CN108121800A

CN108121800A - Information generating method and device based on artificial intelligence

Info

Publication number: CN108121800A
Application number: CN201711396776.XA
Authority: CN
Inventors: 于佃海; 陈立玮; 贺文嵩; 周晓; 刘琼琼
Original assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Current assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Priority date: 2017-12-21
Filing date: 2017-12-21
Publication date: 2018-06-05
Anticipated expiration: 2037-12-21
Also published as: CN108121800B

Abstract

The embodiment of the present application discloses information generating method and device based on artificial intelligence.One specific embodiment of this method includes：It obtains the answer that terminal is sent and obtains request, wherein, which, which obtains request, includes intended application mark and target problem input by user；According to pre-set application identities and the correspondence in the professional question and answer storehouse in professional question and answer storehouse set, in gathering from the specialty question and answer storehouse, determine target specialty question and answer storehouse corresponding with intended application mark, wherein, the target specialty question and answer storehouse includes being related to the default problem of target professional domain, and target specialty question and answer storehouse association is provided with target similarity calculation；According to the keyword in the target problem, from the target specialty question and answer storehouse, choose at least one candidate and preset problem；Problem is preset for each candidate, based on the target similarity calculation, the candidate is generated and presets similarity between problem and the target problem.The embodiment enriches the mode of generation information.

Description

Information generating method and device based on artificial intelligence

Technical field

The invention relates to field of computer technology, and in particular to Internet technical field more particularly to based on people The information generating method and device of work intelligence.

Background technology

Artificial intelligence (Artificial Intelligence), english abbreviation AI.It is research, exploitation for simulating, Theory, method, a new technological sciences of technology and application system for the intelligence of extension and extension people.Artificial intelligence is to calculate One branch of machine science, it attempts to understand essence of intelligence, and produce it is a kind of it is new can be in a manner that human intelligence be similar The intelligence machine made a response, the research in the field include robot, language identification, image identification, natural language processing and specially Family's system etc..

Intelligent Answer System is a kind of new information service system, on the basis of the functions such as knowledge processing, semantics recognition On can analyze user view, fast accurately answer a question for user.Since intelligent Answer System can replace true man with using Family engages in the dialogue, and has the characteristics that the scope of one's knowledge is enriched and replies that speed is fast, therefore receives liking for users.

The content of the invention

The embodiment of the present application proposes information generating method and device based on artificial intelligence.

In a first aspect, the embodiment of the present application provides a kind of information generating method based on artificial intelligence, including：It obtains eventually The answer that end is sent obtains request, wherein, above-mentioned answer acquisition request includes intended application mark and target input by user is asked Topic；According to pre-set application identities and the correspondence in the professional question and answer storehouse in professional question and answer storehouse set, from above-mentioned specialty In the set of question and answer storehouse, target specialty question and answer storehouse corresponding with above-mentioned intended application mark is determined, wherein, above-mentioned target specialty question and answer Storehouse includes being related to the default problem of target professional domain, and above-mentioned target specialty question and answer storehouse association is provided with target similarity calculation mould Type, above-mentioned target similarity calculation is for definite default similarity between problem and target problem；According to above-mentioned target Keyword in problem from above-mentioned target specialty question and answer storehouse, chooses at least one candidate and presets problem；For above-mentioned at least one Each candidate that a candidate is preset in problem presets problem, based on above-mentioned target similarity calculation, generates the candidate and presets Similarity between problem and above-mentioned target problem.

In some embodiments, above-mentioned target similarity calculation includes term vector submodel and similarity calculation submodule Type, wherein, the input of above-mentioned similarity calculation submodel includes the output of above-mentioned term vector submodel, above-mentioned term vector submodel For characterize input text and term vector between correspondence, above-mentioned similarity calculation submodel for characterize term vector pair with Term vector is to the correspondence between the similarity of corresponding input text.

In some embodiments, above-mentioned target similarity calculation for characterize both target problem, default problem with The correspondence between similarity between target problem and default problem；And it above-mentioned is preset for above-mentioned at least one candidate Each candidate in problem presets problem, based on above-mentioned target similarity calculation, generate the candidate preset problem with it is above-mentioned Similarity between target problem, including：The each candidate preset for above-mentioned at least one candidate in problem presets problem, will The candidate presets problem and above-mentioned target problem, imports above-mentioned target similarity calculation, generate the candidate preset problem with Similarity between above-mentioned target problem.

In some embodiments, the above-mentioned each candidate preset for above-mentioned at least one candidate in problem presets problem, The candidate is preset into problem and above-mentioned target problem, imports above-mentioned target similarity calculation, the candidate is generated and presets problem With the similarity between above-mentioned target problem, including：Above-mentioned target problem is imported into above-mentioned term vector submodel, generates above-mentioned mesh Corresponding first term vector of mark problem；The each candidate preset for above-mentioned at least one candidate in problem presets problem, obtains The candidate for advancing with above-mentioned term vector submodel generation presets corresponding second term vector of problem；By above-mentioned first term vector Above-mentioned similarity calculation submodel is imported with second term vector, the candidate is generated and presets between problem and above-mentioned target problem Similarity.

In some embodiments, above-mentioned target specialty question and answer storehouse is further included associates the default answer set with presetting problem, Above-mentioned target similarity calculation is used to characterize target problem, default problem, the default answer set is associated with default problem The correspondence between similarity between three and default problem and above-mentioned target problem；It is and above-mentioned for above-mentioned at least one Each candidate that a candidate is preset in problem presets problem, based on above-mentioned target similarity calculation, generates the candidate and presets Similarity between problem and above-mentioned target problem, including：Each candidate in problem is preset for above-mentioned at least one candidate The candidate is preset problem, problem is preset with the candidate associates the default answer of setting and above-mentioned target problem by default problem, Above-mentioned target similarity calculation is imported, the candidate is generated and presets similarity between problem and above-mentioned target problem.

In some embodiments, the above-mentioned each candidate preset for above-mentioned at least one candidate in problem presets problem, The candidate is preset into problem, problem is preset with the candidate associates the default answer set and above-mentioned target problem, is imported above-mentioned Target similarity calculation generates the candidate and presets similarity between problem and above-mentioned target problem, including：By above-mentioned mesh Mark problem imports above-mentioned term vector submodel, generates corresponding 3rd term vector of above-mentioned target problem；For above-mentioned at least one Each candidate that candidate is preset in problem presets problem, and the candidate that acquisition advances with above-mentioned term vector submodel generation presets Problem and preset corresponding 4th term vector of both answers that problem association sets with the candidate；By above-mentioned 3rd term vector and it is somebody's turn to do 4th term vector imports above-mentioned similarity calculation submodel, generate the candidate preset it is similar between problem and above-mentioned target problem Degree.

In some embodiments, above-mentioned target similarity calculation is through the following steps that training obtained：It obtains logical With sample set, wherein, general sample includes general language material pair and indication information, and indication information is used to indicate the language material of language material centering It expresses same meaning or does not express same meaning；The initial first nerves net pre-established using above-mentioned general sample set, training Network obtains initial nervus opticus network；Target specialty sample set is obtained, wherein, target specialty sample includes target specialty language material Pair and indication information；Using above-mentioned target specialty sample set, it is similar to obtain above-mentioned target for the above-mentioned initial nervus opticus network of training Spend computation model.

In some embodiments, above-mentioned target similarity calculation is through the following steps that training obtained：Obtain mesh The professional sample set of mark, wherein, target specialty sample includes target specialty language material pair and indication information, and indication information is used to indicate language The language material of material centering expresses same meaning or does not express same meaning；Utilize above-mentioned target specialty sample set, training the initial 3rd Neutral net obtains above-mentioned target similarity calculation.

In some embodiments, the above-mentioned each candidate preset for above-mentioned at least one candidate in problem presets problem, Based on above-mentioned target similarity calculation, generate the candidate and preset similarity between problem and above-mentioned target problem, including： The each candidate preset for above-mentioned at least one candidate in problem presets problem, utilizes target similarity calculation, generation The candidate presets the first similarity between problem and above-mentioned target problem；Using general similarity calculation, the time is generated Default the second similarity between problem and above-mentioned target problem of choosing, wherein, above-mentioned general similarity calculation is used to determine Default similarity between problem and target problem；According to default weight, first similarity and second similarity are carried out Weighted sum obtains the candidate and presets similarity between problem and above-mentioned target problem.

