CN107908803A - The response method and device, storage medium, terminal of question and answer interaction - Google Patents
The response method and device, storage medium, terminal of question and answer interaction Download PDFInfo
- Publication number
- CN107908803A CN107908803A CN201711434955.8A CN201711434955A CN107908803A CN 107908803 A CN107908803 A CN 107908803A CN 201711434955 A CN201711434955 A CN 201711434955A CN 107908803 A CN107908803 A CN 107908803A
- Authority
- CN
- China
- Prior art keywords
- answer
- question
- customer problem
- prediction probability
- interaction
- 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
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
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
Abstract
A kind of response method and device, storage medium, terminal of question and answer interaction, the response of question and answer interaction include：Obtain customer problem；Multiple answers for the customer problem are obtained using at least two modes；The multiple answer is screened according to the source parameter of each answer and prediction probability, the source parameter and prediction probability of the answer are determined according to the acquisition modes of the answer；The optimal answer that output screening obtains.The accuracy and continuity that are responded in question and answer interaction can be improved by technical solution of the present invention.
Description
Technical field
The present invention relates to natural language processing technique field, more particularly to a kind of response method of question and answer interaction and device,
Storage medium, terminal.
Background technology
In artificial intelligence technology application field, there are more and more intelligent answer products.Typically, for user
The reply accuracy and reply speed of problem are an important factor for influencing intelligent answer product quality.
A variety of question and answer processing modes exist in the prior art, it is common to have rule-based mode, based on template matches
Mode, the mode based on retrieval, based on production mode etc..Wherein, the mode based on retrieval be by retrieval knowledge storehouse
Some knowledge points produce answer, and knowledge base generally includes multiple knowledge points, its corresponding expansion is asked about in each knowledge point including standard
Exhibition is asked and answer；Answer feedback mechanism based on production is the Automatic generation of information inputted according to active user by word sequence
Arrange the answer of composition.
But in rule-based, template matches, the mode of retrieval, template, example or database have limitation, and
And lack effective language understanding, cause the presence of certain deficiency in the accuracy of answer and flexibility；Based on production mode
Need to establish and training pattern, model complexity height, the stability of answer acquisition process are low.
The content of the invention
Present invention solves the technical problem that it is how to improve the accuracy and continuity responded in question and answer interaction.
In order to solve the above technical problems, the embodiment of the present invention provides a kind of response method of question and answer interaction, question and answer interaction
Response method includes：Obtain customer problem；Multiple answers for the customer problem are obtained using at least two modes；According to
The source parameter and prediction probability of each answer screen the multiple answer, the source parameter of the answer and prediction
Probability is determined according to the acquisition modes of the answer；The optimal answer that output screening obtains.
Optionally, at least two mode is selected from knowledge base, knowledge mapping and learning model.
Optionally, one or more in the following ways determine the prediction probability of answer：If the answer comes from institute
State knowledge base, then calculate the customer problem and the knowledge base Plays and ask and/or extend the semantic similarity asked, using as
The prediction probability of the answer；If the answer comes from the knowledge mapping, according to determining the knowledge mapping
The confidence level of answer determines the prediction probability of the answer；If the answer carrys out self learning model, according to the answer
The sum of conditional probability between adjacent word determines the prediction probability of the answer.
Optionally, the source parameter includes priority, and the priority of the answer from knowledge base, which is higher than, comes from knowledge graph
The priority of the answer of spectrum, the priority of the answer from knowledge mapping are higher than the priority for the answer for carrying out self learning model.
Optionally, the source parameter and prediction probability according to each answer carries out screening bag to the multiple answer
Include：According to the priority sequence of answer, judge whether the prediction probability of each answer is more than given threshold successively, and by head
The secondary prediction probability for judging to obtain is more than the answer of the given threshold as optimal answer.
Optionally, it is described to obtain multiple answers for the customer problem using following using at least two modes
Any two or three of mode obtains the multiple answer：The customer problem is calculated to ask and/or extend with knowledge base Plays
The semantic similarity asked, and determine the first answer from the knowledge base；By the customer problem and knowing in knowledge mapping
Knowledge is matched, and determines the second answer from the knowledge mapping；The customer problem is inputted into learning model, and is determined
The output of the learning model is the 3rd answer.
Optionally, the source parameter and prediction probability according to each answer carries out screening bag to the multiple answer
Include：The importance weight of the answer is determined according to the source parameter of each answer；This is determined according to the prediction probability of each answer
The accuracy weights of answer；Utilize the importance weight of each answer and effective fraction of each answer of accuracy weight computing；
The multiple answer is screened according to effective fraction of each answer.
Optionally, generation of the source parameter of the answer including answer takes.
Optionally, the source parameter and prediction probability according to each answer carries out screening bag to the multiple answer
Include：The corresponding importance weight of the answer is determined according to the source parameter of each answer；Calculate the importance weight of each answer
With the product of prediction probability；The product of each answer business time-consuming with generation is calculated, using effective fraction as each answer；Root
The multiple answer is screened according to effective fraction of each answer.
Optionally, the customer problem is voice；The acquisition customer problem includes：The customer problem is converted into text
This, the optimal answer that the output screening obtains includes：User is sent to after the optimal answer of acquisition is converted to voice.
Optionally, the step of acquisition customer problem is in response to what is performed after the indication message received.
Optionally, the indication message be using professional knowledge storehouse to the customer problem carry out it fails to match when
Send.
Optionally, the indication message is the intention assessment result in the customer problem and default intent classifier
Sent during with success.
Optionally, the intention assessment is the result is that using the intent classifier model that training is completed in advance to the customer problem
Carry out what intention assessment obtained.
The embodiment of the invention also discloses a kind of responding device of question and answer interaction, the responding device of question and answer interaction includes：With
Family problem acquisition module, to obtain customer problem；Answer acquisition module, to be obtained using at least two modes for described
Multiple answers of customer problem；Answer screening module, to according to the source parameter and prediction probability of each answer to described
Multiple answers are screened, and the source parameter and prediction probability of the answer are determined according to the acquisition modes of the answer；Answer
Case output module, to export the optimal answer that screening obtains.
Optionally, at least two mode is selected from knowledge base, knowledge mapping and learning model.
Optionally, the one or more of the answer screening module in the following ways determine the prediction probability of answer：Such as
Answer described in fruit comes from the knowledge base, then calculates the customer problem and ask with the knowledge base Plays and/or extend what is asked
Semantic similarity, using the prediction probability as the answer；If the answer comes from the knowledge mapping, know according to
The confidence level for knowing the answer that collection of illustrative plates determines determines the prediction probability of the answer；If the answer carrys out self learning model,
The prediction probability of the answer is then determined according to the sum of conditional probability between the adjacent word of the answer.
Optionally, the source parameter includes priority, and the priority of the answer from knowledge base, which is higher than, comes from knowledge graph
The priority of the answer of spectrum, the priority of the answer from knowledge mapping are higher than the priority for the answer for carrying out self learning model.
Optionally, the answer screening module includes：First screening unit, to suitable according to the priority of answer height
Sequence, judges whether the prediction probability of each answer is more than given threshold successively, and the prediction probability judged first is more than
The answer of the given threshold is as optimal answer.
