CN110309273A - Answering method and device - Google Patents
Answering method and device Download PDFInfo
- Publication number
- CN110309273A CN110309273A CN201810195536.1A CN201810195536A CN110309273A CN 110309273 A CN110309273 A CN 110309273A CN 201810195536 A CN201810195536 A CN 201810195536A CN 110309273 A CN110309273 A CN 110309273A
- Authority
- CN
- China
- Prior art keywords
- event
- classification
- category
- transition probability
- probability value
- 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.)
- Pending
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
- 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
-
- 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
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
Abstract
The invention discloses a kind of answering method and devices, comprising: the event category for the event that described problem includes is obtained in the problem of inputting from user;Based on the event correlation map being previously obtained, determine at least one object event classification of the event category, wherein, it include the transition probability between multiple event categories and multiple event categories in the event correlation map, the object event classification is that developing stage is located at the event category after the developing stage of the current event classification in the event correlation map, and the current event classification meets preset condition to the transition probability value of the object event classification;Export the corresponding object event data of at least one described object event classification obtained.The present invention is based on the transition probabilities between different event classification, and different types of events can get up, and flexibility greatly improves, and the present invention improves the accuracy of event data output to a certain extent, reduces the operation of user.
Description
Technical field
The present invention relates to technical field of information processing more particularly to a kind of enquirements and answer realized based on question answering system
Answering method and device.
Background technique
Question answering system (Question Answering System, abbreviation QA) is a kind of advanced shape of information retrieval system
Formula, it can answer the problem of user is proposed with natural language with accurate, succinct natural language.
Based on the information being stored in question answering system, to after question answering system input problem, question answering system generally uses user
The following two kinds method obtains and exports corresponding result.
The first is analyzed the problem of input using data statistical approach to user, so that output is related to the problem
The higher problem of other search rates;Second is the method for utilizing knowledge mapping, in advance by some passes in question answering system
Connection information combines, and thus exports the problem of inputting with user and belongs to other related informations of same combination, for example user is defeated
The problem of entering is " Yao Ming ", and question answering system output belongs to other related informations of same combination with " Yao Ming ", such as the work of Yao Ming
Unit information, Yao Ming height and weight information, kinsfolk's information of Yao Ming, friend's information of Yao Ming etc..
Obviously, first method is confined to the feedback that user searches for habit, and second method is confined to preset association letter
Combination is ceased, the application limitation of the two is all larger.
Summary of the invention
In view of the above problems, it proposes on the present invention overcomes the above problem or at least be partially solved in order to provide one kind
State the answering method and device of problem.Technical solution is as follows:
An aspect of of the present present invention provides a kind of answering method, comprising:
The event category for the event that described problem includes is obtained in the problem of inputting from user;
Based on the event correlation map being previously obtained, at least one object event classification of the event category is determined,
In, it include the transition probability between multiple event categories and multiple event categories, the target in the event correlation map
Event category is that developing stage is located at the event after the developing stage of the current event classification in the event correlation map
Classification, and the current event classification meets preset condition to the transition probability value of the object event classification;
Export the corresponding object event data of at least one described object event classification obtained.
Optionally, the event correlation map obtains in the following manner:
Text is inputted for each: being obtained in the input text using preparatory trained Event Distillation model and is included
The event category of multiple events;
Thing according to the time of origin of each event by the early sequence to evening, to the multiple events for including in the input text
Part classification is ranked up;
In every two event after sequence has been calculated into, the event category of time of origin morning is converted into the thing in time of origin evening
The transition probability value of part classification;
Transition probability value between each event category obtained according to the input text for the first quantity, is not worked together
Transition probability value between part classification.
Optionally, the method also includes:
The transition probability value between the set storage different event classification is arranged using the probability of happening.
Optionally, the Event Distillation model obtains as follows:
Using the input text of the second quantity as training sample, the input text as the training sample is marked busy
Part classification;
Using the training sample as the input of Event Distillation model, Event Distillation model is instructed using preset algorithm
Practice, obtains trained Event Distillation model;
Wherein, the event category for the event that the described problem obtained in the problem of inputting from user includes, passes through the instruction
The Event Distillation model realization perfected.
Optionally, the preset algorithm includes: support vector machines algorithm or neural network algorithm.
Optionally, after the corresponding object event data of at least one described object event classification that the output obtains,
The method also includes:
All object event data based on acquisition are determined according to the instruction of user using first object event data as use
The new problem of family input;
The event category for the event that the first object event data includes is obtained from the first object event data,
And using the event category as current event classification, returns and execute the step of determining at least one object event classification.
