CN110019742A - Method and apparatus for handling information - Google Patents
Method and apparatus for handling information Download PDFInfo
- Publication number
- CN110019742A CN110019742A CN201810630153.2A CN201810630153A CN110019742A CN 110019742 A CN110019742 A CN 110019742A CN 201810630153 A CN201810630153 A CN 201810630153A CN 110019742 A CN110019742 A CN 110019742A
- Authority
- CN
- China
- Prior art keywords
- text
- word
- target word
- class
- information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 63
- 230000004044 response Effects 0.000 claims description 40
- 230000009471 action Effects 0.000 claims description 36
- 238000013507 mapping Methods 0.000 claims description 19
- 238000006243 chemical reaction Methods 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 6
- 239000000284 extract Substances 0.000 claims description 5
- 235000013399 edible fruits Nutrition 0.000 claims 1
- 230000006399 behavior Effects 0.000 description 42
- 238000012549 training Methods 0.000 description 26
- 230000006870 function Effects 0.000 description 10
- 238000010801 machine learning Methods 0.000 description 10
- 238000013528 artificial neural network Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 8
- 230000008569 process Effects 0.000 description 8
- 230000006854 communication Effects 0.000 description 7
- 230000015654 memory Effects 0.000 description 7
- 238000012545 processing Methods 0.000 description 7
- 238000004891 communication Methods 0.000 description 6
- 230000002452 interceptive effect Effects 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 5
- 230000003993 interaction Effects 0.000 description 4
- 239000000203 mixture Substances 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 230000000306 recurrent effect Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000009472 formulation Methods 0.000 description 2
- 230000005291 magnetic effect Effects 0.000 description 2
- 239000004065 semiconductor Substances 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 230000003542 behavioural effect Effects 0.000 description 1
- 244000309464 bull Species 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 210000003127 knee Anatomy 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
- G06F16/355—Class or cluster creation or modification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Artificial Intelligence (AREA)
- Human Computer Interaction (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Machine Translation (AREA)
Abstract
The embodiment of the present application discloses the method and apparatus for handling information.One specific embodiment of this method includes: to obtain text to be processed, determines class of service corresponding with the text;Based on the class of service, the target word in the text is determined;History target word is obtained, the class of service, the target word and the history target word is based on, determines goal systems behavior;Execute the goal systems behavior.This embodiment improves the flexibilities that decision is carried out to goal systems behavior.
Description
Technical field
The invention relates to field of computer technology, and in particular to the method and apparatus for handling information.
Background technique
With the development of computer technology, human-computer interaction technology has obtained more and more applications.Man-machine interactive system can
To be the system for imitating the interactive mode of person to person to construct, in general, conversational system can be referred to as.
Existing conversational system is usually that session rules are write in people's formulation, after parsing information transmitted by user, is based on
The information that session rules match corresponding system action (such as feedback a word or progress information search etc.) or fed back.
Before this, it needs manually to expect a variety of clause in advance to carry out information matches.Meanwhile one business scenario of every increase, it needs
Rewrite rule.
Summary of the invention
The embodiment of the present application proposes the method and apparatus for handling information.
In a first aspect, the embodiment of the present application provides a kind of method for handling information, this method comprises: obtaining wait locate
The text of reason determines class of service corresponding with text;Based on class of service, the target word in text is determined;Obtain history
Target word is based on class of service, target word and history target word, determines goal systems behavior;Performance objective system action.
In some embodiments, text to be processed is obtained, comprising: receive the voice messaging of user;To voice messaging into
The conversion of row speech text, generates text to be processed.
In some embodiments, class of service corresponding with text is determined, comprising: enter text into training in advance
Business disaggregated model determines class of service corresponding with text, wherein disaggregated model is used to characterize pair of text and class of service
It should be related to.
In some embodiments, it is based on class of service, determines the target word in text, comprising: in response to determining service class
Do not match with pre-set business classification, characteristic information extracted from text, wherein characteristic information include word sequence vector, word to
Sequence, contextual feature information sequence are measured, word sequence vector is made of the word vector for constituting the word of text, and term vector sequence is by right
The term vector of obtained word is constituted after text participle, and contextual feature information sequence is by word obtained after segmenting to text
The characteristic information of context is constituted;Characteristic information extracted from text is input to target word trained in advance and determines mould
Type obtains the target word in text, wherein target word determines model for determining the target word in text.
In some embodiments, it is based on class of service, determines the target word in text, comprising: in response to determining service class
It is not mismatched with pre-set business classification, determines the string length of text;In response to determining that string length is less than preset length,
Text is matched with the preset target word in preset target set of words, by it is in text, with any preset target word phase
The word matched is determined as the target word of text.
In some embodiments, this method further include: by pre-stored dialog history status information, target word, business
Classification and the system action information for being used to indicate a performed upper system action are summarized, using the information after summarizing as
Current dialogue states information is stored, to be updated to dialog history status information.
In some embodiments, history target word is obtained, class of service, target word and history target word is based on, determines mesh
Mark system behavior, comprising: using class of service, target word and history target word as input information, will input information input to pre-
First trained decision model, obtains goal systems behavior, wherein decision model is used to characterize pair of input information and system action
It should be related to.
In some embodiments, performance objective system action, comprising: in response to determining that goal systems behavior is feedback voice
Information retrieves the candidate answers to match with target word from preset knowledge mapping;Determine be converted into text and candidate
The similarity of answer;In response to determining that similarity is greater than default value, exported candidate answers as target answer.
In some embodiments, performance objective system action, further includes: in response to determining that similarity is not more than present count
Value enters text into end-to-end model trained in advance using candidate answers as the first candidate answers, obtains the second candidate and answer
Case, wherein end-to-end model is used to characterize the corresponding relationship of text and answer;First candidate answers and the second candidate answers are defeated
Enter to order models trained in advance, obtain ranking results, wherein order models are for being ranked up candidate answers;It is based on
Ranking results determine target answer, export target answer.
In some embodiments, performance objective system action, comprising: in response to determining that goal systems behavior is inquiry knowledge
Map retrieves knowledge point associated with target word information from preset knowledge mapping, exports knowledge point information.
Second aspect, the embodiment of the present application provide it is a kind of for handling the device of information, the device include: obtain it is single
Member is configured to obtain text to be processed, determines class of service corresponding with text;First determination unit, is configured to
Based on class of service, the target word in text is determined;Second determination unit is configured to obtain history target word, is based on business
Classification, target word and history target word, determine goal systems behavior;Execution unit is configured to performance objective system action.
In some embodiments, acquiring unit includes: receiving module, is configured to receive the voice messaging of user;Conversion
Module is configured to carry out speech text conversion to voice messaging, generates text to be processed.
In some embodiments, acquiring unit is further configured to: being entered text into business classification trained in advance
Model determines class of service corresponding with text, wherein disaggregated model is used to characterize the corresponding relationship of text and class of service.