In some embodiments, above-mentioned general similarity calculation is through the following steps that training obtained：It obtains logical With sample set, wherein, general sample includes general language material pair and indication information, and indication information is used to indicate the language material of language material centering Whether same meaning is expressed；The initial fourth nerve network pre-established using above-mentioned general sample set, training is obtained above-mentioned logical Use similarity calculation.

In some embodiments, the above method further includes：It is preset from above-mentioned at least one candidate in problem, according to similarity By high order on earth, choose predetermined number candidate and preset problem as problem to be presented.

In some embodiments, the above method further includes：Obtain the default answer for associating and setting with above-mentioned problem to be presented； Above-mentioned problem to be presented and above-mentioned default answer are sent to above-mentioned terminal.

In some embodiments, the above method further includes：Above-mentioned problem to be presented is sent to above-mentioned terminal, wherein, on State terminal and show above-mentioned problem to be presented to user, receive it is above-mentioned it is input by user be used to indicate it is matched with above-mentioned target problem The confirmation message of problem to be presented, and return to above-mentioned confirmation message；Receive above-mentioned confirmation message；It will be indicated by above-mentioned confirmation message Problem to be presented associated by default answer, be back to above-mentioned terminal.

Second aspect, the embodiment of the present application provide a kind of information generation device based on artificial intelligence, above device bag It includes：First acquisition unit, the answer for obtaining terminal transmission obtain request, wherein, above-mentioned answer, which obtains request, includes target Application identities and target problem input by user；Determination unit, for according to pre-set application identities and professional question and answer storehouse The correspondence in the professional question and answer storehouse in set from above-mentioned professional question and answer storehouse set, determines and above-mentioned intended application mark pair The target specialty question and answer storehouse answered, wherein, above-mentioned target specialty question and answer storehouse includes being related to the default problem of target professional domain, above-mentioned The association of target specialty question and answer storehouse is provided with target similarity calculation, and above-mentioned target similarity calculation is default for determining Similarity between problem and target problem；First chooses unit, for the keyword in above-mentioned target problem, from above-mentioned In target specialty question and answer storehouse, choose at least one candidate and preset problem；Generation unit, for pre- for above-mentioned at least one candidate Each candidate in rhetoric question topic presets problem, based on above-mentioned target similarity calculation, generate the candidate preset problem with it is upper State the similarity between target problem.

In some embodiments, above-mentioned target similarity calculation for characterize both target problem, default problem with The correspondence between similarity between target problem and default problem；And above-mentioned generation unit, it is additionally operable to：For above-mentioned Each candidate that at least one candidate is preset in problem presets problem, which is preset problem and above-mentioned target problem, imports Above-mentioned target similarity calculation generates the candidate and presets similarity between problem and above-mentioned target problem.

In some embodiments, above-mentioned generation unit, is additionally operable to：Above-mentioned target problem is imported into above-mentioned term vector submodule Type generates corresponding first term vector of above-mentioned target problem；Each candidate in problem is preset for above-mentioned at least one candidate Default problem, the candidate that acquisition advances with above-mentioned term vector submodel generation preset corresponding second term vector of problem；It will Above-mentioned first term vector and second term vector import above-mentioned similarity calculation submodel, generate the candidate preset problem with it is above-mentioned Similarity between target problem.

In some embodiments, above-mentioned target specialty question and answer storehouse is further included associates the default answer set with presetting problem, Above-mentioned target similarity calculation is used to characterize target problem, default problem, the default answer set is associated with default problem The correspondence between similarity between three and default problem and above-mentioned target problem；And above-mentioned generation unit, also use In：The each candidate preset for above-mentioned at least one candidate in problem presets problem, which is preset problem and the candidate The default answer and above-mentioned target problem that default problem association is set, import above-mentioned target similarity calculation, generation should Candidate presets the similarity between problem and above-mentioned target problem.

In some embodiments, above-mentioned generation unit, is additionally operable to：Above-mentioned target problem is imported into above-mentioned term vector submodule Type generates corresponding 3rd term vector of above-mentioned target problem；Each candidate in problem is preset for above-mentioned at least one candidate Default problem, the candidate that acquisition advances with above-mentioned term vector submodel generation preset problem and preset problem pass with the candidate Join both the answer set corresponding 4th term vectors；Above-mentioned 3rd term vector and the 4th term vector are imported into above-mentioned similarity Submodel is calculated, the candidate is generated and presets similarity between problem and above-mentioned target problem.

In some embodiments, above-mentioned generation unit, is additionally operable to：It is preset for above-mentioned at least one candidate every in problem A candidate presets problem, using target similarity calculation, generates the candidate and presets between problem and above-mentioned target problem First similarity；Using general similarity calculation, generate the candidate preset between problem and above-mentioned target problem second Similarity, wherein, above-mentioned general similarity calculation is for definite default similarity between problem and target problem；According to Default weight, is weighted summation to first similarity and second similarity, obtains the candidate and preset problem and above-mentioned mesh Similarity between mark problem.

In some embodiments, above device further includes：Second chooses unit, for being preset from above-mentioned at least one candidate In problem, according to similarity by high order on earth, choose predetermined number candidate and preset problem as problem to be presented.

In some embodiments, above device further includes：Second acquisition unit closes for obtaining with above-mentioned problem to be presented Join the default answer set；First transmitting element, for above-mentioned problem to be presented and above-mentioned default answer to be sent to above-mentioned end End.

In some embodiments, above device further includes：Second transmitting element, for above-mentioned problem to be presented to be sent to Above-mentioned terminal, wherein, above-mentioned terminal shows above-mentioned problem to be presented to user, receive it is above-mentioned it is input by user be used to indicate with it is upper The confirmation message of the matched problem to be presented of target problem is stated, and returns to above-mentioned confirmation message；Receiving unit, it is above-mentioned for receiving Confirmation message；Returning unit for the default answer associated by the problem to be presented indicated by by above-mentioned confirmation message, is back to Above-mentioned terminal.

The third aspect, the embodiment of the present application provide a kind of server, and above-mentioned server includes：One or more processing Device；Storage device, for storing one or more programs, when said one or multiple programs are by said one or multiple processors During execution so that said one or multiple processors realization such as the method for first aspect.

Fourth aspect, the embodiment of the present application provide a kind of computer readable storage medium, are stored thereon with computer journey Sequence realizes the method such as first aspect when the program is executed by processor.

Information generating method and device provided by the embodiments of the present application based on artificial intelligence, by obtaining terminal transmission Answer obtains request, wherein, above-mentioned answer, which obtains request, includes intended application mark and target problem input by user；According to pre- The application identities first set and the correspondence in the professional question and answer storehouse in professional question and answer storehouse set, from above-mentioned professional question and answer storehouse set In, determine target specialty question and answer storehouse corresponding with above-mentioned intended application mark, wherein, above-mentioned target specialty question and answer storehouse includes being related to The default problem of target professional domain, above-mentioned target specialty question and answer storehouse association are provided with target similarity calculation, above-mentioned mesh Similarity calculation is marked for determining default similarity between problem and target problem；According to the pass in above-mentioned target problem Keyword from above-mentioned target specialty question and answer storehouse, chooses at least one candidate and presets problem；It is preset for above-mentioned at least one candidate Each candidate in problem presets problem, based on above-mentioned target similarity calculation, generate the candidate preset problem with it is above-mentioned Similarity between target problem improves the accuracy of generated information.

Description of the drawings

By reading the detailed description made to non-limiting example made with reference to the following drawings, the application's is other Feature, objects and advantages will become more apparent upon：

Fig. 1 is that this application can be applied to exemplary system architecture figures therein；

Fig. 2 is the flow chart according to one embodiment of the information generating method based on artificial intelligence of the application；

Fig. 3 is the schematic diagram according to an application scenarios of the information generating method based on artificial intelligence of the application；

Fig. 4 is the flow chart according to another embodiment of the information generating method based on artificial intelligence of the application；

Fig. 5 is the flow chart according to another embodiment of the information generating method based on artificial intelligence of the application；

Fig. 6 is the flow chart according to another embodiment of the information generating method based on artificial intelligence of the application；

Fig. 7 is a kind of exemplary process diagram of realization method of the method according to Fig. 6；

Fig. 8 is the flow chart according to another embodiment of the information generating method based on artificial intelligence of the application；

Fig. 9 is a kind of exemplary process diagram of realization method of the method according to Fig. 8；

Figure 10 is the structure diagram according to one embodiment of the information generation device based on artificial intelligence of the application；

Figure 11 is adapted for the structure diagram of the computer system of the server for realizing the embodiment of the present application.