Optionally, the current event classification meets preset condition packet to the transition probability value of the object event classification
Include: the transition probability value of current event classification to the object event classification is maximum, or, current event classification is to the target thing
The transition probability value of part classification is greater than preset threshold.
Another aspect of the present invention provides a kind of question and answer system, comprising:
Event category acquiring unit, the problem of for being inputted from user in obtain the event class of the event that described problem includes
Not;
Object event classification acquiring unit, for determining the event category based on the event correlation map being previously obtained
At least one object event classification, wherein in the event correlation map include multiple event categories and multiple event classes
Transition probability between not, the object event classification are that developing stage is located at the current event in the event correlation map
Event category after the developing stage of classification, and the current event classification is to the transition probability value of the object event classification
Meet preset condition;
Output unit, for exporting the corresponding object event data of at least one object event classification described in acquisition.
Another aspect of the invention provides a kind of storage medium, is stored thereon with program, and described program is executed by processor
The previously described answering method of Shi Shixian.
Another aspect of the invention provides a kind of processor, and the processor is for running program, wherein described program fortune
Previously described answering method is executed when row.
By above-mentioned technical proposal, in answering method and device provided by the invention, obtained from the problem of user's input
The event category for the event that problem includes;Based on the event correlation map being previously obtained, at least the one of the event category is determined
A object event classification, wherein include between multiple event categories and multiple event categories in the event correlation map
Transition probability, the object event classification are the hair that developing stage is located at the current event classification in the event correlation map
Event category after the exhibition stage, and the current event classification is default to the transition probability value of object event classification satisfaction
Condition;And then export the corresponding object event data of at least one described object event classification obtained.The present invention is based on differences
Transition probability between event category can get up different types of events, compared to the prior art in data statistics
The method of method and knowledge mapping, flexibility greatly improve.And the transition probability between different event classification is embodied and is not worked together
The probability of front and back, the object event number exported hereby based on the transition probability between different event classification to user occur between part
According to wanting that the probability of most possible event data understood greatly increases for user, it is defeated to improve event data to a certain extent
Accuracy out reduces the operation of user.
The above description is only an overview of the technical scheme of the present invention, in order to better understand the technical means of the present invention,
And it can be implemented in accordance with the contents of the specification, and in order to allow above and other objects of the present invention, feature and advantage can
It is clearer and more comprehensible, the followings are specific embodiments of the present invention.
Detailed description of the invention
By reading the following detailed description of the preferred embodiment, various other advantages and benefits are common for this field
Technical staff will become clear.The drawings are only for the purpose of illustrating a preferred embodiment, and is not considered as to the present invention
Limitation.And throughout the drawings, the same reference numbers will be used to refer to the same parts.In the accompanying drawings:
Fig. 1 shows a kind of flow chart of answering method provided in an embodiment of the present invention;
Fig. 2 shows the flow charts of the training method of Event Distillation model in the embodiment of the present invention;
Fig. 3 shows the flow chart of the method for the transition probability value between obtaining different event classification in the embodiment of the present invention;
Fig. 4 shows the flow chart of another answering method provided in an embodiment of the present invention;
Fig. 5 shows the structural schematic diagram of question and answer system provided in an embodiment of the present invention.
Specific embodiment
Exemplary embodiments of the present disclosure are described in more detail below with reference to accompanying drawings.Although showing the disclosure in attached drawing
Exemplary embodiment, it being understood, however, that may be realized in various forms the disclosure without should be by embodiments set forth here
It is limited.On the contrary, these embodiments are provided to facilitate a more thoroughly understanding of the present invention, and can be by the scope of the present disclosure
It is fully disclosed to those skilled in the art.
Referring to Fig. 1, it illustrates a kind of flow chart of answering method provided in an embodiment of the present invention, method may include
Following steps:
Step 101, the event category for the event that described problem includes is obtained in the problem of inputting from user.
In practical application, user input the problem of may be a short sentence, a long sentence or a text.It is understood that
Ground, if user is a short sentence the problem of input, the general short sentence only includes a specific event, and thus the present invention obtains
The event category of the event.And if user is a text the problem of input, it can be recorded very in text under normal circumstances
Multiple events, therefore the present invention can be from the event category for obtaining all events for including in the text in the text that user inputs.
For ease of description, the present invention is illustrated for being a short sentence the problem of user inputs.
In the solution of the present invention, user input the problem of be not limited to user propose the problem of or give expression to doubt
It asks, and can further contemplate that one or more events of order and user's narration that the request of user's sending, user provide
Deng as long as can therefrom recognize that user currently pays close attention to or exists the place felt uncertain or can therefrom extract outgoing event.Example
Such as, it is assumed that the short sentence of user's input is " please search neighbouring restaurant ", the then thing for the event that the present invention is got from the short sentence
Part classification is " searching restaurant ".