In some embodiments, the first determination unit includes: extraction module, be configured in response to determine class of service with
Pre-set business classification matches, and characteristic information is extracted from text, wherein characteristic information includes word sequence vector, term vector sequence
Column, contextual feature information sequence, word sequence vector are made of the word vector for constituting the word of text, and term vector sequence is by text
The term vector of obtained word is constituted after participle, and contextual feature information sequence is by the upper and lower of word obtained after segmenting to text
The characteristic information of text is constituted;First input module is configured to characteristic information extracted from text being input to preparatory instruction
Experienced target word determines model, obtains the target word in text, wherein target word determines model for determining the target in text
Word.
In some embodiments, the first determination unit includes: the first determining module, is configured in response to determine service class
It is not mismatched with pre-set business classification, determines the string length of text;Matching module is configured in response to determine character string
Length be less than preset length, text is matched with the preset target word in preset target set of words, by it is in text, with times
The word that one preset target word matches is determined as the target word of text.
In some embodiments, device further include: storage unit is configured to pre-stored dialog history state
It information, target word, class of service and is used to indicate the system action information of a performed upper system action and is summarized, it will
Information after summarizing is stored as current dialogue states information, to be updated to dialog history status information.
In some embodiments, the second determination unit is further configured to: by class of service, target word and history target
Word will input information input to decision model trained in advance, obtain goal systems behavior, wherein decision as input information
Model is used to characterize the corresponding relationship of input information and system action.
In some embodiments, execution unit includes: retrieval module, is configured in response to determine that goal systems behavior is
Voice messaging is fed back, the candidate answers to match with target word are retrieved from preset knowledge mapping;Second determining module is matched
It is set to the similarity for determining be converted into text and candidate answers;First output module is configured in response to determine similar
Degree is greater than default value, exports candidate answers as target answer.
In some embodiments, execution unit further include: the second input module is configured in response to determine similarity not
Greater than default value, using candidate answers as the first candidate answers, enters text into end-to-end model trained in advance, obtain
Second candidate answers, wherein end-to-end model is used to characterize the corresponding relationship of text and answer;Sorting module, be configured to by
First candidate answers and the second candidate answers are input to order models trained in advance, obtain ranking results, wherein order models
For being ranked up to candidate answers;Second output module is configured to determine target answer based on ranking results, exports target
Answer.
In some embodiments, execution unit is further configured to: in response to determining that goal systems behavior is that inquiry is known
Know map, knowledge point associated with target word information is retrieved from preset knowledge mapping, exports knowledge point information.
The third aspect, the embodiment of the present application provide a kind of one or more processors;Storage device is stored thereon with one
A or multiple programs, when one or more programs are executed by one or more processors, so that one or more processors are realized
The method of any embodiment in method such as handling information.
Fourth aspect, the embodiment of the present application provide a kind of computer-readable medium, are stored thereon with computer program, should
The method such as any embodiment in the method for handling information is realized when program is executed by processor.
Method and apparatus provided by the embodiments of the present application for handling information, by determining and obtaining text to be processed
Corresponding class of service is then based on class of service, determines the target word in text, is based on class of service, target word later
With acquired history target word, goal systems behavior, last performance objective system action, without manually formulating are determined
Session rules are matched, and can be performed the next step system action automatically based on class of service and target word, be improved to target
The flexibility of system action progress decision.
Detailed description of the invention
By reading a detailed description of non-restrictive embodiments in the light of the attached drawings below, the application's is other
Feature, objects and advantages will become more apparent upon:
Fig. 1 is that one embodiment of the application can be applied to exemplary system architecture figure therein;
Fig. 2 is the flow chart according to one embodiment of the method for handling information of the application;
Fig. 3 is the schematic diagram according to an application scenarios of the method for handling information of the application;
Fig. 4 is the flow chart according to another embodiment of the method for handling information of the application;
Fig. 5 is the structural schematic diagram according to one embodiment of the device for handling information of the application;
Fig. 6 is adapted for the structural schematic diagram for the computer system for realizing the server of 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, part relevant to related invention is illustrated only in attached drawing.
It should be noted that in the absence of conflict, the features in the embodiments and the embodiments of the present 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 is shown can be using the application for handling the method for information or the example of the device for handling information
Property system architecture 100.
As shown in Figure 1, system architecture 100 may include terminal device 101,102,103, network 104 and server 105.
Network 104 between terminal device 101,102,103 and server 105 to provide the medium of communication link.Network 104 can be with
It is somebody's turn to do including various connections, such as wired, wireless communication link or fiber optic cables etc..
User can be used terminal device 101,102,103 and be interacted by network 104 with server 105, to receive or send out
Send message etc..Various telecommunication customer end applications can be installed, such as interactive voice class is answered on terminal device 101,102,103
With, shopping class application, searching class application, instant messaging tools, mailbox client, social platform software etc..
Terminal device 101,102,103 can be hardware, be also possible to software.When terminal device 101,102,103 is hard
When part, the various electronic equipments of network communication, including but not limited to smart phone, plate are can be with display screen and supported
Computer, pocket computer on knee and desktop computer etc..When terminal device 101,102,103 is software, can install
In above-mentioned cited electronic equipment.Multiple softwares or software module may be implemented into (such as providing distributed clothes in it
Business), single software or software module also may be implemented into.It is not specifically limited herein.
Server 105 can be to provide the server of various services, such as to being installed on terminal device 101,102,103
Interactive voice class application provide support background server.Background server can carry out voice to the voice messaging received
The processing such as text conversion, and analyze etc. processing to the text after convert, generation processing result (such as determine target system
System behavior, and performance objective system action).
It should be noted that server can be hardware, it is also possible to software.When server is hardware, may be implemented
At the distributed server cluster that multiple servers form, individual server also may be implemented into.It, can when server is software
To be implemented as multiple softwares or software module (such as providing Distributed Services), single software or software also may be implemented into
Module.It is not specifically limited herein.
It should be noted that the method provided by the embodiment of the present application for handling information is generally held by server 105
Row, correspondingly, the device for handling information is generally positioned in server 105.
It should be understood that the number of terminal device, network and server in Fig. 1 is only schematical.According to realization need
It wants, can have any number of terminal device, network and server.
With continued reference to Fig. 2, the process of one embodiment of the method for handling information according to the application is shown
200.The method for being used to handle information, comprising the following steps:
Step 201, text to be processed is obtained, determines class of service corresponding with text.
It in the present embodiment, can be with for handling the executing subject (such as server 105 shown in FIG. 1) of the method for information
Text to be processed is obtained first.Wherein, above-mentioned text can be terminal used by a user (such as terminal shown in FIG. 1 set
Standby 101,102,103) above-mentioned executing subject is sent to by wired connection mode or radio connection.It may be noted that
Be, above-mentioned radio connection can include but is not limited to 3G/4G connection, WiFi connection, bluetooth connection, WiMAX connection,
Zigbee connection, UWB (ultra wideband) connection and other currently known or exploitation in the future radio connections.