Specific embodiment

The application is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched The specific embodiment stated is used only for explaining related invention rather than the restriction to the invention.It also should be noted that in order to Convenient for description, illustrated only in attached drawing and invent relevant part with related.

It should be noted that in the case where there is no conflict, the feature in embodiment and embodiment in the application can phase Mutually combination.The application is described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

Fig. 1 shows the information generating method based on artificial intelligence that can apply the application or the letter based on artificial intelligence Cease the exemplary system architecture 100 of the embodiment of generating means.

As shown in Figure 1, system architecture 100 can include terminal device 101,102,103, network 104 and server 105. Network 104 can be to provide the medium of communication link between terminal device 101,102,103 and server 105.Network 104 can include various connection types, such as wired, wireless communication link or fiber optic cables etc..

User can be interacted with using terminal equipment 101,102,103 by network 104 with server 105, to receive or send out Send message etc..Various telecommunication customer end applications can be installed, such as intelligent answer class should on terminal device 101,102,103 With, web browser applications, the application of shopping class, searching class application, instant messaging tools, mailbox client, social platform software Deng.

Terminal device 101,102,103 can be the various electronic equipments for having display screen, include but not limited to intelligent hand Machine, tablet computer, E-book reader, MP3 player (Moving Picture Experts Group Audio Layer III, dynamic image expert's compression standard audio level 3), MP4 (Moving Picture Experts Group Audio Layer IV, dynamic image expert's compression standard audio level 4) player, pocket computer on knee and desktop computer etc. Deng.

Server 105 can be to provide the server of various services, such as to intelligently being asked on terminal device 101,102,103 It answers class application and the background server supported is provided.Background server can obtain the data such as request to the answer received and divide The processing such as analysis, and handling result (such as the problem of matching and/or answer) is fed back into terminal device.

It should be noted that the information generating method based on artificial intelligence that the embodiment of the present application is provided is generally by servicing Device 105 performs, and correspondingly, the information generation device based on artificial intelligence is generally positioned in server 105.

It should be understood that the number of the terminal device, network and server in Fig. 1 is only schematical.According to realization need Will, can have any number of terminal device, network and server.

With continued reference to Fig. 2, it illustrates an implementations of the information generating method based on artificial intelligence according to the application The flow 200 of example.The above-mentioned information generating method based on artificial intelligence, comprises the following steps：

Step 201, obtain the answer that terminal is sent and obtain request.

In the present embodiment, electronic equipment (such as Fig. 1 institutes of the information generating method operation based on artificial intelligence thereon The server shown) the answer acquisition request that terminal is sent can be obtained from local or other electronic equipments.

In the present embodiment, above-mentioned electronic equipment directly can receive above-mentioned answer from other electronic equipments and obtain request, To obtain above-mentioned illness description information；It can also receive after above-mentioned answer obtains request and be stored to local from other electronic equipments, Again above-mentioned illness description information is obtained from local.

In the present embodiment, above-mentioned answer, which obtains request, includes intended application mark and target problem input by user.

In the present embodiment, above-mentioned terminal can be that user inputs the terminal of problem using it.

In the present embodiment, above application mark can be the application that user inputs problem wherein.It is as an example, above-mentioned Using can be the application of medical class, bank's class application etc..User is being used in application, the problem of being pre-set from application Advice window inputs problem.The above problem can be sent to server by above application using terminal.

Optionally, problem input by user can be involved in the problems, such as professional domain.As an example, problem input by user It can be " Patent Law how many in total”.

Step 202, according to pre-set application identities and the corresponding pass in the professional question and answer storehouse in professional question and answer storehouse set System in gathering from professional question and answer storehouse, determines target specialty question and answer storehouse corresponding with intended application mark.

In the present embodiment, during above-mentioned electronic equipment can be gathered according to pre-set application identities with professional question and answer storehouse Professional question and answer storehouse correspondence, from above-mentioned professional question and answer storehouse set, determine and the corresponding mesh of above-mentioned intended application mark The professional question and answer storehouse of mark.

As an example, the professional question and answer storehouse in professional question and answer storehouse set can include but is not limited to one or more of： The professional question and answer storehouse in the professional question and answer storehouse of legal field, the professional question and answer storehouse of medical field and financial field etc..

As an example, the intended application that above-mentioned answer obtains request is identified as the application identities that law class is applied, then can To determine the professional question and answer storehouse of legal field as target specialty question and answer storehouse.

In the present embodiment, above-mentioned target specialty question and answer storehouse includes being related to the default problem of target professional domain.

Optionally, above-mentioned target specialty question and answer storehouse is further included associates the default answer set with presetting problem.

As an example, the professional question and answer storehouse of legal field can include some the default problems for being related to legal field, for example, " how many item of criminal law" " how many item of Patent Law" etc..

In the present embodiment, above-mentioned target specialty question and answer storehouse association is provided with target similarity calculation.

As an example, the professional question and answer storehouse association of legal field is provided with the relevant similarity calculation of legal issue, The professional question and answer storehouse association of medical field is provided with the relevant similarity calculation of medical care problem, the professional question and answer of financial field Storehouse medical treatment association is provided with the relevant similarity calculation of monetary affair.

It should be noted that corresponding a certain field, can utilize the language material in the field pointedly to similarity calculation mould Type is trained.It is thus obtained to have preferable matching feature to this field for the problem that the model in this field, from And improve the matched accuracy of problem.

In the present embodiment, above-mentioned target similarity calculation is for definite default phase between problem and target problem Like degree.

It should be noted that target similarity calculation be not intended to determine it is similar between target problem and answer Degree, but for determining the similarity between default problem and target problem in target specialty question and answer storehouse.

Step 203, from target specialty question and answer storehouse, it is pre- to choose at least one candidate for the keyword in target problem Rhetoric question is inscribed.

In the present embodiment, above-mentioned electronic equipment can be special from above-mentioned target according to the keyword in above-mentioned target problem In industry question and answer storehouse, choose at least one candidate and preset problem.

As an example, cutting word can be carried out using to target problem, then keyword is extracted from the result that cutting word obtains.

" Patent Law in total how many item " can be " Patent Law, how many, item " as an example, target problem.

It is alternatively possible to using the mode of inverted index, from target specialty question and answer storehouse, select in target problem The higher one or more of keyword registration presets problem, and problem is preset as candidate.

As an example, for target problem " Patent Law in total how many item ", three candidates can be selected and preset problem " the item number of Patent Law is how many" " what the effect of Patent Law is" and " number of words of Patent Law is how many”.

Step 204, each candidate preset at least one candidate in problem presets problem, based on target similarity meter Model is calculated, the candidate is generated and presets similarity between problem and target problem.

In the present embodiment, above-mentioned electronic equipment can preset each candidate in problem for above-mentioned at least one candidate Default problem based on above-mentioned target similarity calculation, generates the candidate and presets phase between problem and above-mentioned target problem Like degree.

In some optional realization methods of the present embodiment, preset from above-mentioned at least one candidate in problem, according to phase Like degree by high order on earth, choose predetermined number candidate and preset problem as problem to be presented.

As an example, for target problem " Patent Law in total how many item ", two problems to be presented can be selected " specially The item number of sharp method is how many" and " number of words of Patent Law is how many”.

In some optional realization methods of the present embodiment, the above method further includes：It obtains and above-mentioned problem to be presented Associate the default answer set；Above-mentioned problem to be presented and above-mentioned default answer are sent to above-mentioned terminal.

As an example, above-mentioned electronic equipment can also obtain problem to be presented and " the item number of Patent Law is how many" association The default answer set, such as can be " 67 ".Above-mentioned electronic equipment can also obtain problem to be presented and the " word of Patent Law Number is how many", the default answer of setting is associated, such as can be " 9,000 or so ".

In some optional realization methods of the present embodiment, the above method further includes：Above-mentioned problem to be presented is sent To above-mentioned terminal, wherein, above-mentioned terminal shows above-mentioned problem to be presented to user, receive it is above-mentioned it is input by user be used to indicate with The confirmation message of the above-mentioned matched problem to be presented of target problem, and return to above-mentioned confirmation message；Receive above-mentioned confirmation message；It will The default answer associated by problem to be presented indicated by above-mentioned confirmation message, is back to above-mentioned terminal.

As an example, above-mentioned electronic equipment can " the item number of Patent Law be how many by problem to be presented" and " Patent Law Number of words is how many" it is sent to terminal.The problem to be presented that terminal display receives.Then terminal receives user's input validation letter Breath confirms " the item number of Patent Law is how many" be and the matched problem to be presented of target problem.Terminal will confirm that information is sent to Server.Server receives confirmation message, by " the item number of Patent Law is how many" associated by default answer " 67 ", return To terminal.The above-mentioned default answer " 67 " of terminal display.

As an example, above-mentioned target similarity calculation can be mapping table, above-mentioned target similarity calculation mould Type can correspond to storage problem pair and similarity.Above-mentioned electronic equipment determines that candidate puts up a question in advance after target problem is received After topic, it can search in a large amount of problems pair of storage and preset the problem of problem matches pair with target problem and candidate.It will look into The problem of finding is determined as candidate and presets similarity between problem and above-mentioned target problem to corresponding similarity.

With continued reference to Fig. 3, Fig. 3 is the application scenarios according to the information generating method based on artificial intelligence of the present embodiment One schematic diagram.In the application scenarios of Fig. 3, user initiates an answer acquisition first with terminal 301 to server 302 please Ask 303；Afterwards, server can obtain the content that answer obtains request from the background, for example, intended application mark " application of law class " With target problem " Patent Law how many in total”.Then, above-mentioned server is according to pre-set application identities and professional question and answer The correspondence in the professional question and answer storehouse in the set of storehouse in gathering from professional question and answer storehouse, determines that " law class should with intended application mark With " corresponding target specialty question and answer storehouse " the professional question and answer storehouse of legal field ".After again, above-mentioned server can be according to target problem In keyword " Patent Law, how many, item ", from target specialty question and answer storehouse, choose at least one candidate and preset problem, such as select Three candidates of taking-up preset problem, and " the item number of Patent Law is how many" " what the effect of Patent Law is" and " number of words of Patent Law It is how many”.After again, each candidate that server can be preset in problem at least one candidate presets problem, based on target Similarity calculation generates the candidate and presets similarity between problem and target problem, for example, " Patent Law can be generated Item number be how many" similarity 95% between target problem " Patent Law in total how many item "；" Patent Law can be generated Effect what is" similarity 30% between target problem " Patent Law in total how many item "；" Patent Law can be generated Number of words be how many" similarity 60% between target problem " Patent Law in total how many item ".Finally, above-mentioned server The highest candidate of similarity can be returned to terminal and preset the corresponding answer 304 of problem, for example, " 67 ".

The method that above-described embodiment of the application provides, the answer sent by obtaining terminal obtain request, wherein, it is above-mentioned Answer, which obtains request, includes intended application mark and target problem input by user；According to pre-set application identities and specialty The correspondence in the professional question and answer storehouse in the set of question and answer storehouse from above-mentioned professional question and answer storehouse set, determines and above-mentioned intended application Corresponding target specialty question and answer storehouse is identified, wherein, above-mentioned target specialty question and answer storehouse includes being related to the pre- rhetoric question of target professional domain Topic, above-mentioned target specialty question and answer storehouse association are provided with target similarity calculation, and above-mentioned target similarity calculation is used for Determine default similarity between problem and target problem；According to the keyword in above-mentioned target problem, from above-mentioned target specialty In question and answer storehouse, choose at least one candidate and preset problem；The each candidate preset for above-mentioned at least one candidate in problem is pre- Rhetoric question is inscribed, based on above-mentioned target similarity calculation, generate the candidate preset it is similar between problem and above-mentioned target problem Degree improves the accuracy of generated information.

It please refers to Fig.4, it illustrates one embodiment of the information generating method based on artificial intelligence according to the application Flow 400.The above-mentioned information generating method based on artificial intelligence, comprises the following steps：

Step 401, obtain the answer that terminal is sent and obtain request.

Step 402, according to pre-set application identities and the corresponding pass in the professional question and answer storehouse in professional question and answer storehouse set System in gathering from professional question and answer storehouse, determines target specialty question and answer storehouse corresponding with intended application mark.

In the present embodiment, during above-mentioned electronic equipment can be gathered according to pre-set application identities with professional question and answer storehouse Professional question and answer storehouse correspondence, from professional question and answer storehouse gather in, determine target specialty corresponding with intended application mark ask Answer storehouse.

In the present embodiment, above-mentioned target similarity calculation is used to characterize both target problem, default problem and mesh The correspondence between similarity between mark problem and default problem.

Step 403, from target specialty question and answer storehouse, it is pre- to choose at least one candidate for the keyword in target problem Rhetoric question is inscribed.

Step 404, each default candidate preset at least one candidate in problem presets problem, which is preset Problem and above-mentioned target problem import target similarity calculation, generate the candidate preset problem and above-mentioned target problem it Between similarity.

In some optional realization methods of the present embodiment, above-mentioned target similarity calculation can include term vector Submodel and similarity calculation submodel.

In this realization method, above-mentioned term vector submodel inputs the corresponding pass between text and term vector for characterizing System.

In this realization method, above-mentioned similarity calculation submodel is used to characterize term vector for term vector to corresponding defeated Enter the correspondence between the similarity of text.

In this realization method, above-mentioned term vector submodel can be using samples of text collection, and initial model is instructed It gets.Herein, samples of text includes text and term vector.Initial model can be bag of words, convolutional neural networks Model, length time memory model etc..

As an example, initial model, can refer to model is unbred or does not train completion.Initial model can be with Initial parameter is provided with, parameter can be adjusted constantly in the training process.

In this realization method, above-mentioned similarity calculation submodel can be using term vector sample set, to initial nerve What network was trained.Herein, term vector sample includes the similarity between term vector pair and term vector pair.

In this realization method, initial neutral net can be various neutral nets, for example, convolutional neural networks, cycling Neutral net, shot and long term Memory Neural Networks etc..

As an example, initial neutral net, can refer to neutral net is unbred or does not train completion.Initially Each layer of neutral net can be provided with initial parameter, and parameter can be adjusted constantly in the training process.

As an example, the artificial neural network that initial neutral net can be various types of indisciplines or training is not completed Network or the artificial neural network that a variety of indisciplines or not training is not completed are combined obtained model, for example, initially Neutral net can be unbred convolutional neural networks or unbred Recognition with Recurrent Neural Network, can also be Institute is combined to unbred convolutional neural networks, unbred Recognition with Recurrent Neural Network and unbred full articulamentum Obtained model.

In this realization method, above-mentioned term vector submodel and similarity calculation submodel can train together, can also Separately training.

In this realization method, the input of above-mentioned similarity calculation submodel includes the output of above-mentioned term vector submodel.

As an example, the input of above-mentioned similarity model can include the corresponding with target problem of term vector submodel output Term vector, can also include presetting the corresponding term vector of problem with candidate.It should be noted that it is corresponding to preset problem with candidate Term vector may not be what term vector submodel exported.

In this realization method, step 404 can be accomplished by the following way：By above-mentioned target problem import upper predicate to Quantum model generates corresponding first term vector of above-mentioned target problem；It is preset for above-mentioned at least one candidate every in problem A candidate presets problem, obtain advance with above-mentioned term vector submodel generation the candidate preset corresponding second word of problem to Amount；Above-mentioned first term vector and second term vector are imported into above-mentioned similarity calculation submodel, the candidate is generated and presets problem With the similarity between above-mentioned target problem.

It should be noted that for each default problem in target question and answer storehouse, side shown in the present embodiment can implemented Before method, the term vector that term vector submodel generates the default problem is advanced with, avoids implementing in method shown in the present embodiment Candidate is calculated in the process and presets the term vector of problem, it is possible thereby to improve calculating speed.

Figure 4, it is seen that compared with the corresponding embodiments of Fig. 2, the information based on artificial intelligence in the present embodiment The flow 400 of generation method highlights by the use of candidate and presets problem with above-mentioned target problem as inputting, and calculates the step of similarity Suddenly.The scheme of the present embodiment description can enrich the mode of generation information as a result,.

Fig. 5 is refer to, it illustrates one embodiment of the information generating method based on artificial intelligence according to the application Flow 500.The above-mentioned information generating method based on artificial intelligence, comprises the following steps：

Step 501, obtain the answer that terminal is sent and obtain request.

Step 502, according to pre-set application identities and the corresponding pass in the professional question and answer storehouse in professional question and answer storehouse set System in gathering from professional question and answer storehouse, determines target specialty question and answer storehouse corresponding with intended application mark.

In the present embodiment, above-mentioned target specialty question and answer storehouse is further included associates the default answer set with presetting problem.

In the present embodiment, above-mentioned target similarity calculation is used to characterize target problem, default problem, be put up a question with pre- The correspondence between similarity between the default answer three of topic association setting and default problem and above-mentioned target problem.

Step 503, from target specialty question and answer storehouse, it is pre- to choose at least one candidate for the keyword in target problem Rhetoric question is inscribed.

Step 504, each candidate preset at least one candidate in problem presets problem, which is put up a question in advance Topic presets with the candidate problem and associates the default answer of setting and above-mentioned target problem, imports above-mentioned target similarity calculation Model generates the candidate and presets similarity between problem and above-mentioned target problem.

It should be noted that preset the auxiliary information of problem using default answer as candidate, preset problem with candidate together with Target similarity calculation is inputted, can usually determine to preset problem with the candidate that target problem is increasingly similar.

As an example, the keyword in problem may be repeated in default answer.For example, for target problem, " Patent Law is total How many altogether", by the use of " 67 ", this default answer can improve candidate and preset the problem " item of Patent Law as auxiliary information Number is how many" and target problem " Patent Law how many in total" between similarity.

It should be noted that preset the auxiliary information of problem using default answer as candidate, preset problem with candidate together with Input target similarity calculation, it is also possible to since default answer is more matched with default problem, improve candidate and preset problem Similarity between target problem.

It is this existing although as an example, some default problem meaning difference are larger, but can be answered with same answer The appearance of elephant is usually because answer content is more, covers several different aspects.In this case, it is possible to due to more Suitable answer, and find the problem of more similar.

As an example, the input of above-mentioned similarity model can include the corresponding with target problem of term vector submodel output Term vector, can also include presetting problem and the corresponding term vector of default answer with candidate.It is it should be noted that pre- with candidate It may not be what term vector submodel exported that term vector corresponding with answer is preset is inscribed in rhetoric question.

In some optional realization methods of embodiment, step 504 can be accomplished by the following way：By above-mentioned target Problem imports above-mentioned term vector submodel, generates corresponding 3rd term vector of above-mentioned target problem；For above-mentioned at least one time Each candidate in the default problem of choosing presets problem, and the candidate that acquisition advances with above-mentioned term vector submodel generation puts up a question in advance It inscribes and presets problem with the candidate and associate both the answer set corresponding 4th term vectors；By above-mentioned 3rd term vector and this Four term vectors import above-mentioned similarity calculation submodel, generate the candidate preset it is similar between problem and above-mentioned target problem Degree.

In this realization method, the associated default answer of problem can be preset for each default problem and with this, by this Default problem and the default answer are spliced, and obtain problem answers splicing result.Then computational problem answer splicing result, in advance The term vector of the problem answers splicing result is generated first with term vector submodel.

From figure 5 it can be seen that compared with the corresponding embodiments of Fig. 2, the information based on artificial intelligence in the present embodiment The flow 500 of generation method highlights by the use of candidate and presets problem, default answer and target problem as input, and generation is similar The step of spending.The scheme of the present embodiment description can introduce default answer as auxiliary information as a result, determine similarity, thus Abundant information generating mode, and it is possible to further improve the accuracy of generated information.

Fig. 6 is refer to, it illustrates one embodiment of the information generating method based on artificial intelligence according to the application Flow 600.The above-mentioned information generating method based on artificial intelligence, comprises the following steps：

Step 601, obtain the answer that terminal is sent and obtain request.

Step 602, according to pre-set application identities and the corresponding pass in the professional question and answer storehouse in professional question and answer storehouse set System in gathering from professional question and answer storehouse, determines target specialty question and answer storehouse corresponding with intended application mark.

In the present embodiment, Fig. 7 is refer to, above-mentioned target similarity calculation is trained by flow 700 shown in Fig. 7 It obtains：

Step 701, general sample set is obtained.

In this realization method, general sample includes general language material pair and indication information, and indication information is used to indicate language material The language material of centering expresses same meaning or does not express same meaning.

As an example, general language material is to including, " what day is it today" and " week today is several", indication information is used to indicate two Person expresses same meaning.

Step 702, the initial first nerves network pre-established using general sample set, training obtains initial second god Through network.

In this realization method, initial first nerves network can be various neutral nets, for example, convolutional neural networks, Recognition with Recurrent Neural Network, shot and long term Memory Neural Networks etc..

As an example, initial first nerves network, can refer to initial first nerves network is unbred or does not instruct Practice what is completed.Each layer of initial first nerves network can be provided with initial parameter, and parameter in the training process can be by not Disconnected adjustment.

As an example, the artificial god that initial first nerves network can be various types of indisciplines or training is not completed To a variety of indisciplines or the artificial neural network completed is not trained through network or to be combined obtained model, for example, Initial neutral net can be unbred convolutional neural networks or unbred Recognition with Recurrent Neural Network, may be used also To be to carry out group to unbred convolutional neural networks, unbred Recognition with Recurrent Neural Network and unbred full articulamentum Close obtained model.

Step 703, target specialty sample set is obtained.

Herein, target specialty sample includes target specialty language material pair and indication information.

As an example, to that can include, " the item number of Patent Law is how many to professional language material" and " Patent Law in total how many Item ", indication information are used to indicate the two and express same meaning.

Step 704, using target specialty sample set, the initial nervus opticus network of training obtains target similarity calculation mould Type.

It should be noted that on the basis of universal model, (thermal starting) is finely adjusted to model using professional language material Mode can realize that the sentence of all-purpose language type can either be identified in target similarity calculation, also can be to special The sentence of industry language form is identified.Thus, it is possible to improve the accuracy rate of identification.

Step 603, from target specialty question and answer storehouse, it is pre- to choose at least one candidate for the keyword in target problem Rhetoric question is inscribed.

Step 604, each candidate preset at least one candidate in problem presets problem, by target similarity calculation Model generates the candidate and presets similarity between problem and target problem.

In the present embodiment, above-mentioned electronic equipment can preset each candidate in problem for above-mentioned at least one candidate Default problem, each candidate preset at least one candidate in problem preset problem, are given birth to by target similarity calculation The similarity between problem and target problem is preset into the candidate.

It can be seen from figures 6 and 7 that compared with the corresponding embodiments of Fig. 2, in the present embodiment based on artificial intelligence The flow 600 of information generating method is highlighted recycles target specialty sample the set pair analysis model to be trained first with general sample set The step of.The scheme of the present embodiment description can improve the accuracy rate of the information of target similarity calculation output as a result,.

Fig. 8 is refer to, it illustrates one embodiment of the information generating method based on artificial intelligence according to the application Flow 800.The above-mentioned information generating method based on artificial intelligence, comprises the following steps：

Step 801, obtain the answer that terminal is sent and obtain request.

Step 802, according to pre-set application identities and the corresponding pass in the professional question and answer storehouse in professional question and answer storehouse set System in gathering from professional question and answer storehouse, determines target specialty question and answer storehouse corresponding with intended application mark.

In the present embodiment, Fig. 9 is refer to, above-mentioned target similarity calculation is obtained by the training of flow 900：

Step 901, target specialty sample set is obtained.

Herein, target specialty sample includes target specialty language material pair and indication information, and indication information is used to indicate language material The language material of centering expresses same meaning or does not express same meaning.

Step 902, using target specialty sample set, the initial third nerve network of training obtains target similarity calculation mould Type.

It should be noted that using target specialty sample set, i.e. target specialty language material is trained initial third nerve network, is obtained The target similarity calculation arrived has preferable processing capacity to being involved in the problems, such as that target is professional.

In this realization method, initial third nerve network can be various neutral nets, for example, convolutional neural networks, Recognition with Recurrent Neural Network, shot and long term Memory Neural Networks etc..

As an example, initial third nerve network, can refer to initial third nerve network is unbred or does not instruct Practice what is completed.Each layer of initial third nerve network can be provided with initial parameter, and parameter in the training process can be by not Disconnected adjustment.

As an example, the artificial god that initial third nerve network can be various types of indisciplines or training is not completed To a variety of indisciplines or the artificial neural network completed is not trained through network or to be combined obtained model, for example, Initial neutral net can be unbred convolutional neural networks or unbred Recognition with Recurrent Neural Network, may be used also To be to carry out group to unbred convolutional neural networks, unbred Recognition with Recurrent Neural Network and unbred full articulamentum Close obtained model.

Step 804, each candidate preset at least one candidate in problem presets problem, utilizes target similarity meter Model is calculated, the candidate is generated and presets the first similarity between problem and target problem.

In the present embodiment, above-mentioned electronic equipment can preset each candidate in problem for above-mentioned at least one candidate Default problem using target similarity calculation, generates the candidate and presets the first phase between problem and above-mentioned target problem Like degree.

Step 805, each candidate preset at least one candidate in problem presets problem, utilizes general similarity meter Model is calculated, the candidate is generated and presets the second similarity between problem and above-mentioned target problem.

In the present embodiment, each candidate that above-mentioned electronic equipment can be preset in problem at least one candidate presets Problem using general similarity calculation, generates the candidate and presets the second similarity between problem and target problem.

In the present embodiment, above-mentioned general similarity calculation is for definite default phase between problem and target problem Like degree.

In the present embodiment, above-mentioned general similarity calculation is through the following steps that training obtained：It obtains general Sample set, wherein, general sample includes general language material pair and indication information, and the language material that indication information is used to indicate language material centering is The no same meaning of expression；The initial fourth nerve network pre-established using above-mentioned general sample set, training, is obtained above-mentioned general Similarity calculation.

It should be noted that using general sample set, i.e., general language material trains initial fourth nerve network, obtained target Similarity calculation, to being involved in the problems, such as that all-purpose language has preferable processing capacity.

In this realization method, initial fourth nerve network can be various neutral nets, for example, convolutional neural networks, Recognition with Recurrent Neural Network, shot and long term Memory Neural Networks etc..

As an example, initial fourth nerve network, can refer to initial fourth nerve network is unbred or does not instruct Practice what is completed.Each layer of initial fourth nerve network can be provided with initial parameter, and parameter in the training process can be by not Disconnected adjustment.

As an example, the artificial god that initial fourth nerve network can be various types of indisciplines or training is not completed To a variety of indisciplines or the artificial neural network completed is not trained through network or to be combined obtained model, for example, Initial neutral net can be unbred convolutional neural networks or unbred Recognition with Recurrent Neural Network, may be used also To be to carry out group to unbred convolutional neural networks, unbred Recognition with Recurrent Neural Network and unbred full articulamentum Close obtained model.

Step 806, each candidate preset at least one candidate in problem presets problem, right according to default weight First similarity and second similarity are weighted summation, obtain the candidate preset it is similar between problem and target problem Degree.

In the present embodiment, above-mentioned electronic equipment can preset each candidate in problem for above-mentioned at least one candidate Default problem, according to default weight, is weighted summation to first similarity and second similarity, obtains the candidate and preset Similarity between problem and above-mentioned target problem.

It should be noted that using both universal model and target similarity calculation, final similarity is determined, it can be with The similarity in terms of similarity and professional voice during two problems are taken into account in terms of the all-purpose language, improves generated similarity Accuracy.

From Fig. 8 and Fig. 9 as can be seen that compared with the corresponding embodiments of Fig. 2, in the present embodiment based on artificial intelligence The flow 800 of information generating method is highlighted to both universal model and target similarity calculation, determines final similarity The step of.The scheme of the present embodiment description enriches information generating mode as a result, and it is possible to improve generated similarity Accuracy.

With further reference to Figure 10, as the realization to method shown in above-mentioned each figure, this application provides one kind based on artificial One embodiment of the information generation device of intelligence, the device embodiment is corresponding with embodiment of the method shown in Fig. 2, the device It specifically can be applied in various electronic equipments.

As shown in Figure 10, the above-mentioned information generation device 1000 based on artificial intelligence of the present embodiment includes：First obtains Unit 1001, determination unit 1002, first choose unit 1003 and generation unit 1004.Wherein, first acquisition unit, for obtaining The answer that terminal is sent is taken to obtain request, wherein, above-mentioned answer, which obtains request, includes intended application mark and mesh input by user Mark problem；Determination unit, for according to pre-set application identities and pair in the professional question and answer storehouse in professional question and answer storehouse set It should be related to, from above-mentioned professional question and answer storehouse set, determine target specialty question and answer storehouse corresponding with above-mentioned intended application mark, In, above-mentioned target specialty question and answer storehouse includes being related to the default problem of target professional domain, and above-mentioned target specialty question and answer storehouse association is set Target similarity calculation is equipped with, above-mentioned target similarity calculation is preset for definite between problem and target problem Similarity；First chooses unit, for the keyword in above-mentioned target problem, from above-mentioned target specialty question and answer storehouse, choosing At least one candidate is taken to preset problem；Generation unit, for presetting each candidate in problem for above-mentioned at least one candidate Default problem based on above-mentioned target similarity calculation, generates the candidate and presets phase between problem and above-mentioned target problem Like degree.

In some optional realization methods of the present embodiment, above-mentioned target similarity calculation includes term vector submodule Type and similarity calculation submodel, wherein, the input of above-mentioned similarity calculation submodel includes the defeated of above-mentioned term vector submodel Go out, above-mentioned term vector submodel is for the correspondence between characterization input text and term vector, above-mentioned similarity calculation submodule Type is used to characterize term vector pair with term vector to the correspondence between the similarity of corresponding input text.

In some optional realization methods of the present embodiment, above-mentioned target similarity calculation is asked for characterizing target Correspondence between the similarity of both topic, default problem between target problem and default problem；And above-mentioned generation list Member is additionally operable to：The each candidate preset for above-mentioned at least one candidate in problem presets problem, by the candidate preset problem and Above-mentioned target problem imports above-mentioned target similarity calculation, generates the candidate and presets between problem and above-mentioned target problem Similarity.

In some optional realization methods of the present embodiment, above-mentioned generation unit is additionally operable to：Above-mentioned target problem is led Enter above-mentioned term vector submodel, generate corresponding first term vector of above-mentioned target problem；It is preset for above-mentioned at least one candidate Each candidate in problem presets problem, and the candidate that acquisition advances with above-mentioned term vector submodel generation presets problem correspondence The second term vector；Above-mentioned first term vector and second term vector are imported into above-mentioned similarity calculation submodel, generate the time The default similarity between problem and above-mentioned target problem of choosing.

In some optional realization methods of the present embodiment, above-mentioned target specialty question and answer storehouse further includes and presets problem pass Join the default answer set, above-mentioned target similarity calculation closes for characterizing target problem, default problem and default problem Join the correspondence between the similarity between the default answer three set and default problem and above-mentioned target problem；On and Generation unit is stated, is additionally operable to：The each candidate preset for above-mentioned at least one candidate in problem presets problem, and the candidate is pre- Rhetoric question topic presets with the candidate problem and associates the default answer of setting and above-mentioned target problem, imports above-mentioned target similarity Computation model generates the candidate and presets similarity between problem and above-mentioned target problem.

In some optional realization methods of the present embodiment, above-mentioned generation unit is additionally operable to：Above-mentioned target problem is led Enter above-mentioned term vector submodel, generate corresponding 3rd term vector of above-mentioned target problem；It is preset for above-mentioned at least one candidate Each candidate in problem presets problem, obtain advance with above-mentioned term vector submodel generation the candidate preset problem and with The candidate presets both the answer that problem association is set corresponding 4th term vectors；By above-mentioned 3rd term vector and the 4th word to Amount imports above-mentioned similarity calculation submodel, generates the candidate and presets similarity between problem and above-mentioned target problem.

In some optional realization methods of the present embodiment, above-mentioned target similarity calculation through the following steps that What training obtained：General sample set is obtained, wherein, general sample includes general language material pair and indication information, and indication information is used for Indicate that the language material of language material centering expresses same meaning or do not express same meaning；Using above-mentioned general sample set, training is built in advance Vertical initial first nerves network, obtains initial nervus opticus network；Target specialty sample set is obtained, wherein, target specialty sample This includes target specialty language material pair and indication information；Utilize above-mentioned target specialty sample set, the above-mentioned initial nervus opticus net of training Network obtains above-mentioned target similarity calculation.

In some optional realization methods of the present embodiment, above-mentioned target similarity calculation through the following steps that What training obtained：Target specialty sample set is obtained, wherein, target specialty sample includes target specialty language material pair and indication information, The language material that indication information is used to indicate language material centering expresses same meaning or does not express same meaning；Utilize above-mentioned target specialty sample This collection, the initial third nerve network of training, obtains above-mentioned target similarity calculation.

In some optional realization methods of the present embodiment, above-mentioned generation unit is additionally operable to：For above-mentioned at least one Each candidate that candidate is preset in problem presets problem, using target similarity calculation, generate the candidate preset problem with The first similarity between above-mentioned target problem；Using general similarity calculation, generate the candidate preset problem with it is above-mentioned The second similarity between target problem, wherein, above-mentioned general similarity calculation is asked for determining to preset problem with target Similarity between topic；According to default weight, summation is weighted to first similarity and second similarity, obtains the time The default similarity between problem and above-mentioned target problem of choosing.

In some optional realization methods of the present embodiment, above-mentioned general similarity calculation through the following steps that What training obtained：General sample set is obtained, wherein, general sample includes general language material pair and indication information, and indication information is used for Indicate whether the language material of language material centering expresses same meaning；Using above-mentioned general sample set, the initial 4th pre-established is trained Neutral net obtains above-mentioned general similarity calculation.

In some optional realization methods of the present embodiment, above device further includes：Second chooses unit (not shown), For being preset from above-mentioned at least one candidate in problem, according to similarity by high order on earth, predetermined number candidate is chosen Default problem is as problem to be presented.

In some optional realization methods of the present embodiment, above device further includes：Second acquisition unit (not shown), For obtaining the default answer for associating and setting with above-mentioned problem to be presented；First transmitting element (not shown), for being treated above-mentioned Showing problem and above-mentioned default answer are sent to above-mentioned terminal.

In some optional realization methods of the present embodiment, above device further includes：Second transmitting element (not shown), For above-mentioned problem to be presented to be sent to above-mentioned terminal, wherein, above-mentioned terminal shows above-mentioned problem to be presented to user, receives The above-mentioned confirmation message input by user being used to indicate with the matched problem to be presented of above-mentioned target problem, and return to above-mentioned confirmation Information；Receiving unit (not shown), for receiving above-mentioned confirmation message；Returning unit (not shown), for by above-mentioned confirmation letter Default answer associated by the indicated problem to be presented of breath, is back to above-mentioned terminal.

In the present embodiment, first acquisition unit 1001, determination unit 1002, first choose unit 1003 and generation unit 1004 specific processing and its caused technique effect can correspond to step 201 in embodiment, step 202, step with reference to figure 2 respectively Rapid 203 and the related description of step 204, details are not described herein.

It should be noted that the realization of each unit is thin in the information generation device provided in this embodiment based on artificial intelligence Section and technique effect may be referred to the explanation of other embodiments in the application, and details are not described herein.

Below with reference to Figure 11, it illustrates suitable for being used for realizing the computer system 11 of the server of the embodiment of the present application Structure diagram.Server shown in Figure 11 is only an example, should not be to the function and use scope of the embodiment of the present application Bring any restrictions.

As shown in figure 11, computer system 1100 include central processing unit (CPU) 1101, can according to be stored in only It reads the program in memory (ROM) 1102 or is loaded into from storage part 1108 in random access storage device (RAM) 1103 Program and perform various appropriate actions and processing.In RAM 1103, also it is stored with system 1100 and operates required various journeys Sequence and data.CPU 1101, ROM 1102 and RAM 1103 are connected with each other by bus 1104.Input/output (I/O) interface 1105 are also connected to bus 1104.

I/O interfaces 1105 are connected to lower component：Importation 1106 including keyboard, mouse etc.；Including such as cathode The output par, c 1107 of ray tube (CRT), liquid crystal display (LCD) etc. and loud speaker etc.；Storage part including hard disk etc. 1108；And the communications portion 1109 of the network interface card including LAN card, modem etc..Communications portion 1109 passes through Communication process is performed by the network of such as internet.Driver 1110 is also according to needing to be connected to I/O interfaces 1105.It is detachable to be situated between Matter 1111, such as disk, CD, magneto-optic disk, semiconductor memory etc. are mounted on driver 1110 as needed, so as to Storage part 1108 is mounted into as needed in the computer program read from it.

Particularly, in accordance with an embodiment of the present disclosure, it may be implemented as computer above with reference to the process of flow chart description Software program.For example, embodiment of the disclosure includes a kind of computer program product, including being carried on computer-readable medium On computer program, which includes for the program code of the method shown in execution flow chart.In such reality It applies in example, which can be downloaded and installed from network by communications portion 1109 and/or from detachable media 1111 are mounted.When the computer program is performed by central processing unit (CPU) 1101, perform and limited in the present processes Above-mentioned function.

It should be noted that the above-mentioned computer-readable medium of the application can be computer-readable signal media or meter Calculation machine readable storage medium storing program for executing either the two any combination.Computer readable storage medium for example can be --- but not It is limited to --- electricity, magnetic, optical, electromagnetic, system, device or the device of infrared ray or semiconductor or arbitrary above combination.Meter The more specific example of calculation machine readable storage medium storing program for executing can include but is not limited to：Electrical connection with one or more conducting wires, just It takes formula computer disk, hard disk, random access storage device (RAM), read-only memory (ROM), erasable type and may be programmed read-only storage Device (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), light storage device, magnetic memory device, Or above-mentioned any appropriate combination.In this application, computer readable storage medium can any include or store journey The tangible medium of sequence, the program can be commanded the either device use or in connection of execution system, device.And at this In application, computer-readable signal media can include in a base band or as carrier wave a part propagation data-signal, Wherein carry computer-readable program code.Diversified forms may be employed in the data-signal of this propagation, including but it is unlimited In electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be that computer can Any computer-readable medium beyond storage medium is read, which can send, propagates or transmit and be used for By instruction execution system, device either device use or program in connection.It is included on computer-readable medium Program code can be transmitted with any appropriate medium, be included but not limited to：Wirelessly, electric wire, optical cable, RF etc. or above-mentioned Any appropriate combination.

Flow chart and block diagram in attached drawing, it is illustrated that according to the system of the various embodiments of the application, method and computer journey Architectural framework in the cards, function and the operation of sequence product.In this regard, each box in flow chart or block diagram can generation The part of one module of table, program segment or code, the part of the module, program segment or code include one or more use In the executable instruction of logic function as defined in realization.It should also be noted that it is marked at some as in the realization replaced in box The function of note can also be occurred with being different from the order marked in attached drawing.For example, two boxes succeedingly represented are actually It can perform substantially in parallel, they can also be performed in the opposite order sometimes, this is depending on involved function.Also to note Meaning, the combination of each box in block diagram and/or flow chart and the box in block diagram and/or flow chart can be with holding The dedicated hardware based system of functions or operations as defined in row is realized or can use specialized hardware and computer instruction Combination realize.

Being described in unit involved in the embodiment of the present application can be realized by way of software, can also be by hard The mode of part is realized.Described unit can also be set in the processor, for example, can be described as：A kind of processor bag Include first acquisition unit, determination unit, the first selection unit and generation unit.Wherein, the title of these units is in certain situation Under do not form restriction to the unit in itself, for example, first acquisition unit is also described as, " obtain that terminal sends answers Case obtains the unit of request ".

As on the other hand, present invention also provides a kind of computer-readable medium, which can be Included in device described in above-described embodiment；Can also be individualism, and without be incorporated the device in.Above-mentioned calculating Machine readable medium carries one or more program, when said one or multiple programs are performed by the device so that should Device：It obtains the answer that terminal is sent and obtains request, wherein, above-mentioned answer acquisition request includes intended application mark and user is defeated The target problem entered；The correspondence in the professional question and answer storehouse in being gathered according to pre-set application identities with professional question and answer storehouse, From above-mentioned professional question and answer storehouse set, target specialty question and answer storehouse corresponding with above-mentioned intended application mark is determined, wherein, above-mentioned mesh The professional question and answer storehouse of mark includes being related to the default problem of target professional domain, and above-mentioned target specialty question and answer storehouse association is provided with target phase Like degree computation model, above-mentioned target similarity calculation is for definite default similarity between problem and target problem；Root According to the keyword in above-mentioned target problem, from above-mentioned target specialty question and answer storehouse, choose at least one candidate and preset problem；For Each candidate that above-mentioned at least one candidate is preset in problem presets problem, based on above-mentioned target similarity calculation, generation The candidate presets the similarity between problem and above-mentioned target problem.

The preferred embodiment and the explanation to institute's application technology principle that above description is only the application.People in the art Member should be appreciated that invention scope involved in the application, however it is not limited to the technology that the particular combination of above-mentioned technical characteristic forms Scheme, while should also cover in the case where not departing from foregoing invention design, it is carried out by above-mentioned technical characteristic or its equivalent feature The other technical solutions for being combined and being formed.Such as features described above has similar work(with (but not limited to) disclosed herein The technical solution that the technical characteristic of energy is replaced mutually and formed.

Claims

1. a kind of information generating method based on artificial intelligence, including：

It obtains the answer that terminal is sent and obtains request, wherein, the answer acquisition request includes intended application mark and user is defeated The target problem entered；

According to pre-set application identities and the correspondence in the professional question and answer storehouse in professional question and answer storehouse set, from the specialty In the set of question and answer storehouse, target specialty question and answer storehouse corresponding with intended application mark is determined, wherein, the target specialty question and answer Storehouse includes being related to the default problem of target professional domain, and the target specialty question and answer storehouse association is provided with target similarity calculation mould Type, the target similarity calculation is for definite default similarity between problem and target problem；

According to the keyword in the target problem, from the target specialty question and answer storehouse, choose at least one candidate and put up a question in advance Topic；

The each candidate preset at least one candidate in problem presets problem, based on the target similarity calculation mould Type generates the candidate and presets similarity between problem and the target problem.

2. according to the method described in claim 1, wherein, the target similarity calculation includes term vector submodel and phase Submodel is calculated like degree, wherein, the input of the similarity calculation submodel includes the output of the term vector submodel, described For term vector submodel for characterizing the correspondence between input text and term vector, the similarity calculation submodel is used for table Term vector pair is levied with term vector to the correspondence between the similarity of corresponding input text.

3. according to the method described in claim 2, wherein, the target similarity calculation is for characterizing target problem, pre- Correspondence of both rhetoric question topics between the similarity between target problem and default problem；And

The each candidate preset at least one candidate in problem presets problem, based on the target similarity meter Model is calculated, the candidate is generated and presets similarity between problem and the target problem, including：

The each candidate preset at least one candidate in problem presets problem, which is preset problem and the mesh Mark problem imports the target similarity calculation, generate the candidate preset it is similar between problem and the target problem Degree.

4. according to the method described in claim 3, wherein, each time preset at least one candidate in problem The default problem of choosing, presets problem and the target problem by the candidate, imports the target similarity calculation, generate the time The default similarity between problem and the target problem of choosing, including：

The target problem is imported into the term vector submodel, generates corresponding first term vector of the target problem；

Each candidate in problem is preset at least one candidate and presets problem, acquisition advances with term vector The candidate of model generation presets corresponding second term vector of problem；First term vector and second term vector are imported into institute Similarity calculation submodel is stated, the candidate is generated and presets similarity between problem and the target problem.

5. according to the method described in claim 2, wherein, the target specialty question and answer storehouse is further included associates setting with presetting problem Default answer, the target similarity calculation is for characterizing target problem, default problem, associate setting with default problem Default answer three and default problem and the target problem between similarity between correspondence；And

The each candidate preset at least one candidate in problem presets problem, which is preset problem and the time The default answer and the target problem that the default problem association of choosing is set, import the target similarity calculation, generate The candidate presets the similarity between problem and the target problem.

6. according to the method described in claim 5, wherein, each time preset at least one candidate in problem The candidate is preset problem, problem is preset with the candidate associates the default answer of setting and the target is asked by the default problem of choosing Topic, imports the target similarity calculation, generates the candidate and presets similarity between problem and the target problem, bag It includes：

The target problem is imported into the term vector submodel, generates corresponding 3rd term vector of the target problem；

Each candidate in problem is preset at least one candidate and presets problem, acquisition advances with term vector The candidate of model generation, which presets problem and presets problem with the candidate, associates both the answer set corresponding 4th term vectors； 3rd term vector and the 4th term vector are imported into the similarity calculation submodel, the candidate is generated and presets problem and institute State the similarity between target problem.

7. according to the method any one of claim 1-6, wherein, the target similarity calculation is by following What step was trained：

General sample set is obtained, wherein, general sample includes general language material pair and indication information, and indication information is used to indicate language material The language material of centering expresses same meaning or does not express same meaning；

The initial first nerves network pre-established using the general sample set, training, obtains initial nervus opticus network；

Target specialty sample set is obtained, wherein, target specialty sample includes target specialty language material pair and indication information；

Using the target specialty sample set, the training initial nervus opticus network obtains the target similarity calculation mould Type.

8. according to the method any one of claim 1-6, wherein, the target similarity calculation is by following What step was trained：

Target specialty sample set is obtained, wherein, target specialty sample includes target specialty language material pair and indication information, indication information The language material for being used to indicate language material centering expresses same meaning or does not express same meaning；

Using the target specialty sample set, the initial third nerve network of training obtains the target similarity calculation.

9. according to the method described in claim 8, wherein, each time preset at least one candidate in problem The default problem of choosing, based on the target similarity calculation, generates the candidate and presets between problem and the target problem Similarity, including：

The each candidate preset at least one candidate in problem presets problem, using target similarity calculation, It generates the candidate and presets the first similarity between problem and the target problem；Utilize general similarity calculation, generation The candidate presets the second similarity between problem and the target problem, wherein, the general similarity calculation is used for Determine default similarity between problem and target problem；According to default weight, to first similarity and second similarity Summation is weighted, the candidate is obtained and presets similarity between problem and the target problem.

10. according to the method described in claim 9, wherein, the general similarity calculation is through the following steps that training It obtains：

General sample set is obtained, wherein, general sample includes general language material pair and indication information, and indication information is used to indicate language material Whether the language material of centering expresses same meaning；

The initial fourth nerve network pre-established using the general sample set, training, obtains the general similarity calculation Model.

11. according to the method any one of claim 1-6, wherein, the method further includes：

It is preset from least one candidate in problem, according to similarity by high order on earth, chooses predetermined number candidate Default problem is as problem to be presented.

12. according to the method for claim 11, wherein, the method further includes：

Obtain the default answer for associating and setting with the problem to be presented；

The problem to be presented and the default answer are sent to the terminal.

13. according to the method for claim 11, wherein, the method further includes：

The problem to be presented is sent to the terminal, wherein, the terminal shows the problem to be presented to user, receives The confirmation message input by user being used to indicate with the matched problem to be presented of the target problem, and return to the confirmation Information；

Receive the confirmation message；

The default answer associated by problem to be presented indicated by by the confirmation message, is back to the terminal.

14. a kind of information generation device based on artificial intelligence, including：

First acquisition unit, the answer for obtaining terminal transmission obtain request, wherein, the answer, which obtains request, includes target Application identities and target problem input by user；

Determination unit, for the corresponding pass according to pre-set application identities with the professional question and answer storehouse in professional question and answer storehouse set System from the professional question and answer storehouse set, determines target specialty question and answer storehouse corresponding with intended application mark, wherein, institute Stating target specialty question and answer storehouse includes being related to the default problem of target professional domain, and the target specialty question and answer storehouse association is provided with mesh Similarity calculation is marked, the target similarity calculation is default similar between problem and target problem for determining Degree；

First chooses unit, for the keyword in the target problem, from the target specialty question and answer storehouse, chooses extremely A few candidate presets problem；

Generation unit presets problem, based on the mesh for presetting each candidate in problem at least one candidate Similarity calculation is marked, the candidate is generated and presets similarity between problem and the target problem.

15. a kind of electronic equipment, including：

One or more processors；

Storage device, for storing one or more programs,

When one or more of programs are performed by one or more of processors so that one or more of processors Realize the method as described in any in claim 1-13.

16. a kind of computer readable storage medium, is stored thereon with computer program, wherein, when which is executed by processor Realize the method as described in any in claim 1-13.