It should be noted that event category is a kind of event description being manually named in advance to certain sentences, specifically
Explain the content for seeing and hereinafter recording about Event Distillation model part, details are not described herein by inventor.
Specifically, the realization of the event category for the event that described problem includes is obtained in the problem of present invention is inputted from user
Method can be with are as follows: the problem of inputting user is input in preparatory trained Event Distillation model, utilizes the Event Distillation
Model extraction goes out the event category for the event for including in described problem, and then obtains the event class of the Event Distillation model output
Not.In the following, as shown in connection with fig. 2, the Event Distillation model in the present invention can be specifically using shown in following steps 201- step 202
Method be trained to obtain.
Step 201, the input text quilt using the input text of the second quantity as training sample, as the training sample
It is labeled with event category.
Wherein the specific value of the second quantity can be configured according to actual needs, and which is not limited by the present invention.
It specifically, first can be by way of handmarking, to each in the input text of the second quantity in the present invention
Respectively related event information extraction comes out input text, is labeled respectively to the event category of each event, preferably also
It can be labeled for the element in each event.Wherein event information may include Time To Event, event title, behavior
Main body, object of action and the peculiar categorical data of certain events etc., the corresponding all event information conducts extracted of an event
One event data is stored.
In order to enable Event Distillation model can relatively easily identify each event, optionally, the present invention is for thing
When part is labeled, machine recognizable dynamic guest's syntactic structure can be used to be labeled for each event.Wherein, of the invention
When being labeled to the element in each event and each event, for content is consistent but original text state it is inconsistent more
A event marks identical label, to realize the standardization of event category.For example event A and event B all relate to commodity
Using case of encroachment of right, but the statement of the original text of event A and event B is inconsistent, but the present invention still can mark " quotient for event A and event B
Product use case of encroachment of right " label.
Step 202, using the training sample as the input of Event Distillation model, using preset algorithm to Event Distillation mould
Type is trained, and obtains trained Event Distillation model.
Wherein preset algorithm can be SVM (Support Vector Machine, support vector machines) algorithm or nerve net
Network algorithm etc..
By the way that these are input in Event Distillation model by the training sample manually marked, it will be able to Event Distillation
Model is trained.After having input a certain number of samples, it may be considered that terminating to the training of Event Distillation model.It
Afterwards, when inputting the text without mark into trained Event Distillation model, Event Distillation model can therefrom be identified simultaneously
Extract the event category of each event.
It should be noted that obtaining trained Event Distillation model in the present invention, especially obtain for the first time trained
After Event Distillation model, in order to guarantee the accuracy of Event Distillation model, to guarantee number when subsequent applications Event Distillation model
According to accuracy, a collection of test sample (such as test bill of complaint, test billof defence) can be chosen, by the test specimens one's duty of the selection
It is not input to Event Distillation model, Event Distillation model accuracy is tested and is finely tuned to realize.
Step 102, based on the event correlation map being previously obtained, at least one object event of the event category is determined
Classification, wherein include the transition probability between multiple event categories and multiple event categories, institute in the event correlation map
Stating object event classification is that developing stage is located at after the developing stage of the current event classification in the event correlation map
Event category, and the current event classification meets preset condition to the transition probability value of the object event classification.
Transition probability between different event classification illustrates the probability of happening of next event after event generation, such as
The transition probability value of event A and event B is 80%, and the transition probability value of event A and event C are 60%, after illustrating that event A occurs
The probability that event B can occur is 80%, and the probability that event C can occur is 60%.
Wherein preset condition maximum, transition probability value can be greater than preset threshold etc. for transition probability value.Preset threshold example
For example 75%, value size can flexibly be set according to actual needs, and which is not limited by the present invention.
That is, the problem of present invention can be based on user's input (object of query, request inquiry including user,
The content etc. of request), it is thus understood that user's there is a strong possibility in next step information that property will obtain, thus intelligently by these
Information is communicated to user.
The present invention is based on the transition probabilities between different event classification, and different types of events can get up, and compares
In the method for data statistical approach in the prior art and knowledge mapping, it is no longer limited to feedback or association that user searches for habit
The combination of information, flexibility greatly improve.And the transition probability value between different event classification embody different event between occur
The probability of front and back, hereby based on the object event data that the transition probability value between different event classification is exported to user, for
Family wants that the probability of the most possible event data understood greatly increases, and improves the accurate of event data output to a certain extent
Property, reduce the operation of user.
In the following, as shown in connection with fig. 3, the present invention obtains the implementation method of the transition probability value between different event classification, can have
Body is using method shown in following steps 301- step 304.