After receiving text to be processed, above-mentioned executing subject can determine class of service corresponding with the text.
Above-mentioned class of service can include but is not limited to: product inquires class, after-sale service class, order inquiries class, preferential inquiry class, specific
The preferential inquiry class of product, page jump class, other classes.Herein, various modes be can use and determine class of service.As an example,
The mapping table of keyword and class of service can be previously stored in above-mentioned executing subject.Above-mentioned executing subject can be by institute
The text of acquisition is matched with the keyword in mapping table, and class of service corresponding to the keyword that will match to determines
For class of service corresponding with text to be processed.For example, class of service can if the keyword being matched to is " I wants to buy "
To be product inquiry class.If the keyword being matched to is " return of goods ", class of service can be after-sale service class.If being matched to
Keyword is " order ", then class of service can be order inquiries class.If the keyword being matched to is " preferential ", class of service
It can be preferential inquiry class, etc..
Step 202, it is based on class of service, determines the target word in text.
In the present embodiment, above-mentioned executing subject can determine the target word in text based on identified class of service.
Wherein, target word can be following one or more: product word, brand word, attribute word (such as price, memory, color etc.).It needs
It is noted that target word is also possible to other words, it is not limited to above-mentioned enumerate.Herein, it can use various modes to determine
State the target word in text.
In some optional implementations of the present embodiment, each class of service can correspond to a training in advance
Target word determines model.Wherein, above-mentioned target word determines that model is determined for the target word in text.Above-mentioned executing subject
Target word corresponding with identified class of service can be used and determine model, the target word in above-mentioned text is determined and
It extracts.It specifically may refer to following steps execution:
The first step extracts characteristic information from above-mentioned text.Wherein, features described above information can be by constituting above-mentioned text
Word word vector constitute word sequence vector;It is also possible to be made of the term vector to obtained word after above-mentioned text participle
Term vector sequence.In addition, features described above information can also include above-mentioned word sequence vector and above-mentioned term vector sequence simultaneously.This
Outside, features described above information can also include other features, such as by the context (example to obtained word after above-mentioned text participle
Such as the character string being made of the latter word of the previous word of the word, the word, the word) characteristic information (such as in context
Each word the sequence that is constituted of term vector) constitute contextual feature information sequence.
It should be noted that word vector can be intended to indicate that the vector of the feature of word.Every one-dimensional value of word vector can
There is certain feature that is semantic and grammatically explaining to represent one.Wherein, feature can be for the fundamental to word
The various information that (such as radical, radical, stroke, meaning etc.) is characterized.As an example, can be preparatory in above-mentioned executing subject
It stores 21886 included in Chinese Internal Code Specification (Chinese Internal Code Specification, GBK)
The mapping table of a Chinese character and graphical symbol and word vector, each word vector can have identical dimension.For above-mentioned text
Each of this word, above-mentioned executing subject can find word vector corresponding to the word from above-mentioned mapping table.This
The word vector of place, each word and graphical symbol can be the instruction for having supervision carried out using machine learning method to neural network
Practice and trains obtain or technical staff pre-set based on mass data statistics in advance.
It should be noted that term vector can be intended to indicate that the vector of the feature of word.Every one-dimensional value of term vector can
There is certain feature that is semantic and grammatically explaining to represent one.Feature herein can be for the fundamental to word
The various information that (such as meaning etc.) is characterized.It is each that above-mentioned executing subject can use various term vector generating modes generations
The term vector of entry, it is, for example, possible to use the generations of existing term vector Core Generator (such as word2vec etc.), or utilize instruction
The mode for practicing neural network generates.It is pointed out that above-mentioned word vector generation method and term vector generation method and similarity
Calculation method is the well-known technique studied and applied extensively at present, and details are not described herein.
Second step will be input to mesh corresponding with identified class of service by extracted characteristic information from above-mentioned text
Mark word determines model, obtains the target word in above-mentioned text.It should be noted that target word determines that model can be based on training
Sample, using machine learning method to existing model structure (such as shot and long term memory network (Long Short-Term
Memory, LSTM) or LSTM and CRF (Conditional Random Field algorithm, condition random field algorithm)
Model structure etc. after being combined) carry out Training after it is obtained.Herein, each training sample may include one
The mark of a text and the target word being used to indicate in the text.It can use machine learning method, by the text in training sample
This conduct input is trained used existing model structure, will train using the mark in training sample as output
The model afterwards is determined as target word and determines model.It should be noted that carrying out model training using machine learning method is mesh
The well-known technique of preceding extensive research and application, details are not described herein.
In some optional implementations of the present embodiment, above-mentioned executing subject can be first by identified service class
It is not matched with pre-set business classification.As an example, pre-set business classification may include product inquiry class, after-sale service class,
Order inquiries class, preferential inquiry class, the preferential inquiry class of specific products, page jump class.Above-mentioned executing subject is in response to determining industry
Business classification matches with pre-set business classification, can extract characteristic information from above-mentioned text first.Wherein, characteristic information can be with
Including word sequence vector, term vector sequence, contextual feature information sequence.Then, can will from text extracted feature
Information input to target word trained in advance determines model, obtains the target word in text, wherein above-mentioned target word determines model
The target word being determined in text.Target word herein determine model can be used for the text of various businesses classification into
Row target word is extracted and is determined.
In some optional implementations of the present embodiment, in response to the above-mentioned class of service of determination and pre-set business classification
(i.e. class of service is other classes removed other than pre-set business classification) is mismatched, above-mentioned executing subject can determine above-mentioned text
String length (i.e. the number of character).It is less than preset length (such as 4) in response to the above-mentioned string length of determination, can incites somebody to action
Above-mentioned text is matched with the preset target word in preset target set of words, by above-mentioned text and any preset target
The word that word matches is determined as the target word of above-mentioned text.Herein, preset target set of words can be based on to historical data
Statistics (such as counted according to the frequency of appearance) and summarize.
In some optional implementations of the present embodiment, above-mentioned executing subject can be by pre-stored dialog history
Status information, above-mentioned target word, above-mentioned class of service and the system action letter for being used to indicate a performed upper system action
Breath summarized, stored the information after summarizing as current dialogue states information, with to dialog history status information into
Row updates.Wherein, dialog history status information may include the information such as history target word, legacy system behavioural information.
In some optional implementations of the present embodiment, before determining target word, above-mentioned executing subject can be with
Above-mentioned text is handled.As an example, due to needing technical staff to carry out target word in the generating process of training sample
Mark, therefore, the data that can be marked in advance to technical staff are spot-check, and the mistake found is summarized as rule, benefit
Program is write with the rule.At this point it is possible to carry out data correcting to above-mentioned text using the program.As another example, above-mentioned
Executing subject can safeguard a common wrong word vocabulary.The common wrong word vocabulary includes common wrong word.Above-mentioned execution
Main body can use the common wrong word vocabulary and carry out wrong word identification to above-mentioned text, and wrong words is corrected as correct word
Word.As another example, due to needing technical staff to carry out the mark of target word, different skills in the generating process of training sample
The granularity that art personnel mark target word is different.For example, some technical staff will for text " I will buy the one-piece dress of pink colour "
" pink colour " is labeled as attribute word, and " pink colour " is labeled as attribute word by other technical staff.Therefore, journey can be write in advance
Sequence carries out unification using granularity of the program to the data that technical staff is marked.At this point, this can be used in above-mentioned executing subject
Program handles text to be processed.
