taking each user intention of the first round as a first-layer node in a tree hierarchy to be displayed, and establishing a path between the first-layer node and a root node of the tree hierarchy; step 25, according to the turn sequence of each turn of dialogue in the historical multi-turn dialogue, carrying out statistical analysis on each turn, wherein the statistical analysis on any ith turn comprises, for any first node of the ith layer-1, screening out the current turn dialogue with user intention track conforming to the path from the root node to the first node from each historical multi-turn dialogue, carrying out statistical analysis on the user intention corresponding to each current turn dialogue, and obtaining the statistical analysis result of the current turn dialogue, wherein the statistical analysis result at least comprises a plurality of current turn user intentions; and step 26, taking the intentions of the current round users as a plurality of i-layer nodes, and establishing paths from the i-layer nodes to the first node. Specific implementations of the above steps are described below.
First, in step 21, each historical multi-round dialogue between each user and customer service in a preset time period is obtained. It will be appreciated that the statistical analysis of the user's intent and the presentation of the results may be performed periodically, for example, daily, weekly or monthly, and accordingly, the predetermined period of time may be one day, one week or one month before the current date.
In this embodiment of the present disclosure, a session between a user and a customer service may be referred to as a historical multi-round session, including a session between a user and a robot customer service, and a session between a subsequent user and a manual customer service when the robot customer service cannot solve a user problem.
Then, in step 22, for any one of the historical multi-turn conversations, the user intent corresponding to each of the historical multi-turn conversations is determined using the intent classification model. It will be appreciated that the intent classification model typically first determines the confidence scores for each intent category for each dialog turn, and then determines the intent category for each dialog turn, i.e., the user intent, based on the confidence scores for each intent category.
In the embodiment of the present disclosure, the user intention of the same round of dialogue of different historic rounds of dialogue may be the same or may be different. The user intentions of the ith dialog turn of the different historical multi-turn dialogs may be different, and the user intentions of the (i+1) th dialog turn may be the same or different. The user intentions of the ith dialog of the different historical multi-dialog may be the same or different for the (i+1) th dialog. The following describes a table of correspondence between historical multi-turn conversations and user intention of each turn of conversations shown in the first table.
Table one: user intention corresponding relation table of historical multi-round dialogue and each round dialogue
Referring to the example of Table one, the user intentions of the first-round dialogs of the history multi-round dialogs 1 and 2 are different, and the user intentions of the first-round dialogs of the history multi-round dialogs 1 and 3 are the same. The user intentions of the first dialog of the historical multi-dialog 1 and the first dialog of the historical multi-dialog 2 are different, and the user intentions of the second dialog are the same. The user intentions of the first dialog of the historical multi-dialog 1 and the first dialog of the historical multi-dialog 3 are the same, and the user intentions of the second dialog are different. It will be appreciated that, typically, the user intent identified in the first few of the multiple conversations is a knowledge point element, and the user intent identified in the last conversation is a standard question, and that when the identified user intent is a standard question, the multiple conversations may be ended.
Next, in step 23, statistical analysis is performed on the user intentions corresponding to the first-round dialogs in the historical multi-round dialogs, so as to obtain a statistical analysis result of the first-round dialogs, where the statistical analysis result at least includes the user intentions of the first-round dialogs. It will be appreciated that there may be a first dialog of a plurality of different historical multi-dialog runs corresponding to the same user intent.
For example, the number of historical multi-turn conversations is 1000, the number of predetermined user intentions is 100, and the number of first-turn user intentions included in the statistical analysis result is 20. The statistical analysis result shows the overall distribution condition of the user intentions in the preset time period, namely the attention points or the interest points of the user group, and the statistical analysis is carried out on the user intentions of the historical multi-round conversations, so that the overall distribution change of the user telephone operation caused by external emergencies is conveniently found, the intention classification model is adaptively adjusted according to the change, and the follow-up more accurate determination of the user intentions of the current multi-round conversations is facilitated. It will be appreciated that the emergency event may be an epidemic situation, an earthquake or tsunami, etc.
In one example, the statistical analysis result further includes traffic data and business index data corresponding to each user intention therein, respectively.
Further, the flow data may include a number of user sentences corresponding to the user intent, and/or a proportion of the number of user sentences corresponding to the user intent in the total number of user sentences. For example, the number of historical multi-turn conversations is 1000, that is, the total number of user sentences of the first-turn conversations is 1000, wherein statistical analysis is performed to obtain 150 user sentences corresponding to user intention 1, and the proportion of the number of user sentences corresponding to user intention 1 in the total number of user sentences is 15%.
Further, the traffic index data may include conversion rate. Where conversion generally refers to user behavior of a user that occurs in a predetermined manner, e.g., user behavior such as clicking, purchasing, etc.; conversion refers to the proportion of the user of the historical multi-turn dialog that occurs at a predetermined user behavior, for example, the proportion is 9%. Generally, the higher the conversion, the better.