For the accuracy of the transition probability value between the different event classification that guarantees, the present invention preferably selects first
The input text of quantity, after obtaining the transition probability value between the different event classification obtained for each input text,
In conjunction with for the transition probability value between each obtained different event classification of input text, finally obtain an accuracy compared with
Transition probability value between high different event classification.
Wherein, the specific value of the first quantity can be configured according to actual needs, and which is not limited by the present invention.
Step 301, text is inputted for each: obtaining the input text using preparatory trained Event Distillation model
In include multiple events event category.
Text is inputted for each, the trained event that step 201 to step 202 above obtains is input to and mentions
In modulus type, using the Event Distillation model extraction and the event categories of the multiple events for including in the input text is exported.
Step 302, multiple to include in the input text according to the time of origin of each event by the early sequence to evening
The event category of event is ranked up.
Get input text in include all events event category after, according to the time of origin of each event by
The early sequence to evening, is ranked up the event category for the multiple events for including in the input text.
Step 303, in the every two event after sequence has been calculated into, when the event category of time of origin morning is converted into generation
Between evening event category transition probability value.
By step 303, the conversion that the present invention has obtained between all event categories for including in each input text is general
Rate value.
Step 304, it according to the transition probability value between the obtained each event category of input text for the first quantity, obtains
To the transition probability value between different event classification.
It is all by what is obtained after transition probability value between all event categories for including in obtaining each input text
Transition probability value between all event categories for including in input text is taken into consideration, recalculates, finally obtaining one has
The transition probability value that can indicate the successive probability of happening of different event of high accuracy.
Preferably as the present invention, the present invention can use after obtaining the transition probability value between different event classification
Probability of happening arrangement set stores the transition probability value between the different event classification.It as a result, in practical application, can be straight
It connects according to the transition probability value between the different event classification stored in probability of happening arrangement set, to obtain object event class
Not.And updated if being related to data, for example (such as event A- event B) transition probability value changes between event, event can be directly based upon
The data stored in probabilistic set are updated, convenient and efficient.
Step 103, the corresponding object event data of at least one described object event classification of acquisition are exported.
Wherein, object event data may include the event category and/or event argument of object event.The present invention is obtaining
To after object event classification, the corresponding object event data of the object event classification are exported.
It is still " neighbouring restaurant please be search ", the thing that the present invention is got from the short sentence with the short sentence of aforementioned user input
For the event category of part is " search restaurant ", it is assumed that preset condition is that transition probability value is maximum, based on different event classification it
Between transition probability value can know that the transition probability value from " search restaurant " to " classification of meal taste " is 95%, from " lookup
The transition probability value of restaurant " to " distance is recently " is 90%, the transition probability value from " searching restaurant " to " sequence of popularity value " is
85%, then the present invention can be obtained from " searching restaurant " to next stage transition probability value maximum " classification of meal taste " as target
Event category, and then should " classification of meal taste " corresponding object event data, such as " selection buffet ", " selection Sichuan cuisine ", " selection
The outputs such as Guangdong dishes ", " selection Shandong cuisine " are shown to user.
Therefore, in answering method provided by the invention, the thing for the event that problem includes is obtained from the problem of user's input
Part classification;Based on the event correlation map being previously obtained, at least one object event classification of the event category is determined,
In, it include the transition probability between multiple event categories and multiple event categories, the target in the event correlation map
Event category is that developing stage is located at the event after the developing stage of the current event classification in the event correlation map
Classification, and the current event classification meets preset condition to the transition probability value of the object event classification;And then it exports and obtains
The corresponding object event data of at least one described object event classification taken.The present invention is based on turning between different event classification
Different types of events can be got up by changing probability, compared to the prior art in data statistical approach and knowledge mapping
Method, flexibility greatly improve.And the transition probability value between different event classification embodies and front and back occurs between different event
Probability is thought hereby based on the object event data that the transition probability value between different event classification is exported to user for user
The probability of the most possible event data of solution greatly increases, and improves the accuracy of event data output to a certain extent, subtracts
The operation of user is lacked.
As shown in figure 4, on the basis of the above embodiments, after step 103, the present invention may also include:
Step 104, all object event data based on acquisition are determined according to the instruction of user by first object event number
According to the new problem inputted as user;
Step 105, the event that the first object event data includes is obtained from the first object event data
Event category, and using the event category as current event classification, return to step 102.
In the embodiment of the present invention, based on all object event data of output display, user can therefrom choose a mesh
Mark event data (such as first object event data) problem new as one is putd question to.