Step 203, history target word is obtained, class of service, target word and history target word is based on, determines goal systems row
For.
In the present embodiment, the available history target word of above-mentioned executing subject.Wherein, above-mentioned history target word can be
The extracted target word from history text transmitted by above-mentioned user.Above-mentioned history text can be the text in preset duration
(such as within 2 minutes) are also possible to the text of preset times (such as nearly 3 times).Then, above-mentioned executing subject can be based on upper
Class of service, above-mentioned target word and above-mentioned history target word are stated, determines goal systems behavior.System action can be human-computer interaction
Operation performed by system, for example, feedback user a word, calling interface carry out information search, terminate dialogue, inquiry knowledge graph
Spectrum etc..Herein, various modes be can use and determine goal systems behavior.As an example, can be deposited in advance in above-mentioned executing subject
Contain the decision rule of technical staff's formulation to match from different classs of service.Above-mentioned executing subject can be based on corresponding
Decision rule, and the classification of above-mentioned target word and the classification of above-mentioned history target word are combined, determine goal systems behavior.
As an example, whether can determine first in above-mentioned target word and above-mentioned history target word comprising product word.If no
It include that can determine that goal systems behavior is feedback user a word, for inquiring product word required for user.Such as it " asks
What product of your shopping guide can be helped by asking? ".If comprising, can determine in above-mentioned target word and above-mentioned history target word whether include
Brand word.If can determine that system action calls searching interface to carry out information search comprising brand word.If not including brand word,
It can determine that goal systems behavior is feedback user a word, for inquiring brand required for user.Such as " it may I ask your needs
What brand? ".
It should be noted that the determination of other modes progress goal systems behavior can also be used.For example, can use dynamic
State planning mode solves optimal system behavior, and optimal system behavior is determined as goal systems behavior.
Step 204, performance objective system action.
In the present embodiment, above-mentioned executing subject can be with performance objective system action.For example, feedback user a word, tune
Information search is carried out with interface, terminates to talk with, inquire knowledge mapping etc..
With continued reference to the signal that Fig. 3, Fig. 3 are according to the application scenarios of the method for handling information of the present embodiment
Figure.In the application scenarios of Fig. 3, the interactive voice class application installed in user's using terminal equipment 301, with server 302
The voice interactive system run carries out human-computer interaction.User 301 typing of using terminal equipment first voice messaging.Terminal is set
After the voice messaging is converted to text by standby 301, the text (such as " I thinks bull's machine ") is had sent to server 302.Service
After device 302 gets the text, it is first determined class of service (such as product inquiry class) corresponding with above-mentioned text.Then base
In the class of service, the target word (such as product word " mobile phone ") in the text is determined.History target word is obtained later, and being based on should
Class of service, the target word and the history target word, determine goal systems behavior.Herein, it can determine that the target word is gone through with this
It whether include product word in history target word, if not including, it is determined that goal systems behavior is that feedback user in short (such as " is asked
Ask what brand you need ").Finally, executing the goal systems behavior.
The method provided by the above embodiment of the application, by determining service class corresponding with text to be processed is obtained
Not, it is then based on class of service, determines the target word in text, be based on above-mentioned class of service, above-mentioned target word later and is obtained
The history target word taken, determines goal systems behavior, finally executes above-mentioned goal systems behavior, without manually formulating pair
Words rule is matched, and can be performed the next step system action automatically based on class of service and target word, be improved to target system
The flexibility of system behavior progress decision.
With further reference to Fig. 4, it illustrates the processes 400 of another embodiment of the method for handling information.The use
In the process 400 of the method for processing information, comprising the following steps:
Step 401, the voice messaging of user is received.
It in the present embodiment, can be with for handling the executing subject (such as server 105 shown in FIG. 1) of the method for information
The voice messaging of above-mentioned user is received first.Herein, user can use terminal (such as terminal device shown in FIG. 1 101,
102,103) voice messaging is sent.
Step 402, speech text conversion is carried out to voice messaging, generates text to be processed.
In the present embodiment, above-mentioned executing subject can carry out speech text conversion to above-mentioned voice messaging, generate wait locate
The text of reason.Herein, various existing speech text conversion methods and tool be can use, voice messaging is converted to text.
Step 403, it enters text into business disaggregated model trained in advance, determines class of service corresponding with text.
In the present embodiment, above-mentioned executing subject can by above-mentioned text input to business disaggregated model trained in advance,
Determine class of service corresponding with above-mentioned text.Wherein, for characterizing, text is corresponding with class of service to close above-mentioned disaggregated model
System.As an example, above-mentioned disaggregated model can be technical staff counted based on mass data and pre-establish for characterizing text
The mapping table of this and the corresponding relationship of class of service.As another example, business disaggregated model can be based on training sample
This, using machine learning method to the existing model (such as convolutional neural networks (Convolutional with classification feature
Neural Network, CNN), support vector machines (Support Vector Machine, SVM) etc.) carry out Training after
It is obtained.Herein, each training sample may include text class of service corresponding with for characterizing the text
Mark.It can use machine learning method, using the text in training sample as input, using the mark in training sample as defeated
Out, the existing model with classification feature is trained, the model after training is determined as business disaggregated model.It needs
Illustrate, carrying out model training using machine learning method is the well-known technique studied and applied extensively at present, herein no longer
It repeats.
Step 404, in response to determining that class of service matches with pre-set business classification, characteristic information is extracted from text.
In the present embodiment, above-mentioned executing subject can first carry out identified class of service and pre-set business classification
Matching.As an example, pre-set business classification may include product inquiry class, after-sale service class, order inquiries class, preferential inquiry
The preferential inquiry class of class, specific products, page jump class.Above-mentioned executing subject is in response to determining class of service and pre-set business classification
Match, characteristic information can be extracted from above-mentioned text.Wherein, characteristic information may include word sequence vector, term vector sequence
Column, contextual feature information sequence.Herein, word sequence vector can be made of the word vector for constituting the word of text, term vector sequence
Column can be made of the term vector of word obtained after segmenting to text, and contextual feature information sequence can be by segmenting text
The characteristic information of the context of obtained word is constituted afterwards.
Step 405, characteristic information extracted from text is input to target word trained in advance and determines model, obtained
Target word in text.
In the present embodiment, above-mentioned executing subject can will be input to preparatory training by extracted characteristic information from text
Target word determine model, obtain the target word in text, wherein above-mentioned target word determines that model is determined in text
Target word.