Optionally, the answer acquisition module includes：First answer acquiring unit, to calculate the customer problem with knowing
Know storehouse Plays and ask and/or extend the semantic similarity asked, and determine the first answer from the knowledge base；Second answer obtains
Unit is taken, the customer problem to be matched with the knowledge in knowledge mapping, and is determined from the knowledge mapping
Second answer；3rd answer acquiring unit, the customer problem is inputted learning model, and determines the learning model
Export as the 3rd answer.
Optionally, the answer screening module includes：First importance weight determination unit, to according to each answer
Source parameter determines the importance weight of the answer；Accuracy weights determination unit, to the prediction probability according to each answer
Determine the accuracy weights of the answer；First effective score calculating unit, to utilize the importance weight and standard of each answer
Effective fraction of each answer of true property weight computing；Second screening unit, to according to effective fraction of each answer to described
Multiple answers are screened.
Optionally, generation of the source parameter of the answer including answer takes.
Optionally, the answer screening module includes：Second importance weight determination unit, to according to each answer
Source parameter determines the corresponding importance weight of the answer；Product computing unit, to calculate the importance weight of each answer
With the product of prediction probability；Second effective score calculating unit, to calculate the product of each answer business time-consuming with generation, with
Effective fraction as each answer；Third filtering unit, to according to effective fraction of each answer to the multiple answer
Screened.
Optionally, the customer problem is voice；The customer problem acquisition module includes：First voice converting unit,
The customer problem is converted to text, the answer output module includes：Second voice converting unit, that will obtain
The optimal answer be converted to voice after be sent to user.
Optionally, the customer problem acquisition module after the indication message that receives in response to performing.
Optionally, the indication message be using professional knowledge storehouse to the customer problem carry out it fails to match when
Send.
Optionally, the indication message is the intention assessment result in the customer problem and default intent classifier
Sent during with success.
Optionally, the intention assessment is the result is that using the intent classifier model that training is completed in advance to the customer problem
Carry out what intention assessment obtained.
The embodiment of the invention also discloses a kind of storage medium, is stored thereon with computer instruction, the computer instruction
The step of response method of the question and answer interaction is performed during operation.
The embodiment of the invention also discloses a kind of terminal, including memory and processor, being stored with the memory can
The computer instruction run on the processor, the processor perform the question and answer interaction when running the computer instruction
Response method the step of.
Compared with prior art, the technical solution of the embodiment of the present invention has the advantages that：
Technical solution of the present invention obtains customer problem；Obtained using at least two modes for the multiple of the customer problem
Answer；The multiple answer is screened according to the source parameter of each answer and prediction probability, the source of the answer
Parameter and prediction probability are determined according to the acquisition modes of the answer；The optimal answer that output screening obtains.The technology of the present invention
Scheme obtains multiple answers using at least two modes, and optimal answer output is then screened in multiple answers；Answered due to obtaining
The mode of case is different, therefore can lift the rich of answer from multiple angles, can to avoid using single mode obtain less than
The situation of answer, ensures the sustainability interacted with user's question and answer, lifts user experience.In addition, according to each from multiple answers
The source parameter and prediction probability of a answer choose optimal answer, it is ensured that it is accurate that answer is replied for customer problem
Property.
Further, at least two mode is selected from knowledge base, knowledge mapping and learning model.Technical solution of the present invention profit
Answer is obtained with knowledge base, knowledge mapping and learning model, is obtained based on different technical principles, therefore can into one
Step lifts the rich of answer；In addition, the accuracy of the answer for customer problem got by above-mentioned three kinds of approach compared with
Height, the accuracy higher of the optimal answer obtained after further screening, further improves what is responded in question and answer interaction
Accuracy.
Further, the source parameter and prediction probability according to each answer carries out screening bag to the multiple answer
Include：The corresponding importance weight of the answer is determined according to the source parameter of each answer；Calculate the importance weight of each answer
With the product of prediction probability；The product of each answer business time-consuming with generation is calculated, using effective fraction as each answer；Root
The multiple answer is screened according to effective fraction of each answer.In technical solution of the present invention, when determining optimal answer,
The generation of the acquisition modes of answer, the prediction probability of answer and answer is taken as considerations；The acquisition modes of answer,
The prediction probability of answer can influence the accuracy of answer, and the generation of answer is time-consuming can to influence interactive promptness, will be above-mentioned
Factor is combined and considered, and can be taken into account the reply accuracy and promptness with user mutual, further be lifted user experience.
Brief description of the drawings
Fig. 1 is a kind of flow chart of the response method of question and answer interaction of the embodiment of the present invention；
Fig. 2 is a kind of schematic diagram of embodiment of step S102 shown in Fig. 1；
Fig. 3 is a kind of flow chart of embodiment of step S103 shown in Fig. 1；
Fig. 4 is the flow chart of another embodiment of step S103 shown in Fig. 1；
Fig. 5 is a kind of structure diagram of the responding device of question and answer interaction of the embodiment of the present invention；
Fig. 6 is a kind of structure diagram of embodiment of answer acquisition module shown in Fig. 5 502；
Fig. 7 is a kind of structure diagram of embodiment of answer screening module shown in Fig. 5 503；
Fig. 8 is the structure diagram of another embodiment of answer screening module shown in Fig. 5 503.
Embodiment
As described in the background art, the prior art rule-based, template matches, retrieval mode in, template, example or
Database has limitation, and lacks effective language understanding, causes to exist in the accuracy of answer and flexibility certain
Deficiency；Need to establish based on production mode and training pattern, model complexity height, answer acquisition process stability are low.
Technical solution of the present invention obtains multiple answers using at least two modes, and optimal answer then is screened in multiple answers
Case exports；Since the mode for obtaining answer is different, the rich of answer can be lifted from multiple angles, can be to avoid use
Single mode obtains the situation less than answer, ensures the sustainability interacted with user's question and answer, lifts user experience.In addition, from
Optimal answer is chosen according to the source parameter and prediction probability of each answer in multiple answers, it is ensured that answer is directed to user
The accuracy that problem is replied.
It is understandable to enable the above objects, features and advantages of the present invention to become apparent, below in conjunction with the accompanying drawings to the present invention
Specific embodiment be described in detail.
Fig. 1 is a kind of flow chart of the response method of question and answer interaction of the embodiment of the present invention.
The response method of the interaction of question and answer shown in Fig. 1 can be used for question answering system, and the response method of the question and answer interaction can be with
Comprise the following steps：
Step S101：Obtain customer problem；
Step S102：Multiple answers for the customer problem are obtained using at least two modes；
Step S103：The multiple answer is screened according to the source parameter of each answer and prediction probability, institute
Stating the source parameter of answer and prediction probability is determined according to the acquisition modes of the answer；
Step S104：The optimal answer that output screening obtains.
In the present embodiment, step S101 can use any enforceable mode to obtain customer problem, such as can be direct
Customer problem is collected from outside, customer problem can also be obtained by way of interface calling.Customer problem can possess
It is semantic.Specifically, customer problem can be the forms such as voice, text.