Still by taking preceding example as an example, it is assumed that the object event data of output include " selection buffet ", " selection Sichuan cuisine ",
" selection Guangdong dishes ", " selection Shandong cuisine " etc., user chooses " selection Sichuan cuisine " by the modes such as clicking, then the present invention will be with " selection
The new problem that Sichuan cuisine " is proposed as user continues return step 102, obtains and exports corresponding with " selection Sichuan cuisine "
Object event data.
The embodiment of the present invention can continue the object event data that system exports as the new problem putd question to next time
It is putd question to, realizes the secondary optimization putd question to, searched for.
Based on a kind of answering method provided by the invention above, the present invention also provides a kind of question and answer systems, as shown in figure 5,
Include:
Event category acquiring unit 100, the problem of for being inputted from user in obtain the thing of the event that described problem includes
Part classification;
Object event classification acquiring unit 200, for determining the event class based on the event correlation map being previously obtained
At least one other object event classification, wherein include multiple event categories and multiple events in the event correlation map
Transition probability between classification, the object event classification are that developing stage is located at the current thing in the event correlation map
Event category after the developing stage of part classification, and the current event classification is to the transition probability of the object event classification
Value meets preset condition;
Output unit 300, for exporting the corresponding object event data of at least one object event classification described in acquisition.
The question and answer system includes processor and memory, and above-mentioned event category acquiring unit 100, object event classification obtain
It takes unit 200 and output unit 300 etc. to store in memory as program unit, memory is stored in by processor execution
In above procedure unit realize corresponding function.
Include kernel in processor, is gone in memory to transfer corresponding program unit by kernel.Kernel can be set one
Or more, question and answer are realized by adjusting kernel parameter.
Memory may include the non-volatile memory in computer-readable medium, random access memory (RAM) and/
Or the forms such as Nonvolatile memory, if read-only memory (ROM) or flash memory (flash RAM), memory include that at least one is deposited
Store up chip.
The embodiment of the invention provides a kind of storage mediums, are stored thereon with program, real when which is executed by processor
The existing answering method.
The embodiment of the invention provides a kind of processor, the processor is for running program, wherein described program operation
Answering method described in Shi Zhihang.
The embodiment of the invention provides a kind of equipment, equipment include processor, memory and storage on a memory and can
The program run on a processor, processor perform the steps of when executing program
The event category for the event that described problem includes is obtained in the problem of inputting from user;
Based on the event correlation map being previously obtained, at least one object event classification of the event category is determined,
In, it include the transition probability between multiple event categories and multiple event categories, the target in the event correlation map
Event category is that developing stage is located at the event after the developing stage of the current event classification in the event correlation map
Classification, and the current event classification meets preset condition to the transition probability value of the object event classification;
Export the corresponding object event data of at least one described object event classification obtained.
Optionally, the event correlation map obtains in the following manner:
Text is inputted for each: being obtained in the input text using preparatory trained Event Distillation model and is included
The event category of multiple events;
Thing according to the time of origin of each event by the early sequence to evening, to the multiple events for including in the input text
Part classification is ranked up;
In every two event after sequence has been calculated into, the event category of time of origin morning is converted into the thing in time of origin evening
The transition probability value of part classification;
Transition probability value between each event category obtained according to the input text for the first quantity, is not worked together
Transition probability value between part classification.
Optionally, the method also includes:
The transition probability value between the set storage different event classification is arranged using the probability of happening.
Optionally, the Event Distillation model obtains as follows:
Using the input text of the second quantity as training sample, the input text as the training sample is marked busy
Part classification;
Using the training sample as the input of Event Distillation model, Event Distillation model is instructed using preset algorithm
Practice, obtains trained Event Distillation model;
Wherein, the event category for the event that the described problem obtained in the problem of inputting from user includes, passes through the instruction
The Event Distillation model realization perfected.
Optionally, the preset algorithm includes: support vector machines algorithm or neural network algorithm.
Optionally, after the corresponding object event data of at least one described object event classification that the output obtains,
The method also includes:
All object event data based on acquisition are determined according to the instruction of user using first object event data as use
The new problem of family input;
The event category for the event that the first object event data includes is obtained from the first object event data,
And using the event category as current event classification, returns and execute the step of determining at least one object event classification.
Optionally, the current event classification meets preset condition packet to the transition probability value of the object event classification
Include: the transition probability value of current event classification to the object event classification is maximum, or, current event classification is to the target thing
The transition probability value of part classification is greater than preset threshold.Equipment herein can be server, PC, PAD, mobile phone etc..