Step 406, in response to determining that class of service and pre-set business classification mismatch, the string length of text is determined.
In the present embodiment, in response to the above-mentioned class of service of determination, (i.e. class of service is with pre-set business classification mismatch
Remove other classes other than pre-set business classification), above-mentioned executing subject can determine string length (the i.e. character of above-mentioned text
Number).
Step 407, in response to determine string length be less than preset length, by text with it is pre- in preset target set of words
It sets target word to be matched, the word that in text and any preset target word matches is determined as to the target word of text.
In the present embodiment, above-mentioned executing subject is less than preset length (such as 4) in response to the above-mentioned string length of determination,
Above-mentioned text can be matched with the preset target word in preset target set of words, by it is in above-mentioned text, with it is any pre-
Set the target word that the word that target word matches is determined as above-mentioned text.Herein, preset target set of words can be based on to history
The statistics (such as being counted according to the frequency of appearance) of data and summarize.
Step 408, using class of service, target word and history target word as input information, will input information input to pre-
First trained decision model, obtains goal systems behavior.
In the present embodiment, above-mentioned executing subject can be by above-mentioned class of service, above-mentioned target word and above-mentioned history target
Word obtains goal systems behavior by above-mentioned input information input to decision model trained in advance as input information.Wherein,
Above-mentioned decision model can be used for characterizing the corresponding relationship of input information and system action.As an example, above-mentioned decision model can
With technical staff is counted based on mass data and is formulated corresponding with the corresponding relationship of system action for characterizing input information
Relation table.As another example, decision model can be based on training sample, using machine learning method to the mind pre-established
Through network (for example, the three-layer neural network being made of input layer, hidden layer and output layer, wherein output layer can be used
Sigmoid function or softmax function etc.) carry out Training after it is obtained.Herein, each training sample can be with
It is marked including one group of input information (including class of service, target word, history target word) and system action.Training sample can be
It is obtained using passing through in data caused by the history interaction of user and system.Above-mentioned executing subject can use machine learning side
Method will enter information as input in training sample, using the mark in training sample as output, to above-mentioned neural network into
Row training, is determined as decision model for the model after training.It should be noted that carrying out model instruction using machine learning method
White silk is the well-known technique studied and applied extensively at present, and details are not described herein.
Step 409, performance objective system action.
In the present embodiment, above-mentioned executing subject can be with performance objective system action.For example, feedback user a word, tune
Information search is carried out with interface, terminates to talk with, inquire knowledge mapping etc..
It is feedback voice in response to the above-mentioned goal systems behavior of determination in some optional implementations of the present embodiment
Information, above-mentioned executing subject can retrieve the candidate answers to match with above-mentioned target word first from preset knowledge mapping.
Then, using various similarity calculating methods (such as Euclidean distance, cosine similarity etc.) determine be converted into text with it is upper
State the similarity of candidate answers.It is greater than default value in response to the above-mentioned similarity of determination, is answered using above-mentioned candidate answers as target
Case is exported.
In some optional implementations of the present embodiment, it is not more than above-mentioned present count in response to the above-mentioned similarity of determination
Value, above-mentioned executing subject can be first using above-mentioned candidate answers as the first candidate answers, by above-mentioned text input to preparatory instruction
Experienced end-to-end model (sequence to sequence, seq2seq), obtains the second candidate answers.Wherein, above-mentioned end-to-end
Model can be used for characterizing the corresponding relationship of text and answer.In practice, end-to-end model can by encoder (encoder) and
Decoder (decoder) two parts composition.Encoder can be the Recognition with Recurrent Neural Network (Recurrent an of several layers
Neural Network, RNN).Decoder can be one and the same or similar Recognition with Recurrent Neural Network network of coder structure.
Later, above-mentioned first candidate answers and above-mentioned second candidate answers can be input to order models trained in advance, arranged
Sequence result.Wherein, above-mentioned order models can be used for being ranked up candidate answers.In practice, base is can be used in order models
In the gradient boosted tree (Gradient Boosting Decison Tree, GBDT) that historical data is fitted in advance.Sort mould
Type can export the probability of each system action, and the probability of each system action is ranked up from big to small.Finally, can be with base
Target answer is determined in above-mentioned ranking results, exports above-mentioned target answer.For example, first answer of sorting is determined as target and answers
Case.
It is inquiry knowledge in response to the above-mentioned goal systems behavior of determination in some optional implementations of the present embodiment
Map, above-mentioned executing subject can retrieve knowledge point associated with above-mentioned target word information from preset knowledge mapping, defeated
Above-mentioned knowledge point information out.
Figure 4, it is seen that the method for handling information compared with the corresponding embodiment of Fig. 2, in the present embodiment
Process 400 the step of highlighting the determination step to target word and goal systems behavior is determined based on decision model.As a result,
The scheme of the present embodiment description, can also be for difference while improving the flexibility to goal systems behavior progress decision
Type of service determines target word using different modes, keeps identified target word more targeted.
With further reference to Fig. 5, as the realization to method shown in above-mentioned each figure, this application provides one kind for handling letter
One embodiment of the device of breath, the Installation practice is corresponding with embodiment of the method shown in Fig. 2, which can specifically answer
For in various electronic equipments.
As shown in figure 5, being used to handle the device 500 of information described in the present embodiment includes: acquiring unit 501, it is configured
At text to be processed is obtained, class of service corresponding with above-mentioned text is determined;First determination unit 502, is configured to base
In above-mentioned class of service, the target word in above-mentioned text is determined;Second determination unit 503, is configured to obtain history target word,
Based on above-mentioned class of service, above-mentioned target word and above-mentioned history target word, goal systems behavior is determined;Execution unit 504 is matched
It is set to and executes above-mentioned goal systems behavior.
In some optional implementations of the present embodiment, above-mentioned acquiring unit 501 may include receiving module and turn
Change the mold block (not shown).Wherein, above-mentioned receiving module may be configured to receive the voice messaging of above-mentioned user.Above-mentioned turn
Mold changing block may be configured to carry out speech text conversion to above-mentioned voice messaging, generate text to be processed.
In some optional implementations of the present embodiment, above-mentioned acquiring unit 501 can be further configured to: will
Above-mentioned text input determines class of service corresponding with above-mentioned text, wherein above-mentioned point to business disaggregated model trained in advance
Class model is used to characterize the corresponding relationship of text and class of service.
In some optional implementations of the present embodiment, above-mentioned first determination unit 502 may include extraction module
With the first input module (not shown).Wherein, said extracted module may be configured in response to the above-mentioned service class of determination
Do not match with pre-set business classification, extract characteristic information from above-mentioned text, wherein features described above information includes word vector sequence
Column, term vector sequence, contextual feature information sequence, above-mentioned word sequence vector is by constituting the word vector structure of the word of above-mentioned text
At above-mentioned term vector sequence is made of the term vector of obtained word after segmenting to above-mentioned text, above-mentioned contextual feature information
Sequence is made of the characteristic information to the context of obtained word after above-mentioned text participle.Above-mentioned first input module can be by
It is configured to for characteristic information extracted from above-mentioned text being input to target word trained in advance and determines model, obtain above-mentioned text
Target word in this, wherein above-mentioned target word determines model for determining the target word in text.