In the specific implementation of step S102, at least two answers can be obtained using at least two modes.Specifically,
The mode for obtaining the answer of customer problem has a variety of, for example, being matched using knowledge base to customer problem, and will be asked with user
Inscribing the standard to match asks corresponding answer as the answer for customer problem；It can also take notice of map space using searching algorithm
In scan for, and using the answer searched as customer problem answer.Wherein, it is intended that space can be it is preset,
It can also be constantly replenished by on-line study.Alternatively, customer problem can also be encoded using deep learning model
Identify and decode the corresponding answer of output.
Due to obtaining answer using various ways, every kind of acquisition modes obtain answer using different principle, therefore can be with
Lift the rich of answer.
Furthermore, when obtaining the answer of customer problem using single mode, it may appear that the situation of answer can not be obtained,
Such as the answer not matched in knowledge base with customer problem, there is no the sheet to match with customer problem in knowledge mapping
Body.So, answer is obtained by using various ways, can ensures the stability that answer obtains to avoid the above situation.
In order to determine the more accurately answer for customer problem, in step s 103, multiple answers are screened.Sieve
The foundation of choosing is the source parameter and prediction probability of each answer.
Wherein, the source parameter of the answer and prediction probability are determined according to the acquisition modes of the answer.Due to adopting
The accuracy and importance of the answer obtained with different acquisition modes have differences, therefore characterize difference using source parameter
The accuracy and importance of answer, and then when screening answer, accuracy or importance can be gone out more using source choice of parameters
High answer, to improve the accuracy of optimal answer.Come from for example, the accuracy of the answer obtained using database mode is higher than
The answer of knowledge mapping, thus, the source parameter of the answer obtained using database mode are different from answering from knowledge mapping
The source parameter of case.
The prediction probability of answer can characterize the accuracy that answer replies customer problem.Prediction probability is bigger, its
The accuracy of corresponding answer is higher.Thus, the answer of accuracy higher can be filtered out using the prediction probability of answer, to improve
The accuracy of optimal answer.
And then in step S104, export the optimal answer that step S103 is screened.Specifically, can be answered optimal
Case directly output is presented to user, such as is presented to user in a manner of text, voice etc..It can also be carried out based on optimal answer
He operates interacting for realization and user.For example, sequence of operations is performed based on optimal answer.
The embodiment of the present invention obtains multiple answers using at least two modes, and optimal answer is then screened in multiple answers
Output.Since the mode for obtaining answer is different, the rich of answer can be lifted from multiple angles, can be to avoid using single
One mode obtains the situation less than answer, ensures the sustainability interacted with user's question and answer, lifts user experience.In addition, from more
Optimal answer is chosen according to the source parameter and prediction probability of each answer in a answer, it is ensured that answer is asked for user
Inscribe the accuracy replied.
The response method of question and answer interaction in the present embodiment can independently execute, the finger independent of other computer instructions
Show.
In a preferred embodiment of the invention, at least two mode is selected from knowledge base, knowledge mapping and study mould
Type.
In the present embodiment, knowledge base can be utilized to obtain answer.Knowledge base includes problem and answer, by by customer problem
Matched with problem in knowledge base to obtain answer.
The present embodiment can also utilize knowledge mapping to obtain answer.Knowledge mapping is a kind of semantic network, including node and
The side of link node.Node represents entity or concept, while the various semantic relations between representing entity/concept.Specifically, know
The data known in collection of illustrative plates are stored with triple data mode, i.e.,：<Entity A, relation, entity B>, such as：<Liu Dehua, goes out
The dried rhizome of rehmannia, Hong Kong>.If customer problem is：" where the birthplace of Liu Dehua is", answer is obtained as " perfume using knowledge mapping
Port ", its prediction probability are 0.9986.
It should be noted that the network structure of knowledge mapping can also be any enforceable mode in the prior art, this
Inventive embodiments are without limitation.
The present embodiment can also utilize learning model to obtain answer.Learning model can be deep learning model or engineering
Model is practised, such as can be shot and long term memory models (long-short term memory, LSTM).The problem of for input,
Learning model can automatically generate answer according to neutral net.
The prediction probability of answer is determined it is possible to further one or more in the following ways：If the answer
From the knowledge base, then calculate the customer problem and ask and/or extend the semantic similarity asked with the knowledge base Plays,
Using the prediction probability as the answer；If the answer comes from the knowledge mapping, determined according to the knowledge mapping
The confidence level of the answer determine the prediction probability of the answer；If the answer carrys out self learning model, according to
The sum of conditional probability between the adjacent word of answer determines the prediction probability of the answer.
In specific implementation, when obtaining answer using knowledge storehouse matching, customer problem and the knowledge base can be got the bid
The prediction probability that the semantic similarity asked can represent gained answer is asked and/or extended to standard；Answer is being obtained using knowledge mapping
When, knowledge mapping has the marking of confidence level for its definite answer, thus, can be with according to the marking of the confidence level to answer
Determine its prediction probability；When the production mode based on deep learning model obtains answer, the prediction probability of answer can be
The sum of conditional probability in answer between front and rear word.
In a kind of embodiment of the present invention, as shown in Fig. 2, step S102 can include following at least two step
Suddenly：
Step S201：Calculate the customer problem and ask and/or extend the semantic similarity asked with knowledge base Plays, and really
Fixed the first answer from the knowledge base；
Step S202：The customer problem is matched with the knowledge in knowledge mapping, and determines to come from the knowledge
Second answer of collection of illustrative plates；
Step S203：The customer problem is inputted into learning model, and determines that the output of the learning model is answered for the 3rd
Case.
It is that customer problem is asked and/or expanded with the standard in knowledge base when obtaining answer using knowledge base in the present embodiment
Exhibition asks and matched, and the standard that given threshold is reached if there is the semantic similarity with customer problem is asked or extend and asks, then general
The standard, which is asked or extended, asks corresponding answer as the first answer.
It is to be matched customer problem with the triple data in knowledge mapping when obtaining answer using knowledge mapping,
If there is the triple data to match with customer problem, then using the triple data interior joint as the second answer.
It is to input customer problem to the learning model when obtaining answer using learning model, which can be certainly
Dynamic generation is directed to the answer of customer problem, and using this as the 3rd answer.
Further, the source parameter includes priority, it is preferable that the priority of the answer from knowledge base is higher than next
From the priority of the answer of knowledge mapping, the priority of the answer from knowledge mapping is higher than the excellent of the answer that carrys out self learning model
First level.
In the present embodiment, in view of the problem of in knowledge base and answer is pre-configured, utilize what knowledge place obtained
The accuracy of answer is higher.Learning model needs to be trained in advance, and the training effect of learning model is trained used
The influence of language material, the accuracy using the answer of learning model generation are relatively low.The answer got using knowledge mapping it is accurate
Property falls between.Thus, the priority of the corresponding source parameter of above-mentioned three kinds of modes is followed successively by knowledge base, knows from big to small
Know collection of illustrative plates and learning model.
It should be noted that according to the difference of practical application scene, may also be different to the demand of answer, therefore come
The priority of source parameter can also carry out the configuration of adaptability according to concrete application scene, and the embodiment of the present invention does not limit this
System.
In a specific embodiment of the invention, step S103 shown in Fig. 1 may comprise steps of：According to the excellent of answer
First level sequence, judges whether the prediction probability of each answer is more than given threshold successively, and pre- by what is judged first
Survey probability and be more than the answer of the given threshold as optimal answer.