Present invention also provides a kind of computer program products, when executing on data processing equipment, are adapted for carrying out just
The program of beginningization there are as below methods step:
The event category for the event that described problem includes is obtained in the problem of inputting from user;
Based on the event correlation map being previously obtained, at least one object event classification of the event category is determined,
In, it include the transition probability between multiple event categories and multiple event categories, the target in the event correlation map
Event category is that developing stage is located at the event after the developing stage of the current event classification in the event correlation map
Classification, and the current event classification meets preset condition to the transition probability value of the object event classification;
Export the corresponding object event data of at least one described object event classification obtained.
Optionally, the event correlation map obtains in the following manner:
Text is inputted for each: being obtained in the input text using preparatory trained Event Distillation model and is included
The event category of multiple events;
Thing according to the time of origin of each event by the early sequence to evening, to the multiple events for including in the input text
Part classification is ranked up;
In every two event after sequence has been calculated into, the event category of time of origin morning is converted into the thing in time of origin evening
The transition probability value of part classification;
Transition probability value between each event category obtained according to the input text for the first quantity, is not worked together
Transition probability value between part classification.
Optionally, the method also includes:
The transition probability value between the set storage different event classification is arranged using the probability of happening.
Optionally, the Event Distillation model obtains as follows:
Using the input text of the second quantity as training sample, the input text as the training sample is marked busy
Part classification;
Using the training sample as the input of Event Distillation model, Event Distillation model is instructed using preset algorithm
Practice, obtains trained Event Distillation model;
Wherein, the event category for the event that the described problem obtained in the problem of inputting from user includes, passes through the instruction
The Event Distillation model realization perfected.
Optionally, the preset algorithm includes: support vector machines algorithm or neural network algorithm.
Optionally, after the corresponding object event data of at least one described object event classification that the output obtains,
The method also includes:
All object event data based on acquisition are determined according to the instruction of user using first object event data as use
The new problem of family input;
The event category for the event that the first object event data includes is obtained from the first object event data,
And using the event category as current event classification, returns and execute the step of determining at least one object event classification.
Optionally, the current event classification meets preset condition packet to the transition probability value of the object event classification
Include: the transition probability value of current event classification to the object event classification is maximum, or, current event classification is to the target thing
The transition probability value of part classification is greater than preset threshold.
It should be understood by those skilled in the art that, embodiments herein can provide as method, system or computer program
Product.Therefore, complete hardware embodiment, complete software embodiment or reality combining software and hardware aspects can be used in the application
Apply the form of example.Moreover, it wherein includes the computer of computer usable program code that the application, which can be used in one or more,
The computer program implemented in usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) produces
The form of product.
The application is referring to method, the process of equipment (system) and computer program product according to the embodiment of the present application
Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions
The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs
Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce
A raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for real
The device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates,
Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or
The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or
The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one
The step of function of being specified in a box or multiple boxes.
In a typical configuration, calculating equipment includes one or more processors (CPU), input/output interface, net
Network interface and memory.
Memory may include the non-volatile memory in computer-readable medium, random access memory (RAM) and/
Or the forms such as Nonvolatile memory, such as read-only memory (ROM) or flash memory (flash RAM).Memory is computer-readable Jie
The example of matter.
Computer-readable medium includes permanent and non-permanent, removable and non-removable media can be by any method
Or technology come realize information store.Information can be computer readable instructions, data structure, the module of program or other data.
The example of the storage medium of computer includes, but are not limited to phase change memory (PRAM), static random access memory (SRAM), moves
State random access memory (DRAM), other kinds of random access memory (RAM), read-only memory (ROM), electric erasable
Programmable read only memory (EEPROM), flash memory or other memory techniques, read-only disc read only memory (CD-ROM) (CD-ROM),
Digital versatile disc (DVD) or other optical storage, magnetic cassettes, tape magnetic disk storage or other magnetic storage devices
Or any other non-transmission medium, can be used for storage can be accessed by a computing device information.As defined in this article, it calculates
Machine readable medium does not include temporary computer readable media (transitory media), such as the data-signal and carrier wave of modulation.
It should also be noted that, the terms "include", "comprise" or its any other variant are intended to nonexcludability
It include so that the process, method, commodity or the equipment that include a series of elements not only include those elements, but also to wrap
Include other elements that are not explicitly listed, or further include for this process, method, commodity or equipment intrinsic want
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including element
There is also other identical elements in process, method, commodity or equipment.
It will be understood by those skilled in the art that embodiments herein can provide as method, system or computer program product.
Therefore, complete hardware embodiment, complete software embodiment or embodiment combining software and hardware aspects can be used in the application
Form.It is deposited moreover, the application can be used to can be used in the computer that one or more wherein includes computer usable program code
The shape for the computer program product implemented on storage media (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.)
Formula.
The above is only embodiments herein, are not intended to limit this application.To those skilled in the art,
Various changes and changes are possible in this application.It is all within the spirit and principles of the present application made by any modification, equivalent replacement,
Improve etc., it should be included within the scope of the claims of this application.