In some optional implementations of the present embodiment, above-mentioned first determination unit 502 may include first determining
Module and matching module (not shown).Wherein, above-mentioned first determining module may be configured in response to the above-mentioned industry of determination
Classification of being engaged in and pre-set business classification mismatch, and determine the string length of above-mentioned text.Above-mentioned matching module may be configured to
It is less than preset length in response to the above-mentioned string length of determination, by the preset target word in above-mentioned text and preset target set of words
It is matched, the word that in above-mentioned text and any preset target word matches is determined as to the target word of above-mentioned text.
In some optional implementations of the present embodiment, which can also include that storage unit (is not shown in figure
Out).Wherein, said memory cells may be configured to pre-stored dialog history status information, above-mentioned target word, above-mentioned
Class of service and the system action information for being used to indicate a performed upper system action are summarized, by the information after summarizing
It is stored as current dialogue states information, to be updated to dialog history status information.
In some optional implementations of the present embodiment, above-mentioned second determination unit 503 can be further configured
At using above-mentioned class of service, above-mentioned target word and above-mentioned history target word as input information, extremely by above-mentioned input information input
Trained decision model in advance, obtains goal systems behavior, wherein above-mentioned decision model is for characterizing input information and system row
For corresponding relationship.
In some optional implementations of the present embodiment, above-mentioned execution unit 504 may include retrieval module, second
Determining module and the first output module (not shown).Wherein, above-mentioned retrieval module may be configured in response in determination
Goal systems behavior is stated as feedback voice messaging, the candidate to match with above-mentioned target word is retrieved from preset knowledge mapping and answers
Case.Above-mentioned second determining module may be configured to determine the similarity of be converted into text and above-mentioned candidate answers.It is above-mentioned
First output module may be configured to be greater than default value in response to the above-mentioned similarity of determination, using above-mentioned candidate answers as mesh
Mark answer is exported.
In some optional implementations of the present embodiment, above-mentioned execution unit 504 can also include the second input mould
Block, sorting module and the second output module (not shown).Wherein, above-mentioned second input module may be configured in response to
Determine that above-mentioned similarity is not more than above-mentioned default value, it is using above-mentioned candidate answers as the first candidate answers, above-mentioned text is defeated
Enter to end-to-end model trained in advance, obtain the second candidate answers, wherein above-mentioned end-to-end model is for characterizing text and answering
The corresponding relationship of case.Above-mentioned sorting module may be configured to input above-mentioned first candidate answers and above-mentioned second candidate answers
To order models trained in advance, ranking results are obtained, wherein above-mentioned order models are for being ranked up candidate answers.On
It states the second output module and is configured to above-mentioned ranking results and determine target answer, export above-mentioned target answer.
In some optional implementations of the present embodiment, above-mentioned execution unit 504 can be further configured to ring
It, for inquiry knowledge mapping, should be retrieved from preset knowledge mapping related to above-mentioned target word in determining above-mentioned goal systems behavior
The knowledge point information of connection exports above-mentioned knowledge point information.
The device provided by the above embodiment of the application is determined by acquiring unit 501 and obtains text phase to be processed
Corresponding class of service, then the first determination unit 502 is based on class of service, determines the target word in text, and second really later
Order member 503 determines goal systems behavior, most based on above-mentioned class of service, above-mentioned target word and acquired history target word
Execution unit 504 executes above-mentioned goal systems behavior afterwards, is matched without manually formulating session rules, can be based on
Class of service and target word perform the next step system action automatically, improve the flexibility that decision is carried out to goal systems behavior.
Below with reference to Fig. 6, it illustrates the computer systems 600 for the server for being suitable for being used to realize the embodiment of the present application
Structural schematic diagram.Server shown in Fig. 6 is only an example, should not function and use scope band to the embodiment of the present application
Carry out any restrictions.
As shown in fig. 6, computer system 600 includes central processing unit (CPU) 601, it can be read-only according to being stored in
Program in memory (ROM) 602 or be loaded into the program in random access storage device (RAM) 603 from storage section 608 and
Execute various movements appropriate and processing.In RAM 603, also it is stored with system 600 and operates required various programs and data.
CPU 601, ROM 602 and RAM 603 are connected with each other by bus 604.Input/output (I/O) interface 605 is also connected to always
Line 604.
I/O interface 605 is connected to lower component: the importation 606 including keyboard, mouse etc.;It is penetrated including such as cathode
The output par, c 607 of spool (CRT), liquid crystal display (LCD) etc. and loudspeaker etc.;Storage section 608 including hard disk etc.;
And the communications portion 609 of the network interface card including LAN card, modem etc..Communications portion 609 via such as because
The network of spy's net executes communication process.Driver 610 is also connected to I/O interface 605 as needed.Detachable media 611, such as
Disk, CD, magneto-optic disk, semiconductor memory etc. are mounted on as needed on driver 610, in order to read from thereon
Computer program be mounted into storage section 608 as needed.
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 comprising be carried on computer-readable medium
On computer program, which includes the program code for method shown in execution flow chart.In such reality
It applies in example, which can be downloaded and installed from network by communications portion 609, and/or from detachable media
611 are mounted.When the computer program is executed by central processing unit (CPU) 601, limited in execution the present processes
Above-mentioned function.It should be noted that computer-readable medium described herein can be computer-readable signal media or
Computer readable storage medium either the two any combination.Computer readable storage medium for example can be --- but
Be not limited to --- electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor system, device or device, or any above combination.
The more specific example of computer readable storage medium can include but is not limited to: have one or more conducting wires electrical connection,
Portable computer diskette, hard disk, random access storage device (RAM), read-only memory (ROM), erasable type may be programmed read-only deposit
Reservoir (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), light storage device, magnetic memory
Part or above-mentioned any appropriate combination.In this application, computer readable storage medium, which can be, any include or stores
The tangible medium of program, the program can be commanded execution system, device or device use or in connection.And
In the application, computer-readable signal media may include in a base band or the data as the propagation of carrier wave a part are believed
Number, wherein carrying computer-readable program code.The data-signal of this propagation can take various forms, including but not
It is limited to electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be computer
Any computer-readable medium other than readable storage medium storing program for executing, the computer-readable medium can send, propagate or transmit use
In by the use of instruction execution system, device or device or program in connection.Include on computer-readable medium
Program code can transmit with any suitable medium, including but not limited to: wireless, electric wire, optical cable, RF etc., Huo Zheshang
Any appropriate combination stated.