In one example, prior to the statistical analysis, the method further comprises:
acquiring a confidence score based on which the intention classification model determines the intention of the user;
the statistical analysis includes:
determining a user statement in a current dialog in a preset number of historical multi-dialog rounds with the lowest confidence score as a low confidence sample corresponding to the user intention for any user intention in all user intentions included in the statistical analysis result; the statistical analysis result also comprises a low confidence sample corresponding to each user intention.
For example, the intent classification model may score the user sentences below each user intent one by one to obtain (user sentences, scores) pairs, and then rank all (user sentences, scores) pairs in ascending order of scores to obtain a list of [ (user sentences 0, scores 0) … (user sentences n, scores n) ] as L, taking the first m elements in L as low confidence samples. Where n is the number of user sentences below the user intent and m is a preset value.
In one example, the statistical analysis includes:
dividing each historical multi-round dialogue of the user and the manual customer service into a plurality of class clusters by using a clustering mode;
matching any user intention in the user intentions included in the statistical analysis result to a target class cluster in the plurality of class clusters by using a text matching mode;
extracting conversation sentences of high-conversion-rate manual customer service from a plurality of historical multi-round conversations included in the target class cluster, wherein the conversation sentences are used as recommendation strategies of conversation sentences of robot customer service in multi-round conversations of a user and the robot customer service; the statistical analysis result also comprises recommendation strategies corresponding to the intentions of the users.
It will be appreciated that the process of multiple rounds of dialogue interactions may be seen as understanding the user's sentences, deriving the user's intent, and then selecting the corresponding strategy for answer. Generally, when a policy is formulated for a certain user intention, only the service experience of a worker can be used for judging, and no reference point exists.
And in step 24, taking each user intention of the first round as a first-layer node in the tree hierarchy to be displayed, and establishing a path between the first-layer node and a root node of the tree hierarchy. It will be appreciated that each first level node corresponds to the user intent of one first round, the number of user intents of the first round being the same as the number of first level nodes.
In one example, the statistical analysis result further includes traffic data and business index data corresponding to each user intention therein, respectively; the method further comprises the steps of:
and taking the flow data and the business index data corresponding to the intention of each user as node additional information which can be displayed of the corresponding node.
Further, the tree hierarchy is a dynamic structure, and the method further comprises:
when the business index data is changed from the business index data of the first index to the business index data of the second index, the business index data included in the node additional information is changed; or when the business index data of the third index is newly added, inserting the business index data of the third index into the node additional information; or deleting the business index data of the fourth index in the node additional information when the business index data cancel the business index data of the fourth index.
In one example, the statistical analysis result further includes a low confidence sample corresponding to each user intention as node additional information of the corresponding node available for presentation.
In one example, the statistical analysis result further includes a recommendation policy corresponding to each user intention, which is node additional information of the corresponding node that is available for presentation.
Next, in step 25, according to the turn sequence of each turn of the dialogs in the historical multi-turn dialogs, performing statistical analysis on each turn, where the statistical analysis on any ith turn includes, for any first node of the ith-1 layer, screening out current turn dialogs with user intention tracks conforming to paths from the root node to the first node from the historical multi-turn dialogs, and performing statistical analysis on user intention corresponding to each current turn dialog, so as to obtain a statistical analysis result of the current turn dialogs, where the statistical analysis result at least includes a plurality of current turn user intentions. It will be appreciated that the above i is greater than or equal to 2, the first level node is established for statistical analysis of each historical multi-round session, and the subsequent second or third level node is established for statistical analysis of a subset of each historical multi-round session.
In the embodiment of the present disclosure, the statistical analysis for any ith round is similar to the statistical analysis for the first round of dialogue, and the content included in the obtained statistical analysis result is similar, and only the range of the statistical historical multi-round dialogue is different, which is not described herein.
Finally, in step 26, the current round user intents are taken as a plurality of i-layer nodes, and paths from the i-layer nodes to the first node are established. It will be appreciated that the process of refining user intent as dialog turns increase may be embodied by paths between nodes.
In one example, the method further comprises: and sequentially displaying the nodes of each level in the tree hierarchy chart according to the gradual operation of the user.
Further, the step-by-step operation according to the user sequentially displays the nodes of each level in the tree hierarchy chart, including:
receiving a first operation of a user on any second node in the tree hierarchy, wherein the first operation is used for indicating to expand a next layer node of the second node;
a next level node connected to the second node through the path is shown.
Further, according to the step-by-step operation of the user, the method sequentially displays the nodes of each level in the tree hierarchy chart, and further includes:
receiving a second operation of a user on the second node, wherein the second operation is used for indicating to pack up the unfolded next-layer node of the second node;
the next level node that has been shown is collapsed.