It is preferential to choose the high answer of priority in the present embodiment.That is, first judge the pre- of the higher answer of priority
Survey whether probability is more than given threshold, if it is, using the answer as optimal answer.Otherwise, continue to judge next priority
The prediction probability of answer whether be more than given threshold, until filtering out optimal answer.The quantity of answer have it is multiple, so as to ensure
Optimal answer can be obtained.
As shown in figure 3, in another specific embodiment of the invention, step S103 shown in Fig. 1 may comprise steps of：
Step S301：The importance weight of the answer is determined according to the source parameter of each answer；
Step S302：The accuracy weights of the answer are determined according to the prediction probability of each answer；
Step S303：Utilize the importance weight of each answer and effective fraction of each answer of accuracy weight computing；
Step S304：The multiple answer is screened according to effective fraction of each answer.
Relative to the source parameter that answer is first considered in previous embodiment, the prediction probability of answer is considered further that；The present embodiment
Middle be carried out at the same time the source parameter and prediction probability of answer is considered.
In the step S301 and step S302 of the present embodiment, it can be determined according to the source parameter of answer corresponding important
Property weights, that is to say, that the acquisition modes of answer can possess different importance weights.Furthermore, different acquisition sides
The accuracy for the answer that formula is got and the importance weight positive correlation of the answer.It can be determined pair according to the prediction probability of answer
The accuracy weights answered.The size of accuracy weights and the size positive correlation of prediction probability.
In the step S303 and step S304 of the present embodiment, the importance weight and accuracy weights of each answer are utilized
Effective fraction of answer is calculated.Effective fraction of answer can be with the accuracy of the comprehensive characterization answer.Effective fraction is higher,
The accuracy of answer is higher.Optimal answer is the effectively highest answer of fraction in each answer.
Specifically, effective fraction can be that importance weight and accuracy weights are obtained by any enforceable mathematical operation
Arrive, can be the sum of importance weight and accuracy weights or importance weight and the product of accuracy weights, this hair
Bright embodiment is without limitation.
Further, generation of the source parameter of the answer including answer takes.Specifically, the generation consumption of answer
When represent response speed, it will influence reply promptness to customer problem.The generation of answer take it is longer, represent obtain this answer
Case the time it takes is longer, and the time that user waits is also longer, so as to reduce user experience.Thus, the generation consumption of answer
When can be used for the selection process of optimal answer.
As shown in figure 4, in another specific embodiment of the invention, step S103 shown in Fig. 1 may comprise steps of：
Step S401：The corresponding importance weight of the answer is determined according to the source parameter of each answer；
Step S402：Calculate the importance weight of each answer and the product of prediction probability；
Step S403：The product of each answer business time-consuming with generation is calculated, using effective fraction as each answer；
Step S404：The multiple answer is screened according to effective fraction of each answer.
Considered relative to the source parameter and prediction probability of answer is carried out at the same time in embodiment illustrated in fig. 3.It is of the invention real
Example is applied also using the time-consuming considerations as optimal answer of generation of answer.
In specific implementation, generation is time-consuming negatively correlated with effective fraction.Generation is time-consuming longer, and the validity of answer is lower.
, can be after the product of importance weight and prediction probability be obtained, by this when effective fraction of answer is calculated in step S403
Product takes with generation and is business, to obtain effective fraction.
Step S401, more embodiments of step S402 and step S404, which can refer in embodiment illustrated in fig. 3, walks
Rapid S301, step S302 and step S304.
In a concrete application scene of the invention, the customer problem can be voice.Then step S101 shown in Fig. 1 can
With including：The customer problem is converted into text.That is, it is necessary to after converting voice data into text, then after performing
Continuous step.Why text is converted speech into, asked and/or expanded with standard to calculate customer problem in subsequent step
Open up the semantic similarity asked.
And then step S104 shown in Fig. 1 can include：Use is sent to after the optimal answer of acquisition is converted to voice
Family.In other words, in order to ensure the uniformity with user mutual, when user is interacted using this mode of voice, feed back to
The optimal answer of user also uses voice.Thus, when optimal answer is textual form, optimal answer is converted to after voice again
Export to user.
In another concrete application scene of the invention, described the step of obtaining customer problem, is in response to cut in what is received
Change after configured information what is performed.
In the present embodiment, the response method of question and answer interaction can be used in combination with other interactions.That is, only
When other interactions indicate to perform the response method step of the present embodiment, the response method step of the present embodiment can be just performed
Suddenly.
Further, the indication message be using professional knowledge storehouse to the customer problem carry out it fails to match
When send.It is professional due to question and answer in the question and answer interaction of professional domain is carried out using professional knowledge storehouse, may
There is the situation that can not obtain answer, therefore indicate that switching performs the response method step of the present embodiment by indication message
Suddenly, it can ensure to get answer, to realize the continuity of interaction, lift user experience.For example, using professional knowledge storehouse to institute
State customer problem to carry out when it fails to match, by performing the response method step of the present embodiment, the optimal answer got is " you
It is good ", into the pattern of chat, realize interactive continuity.
Further, the indication message is the intention assessment result in the customer problem and default intent classifier
Sent during successful match.
In the present embodiment, refer to relative to switching is sent when it fails to match to customer problem progress using professional knowledge storehouse
Show information, the embodiment of the present invention divides customer problem before being matched using professional knowledge storehouse to the customer problem
Class, if the intention assessment result of customer problem and default intent classifier successful match, send indication message, with instruction
Switching performs the response method step of the present embodiment.For example, default intent classifier is chat classification, the intention assessment of customer problem
When as a result to chat classification, then indication message is sent, switch the response method step for performing the present embodiment with instruction.
Further, the intention assessment is the result is that ask the user using the intent classifier model that training is completed in advance
Topic carries out what intention assessment obtained.
In specific implementation, the question and answer language material that can advance with accumulation trains to obtain intent classifier model.What training was completed
Intent classifier model can carry out intention assessment to the customer problem.For example user inputs " I am somewhat unhappy ", it is intended that point
Class model can be classified as chatting classification.Then indication message is sent, switches the sound for performing the present embodiment with instruction
Induction method step, pattern is chatted into machine, and exports optimal answer " small i fools you, and say cheese, the people of smile most has
Glamour！”.
It should be noted that intent classifier model can use existing any enforceable sorting algorithm, the present invention is implemented
Example is without limitation.
As shown in figure 5, the responding device 50 of question and answer interaction can include customer problem acquisition module 501, answer obtains mould
Block 502, answer screening module 503 and answer output module 504.
Wherein, customer problem acquisition module 501 is obtaining customer problem；Answer acquisition module 502 is to using at least
Two ways obtains multiple answers for the customer problem；Answer screening module 503 is to the source according to each answer
Parameter and prediction probability screen the multiple answer, and the source parameter and prediction probability of the answer are answered according to this
What the acquisition modes of case determined；Answer output module 504 is exporting the obtained optimal answer of screening.
Due to obtaining answer using various ways, every kind of acquisition modes obtain answer using different principle, therefore can be with
Lift the rich of answer.
Furthermore, when obtaining the answer of customer problem using single mode, it may appear that the situation of answer can not be obtained,
Such as the answer not matched in knowledge base with customer problem, there is no the knot to match with customer problem in knowledge mapping
Point.So, answer is obtained by using various ways, can ensures the stability that answer obtains to avoid the above situation.