Flow chart and block diagram in attached drawing are illustrated according to the system of the various embodiments of the application, method and computer journey
The architecture, function and operation in the cards of sequence product.In this regard, each box in flowchart or block diagram can generation
A part of one module, program segment or code of table, a part of the module, program segment or code include one or more use
The executable instruction of the logic function as defined in realizing.It should also be noted that in some implementations as replacements, being marked in box
The function of note can also occur in a different order than that indicated in the drawings.For example, two boxes succeedingly indicated are actually
It can be basically executed in parallel, they can also be executed in the opposite order sometimes, and this depends on the function involved.Also it to infuse
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 also can be set in the processor, for example, can be described as: a kind of processor packet
Include acquiring unit, the first determination unit, the second determination unit and execution unit.Wherein, the title of these units is in certain situation
Under do not constitute restriction to the unit itself, for example, acquiring unit is also described as " obtaining the list of text to be processed
Member ".
As on the other hand, present invention also provides a kind of computer-readable medium, which be can be
Included in device described in above-described embodiment;It is also possible to individualism, and without in the supplying device.Above-mentioned calculating
Machine readable medium carries one or more program, when said one or multiple programs are executed by the device, so that should
Device: obtaining text to be processed, determines class of service corresponding with the text;Based on the class of service, institute is determined
State the target word in text;History target word is obtained, the class of service, the target word and the history target word are based on,
Determine goal systems behavior;Execute the goal systems behavior.
Above description is only the preferred embodiment of the application and the explanation to institute's application technology principle.Those skilled in the art
Member is it should be appreciated that invention scope involved in the application, however it is not limited to technology made of the specific combination of above-mentioned technical characteristic
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
Any combination and the other technical solutions formed.Such as features described above has similar function with (but being not limited to) disclosed herein
Can technical characteristic replaced mutually and the technical solution that is formed.
Claims (13)
1. a kind of method for handling information, comprising:
Text to be processed is obtained, determines class of service corresponding with the text;
Based on the class of service, the target word in the text is determined;
History target word is obtained, the class of service, the target word and the history target word is based on, determines goal systems row
For;
Execute the goal systems behavior.
2. the method according to claim 1 for handling information, wherein described to obtain text to be processed, comprising:
Receive the voice messaging of the user;
Speech text conversion is carried out to the voice messaging, generates text to be processed.
3. the method according to claim 1 for handling information, wherein determination industry corresponding with the text
Business classification, comprising:
By the text input to business disaggregated model trained in advance, class of service corresponding with the text is determined, wherein
The disaggregated model is used to characterize the corresponding relationship of text and class of service.
4. the method according to claim 1 for handling information, wherein it is described to be based on the class of service, determine institute
State the target word in text, comprising:
Match in response to the determination class of service with pre-set business classification, extract characteristic information from the text, wherein
The characteristic information includes word sequence vector, term vector sequence, contextual feature information sequence, and the word sequence vector is by constituting
The word vector of the word of the text is constituted, and the term vector sequence is by the term vector structure to obtained word after text participle
At the contextual feature information sequence is made of the characteristic information of the context of obtained word after segmenting to the text;
Characteristic information extracted from the text is input to target word trained in advance and determines model, obtains the text
In target word, wherein the target word determines model for determining target word in text.
5. the method according to claim 1 for handling information, wherein it is described to be based on the class of service, determine institute
State the target word in text, comprising:
It is mismatched in response to the determination class of service and pre-set business classification, determines the string length of the text;
It is less than preset length in response to the determination string length, by the preset mesh in the text and preset target set of words
Mark word is matched, and the word that in the text and any preset target word matches is determined as to the target word of the text.
6. the method according to claim 1 for handling information, wherein the method also includes:
By pre-stored dialog history status information, the target word, the class of service and it is used to indicate performed upper
The system action information of one system action is summarized, and is deposited the information after summarizing as current dialogue states information
Storage, to be updated to dialog history status information.
7. the method according to claim 1 for handling information, wherein the acquisition history target word, based on described
Class of service, the target word and the history target word, determine goal systems behavior, comprising:
Using the class of service, the target word and the history target word as input information, by the input information input
To decision model trained in advance, goal systems behavior is obtained, wherein the decision model is for characterizing input information and system
The corresponding relationship of behavior.
8. the method according to claim 1 for handling information, wherein described to execute the goal systems behavior, packet
It includes:
It is feedback voice messaging in response to the determination goal systems behavior, retrieval and the target from preset knowledge mapping
The candidate answers that word matches;
Determine the similarity of be converted into text and the candidate answers;
It is greater than default value in response to the determination similarity, is exported the candidate answers as target answer.
9. the method according to claim 8 for handling information, wherein it is described to execute the goal systems behavior, also
Include:
It is not more than the default value in response to the determination similarity, it, will using the candidate answers as the first candidate answers
The text input obtains the second candidate answers to end-to-end model trained in advance, wherein the end-to-end model is used for table
It solicits articles the corresponding relationship of this and answer;
First candidate answers and second candidate answers are input to order models trained in advance, obtain sequence knot
Fruit, wherein the order models are for being ranked up candidate answers;
Target answer is determined based on the ranking results, exports the target answer.
10. the method according to claim 1 for handling information, wherein described to execute the goal systems behavior, packet
It includes:
It is inquiry knowledge mapping in response to the determination goal systems behavior, retrieval and the target from preset knowledge mapping
The associated knowledge point information of word, exports the knowledge point information.
11. a kind of for handling the device of information, comprising:
Acquiring unit is configured to obtain text to be processed, determines class of service corresponding with the text;
First determination unit is configured to determine the target word in the text based on the class of service;
Second determination unit is configured to obtain history target word, is based on the class of service, the target word and the history
Target word determines goal systems behavior;
Execution unit is configured to execute the goal systems behavior.
12. a kind of server, comprising:
One or more processors;
Storage device is stored thereon with one or more programs,
When one or more of programs are executed by one or more of processors, so that one or more of processors are real
The now method as described in any in claim 1-10.