Further, the step-by-step operation according to the user sequentially displays the nodes of each level in the tree hierarchy chart, including:
receiving third operation of a user on any second node in the tree hierarchy, wherein the third operation is used for indicating node additional information for expanding the second node;
node-attached information of the second node is displayed in a predetermined area associated with the second node.
Further, according to the step-by-step operation of the user, the method sequentially displays the nodes of each level in the tree hierarchy chart, and further includes:
receiving a fourth operation of a user on the second node, wherein the fourth operation is used for indicating unfolded node additional information of the second node to be folded;
folding the displayed node additional information.
It may be appreciated that the first operation, the second operation, the third operation, or the fourth operation may be a mouse click operation, or a touch operation on a touch screen.
Firstly, acquiring historical multi-round conversations of each user and customer service in a preset time period by the method provided by the embodiment of the specification; then, aiming at any one of the historical multi-round conversations, determining the user intention corresponding to each round of conversations in the historical multi-round conversations by using an intention classification model; then, carrying out statistical analysis on user intentions corresponding to the first-round conversations in the historical multi-round conversations respectively to obtain statistical analysis results of the first-round conversations, wherein the statistical analysis results at least comprise the user intentions of the first-round conversations; respectively taking each user intention of a first round as a first-layer node in a tree hierarchy to be displayed, and establishing paths of the first-layer node and a root node of the tree hierarchy; then, according to the turn sequence of each turn of dialogue in the historical multi-turn dialogue, carrying out statistical analysis on each turn, wherein the statistical analysis on any ith turn comprises, for any first node of the ith-1 layer, screening out the current turn dialogue with user intention track conforming to the path from the root node to the first node from each historical multi-turn dialogue, carrying out statistical analysis on the user intention corresponding to each current turn of dialogue, and obtaining the statistical analysis result of the current turn of dialogue, wherein the statistical analysis result at least comprises a plurality of current turn of user intention; and taking the intentions of the current round users as a plurality of i-layer nodes, and establishing paths from the i-layer nodes to any first node of the i-1 layer. From the above, in the embodiment of the present disclosure, by performing statistical analysis on the user intentions corresponding to each dialog in the historical multi-dialog, determining each level node of the tree-like hierarchical structure to be displayed, and constructing the path between each level node, the embodiment of the present disclosure can provide effective information for more accurately identifying the user intentions, thereby improving the working efficiency.
In the embodiment of the specification, statistical analysis is performed based on a large number of historical dialogue logs of users and customer service, and a viewing framework is constructed according to the statistical analysis result. The viewing framework is a visual framework, provides a display scheme for related indexes and key points sensitive to business and algorithm, and is in the form of a tree hierarchy chart. The tree hierarchy structure is composed of nodes, each node corresponds to a corresponding user intention, the user intention can be simply called intention, each node is provided with node additional information which can be displayed as main information of the node, the node additional information comprises flow, conversion rate, low-score user speaking operation, recommendation strategies and the like, and optionally, timestamp information can be further included. The inspection frame is incremental, and can support changing, inserting and deleting specific nodes along with the change of the business indexes, and the tree hierarchy structure of the inspection frame is expanded into a dynamic tree diagram structure, so that node addition, deletion and modification are supported, and the expansion performance is ensured. The inspection framework has good mobility, and can be rapidly switched among different scene services without resistance connection.
FIG. 3 illustrates a view frame construction schematic diagram in accordance with one embodiment. Referring to fig. 3, the statistical analysis for the historical dialog log mainly includes four parts, data statistics, intent scoring, clustering, and text matching, respectively. In the data statistics part, data statistics can be performed through structured query language (structured query language, SQL) to obtain traffic index data such as traffic or traffic ratio and conversion rate of different users intention in different rounds in user session operation in a certain time period, the traffic index data is used as node information in a viewing frame node, the node information supports incremental expansion, and the traffic index data is changed according to different indexes of different services. In the intent scoring part, the user utterances under each intent can be scored one by one according to the intent classification model to obtain (user utterances, scores) pairs, and then a preset number of low-scoring user utterances are screened out as low-confidence samples and integrated into node information. In the clustering part, each historical multi-round dialogue of the user and the manual customer service can be divided into a plurality of class clusters. In a text matching part, matching any user intention in all user intentions included in the statistical analysis result to a target class cluster in the plurality of class clusters by using a text matching mode; and extracting conversation sentences of the high-conversion-rate manual customer service from a plurality of historical multi-round conversations included in the target class cluster, and integrating the conversation sentences into node information as a recommendation strategy of the conversation sentences of the robot customer service in the multi-round conversations of the user and the robot customer service.
FIG. 4 illustrates a view frame morphology diagram, according to one embodiment. Referring to fig. 4, a root node is first shown, then a first-layer node, namely, a node 1, a node 2, a node 3 and a node 4 in the graph, can be shown according to a clicking operation of a user, can simultaneously show main information of the node, such as user intention, flow, conversion rate and the like corresponding to the node, supports page skipping according to the operation of the user, and shows auxiliary information of the node, such as low-score user speaking and recommending strategies, through the page after skipping. Similarly, the second-layer nodes may be displayed according to a click operation of the user, for example, if the user clicks on the node 1, the displayed second-layer nodes are the node 11, the node 12, the node 13, and the node 14 in the graph. It may be appreciated that the first level node corresponds to the user intent of the first dialog of the plurality of dialogues, the second level node corresponds to the user intent of the second dialog of the plurality of dialogues, and accordingly, if the plurality of dialogues further includes the third dialog, the third level node may also be included in the presentation frame.
Because the complexity of user group speaking operation is often insufficient in fixed intention quantity, the requirements on intention splitting and intention refining are met, and the inspection framework provided by the embodiment of the specification simultaneously characterizes the flow state distribution of the nodes by showing key indexes of service scenes through node information, and can conveniently split the nodes in finer granularity from the service angle through the key indexes and the flow states of the nodes, so that the requirement on manpower is greatly reduced. The framework has good expansibility, can incrementally adjust the information quantity borne by the nodes along with the change of the service, has good mobility of multi-service scenes, and greatly reduces the time cost of manpower in the scene migration process.
According to another aspect, the embodiment further provides a device for statistical analysis and result display of user intention, wherein the user intention is the intention expressed by user sentences in a plurality of historical conversations of a user and customer service, and the device is used for executing the statistical analysis and result display method of the user intention provided by the embodiment of the specification. FIG. 5 shows a schematic block diagram of a statistical analysis of user intent and results presentation device, according to one embodiment. As shown in fig. 5, the apparatus 500 includes:
a first obtaining unit 51, configured to obtain each historical multi-round dialogue between each user and customer service in a preset period of time;
an intention determining unit 52, configured to determine, for any one of the historical multi-turn dialogs acquired by the first acquiring unit 51, a user intention corresponding to each of the historical multi-turn dialogs by using an intention classification model;
a first statistics unit 53, configured to perform statistical analysis on user intentions corresponding to first-round dialogs in each of the historical multi-round dialogs determined by the intent determining unit 52, to obtain a statistical analysis result of the first-round dialogs, where the statistical analysis result includes at least each user intention of the first-round dialog;
a first path establishing unit 54, configured to establish paths between a first-level node and a root node of a tree hierarchy of the tree hierarchy to be displayed by using each user intention of the first round obtained by the first statistics unit 53 as a first-level node in the tree hierarchy to be displayed;
A second statistics unit 55, configured to perform a statistical analysis on each turn according to a turn sequence of each turn of the sessions in the historical multi-turn sessions, where the statistical analysis on any ith turn includes, for any first node of the ith-1 layer, screening out a current turn of the session with a user intention track conforming to a path from the root node to the first node from each historical multi-turn session acquired by the first acquisition unit 51, and performing a statistical analysis on user intention corresponding to each current turn of the session, to obtain a statistical analysis result of the current turn of the session, where the statistical analysis result includes at least a plurality of current turn user intentions;
a second path establishing unit 56, configured to establish paths from the plurality of i-th layer nodes to the first node by using the plurality of current round user intentions obtained by the second statistics unit 55 as a plurality of i-th layer nodes.
Optionally, as an embodiment, the statistical analysis result further includes flow data and business index data corresponding to each user intention therein; the apparatus further comprises:
and the additional information determining unit is used for taking the flow data and the business index data corresponding to the intention of each user as the node additional information which can be displayed of the corresponding node.
Optionally, as an embodiment, the apparatus further includes:
and the display unit is used for sequentially displaying the nodes of each level in the tree-like hierarchical structure chart according to the gradual operation of the user.
Further, the flow data includes the number of user sentences corresponding to the user intention, and/or the proportion of the number of user sentences corresponding to the user intention in the total number of user sentences.
Further, the business index data includes conversion rate.
Optionally, as an embodiment, the apparatus further includes:
a second obtaining unit configured to obtain, before statistical analysis by the first statistical unit 53 or the second statistical unit 55, a confidence score on which the intention classification model is based when determining a user intention;
the first statistics unit 53 or the second statistics unit 55 is specifically configured to determine, for any user intention of the user intentions included in the statistical analysis result, a user sentence in a current dialog among a preset number of historical multi-dialog with a lowest confidence score, as a low confidence sample corresponding to the user intention; the statistical analysis result also comprises a low confidence coefficient sample corresponding to each user intention, and the low confidence coefficient sample is used as node additional information of a corresponding node for display.
Optionally, as an embodiment, the first statistics unit 53 or the second statistics unit 55 includes:
the clustering subunit is used for dividing each historical multi-round dialogue of the user and the manual customer service into a plurality of class clusters by using a clustering mode;
the matching subunit is used for matching any user intention in the user intentions included in the statistical analysis result to a target class cluster in a plurality of class clusters divided by the clustering subunit in a text matching mode;
the extraction subunit is used for extracting conversation sentences of the high-conversion-rate manual customer service from a plurality of historical multi-round conversations included in the target class cluster matched by the matching subunit, and the conversation sentences are used as recommendation strategies of the conversation sentences of the robot customer service in the multi-round conversations of the user and the robot customer service; the statistical analysis result also comprises a recommendation strategy corresponding to each user intention, and the recommendation strategy is used as node additional information which can be displayed for the corresponding node.
Further, the tree hierarchy is a dynamic structure, and the apparatus further includes:
a dynamic adjustment unit, configured to change the business index data included in the node additional information when the business index data is changed from the business index data of the first index to the business index data of the second index; or when the business index data of the third index is newly added, inserting the business index data of the third index into the node additional information; or deleting the business index data of the fourth index in the node additional information when the business index data cancel the business index data of the fourth index.
Further, the display unit includes:
a first receiving subunit, configured to receive a first operation of a user on an arbitrary second node in the tree hierarchy, where the first operation is used to instruct to expand a node of a next layer of the second node;
a presentation subunit for presenting a next level node connected to the second node through the path.
Further, the display unit further includes:
a second receiving subunit, configured to receive a second operation of the second node by a user, where the second operation is used to instruct to stow a deployed next-layer node of the second node;
and the stowing subunit is used for stowing the displayed next-level node.
Further, the display unit includes:
a third receiving subunit, configured to receive a third operation of a user on an arbitrary second node in the tree hierarchy, where the third operation is used to instruct to expand node additional information of the second node;
and a display subunit configured to display node-attached information of the second node in a predetermined area associated with the second node.
Further, the display unit further includes:
A fourth receiving subunit, configured to receive a fourth operation of the second node by a user, where the fourth operation is used to instruct folding of expanded node additional information of the second node;
and the folding subunit is used for folding the displayed node additional information.
With the apparatus provided in the embodiment of the present specification, first, the first obtaining unit 51 obtains each history multi-round dialogue of each user and customer service in a preset period of time; then, the intention determining unit 52 determines, for any one of the historical multi-turn conversations, the user intention corresponding to each of the historical multi-turn conversations, respectively, using the intention classification model; next, the first statistics unit 53 performs statistical analysis on the user intentions corresponding to the first-round conversations in the historical multi-round conversations respectively to obtain statistical analysis results of the first-round conversations, wherein the statistical analysis results at least comprise the user intentions of the first-round conversations; the first path establishing unit 54 then uses the first user intentions as the first level nodes in the tree hierarchy to be displayed, and establishes paths between the first level nodes and the root nodes of the tree hierarchy; next, the second statistics unit 55 performs statistical analysis on each turn according to the turn sequence of each turn of the dialogs in the historical multi-turn dialogs, wherein the statistical analysis on any ith turn includes, for any first node of the ith layer-1, screening out current turn dialogs with user intention tracks conforming to paths from the root node to the first node from each historical multi-turn dialogs, and performing statistical analysis on user intention corresponding to each current turn dialog to obtain a statistical analysis result of the current turn dialogs, wherein the statistical analysis result at least comprises a plurality of current turn user intentions; the second path establishing unit 56 establishes paths from the plurality of i-layer nodes to any of the i-1-layer first nodes, using the plurality of current round user intents as a plurality of i-layer nodes. From the above, in the embodiment of the present disclosure, by performing statistical analysis on the user intentions corresponding to each dialog in the historical multi-dialog, determining each level node of the tree-like hierarchical structure to be displayed, and constructing the path between each level node, the embodiment of the present disclosure can provide effective information for more accurately identifying the user intentions, thereby improving the working efficiency.
According to an embodiment of another aspect, there is also provided a computer-readable storage medium having stored thereon a computer program which, when executed in a computer, causes the computer to perform the method described in connection with fig. 2.
According to an embodiment of yet another aspect, there is also provided a computing device including a memory having executable code stored therein and a processor that, when executing the executable code, implements the method described in connection with fig. 2.
Those skilled in the art will appreciate that in one or more of the examples described above, the functions described in the present invention may be implemented in hardware, software, firmware, or any combination thereof. When implemented in software, these functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium.
The foregoing embodiments have been provided for the purpose of illustrating the general principles of the present invention in further detail, and are not to be construed as limiting the scope of the invention, but are merely intended to cover any modifications, equivalents, improvements, etc. based on the teachings of the invention.
Claims (26)
1. A statistical analysis and result presentation method of user intent expressed by user sentences in a historical multi-round conversation of a user with customer service, the method comprising:
acquiring historical multi-round conversations of each user and customer service in a preset time period;
determining user intentions corresponding to each dialog in each history multi-turn dialog by using an intention classification model according to any history multi-turn dialog in each history multi-turn dialog;
carrying out statistical analysis on user intentions corresponding to the first-round conversations in each historical multi-round conversation respectively to obtain statistical analysis results of the first-round conversations, wherein the statistical analysis results at least comprise the user intentions of the first-round conversations;
taking each user intention of a first round as a first-layer node in a tree hierarchy to be displayed, and establishing paths of the first-layer node and a root node of the tree hierarchy;
according to the turn sequence of each turn of dialogue in the historical multi-turn dialogue, carrying out statistical analysis on each turn, wherein the statistical analysis on any ith turn comprises, for any first node of the ith-1 layer, screening out current turn dialogue with user intention track conforming to the path from the root node to the first node from each historical multi-turn dialogue, carrying out statistical analysis on user intention corresponding to each current turn dialogue, and obtaining statistical analysis results of the current turn dialogue, wherein the statistical analysis results at least comprise a plurality of current turn user intentions;
And taking the intentions of the current round users as a plurality of i-layer nodes, and establishing paths from the i-layer nodes to the first node.
2. The method of claim 1, wherein the statistical analysis result further comprises traffic data and traffic index data corresponding to each user intention therein, respectively; the method further comprises the steps of:
and taking the flow data and the business index data corresponding to the intention of each user as node additional information which can be displayed of the corresponding node.
3. The method of claim 1, wherein the method further comprises: and sequentially displaying the nodes of each level in the tree hierarchy chart according to the gradual operation of the user.
4. The method of claim 2, wherein the traffic data includes a number of user sentences corresponding to user intentions and/or a proportion of the number of user sentences corresponding to user intentions to a total number of user sentences.
5. The method of claim 2, wherein the traffic metric data comprises conversion rate.
6. The method of claim 1, wherein prior to the statistical analysis, the method further comprises:
acquiring a confidence score based on which the intention classification model determines the intention of the user;
The statistical analysis includes:
determining a user statement in a current dialog in a preset number of historical multi-dialog rounds with the lowest confidence score as a low confidence sample corresponding to the user intention for any user intention in all user intentions included in the statistical analysis result; the statistical analysis result also comprises a low confidence coefficient sample corresponding to each user intention, and the low confidence coefficient sample is used as node additional information of a corresponding node for display.
7. The method of claim 1, wherein the statistical analysis comprises:
dividing each historical multi-round dialogue of the user and the manual customer service into a plurality of class clusters by using a clustering mode;
matching any user intention in the user intentions included in the statistical analysis result to a target class cluster in the plurality of class clusters by using a text matching mode;
extracting conversation sentences of high-conversion-rate manual customer service from a plurality of historical multi-round conversations included in the target class cluster, wherein the conversation sentences are used as recommendation strategies of conversation sentences of robot customer service in multi-round conversations of a user and the robot customer service; the statistical analysis result also comprises a recommendation strategy corresponding to each user intention, and the recommendation strategy is used as node additional information which can be displayed for the corresponding node.
8. The method of claim 2, wherein the tree hierarchy is a dynamic structure, the method further comprising:
when the business index data is changed from the business index data of the first index to the business index data of the second index, the business index data included in the node additional information is changed; or when the business index data of the third index is newly added, inserting the business index data of the third index into the node additional information; or deleting the business index data of the fourth index in the node additional information when the business index data cancel the business index data of the fourth index.
9. A method according to claim 3, wherein said sequentially exposing nodes of each level in said tree hierarchy according to a stepwise operation by a user comprises:
receiving a first operation of a user on any second node in the tree hierarchy, wherein the first operation is used for indicating to expand a next layer node of the second node;
a next level node connected to the second node through the path is shown.
10. The method of claim 9, wherein the sequentially exposing the nodes of each level in the tree hierarchy according to the step-by-step operation of the user further comprises:
Receiving a second operation of a user on the second node, wherein the second operation is used for indicating to pack up the unfolded next-layer node of the second node;
the next level node that has been shown is collapsed.
11. A method according to claim 3, wherein said sequentially exposing nodes of each level in said tree hierarchy according to a stepwise operation by a user comprises:
receiving third operation of a user on any second node in the tree hierarchy, wherein the third operation is used for indicating node additional information for expanding the second node;
node-attached information of the second node is displayed in a predetermined area associated with the second node.
12. The method of claim 11, wherein the sequentially exposing the nodes of each level in the tree hierarchy according to the step-by-step operation of the user further comprises:
receiving a fourth operation of a user on the second node, wherein the fourth operation is used for indicating unfolded node additional information of the second node to be folded;
folding the displayed node additional information.
13. A statistical analysis and results presentation device of user intent, the user intent being an intent expressed by a user sentence in a historical multi-turn conversation of a user with a customer service, the device comprising:
The first acquisition unit is used for acquiring historical multi-round conversations between each user and customer service in a preset time period;
the intention determining unit is used for determining the user intention corresponding to each dialog in each history multi-turn dialog by using the intention classifying model according to any history multi-turn dialog in each history multi-turn dialog acquired by the first acquiring unit;
the first statistical unit is used for carrying out statistical analysis on the user intentions respectively corresponding to the first-round conversations in the historical multi-round conversations determined by the intent determining unit to obtain statistical analysis results of the first-round conversations, wherein the statistical analysis results at least comprise the user intentions of the first-round conversations;
the first path establishing unit is used for respectively taking each user intention of the first round obtained by the first statistical unit as a first layer node in the tree hierarchy to be displayed and establishing a path between the first layer node and a root node of the tree hierarchy;
the second statistical unit is used for carrying out statistical analysis on each round according to the round sequence of each round of dialogue in the historical multi-round dialogue, wherein the statistical analysis on any ith round comprises screening out current round dialogue with user intention track conforming to the path from the root node to the first node from the historical multi-round dialogue for any first node of the ith layer-1, and carrying out statistical analysis on the user intention corresponding to each current round dialogue respectively to obtain the statistical analysis result of the current round dialogue, wherein the statistical analysis result at least comprises a plurality of current round user intentions;
And the second path establishing unit is used for taking the current round user intentions obtained by the second statistical unit as a plurality of i-layer nodes and establishing paths from the i-layer nodes to the first node.
14. The apparatus of claim 13, wherein the statistical analysis result further comprises traffic data and traffic index data corresponding to respective user intents therein; the apparatus further comprises:
and the additional information determining unit is used for taking the flow data and the business index data corresponding to the intention of each user as the node additional information which can be displayed of the corresponding node.
15. The apparatus of claim 13, wherein the apparatus further comprises:
and the display unit is used for sequentially displaying the nodes of each level in the tree-like hierarchical structure chart according to the gradual operation of the user.
16. The apparatus of claim 14, wherein the traffic data includes a number of user sentences corresponding to user intents and/or a proportion of the number of user sentences corresponding to user intents to a total number of user sentences.
17. The apparatus of claim 14, wherein the traffic metric data comprises conversion rate.
18. The apparatus of claim 13, wherein the apparatus further comprises:
A second obtaining unit, configured to obtain, before the statistical analysis by the first statistical unit or the second statistical unit, a confidence score based on which the intention classification model determines the intention of the user;
the first statistical unit or the second statistical unit is specifically configured to determine, for any user intention in each user intention included in the statistical analysis result, a user sentence in a current dialog among a preset number of historical multi-dialog rounds with a lowest confidence score, as a low confidence sample corresponding to the user intention; the statistical analysis result also comprises a low confidence coefficient sample corresponding to each user intention, and the low confidence coefficient sample is used as node additional information of a corresponding node for display.
19. The apparatus of claim 13, wherein the first statistical unit or the second statistical unit comprises:
the clustering subunit is used for dividing each historical multi-round dialogue of the user and the manual customer service into a plurality of class clusters by using a clustering mode;
the matching subunit is used for matching any user intention in the user intentions included in the statistical analysis result to a target class cluster in a plurality of class clusters divided by the clustering subunit in a text matching mode;
The extraction subunit is used for extracting conversation sentences of the high-conversion-rate manual customer service from a plurality of historical multi-round conversations included in the target class cluster matched by the matching subunit, and the conversation sentences are used as recommendation strategies of the conversation sentences of the robot customer service in the multi-round conversations of the user and the robot customer service; the statistical analysis result also comprises a recommendation strategy corresponding to each user intention, and the recommendation strategy is used as node additional information which can be displayed for the corresponding node.
20. The apparatus of claim 14, wherein the tree hierarchy is a dynamic structure, the apparatus further comprising:
a dynamic adjustment unit, configured to change the business index data included in the node additional information when the business index data is changed from the business index data of the first index to the business index data of the second index; or when the business index data of the third index is newly added, inserting the business index data of the third index into the node additional information; or deleting the business index data of the fourth index in the node additional information when the business index data cancel the business index data of the fourth index.
21. The apparatus of claim 15, wherein the presentation unit comprises:
a first receiving subunit, configured to receive a first operation of a user on an arbitrary second node in the tree hierarchy, where the first operation is used to instruct to expand a node of a next layer of the second node;
a presentation subunit for presenting a next level node connected to the second node through the path.
22. The apparatus of claim 21, wherein the display unit further comprises:
a second receiving subunit, configured to receive a second operation of the second node by a user, where the second operation is used to instruct to stow a deployed next-layer node of the second node;
and the stowing subunit is used for stowing the displayed next-level node.
23. The apparatus of claim 15, wherein the presentation unit comprises:
a third receiving subunit, configured to receive a third operation of a user on an arbitrary second node in the tree hierarchy, where the third operation is used to instruct to expand node additional information of the second node;
and a display subunit configured to display node-attached information of the second node in a predetermined area associated with the second node.
24. The apparatus of claim 23, wherein the display unit further comprises:
a fourth receiving subunit, configured to receive a fourth operation of the second node by a user, where the fourth operation is used to instruct folding of expanded node additional information of the second node;
and the folding subunit is used for folding the displayed node additional information.
25. A computer readable storage medium having stored thereon a computer program which, when executed in a computer, causes the computer to perform the method of any of claims 1-12.
26. A computing device comprising a memory having executable code stored therein and a processor which, when executing the executable code, implements the method of any of claims 1-12.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011034837.XA CN112115249B (en) | 2020-09-27 | 2020-09-27 | Statistical analysis and result display method and device for user intention |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011034837.XA CN112115249B (en) | 2020-09-27 | 2020-09-27 | Statistical analysis and result display method and device for user intention |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112115249A CN112115249A (en) | 2020-12-22 |
| CN112115249B true CN112115249B (en) | 2023-11-14 |
Family
ID=73797823
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011034837.XA Active CN112115249B (en) | 2020-09-27 | 2020-09-27 | Statistical analysis and result display method and device for user intention |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112115249B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113449093B (en) * | 2021-07-08 | 2024-04-16 | 京东科技控股股份有限公司 | Intelligent response method and device and electronic equipment |
| CN113641897B (en) * | 2021-07-28 | 2024-04-19 | 上海明略人工智能(集团)有限公司 | Recommendation method and device based on session text, electronic equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019011356A1 (en) * | 2017-07-14 | 2019-01-17 | Cognigy Gmbh | Method for conducting dialog between human and computer |
| CN110637339A (en) * | 2017-05-15 | 2019-12-31 | 苹果公司 | Optimizing dialog policy decisions for a digital assistant using implicit feedback |
| CN111400480A (en) * | 2020-04-21 | 2020-07-10 | 支付宝(杭州)信息技术有限公司 | User intention recognition method and device for multi-turn conversation |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10216832B2 (en) * | 2016-12-19 | 2019-02-26 | Interactions Llc | Underspecification of intents in a natural language processing system |
-
2020
- 2020-09-27 CN CN202011034837.XA patent/CN112115249B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110637339A (en) * | 2017-05-15 | 2019-12-31 | 苹果公司 | Optimizing dialog policy decisions for a digital assistant using implicit feedback |
| WO2019011356A1 (en) * | 2017-07-14 | 2019-01-17 | Cognigy Gmbh | Method for conducting dialog between human and computer |
| CN111400480A (en) * | 2020-04-21 | 2020-07-10 | 支付宝(杭州)信息技术有限公司 | User intention recognition method and device for multi-turn conversation |
Non-Patent Citations (1)
| Title |
|---|
| 一种基于记忆网络的多轮对话下的意图识别方法;杨成彪;吕荣荣;吴刚;;电子技术与软件工程(10);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112115249A (en) | 2020-12-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10679624B2 (en) | Personal directory service | |
| US8041568B2 (en) | Business listing search | |
| US7890326B2 (en) | Business listing search | |
| US9282162B2 (en) | Processing user profiles of users in an electronic community | |
| US9251251B2 (en) | Dynamic language model | |
| US7451161B2 (en) | Compatibility scoring of users in a social network | |
| US11238034B2 (en) | Recommendation platform for structured queries | |
| CN112115249B (en) | Statistical analysis and result display method and device for user intention | |
| US7792808B2 (en) | More efficient search algorithm (MESA) using virtual search parameters | |
| US20140279803A1 (en) | Disambiguating data using contextual and historical information | |
| US20190333499A1 (en) | Pronunciation analysis and correction feedback | |
| CN114723528A (en) | Commodity personalized recommendation method and system based on knowledge graph | |
| CN104036004B (en) | Search for error correction method and search error correction device | |
| US11226972B2 (en) | Ranking collections of document passages associated with an entity name by relevance to a query | |
| US20110219299A1 (en) | Method and system of providing completion suggestion to a partial linguistic element | |
| CN110659349A (en) | Log query method, device, equipment and computer readable storage medium | |
| CN105843889A (en) | Credibility based big data and general data oriented data collection method and system | |
| KR101231438B1 (en) | Search reasult providing system and method for providing loan word pronunciation search service | |
| US20230143079A1 (en) | Systems and methods for discovering and quantifying principal factors or dominant paths in interconnected journey data | |
| US8738612B1 (en) | Resolving ambiguous queries | |
| CA2603778A1 (en) | Method and system for searching and ranking entries stored in a directory | |
| WO2019179208A1 (en) | Drop-down suggestion list | |
| US20120158695A1 (en) | More efficient search algorithm (MESA) using: integrated system optimizer | |
| CN109558531A (en) | News information method for pushing, device and computer equipment | |
| US20140165002A1 (en) | Method and system using natural language processing for multimodal voice configurable input menu elements |
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 |