The embodiment of the present invention obtains multiple answers using at least two modes, and optimal answer is then screened in multiple answers
Output；Since the mode for obtaining answer is different, the rich of answer can be lifted from multiple angles, can be to avoid using single
One mode obtains the situation less than answer, ensures the sustainability interacted with user's question and answer, lifts user experience.In addition, from more
Optimal answer is chosen according to the source parameter and prediction probability of each answer in a answer, it is ensured that answer is asked for user
Inscribe the accuracy replied.
The responding device 50 of question and answer interaction in the present embodiment can independently execute, independent of other computer instructions
Instruction.
Preferably, at least two mode is selected from knowledge base, knowledge mapping and learning model.
Further, the source parameter includes priority, and the priority of the answer from knowledge base, which is higher than, comes from knowledge
The priority of the answer of collection of illustrative plates, the priority of the answer from knowledge mapping are higher than the priority for the answer for carrying out self learning model.
In a kind of embodiment of the present invention, as shown in fig. 6, the answer acquisition module 502 can include first
Answer acquiring unit 5021, the semantic similarity asked is asked and/or extends to calculate the customer problem with knowledge base Plays,
And determine the first answer from the knowledge base；
Second answer acquiring unit 5022, the customer problem to be matched with the knowledge in knowledge mapping, and
Determine the second answer from the knowledge mapping；
3rd answer acquiring unit 5023, the customer problem is inputted learning model, and determines the study mould
The output of type is the 3rd answer.
It is that customer problem is asked and/or expanded with the standard in knowledge base when obtaining answer using knowledge base in the present embodiment
Exhibition asks and matched, and the standard that given threshold is reached if there is the semantic similarity with customer problem is asked or extend and asks, then general
The standard, which is asked or extended, asks corresponding answer as the first answer.
It is to be matched customer problem with the triple data in knowledge mapping when obtaining answer using knowledge mapping,
If there is the triple data to match with customer problem, then using the triple data interior joint as the second answer.
It is to input customer problem to the learning model when obtaining answer using learning model, which can be certainly
Dynamic generation is directed to the answer of customer problem, and using this as the 3rd answer.
In a specific embodiment of the invention, answer screening module 503 includes the first screening unit (not shown), to
According to the priority sequence of answer, judge whether the prediction probability of each answer is more than given threshold successively, and will first
The prediction probability for judging to obtain is more than the answer of the given threshold as optimal answer.
In another specific embodiment of the invention, as shown in fig. 7, answer screening module 503 can include：
First importance weight determination unit 5031, the important of the answer is determined to the source parameter according to each answer
Property weights；
Accuracy weights determination unit 5032, to determine that the accuracy of the answer is weighed according to the prediction probability of each answer
Value；
First effective score calculating unit 5033, to utilize the importance weight and accuracy weight computing of each answer
Effective fraction of each answer；
Second screening unit 5034, screens the multiple answer to effective fraction according to each answer.
Relative to the source parameter that answer is first considered in previous embodiment, the prediction probability of answer is considered further that；The present embodiment
Middle be carried out at the same time the source parameter and prediction probability of answer is considered.
Further, generation of the source parameter of the answer including answer takes.
As shown in figure 8, in another specific embodiment of the invention, answer screening module 503 can include：
Second importance weight determination unit 5035, to determine that the answer is corresponding according to the source parameter of each answer
Importance weight；
Product computing unit 5036, to calculate the product of the importance weight of each answer and prediction probability；
Second effective score calculating unit 5037, to calculate the product of each answer business time-consuming with generation, using as
Effective fraction of each answer；
Third filtering unit 5038, screens the multiple answer to effective fraction according to each answer.
Considered relative to the source parameter and prediction probability of answer is carried out at the same time in embodiment illustrated in fig. 7.It is of the invention real
Example is applied also using the time-consuming considerations as optimal answer of generation of answer.
In a concrete application scene of the invention, the customer problem is voice；The customer problem acquisition module 501
Including：First voice converting unit (not shown), the customer problem is converted to text, the answer output module
504 include：Second voice converting unit (not shown), to be sent to use after the optimal answer of acquisition is converted to voice
Family.
That is, it is necessary to after converting voice data into text, then follow-up step is performed.Why digitize the speech into
It is to ask and/or extend the semantic similarity asked with standard to calculate customer problem in subsequent step for text.
In another concrete application scene of the invention, the customer problem acquisition module 501 is cut in response to what is received
Performed after changing configured information.
Further, the indication message be using professional knowledge storehouse to the customer problem carry out it fails to match
When send.It is professional due to question and answer in the question and answer interaction of professional domain is carried out using professional knowledge storehouse, may
There is the situation that can not obtain answer, therefore indicate that switching performs the response method step of the present embodiment by indication message
Suddenly, it can ensure to get answer, to realize the continuity of interaction, lift user experience.
For example, using professional knowledge storehouse to the customer problem carry out it fails to match when, by the sound for performing the present embodiment
Induction method step, the optimal answer got are " hello ", into the pattern of chat, realize interactive continuity.
Further, the indication message is the intention assessment result in the customer problem and default intent classifier
Sent during successful match.Relative to using professional knowledge storehouse switching instruction is sent to the customer problem progress when it fails to match
Information, the embodiment of the present invention divide customer problem before being matched using professional knowledge storehouse to the customer problem
Class, if the intention assessment result of customer problem and default intent classifier successful match, send indication message, with instruction
Switching performs the response method step of the present embodiment.
For example, default intent classifier when the intention assessment result of customer problem is chats classification, then sends to chat classification
Indication message, switches the response method step for performing the present embodiment with instruction.
Further, the intention assessment is the result is that ask the user using the intent classifier model that training is completed in advance
Topic carries out what intention assessment obtained.
On operation principle, the more contents of working method of the responding device 50 of question and answer interaction, Fig. 1 is referred to
Associated description into Fig. 4, which is not described herein again.
The embodiment of the invention also discloses a kind of storage medium, is stored thereon with computer instruction, the computer instruction
The step of response method of the question and answer interaction shown in Fig. 1 to Fig. 4 any embodiment can be performed during operation.The storage medium
It can include ROM, RAM, disk or CD etc..The storage medium can also include non-volatility memorizer (non-
) or non-transient (non-transitory) memory etc. volatile.
The embodiment of the invention also discloses a kind of terminal, the terminal can include memory and processor, the storage
The computer instruction that can be run on the processor is stored with device.The processor can be with when running the computer instruction
The step of performing the response method of the question and answer interaction shown in Fig. 1 to Fig. 4 any embodiment.The terminal includes but not limited to hand
The terminal devices such as machine, computer, tablet computer.
Although present disclosure is as above, the present invention is not limited to this.Any those skilled in the art, are not departing from this
In the spirit and scope of invention, it can make various changes or modifications, therefore protection scope of the present invention should be with claim institute
Subject to the scope of restriction.
Claims (30)
- A kind of 1. response method of question and answer interaction, it is characterised in that including：Obtain customer problem；Multiple answers for the customer problem are obtained using at least two modes；The multiple answer is screened according to the source parameter of each answer and prediction probability, the source ginseng of the answer Number and prediction probability are determined according to the acquisition modes of the answer；The optimal answer that output screening obtains.
- 2. the response method of question and answer interaction according to claim 1, it is characterised in that at least two mode, which is selected from, to be known Know storehouse, knowledge mapping and learning model.
- 3. the response method of question and answer interaction according to claim 2, it is characterised in that one kind or more in the following ways The prediction probability of the definite answer of kind：If the answer comes from the knowledge base, calculate the customer problem and ask and/or expand with the knowledge base Plays The semantic similarity asked is opened up, using the prediction probability as the answer；If the answer comes from the knowledge mapping, the confidence level of the answer determined according to the knowledge mapping determines The prediction probability of the answer；If the answer carrys out self learning model, according to determining the sum of conditional probability between the adjacent word of the answer The prediction probability of answer.
- 4. the response method of question and answer interaction according to claim 2, it is characterised in that the source parameter includes preferential Level, the priority of the answer from knowledge base are higher than the priority of the answer from knowledge mapping, the answer from knowledge mapping Priority be higher than the priority of the answer for carrying out self learning model.
- 5. the response method of question and answer interaction according to claim 4, it is characterised in that the source according to each answer Parameter and prediction probability carry out screening to the multiple answer to be included：According to the priority sequence of answer, judge whether the prediction probability of each answer is more than given threshold successively, and will The prediction probability judged first is more than the answer of the given threshold as optimal answer.
- 6. the response method of question and answer interaction according to claim 2, it is characterised in that described to be obtained using at least two modes To multiple answers for the customer problem the multiple answer is obtained using following any two or three of mode：Calculate the customer problem and ask and/or extend the semantic similarity asked with knowledge base Plays, and determine to know from described Know first answer in storehouse；The customer problem is matched with the knowledge in knowledge mapping, and determines second answering from the knowledge mapping Case；The customer problem is inputted into learning model, and the output of the definite learning model is the 3rd answer.
- 7. the response method of question and answer interaction according to claim 1, it is characterised in that the source according to each answer Parameter and prediction probability carry out screening to the multiple answer to be included：The importance weight of the answer is determined according to the source parameter of each answer；The accuracy weights of the answer are determined according to the prediction probability of each answer；Utilize the importance weight of each answer and effective fraction of each answer of accuracy weight computing；The multiple answer is screened according to effective fraction of each answer.
- 8. the response method of question and answer interaction according to claim 1, it is characterised in that the source parameter of the answer includes The generation of answer takes.
- 9. the response method of question and answer interaction according to claim 8, it is characterised in that the source according to each answer Parameter and prediction probability carry out screening to the multiple answer to be included：The corresponding importance weight of the answer is determined according to the source parameter of each answer；Calculate the importance weight of each answer and the product of prediction probability；The product of each answer business time-consuming with generation is calculated, using effective fraction as each answer；The multiple answer is screened according to effective fraction of each answer.
- 10. the response method of question and answer interaction according to claim 1, it is characterised in that the customer problem is voice；Institute Stating acquisition customer problem includes：The customer problem is converted into text, the optimal answer that the output screening obtains includes：By the described optimal of acquisition Answer is sent to user after being converted to voice.
- 11. the response method of question and answer interaction according to claim 1, it is characterised in that the step for obtaining customer problem Suddenly it is in response to what is performed after the indication message received.
- 12. the response method of question and answer according to claim 11 interaction, it is characterised in that the indication message be Carry out what is sent when it fails to match to the customer problem using professional knowledge storehouse.
- 13. the response method of question and answer according to claim 11 interaction, it is characterised in that the indication message be Sent when the intention assessment result of the customer problem and default intent classifier successful match.
- 14. the response method of question and answer interaction according to claim 13, it is characterised in that the intention assessment is the result is that profit The intent classifier model completed with advance training carries out what intention assessment obtained to the customer problem.
- A kind of 15. responding device of question and answer interaction, it is characterised in that including：Customer problem acquisition module, to obtain customer problem；Answer acquisition module, to obtain multiple answers for the customer problem using at least two modes；Answer screening module, sieves the multiple answer to the source parameter according to each answer and prediction probability Choosing, the source parameter and prediction probability of the answer are determined according to the acquisition modes of the answer；Answer output module, to export the optimal answer that screening obtains.
- 16. the responding device of question and answer interaction according to claim 15, it is characterised in that at least two mode is selected from Knowledge base, knowledge mapping and learning model.
- 17. the responding device of question and answer interaction according to claim 16, it is characterised in that the answer screening module uses The one or more of in the following manner determine the prediction probability of answer：If the answer comes from the knowledge base, calculate the customer problem and ask and/or expand with the knowledge base Plays The semantic similarity asked is opened up, using the prediction probability as the answer；If the answer comes from the knowledge mapping, the confidence level of the answer determined according to the knowledge mapping determines The prediction probability of the answer；If the answer carrys out self learning model, according to determining the sum of conditional probability between the adjacent word of the answer The prediction probability of answer.
- 18. the responding device of question and answer interaction according to claim 16, it is characterised in that the source parameter includes preferential Level, the priority of the answer from knowledge base are higher than the priority of the answer from knowledge mapping, the answer from knowledge mapping Priority be higher than the priority of the answer for carrying out self learning model.
- 19. the responding device of question and answer interaction according to claim 18, it is characterised in that the answer screening module bag Include：First screening unit, whether according to the priority sequence of answer, to judge the prediction probability of each answer successively It is more than the answer of the given threshold as optimal answer more than given threshold, and using the prediction probability judged first.
- 20. the responding device of question and answer interaction according to claim 16, it is characterised in that the answer acquisition module bag Include：First answer acquiring unit, the semantic phase asked is asked and/or extends to calculate the customer problem with knowledge base Plays Like degree, and determine the first answer from the knowledge base；Second answer acquiring unit, the customer problem to be matched with the knowledge in knowledge mapping, and determines to come from Second answer of the knowledge mapping；3rd answer acquiring unit, the customer problem is inputted learning model, and determines the output of the learning model For the 3rd answer.
- 21. the responding device of question and answer interaction according to claim 15, it is characterised in that the answer screening module bag Include：First importance weight determination unit, to determine the importance weight of the answer according to the source parameter of each answer；Accuracy weights determination unit, to determine the accuracy weights of the answer according to the prediction probability of each answer；First effective score calculating unit, to utilize the importance weight of each answer and each answer of accuracy weight computing Effective fraction；Second screening unit, screens the multiple answer to effective fraction according to each answer.
- 22. the responding device of question and answer interaction according to claim 15, it is characterised in that the source parameter bag of the answer The generation for including answer takes.
- 23. the responding device of question and answer interaction according to claim 22, it is characterised in that the answer screening module bag Include：Second importance weight determination unit, to determine that the corresponding importance of the answer is weighed according to the source parameter of each answer Value；Product computing unit, to calculate the product of the importance weight of each answer and prediction probability；Second effective score calculating unit, to calculate the product of each answer business time-consuming with generation, to be used as each answer Effective fraction；Third filtering unit, screens the multiple answer to effective fraction according to each answer.
- 24. the responding device of question and answer interaction according to claim 15, it is characterised in that the customer problem is voice； The customer problem acquisition module includes：First voice converting unit, the customer problem is converted to text, the answer output module includes：Second language Sound converting unit, to be sent to user after the optimal answer of acquisition is converted to voice.
- 25. the responding device of question and answer interaction according to claim 15, it is characterised in that the customer problem acquisition module In response to being performed after the indication message that receives.
- 26. the responding device of question and answer according to claim 25 interaction, it is characterised in that the indication message be Carry out what is sent when it fails to match to the customer problem using professional knowledge storehouse.
- 27. the responding device of question and answer according to claim 25 interaction, it is characterised in that the indication message be Sent when the intention assessment result of the customer problem and default intent classifier successful match.
- 28. the responding device of question and answer interaction according to claim 27, it is characterised in that the intention assessment is the result is that profit The intent classifier model completed with advance training carries out what intention assessment obtained to the customer problem.
- 29. a kind of storage medium, is stored thereon with computer instruction, it is characterised in that is performed during the computer instruction operation The step of response method that question and answer any one of claim 1 to 14 interact.
- 30. a kind of terminal, including memory and processor, the meter that can be run on the processor is stored with the memory Calculation machine instructs, it is characterised in that perform claim requires any one of 1 to 14 institute when the processor runs the computer instruction The step of stating the response method of question and answer interaction.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711434955.8A CN107908803B (en) | 2017-12-26 | 2017-12-26 | Question-answer interaction response method and device, storage medium and terminal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711434955.8A CN107908803B (en) | 2017-12-26 | 2017-12-26 | Question-answer interaction response method and device, storage medium and terminal |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107908803A true CN107908803A (en) | 2018-04-13 |
| CN107908803B CN107908803B (en) | 2020-10-27 |
Family
ID=61871407
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711434955.8A Active CN107908803B (en) | 2017-12-26 | 2017-12-26 | Question-answer interaction response method and device, storage medium and terminal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107908803B (en) |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108804627A (en) * | 2018-05-31 | 2018-11-13 | 科大讯飞股份有限公司 | Information acquisition method and device |
| CN108920530A (en) * | 2018-06-08 | 2018-11-30 | 泰康保险集团股份有限公司 | A kind of information processing method, device, storage medium and electronic equipment |
| CN108920539A (en) * | 2018-06-12 | 2018-11-30 | 广东小天才科技有限公司 | A kind of method and private tutor's machine for searching for problem answers |
| CN108932289A (en) * | 2018-05-23 | 2018-12-04 | 北京华建蓝海科技有限责任公司 | One kind being based on the problem of information extraction and deep learning answer treatment method and system |
| CN109033223A (en) * | 2018-06-29 | 2018-12-18 | 北京百度网讯科技有限公司 | For method, apparatus, equipment and computer readable storage medium across type session |
| CN109033277A (en) * | 2018-07-10 | 2018-12-18 | 广州极天信息技术股份有限公司 | Class brain system, method, equipment and storage medium based on machine learning |
| CN109213847A (en) * | 2018-09-14 | 2019-01-15 | 广州神马移动信息科技有限公司 | Layered approach and its device, electronic equipment, the computer-readable medium of answer |
| CN109460462A (en) * | 2018-11-15 | 2019-03-12 | 中通天鸿（北京）通信科技股份有限公司 | A kind of Chinese Similar Problems generation System and method for |
| CN109710818A (en) * | 2018-12-28 | 2019-05-03 | 咪咕文化科技有限公司 | The determination method of answer weight, answer determine method, apparatus and storage medium |
| CN109739969A (en) * | 2018-12-29 | 2019-05-10 | 联想(北京)有限公司 | Answer generation method and intelligent conversational system |
| CN109815321A (en) * | 2018-12-26 | 2019-05-28 | 出门问问信息科技有限公司 | Question answering method, device, equipment and storage medium |
| CN110019742A (en) * | 2018-06-19 | 2019-07-16 | 北京京东尚科信息技术有限公司 | Method and apparatus for handling information |
| CN110390003A (en) * | 2019-06-19 | 2019-10-29 | 北京百度网讯科技有限公司 | Question and answer processing method and system, computer equipment and readable medium based on medical treatment |
| CN110619042A (en) * | 2019-03-13 | 2019-12-27 | 北京航空航天大学 | Neural network-based teaching question and answer system and method |
| WO2020006835A1 (en) * | 2018-07-03 | 2020-01-09 | 平安科技（深圳）有限公司 | Customer service method, apparatus, and device for engaging in multiple rounds of question and answer, and storage medium |
| CN111460116A (en) * | 2020-03-19 | 2020-07-28 | 北京百度网讯科技有限公司 | Question answering method, question answering system, electronic equipment and storage medium |
| CN111611354A (en) * | 2019-02-26 | 2020-09-01 | 北京嘀嘀无限科技发展有限公司 | Man-machine conversation control method, device, server and readable storage medium |
| CN111767374A (en) * | 2019-03-29 | 2020-10-13 | 北京搜狗科技发展有限公司 | Data processing method, device and machine readable medium |
| CN111949783A (en) * | 2020-08-12 | 2020-11-17 | 中国银行股份有限公司 | Question and answer result generation method and device in knowledge base |
| CN112037905A (en) * | 2020-07-16 | 2020-12-04 | 朱卫国 | Medical question answering method, equipment and storage medium |
| CN112309587A (en) * | 2020-11-26 | 2021-02-02 | 微医云(杭州)控股有限公司 | On-line inquiry method, system, server and storage medium |
| CN112632255A (en) * | 2020-12-29 | 2021-04-09 | 海信视像科技股份有限公司 | Method and device for obtaining question and answer results |
| WO2021169640A1 (en) * | 2020-02-25 | 2021-09-02 | 京东方科技集团股份有限公司 | Query device and method, apparatus, and storage medium |
| CN113836287A (en) * | 2021-10-11 | 2021-12-24 | 深圳追一科技有限公司 | Method and device for determining question and answer result, computer equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101097573A (en) * | 2006-06-28 | 2008-01-02 | 腾讯科技（深圳）有限公司 | Automatically request-answering system and method |
| CN101431573A (en) * | 2007-11-08 | 2009-05-13 | 上海赢思软件技术有限公司 | Method and equipment for implementing automatic customer service through human-machine interaction technology |
| CN101520802A (en) * | 2009-04-13 | 2009-09-02 | 腾讯科技（深圳）有限公司 | Question-answer pair quality evaluation method and system |
| CN102110128A (en) * | 2009-12-28 | 2011-06-29 | 日电（中国）有限公司 | Traffic information conversion method and device based on traffic information element knowledge base |
-
2017
- 2017-12-26 CN CN201711434955.8A patent/CN107908803B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101097573A (en) * | 2006-06-28 | 2008-01-02 | 腾讯科技（深圳）有限公司 | Automatically request-answering system and method |
| CN101431573A (en) * | 2007-11-08 | 2009-05-13 | 上海赢思软件技术有限公司 | Method and equipment for implementing automatic customer service through human-machine interaction technology |
| CN101520802A (en) * | 2009-04-13 | 2009-09-02 | 腾讯科技（深圳）有限公司 | Question-answer pair quality evaluation method and system |
| CN102110128A (en) * | 2009-12-28 | 2011-06-29 | 日电（中国）有限公司 | Traffic information conversion method and device based on traffic information element knowledge base |
Non-Patent Citations (2)
| Title |
|---|
| 曾帅 等: "《面向知识自动化的自动问答研究进展》", 《自动化学报》 * |
| 钱宏泽: "《基于中草药语义网的自动问答系统的研究与实现》", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108932289A (en) * | 2018-05-23 | 2018-12-04 | 北京华建蓝海科技有限责任公司 | One kind being based on the problem of information extraction and deep learning answer treatment method and system |
| CN108804627A (en) * | 2018-05-31 | 2018-11-13 | 科大讯飞股份有限公司 | Information acquisition method and device |
| CN108804627B (en) * | 2018-05-31 | 2021-04-06 | 科大讯飞股份有限公司 | Information acquisition method and device |
| CN108920530A (en) * | 2018-06-08 | 2018-11-30 | 泰康保险集团股份有限公司 | A kind of information processing method, device, storage medium and electronic equipment |
| CN108920530B (en) * | 2018-06-08 | 2022-01-28 | 泰康保险集团股份有限公司 | Information processing method and device, storage medium and electronic equipment |
| CN108920539A (en) * | 2018-06-12 | 2018-11-30 | 广东小天才科技有限公司 | A kind of method and private tutor's machine for searching for problem answers |
| CN110019742A (en) * | 2018-06-19 | 2019-07-16 | 北京京东尚科信息技术有限公司 | Method and apparatus for handling information |
| CN109033223A (en) * | 2018-06-29 | 2018-12-18 | 北京百度网讯科技有限公司 | For method, apparatus, equipment and computer readable storage medium across type session |
| WO2020006835A1 (en) * | 2018-07-03 | 2020-01-09 | 平安科技（深圳）有限公司 | Customer service method, apparatus, and device for engaging in multiple rounds of question and answer, and storage medium |
| CN109033277A (en) * | 2018-07-10 | 2018-12-18 | 广州极天信息技术股份有限公司 | Class brain system, method, equipment and storage medium based on machine learning |
| CN109213847A (en) * | 2018-09-14 | 2019-01-15 | 广州神马移动信息科技有限公司 | Layered approach and its device, electronic equipment, the computer-readable medium of answer |
| CN109460462A (en) * | 2018-11-15 | 2019-03-12 | 中通天鸿（北京）通信科技股份有限公司 | A kind of Chinese Similar Problems generation System and method for |
| CN109460462B (en) * | 2018-11-15 | 2021-10-19 | 中通天鸿（北京）通信科技股份有限公司 | Chinese similarity problem generation system and method |
| CN109815321B (en) * | 2018-12-26 | 2020-12-11 | 出门问问信息科技有限公司 | Question answering method, device, equipment and storage medium |
| CN109815321A (en) * | 2018-12-26 | 2019-05-28 | 出门问问信息科技有限公司 | Question answering method, device, equipment and storage medium |
| CN109710818A (en) * | 2018-12-28 | 2019-05-03 | 咪咕文化科技有限公司 | The determination method of answer weight, answer determine method, apparatus and storage medium |
| CN109739969A (en) * | 2018-12-29 | 2019-05-10 | 联想(北京)有限公司 | Answer generation method and intelligent conversational system |
| CN111611354A (en) * | 2019-02-26 | 2020-09-01 | 北京嘀嘀无限科技发展有限公司 | Man-machine conversation control method, device, server and readable storage medium |
| CN111611354B (en) * | 2019-02-26 | 2023-09-29 | 北京嘀嘀无限科技发展有限公司 | Man-machine conversation control method and device, server and readable storage medium |
| CN110619042A (en) * | 2019-03-13 | 2019-12-27 | 北京航空航天大学 | Neural network-based teaching question and answer system and method |
| CN111767374A (en) * | 2019-03-29 | 2020-10-13 | 北京搜狗科技发展有限公司 | Data processing method, device and machine readable medium |
| CN110390003A (en) * | 2019-06-19 | 2019-10-29 | 北京百度网讯科技有限公司 | Question and answer processing method and system, computer equipment and readable medium based on medical treatment |
| WO2021169640A1 (en) * | 2020-02-25 | 2021-09-02 | 京东方科技集团股份有限公司 | Query device and method, apparatus, and storage medium |
| CN111460116A (en) * | 2020-03-19 | 2020-07-28 | 北京百度网讯科技有限公司 | Question answering method, question answering system, electronic equipment and storage medium |
| CN112037905A (en) * | 2020-07-16 | 2020-12-04 | 朱卫国 | Medical question answering method, equipment and storage medium |
| CN111949783A (en) * | 2020-08-12 | 2020-11-17 | 中国银行股份有限公司 | Question and answer result generation method and device in knowledge base |
| CN112309587A (en) * | 2020-11-26 | 2021-02-02 | 微医云(杭州)控股有限公司 | On-line inquiry method, system, server and storage medium |
| CN112632255A (en) * | 2020-12-29 | 2021-04-09 | 海信视像科技股份有限公司 | Method and device for obtaining question and answer results |
| CN113836287A (en) * | 2021-10-11 | 2021-12-24 | 深圳追一科技有限公司 | Method and device for determining question and answer result, computer equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107908803B (en) | 2020-10-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107908803A (en) | The response method and device, storage medium, terminal of question and answer interaction | |
| CN110019729A (en) | Intelligent answer method and storage medium, terminal | |
| CN110019838A (en) | Intelligent Answer System and intelligent terminal | |
| CN110019728A (en) | Automatic interaction method and storage medium, terminal | |
| CN104778173B (en) | Target user determination method, device and equipment | |
| CN108052979A (en) | The method, apparatus and equipment merged to model predication value | |
| CN108833382A (en) | Network information detection method, device, storage medium and computer equipment | |
| CN104915354A (en) | Multimedia file pushing method and device | |
| CN108920510A (en) | Automatic chatting method, device and electronic equipment | |
| WO2017143773A1 (en) | Crowdsourcing learning method and device | |
| CN109960814B (en) | Model parameter searching method and device | |
| CN110223358A (en) | Visible pattern design method, training method, device, system and storage medium | |
| CN108805611A (en) | Advertisement screening technique and device | |
| CN110445939B (en) | Capacity resource prediction method and device | |
| CN110335596A (en) | Products Show method, apparatus, equipment and storage medium based on speech recognition | |
| CN109242522A (en) | The foundation of target user's identification model, target user's recognition methods and device | |
| CN107203883A (en) | A kind of risk control method and equipment | |
| CN115114421A (en) | Question-answer model training method | |
| EP4273750A1 (en) | Data processing method and apparatus, computing device, and test simplification device | |
| CN106897282A (en) | The sorting technique and equipment of a kind of customer group | |
| CN116013228A (en) | Music generation method and device, electronic equipment and storage medium thereof | |
| CN110019730A (en) | Automatic interaction system and intelligent terminal | |
| CN107976919B (en) | A kind of Study of Intelligent Robot Control method, system and electronic equipment | |
| CN115543090A (en) | Topic transfer method, topic transfer device, electronic equipment and storage medium | |
| CN108428234A (en) | Interactive Segmentation performance optimization method based on image segmentation result evaluation |
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 |