Claims (10)
1. a kind of answering method characterized by comprising
The event category for the event that described problem includes is obtained in the problem of inputting from user;
Based on the event correlation map being previously obtained, at least one object event classification of the event category is determined, wherein institute
It states in event correlation map including the transition probability between multiple event categories and multiple event categories, the object event class
Developing stage it Wei not be located at the event category after the developing stage of the current event classification in the event correlation map, and
The current event classification meets preset condition to the transition probability value of the object event classification;
Export the corresponding object event data of at least one described object event classification obtained.
2. the method according to claim 1, wherein the event correlation map obtains in the following manner:
Text is inputted for each: include in the input text multiple are obtained using preparatory trained Event Distillation model
The event category of event;
Event class according to the time of origin of each event by the early sequence to evening, to the multiple events for including in the input text
It is not ranked up;
In every two event after sequence has been calculated into, the event category of time of origin morning is converted into the event class in time of origin evening
Other transition probability value;
Transition probability value between each event category obtained according to the input text for the first quantity, obtains different event class
Other transition probability value.
3. according to the method described in claim 2, it is characterized in that, the method also includes:
The transition probability value between the set storage different event classification is arranged using the probability of happening.
4. according to the method described in claim 2, it is characterized in that, the Event Distillation model obtains as follows:
Using the input text of the second quantity as training sample, the input text as the training sample is labeled with event class
Not;
Using the training sample as the input of Event Distillation model, Event Distillation model is trained using preset algorithm,
Obtain trained Event Distillation model;
Wherein, the event category for the event that the described problem obtained in the problem of inputting from user includes, is trained by described
Event Distillation model realization.
5. according to the method described in claim 4, it is characterized in that, the preset algorithm include: support vector machines algorithm or
Neural network algorithm.
6. the method according to claim 1, wherein at least one described object event class that the output obtains
After not corresponding object event data, the method also includes:
All object event data based on acquisition, it is according to the instruction of user determination that first object event data is defeated as user
The new problem entered;
The event category for the event that the first object event data includes is obtained from the first object event data, and will
The event category returns as current event classification and executes the step of determining at least one object event classification.
7. method according to claim 1-6, which is characterized in that the current event classification is to the target thing
It includes: transition probability value of the current event classification to the object event classification that the transition probability value of part classification, which meets preset condition,
Maximum, or, current event classification is greater than preset threshold to the transition probability value of the object event classification.
8. a kind of question and answer system characterized by comprising
Event category acquiring unit, the problem of for being inputted from user in obtain the event category of the event that described problem includes;
Object event classification acquiring unit, for determining the event category extremely based on the event correlation map being previously obtained
A few object event classification, wherein include in the event correlation map multiple event categories and multiple event categories it
Between transition probability, the object event classification be the event correlation map in developing stage be located at the current event classification
Developing stage after event category, and the current event classification to the transition probability value of the object event classification meet
Preset condition;
Output unit, for exporting the corresponding object event data of at least one object event classification described in acquisition.
9. a kind of storage medium, which is characterized in that be stored thereon with program, realize that right is wanted when described program is executed by processor
Answering method described in asking any one of 1 to 7.
10. a kind of processor, which is characterized in that the processor is for running program, wherein right of execution when described program is run
Benefit require any one of 1 to 7 described in answering method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810195536.1A CN110309273A (en) | 2018-03-09 | 2018-03-09 | Answering method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810195536.1A CN110309273A (en) | 2018-03-09 | 2018-03-09 | Answering method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110309273A true CN110309273A (en) | 2019-10-08 |
Family
ID=68073311
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810195536.1A Pending CN110309273A (en) | 2018-03-09 | 2018-03-09 | Answering method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110309273A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110955764A (en) * | 2019-11-19 | 2020-04-03 | 百度在线网络技术(北京)有限公司 | Scene knowledge graph generation method, man-machine conversation method and related equipment |
CN111459959A (en) * | 2020-03-31 | 2020-07-28 | 北京百度网讯科技有限公司 | Method and apparatus for updating event set |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080116449A1 (en) * | 2006-11-02 | 2008-05-22 | Macready William G | Processing relational database problems using analog processors |
CN105094315A (en) * | 2015-06-25 | 2015-11-25 | 百度在线网络技术(北京)有限公司 | Method and apparatus for smart man-machine chat based on artificial intelligence |
US20160086097A1 (en) * | 2014-09-24 | 2016-03-24 | Nec Laboratories America, Inc. | Automatic Discovery of Message Ordering Invariants in Heterogeneous Logs |
CN106156299A (en) * | 2016-06-29 | 2016-11-23 | 北京小米移动软件有限公司 | The subject content recognition methods of text message and device |
CN106649694A (en) * | 2016-12-19 | 2017-05-10 | 北京云知声信息技术有限公司 | Method and device for identifying user's intention in voice interaction |
CN106934012A (en) * | 2017-03-10 | 2017-07-07 | 上海数眼科技发展有限公司 | A kind of question answering in natural language method and system of knowledge based collection of illustrative plates |
CN107122416A (en) * | 2017-03-31 | 2017-09-01 | 北京大学 | A kind of Chinese event abstracting method |
-
2018
- 2018-03-09 CN CN201810195536.1A patent/CN110309273A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080116449A1 (en) * | 2006-11-02 | 2008-05-22 | Macready William G | Processing relational database problems using analog processors |
US20160086097A1 (en) * | 2014-09-24 | 2016-03-24 | Nec Laboratories America, Inc. | Automatic Discovery of Message Ordering Invariants in Heterogeneous Logs |
CN105094315A (en) * | 2015-06-25 | 2015-11-25 | 百度在线网络技术(北京)有限公司 | Method and apparatus for smart man-machine chat based on artificial intelligence |
CN106156299A (en) * | 2016-06-29 | 2016-11-23 | 北京小米移动软件有限公司 | The subject content recognition methods of text message and device |
CN106649694A (en) * | 2016-12-19 | 2017-05-10 | 北京云知声信息技术有限公司 | Method and device for identifying user's intention in voice interaction |
CN106934012A (en) * | 2017-03-10 | 2017-07-07 | 上海数眼科技发展有限公司 | A kind of question answering in natural language method and system of knowledge based collection of illustrative plates |
CN107122416A (en) * | 2017-03-31 | 2017-09-01 | 北京大学 | A kind of Chinese event abstracting method |
Non-Patent Citations (3)
Title |
---|
ZHONGYANG LI等: "EEG: Knowledge Base for Event Evolutionary Principles and Patterns"", 《CHINESE NATIONAL CONFERENCE ON SOCIAL MEDIA PROCESSING》 * |
李忠阳等: "事理图谱:事件演化的规律和模式", 《哈工大SCIR》 * |
杨文: "从知识图谱到事理图谱", 《HTTPS://WWW.LEIPHONE.COM/NEWS/201711/FX6MGS9WRPBSHNIG.HTML》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110955764A (en) * | 2019-11-19 | 2020-04-03 | 百度在线网络技术(北京)有限公司 | Scene knowledge graph generation method, man-machine conversation method and related equipment |
CN111459959A (en) * | 2020-03-31 | 2020-07-28 | 北京百度网讯科技有限公司 | Method and apparatus for updating event set |
CN111459959B (en) * | 2020-03-31 | 2023-06-30 | 北京百度网讯科技有限公司 | Method and apparatus for updating event sets |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US20200210647A1 (en) | Automated Summarization of Extracted Insight Data | |
US11841854B2 (en) | Differentiation of search results for accurate query output | |
US20220342926A1 (en) | User interface for context labeling of multimedia items | |
WO2019192261A1 (en) | Payment mode recommendation method and device and equipment | |
CN103488766B (en) | application program searching method and device | |
CN106557480B (en) | Method and device for realizing query rewriting | |
CN103886022B (en) | A kind of query facility and its method carrying out paging query based on major key field | |
CN104978356B (en) | A kind of recognition methods of synonym and device | |
CN111125086B (en) | Method, device, storage medium and processor for acquiring data resources | |
CN104298679A (en) | Application service recommendation method and device | |
CN108416616A (en) | The sort method and device of complaints and denunciation classification | |
CN111930848B (en) | Data partition storage method, device and system | |
CN111061954B (en) | Search result sorting method and device and storage medium | |
US11928142B2 (en) | Information processing apparatus and information processing method | |
JP6728178B2 (en) | Method and apparatus for processing search data | |
CN110795613B (en) | Commodity searching method, device and system and electronic equipment | |
CN106033455B (en) | Method and equipment for processing user operation information | |
KR20200121744A (en) | Method and device for processing user personal, server and storage medium | |
CN105164672A (en) | Content classification | |
JP6377917B2 (en) | Image search apparatus and image search program | |
CN110309273A (en) | Answering method and device | |
CN117033744A (en) | Data query method and device, storage medium and electronic equipment | |
CN108241652A (en) | Keyword clustering method and device | |
CN111723273A (en) | Smart cloud retrieval system and method | |
CN106776654B (en) | Data searching method and device |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191008 |
|
RJ01 | Rejection of invention patent application after publication |