13. a kind of computer-readable medium, is stored thereon with computer program, wherein the realization when program is executed by processor
Method as described in any in claim 1-10.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810630153.2A CN110019742B (en) | 2018-06-19 | 2018-06-19 | Method and device for processing information |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810630153.2A CN110019742B (en) | 2018-06-19 | 2018-06-19 | Method and device for processing information |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110019742A true CN110019742A (en) | 2019-07-16 |
CN110019742B CN110019742B (en) | 2024-05-17 |
Family
ID=67188315
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810630153.2A Active CN110019742B (en) | 2018-06-19 | 2018-06-19 | Method and device for processing information |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110019742B (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110874438A (en) * | 2019-11-19 | 2020-03-10 | 中国银行股份有限公司 | Function navigation method and device |
CN110909725A (en) * | 2019-10-18 | 2020-03-24 | 平安科技(深圳)有限公司 | Method, device and equipment for recognizing text and storage medium |
CN111199430A (en) * | 2020-01-08 | 2020-05-26 | 中信银行股份有限公司 | Service solution generation method and device and electronic equipment |
CN111783810A (en) * | 2019-09-24 | 2020-10-16 | 北京沃东天骏信息技术有限公司 | Method and apparatus for determining attribute information of user |
WO2021068493A1 (en) * | 2019-10-11 | 2021-04-15 | 京东数字科技控股有限公司 | Method and apparatus for processing information |
CN112908319A (en) * | 2019-12-04 | 2021-06-04 | 海信视像科技股份有限公司 | Method and equipment for processing information interaction |
CN113033170A (en) * | 2021-04-23 | 2021-06-25 | 中国平安人寿保险股份有限公司 | Table standardization processing method, device, equipment and storage medium |
CN114861680A (en) * | 2022-05-27 | 2022-08-05 | 马上消费金融股份有限公司 | Conversation processing method and device |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160063115A1 (en) * | 2014-08-27 | 2016-03-03 | Facebook, Inc. | Blending by Query Classification on Online Social Networks |
CN105893465A (en) * | 2016-03-28 | 2016-08-24 | 北京京东尚科信息技术有限公司 | Automatic question answering method and device |
CN107908803A (en) * | 2017-12-26 | 2018-04-13 | 上海智臻智能网络科技股份有限公司 | The response method and device, storage medium, terminal of question and answer interaction |
CN107958091A (en) * | 2017-12-28 | 2018-04-24 | 北京贝塔智投科技有限公司 | A kind of NLP artificial intelligence approaches and interactive system based on financial vertical knowledge mapping |
CN108022586A (en) * | 2017-11-30 | 2018-05-11 | 百度在线网络技术(北京)有限公司 | Method and apparatus for controlling the page |
CN108121824A (en) * | 2018-01-12 | 2018-06-05 | 北京融快线科技有限公司 | A kind of chat robots and system towards financial service |
-
2018
- 2018-06-19 CN CN201810630153.2A patent/CN110019742B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160063115A1 (en) * | 2014-08-27 | 2016-03-03 | Facebook, Inc. | Blending by Query Classification on Online Social Networks |
CN105893465A (en) * | 2016-03-28 | 2016-08-24 | 北京京东尚科信息技术有限公司 | Automatic question answering method and device |
CN108022586A (en) * | 2017-11-30 | 2018-05-11 | 百度在线网络技术(北京)有限公司 | Method and apparatus for controlling the page |
CN107908803A (en) * | 2017-12-26 | 2018-04-13 | 上海智臻智能网络科技股份有限公司 | The response method and device, storage medium, terminal of question and answer interaction |
CN107958091A (en) * | 2017-12-28 | 2018-04-24 | 北京贝塔智投科技有限公司 | A kind of NLP artificial intelligence approaches and interactive system based on financial vertical knowledge mapping |
CN108121824A (en) * | 2018-01-12 | 2018-06-05 | 北京融快线科技有限公司 | A kind of chat robots and system towards financial service |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111783810B (en) * | 2019-09-24 | 2023-12-08 | 北京沃东天骏信息技术有限公司 | Method and device for determining attribute information of user |
CN111783810A (en) * | 2019-09-24 | 2020-10-16 | 北京沃东天骏信息技术有限公司 | Method and apparatus for determining attribute information of user |
WO2021068493A1 (en) * | 2019-10-11 | 2021-04-15 | 京东数字科技控股有限公司 | Method and apparatus for processing information |
CN110909725B (en) * | 2019-10-18 | 2023-09-19 | 平安科技(深圳)有限公司 | Method, device, equipment and storage medium for recognizing text |
CN110909725A (en) * | 2019-10-18 | 2020-03-24 | 平安科技(深圳)有限公司 | Method, device and equipment for recognizing text and storage medium |
CN110874438B (en) * | 2019-11-19 | 2023-04-07 | 中国银行股份有限公司 | Function navigation method and device |
CN110874438A (en) * | 2019-11-19 | 2020-03-10 | 中国银行股份有限公司 | Function navigation method and device |
CN112908319B (en) * | 2019-12-04 | 2022-10-25 | 海信视像科技股份有限公司 | Method and equipment for processing information interaction |
CN112908319A (en) * | 2019-12-04 | 2021-06-04 | 海信视像科技股份有限公司 | Method and equipment for processing information interaction |
CN111199430A (en) * | 2020-01-08 | 2020-05-26 | 中信银行股份有限公司 | Service solution generation method and device and electronic equipment |
CN113033170A (en) * | 2021-04-23 | 2021-06-25 | 中国平安人寿保险股份有限公司 | Table standardization processing method, device, equipment and storage medium |
CN113033170B (en) * | 2021-04-23 | 2023-08-04 | 中国平安人寿保险股份有限公司 | Form standardization processing method, device, equipment and storage medium |
CN114861680A (en) * | 2022-05-27 | 2022-08-05 | 马上消费金融股份有限公司 | Conversation processing method and device |
CN114861680B (en) * | 2022-05-27 | 2023-07-25 | 马上消费金融股份有限公司 | Dialogue processing method and device |
Also Published As
Publication number | Publication date |
---|---|
CN110019742B (en) | 2024-05-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110019742A (en) | Method and apparatus for handling information | |
CN108305626A (en) | The sound control method and device of application program | |
CN109299458A (en) | Entity recognition method, device, equipment and storage medium | |
CN111428010B (en) | Man-machine intelligent question-answering method and device | |
CN106407178A (en) | Session abstract generation method and device | |
CN109190124B (en) | Method and apparatus for participle | |
US11580299B2 (en) | Corpus cleaning method and corpus entry system | |
US20200320978A1 (en) | System and Method For Training A Virtual Agent To Identify A User's Intent From A Conversation | |
CN108121800A (en) | Information generating method and device based on artificial intelligence | |
CN110046254A (en) | Method and apparatus for generating model | |
CN110347863A (en) | Talk about art recommended method and device and storage medium | |
CN108933730A (en) | Information-pushing method and device | |
CN109766418A (en) | Method and apparatus for output information | |
CN109739605A (en) | The method and apparatus for generating information | |
CN109299477A (en) | Method and apparatus for generating text header | |
CN109635094A (en) | Method and apparatus for generating answer | |
CN109325121A (en) | Method and apparatus for determining the keyword of text | |
CN109582954A (en) | Method and apparatus for output information | |
CN109858045A (en) | Machine translation method and device | |
CN110288995A (en) | Exchange method, device, storage medium and electronic equipment based on speech recognition | |
CN108280200A (en) | Method and apparatus for pushed information | |
CN109895781A (en) | Method for controlling a vehicle and device | |
CN110162675A (en) | Generation method, device, computer-readable medium and the electronic equipment of answer statement | |
CN109978465A (en) | Source of goods recommended method, device, electronic equipment, storage medium | |
CN109739969A (en) | Answer generation method and intelligent conversational system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |