CN108573046A - A kind of user instruction treatment method and device based on AI systems - Google Patents
A kind of user instruction treatment method and device based on AI systems Download PDFInfo
- Publication number
- CN108573046A CN108573046A CN201810349754.6A CN201810349754A CN108573046A CN 108573046 A CN108573046 A CN 108573046A CN 201810349754 A CN201810349754 A CN 201810349754A CN 108573046 A CN108573046 A CN 108573046A
- Authority
- CN
- China
- Prior art keywords
- agent
- user instruction
- reply
- systems
- instruction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/284—Lexical analysis, e.g. tokenisation or collocates
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
Abstract
A kind of user instruction treatment method based on AI systems, including receive the user instruction of input；The method further includes the first response steps, and first response steps include：Classifying step is based on grader, and semantic analysis is carried out to the user instruction, determines that user instruction is classified；First Agent matching steps, classify according to user instruction, in multiple Agent of AI systems, determine and highest first Agent of the user instruction matching degree；Be intended to matching step, by the semantic analysis to user instruction, determine under the first Agent with the intention corresponding to the user instruction；First return phase, by the first Agent based on the intention corresponding to the user instruction, is carried out first to the user instruction and replied by closed domain AI systems.The present processes and device can promote AI efficiency and accuracy rate to the Accurate classification of reception command content, Rapid matching corresponding function module.
Description
Technical field
This application involves artificial intelligence fields, more particularly to a kind of user instruction treatment method and dress based on AI systems
It sets.
Background technology
Scientific and technological high speed development it is present, various advanced, intelligent inventions emerge one after another, and people are constantly made a spurt by new things
It, is also constantly seeking new breakthrough.The appearance of artificial intelligence is undoubtedly the much progress in human history, it is greatly accelerated
The paces of development in science and technology change people's living habit, have liberated labour etc..AI (Artificial Intelligence,
Artificial intelligence) refer to the technology that the think of peace-keeping operations of the mankind are simulated using modernization instruments such as computers, with AI technologies
It increasingly improves, AI technologies have been applied to the various aspects of production and living.It is research, develops for simulating, extending and extend
Theory, a new technological sciences of method, technology and application system for the intelligence of people.
AI systems are applied to different scenes, for example, the intelligent robot equipped with AI systems, realizes that machine person to person is direct
Exchange, while natural language understanding is used for according to the Agent in robot different application environment setting AI systems, realized different
The dialogue of environment.In the prior art, closed domain AI systems are made of multiple Agent (natural language understanding module), and each Agent is mutual
Phase is independent, one closed domain of composition, responsible one kind function, an as function module, mutual indepedent between different Agent.When
When customizing a certain new function, need to re-define in new Agent in other Agent in Intent of defined mistake etc.
Hold, durability is poor between module, and can lead to additional design and training.Meanwhile system is excessively lengthy and jumbled, leads to efficiency
Decline.
Invention content
The application's aims to overcome that the above problem or solves or extenuate to solve the above problems at least partly.
According to the one side of the application, a kind of user instruction treatment method based on AI systems is provided, including receive
The user instruction of input；The method further includes the first response steps, and first response steps include：Classifying step is based on
Grader carries out semantic analysis to the user instruction, determines that user instruction is classified；First Agent matching steps, according to user
Instruction classification determines and highest first Agent of the user instruction matching degree in multiple Agent of AI systems；Intention
The intention in the first Agent and corresponding to the user instruction is determined by the semantic analysis to user instruction with step
(Intent)；First return phase is based on by the first Agent corresponding to the user instruction by closed domain AI systems
Intention, to the user instruction carry out first reply.Each Agent includes multiple Intent (mappings of language and behavior
Relationship), Entity (set of parameter in input content) and Context (context environmental).The instruction that user sends out is received,
The instruction that user sends out can be natural language, behavior or word etc.；User instruction is analyzed first, and the semanteme of instruction is carried out
Instruction is subdivided into different function module or function classification, as by analysis, the intention of analysis instruction to this instruction classification
Agent, the first Agent matching steps are classified according to user instruction and are found and the instruction in multiple Agent with the intention instructed
The highest Agent of matching degree is the first Agent, it is intended that matching step is found corresponding with instruction intention in the first Agent
It is intended to (Intent), according to corresponding to Intent in this first Agent, the Entity for including by the Intent and the instruction
Corresponding reply content is matched, closed domain AI systems provide the first reply according to corresponding Intent under the first Agent.Thus
Primary exchange is completed, in this process, the addition of classifying step makes before replying each instruction, to the instruction that receives
Intention is analyzed, and is found, is matched in existing multiple Agent in systems, finds the highest Agent of matching degree, and
And Intent corresponding with the intention in the Agent is found, each Agent can have the chance of recycling, complete to exchange
While there is no new Agent to generate, while the content that other Agent were defined will not be stored, to receiving command content
Accurate classification, Rapid matching corresponding function module promote AI efficiency and accuracy rate.
Above-mentioned user instruction treatment method further includes after first response steps：First recording step, storage described the
One reply and corresponding first Agent of first reply.
Above-mentioned user instruction treatment method, further include second response steps parallel with first response steps, compared with
Step and the second recording step, second response steps include：Second matching step, in the user instruction, that there are sessions is pre-
In the case of depositing Agent, the session is enabled to prestore Agent as the 2nd Agent；Second return phase, by closed domain AI systems,
If the 2nd Agent is identical as the first Agent, the first reply is provided；If the 2nd Agent and described first
Agent is different, and carrying out second to the user instruction by the 2nd Agent replys；The comparison step, to described first time
It is multiple and described second is replied and compared based on weight, select weight it is larger reply as the user instruction corresponding time
It is multiple；If replied without second, directly gives described first and reply.It is pre- if there is no session in above-mentioned second matching step
Agent is deposited, then the second matching step terminates, and system provides described first according to the first Agent and replys；Second record
Step, store the user instruction and the larger reply of the weight and it is described reply for Agent.System operation it
Afterwards, all there are one independent conversation histories for each equipment.Before not engaging in the dialogue, it is not present in the corresponding conversation history of equipment
Session prestores Agent, after only carrying out dialogue, matched Agent in this session can be stored in conversation history, as next time
The session of dialogue prestores Agent, can be called when next time, dialogue sent out request to improve accuracy.Each Agent
A corresponding output is replied.If there is no session to prestore Agent in conversation history, i.e. the first time request of the session, then should
Only there are one answers to reply for secondary session, and both the first of the first response steps was replied；If there are sessions to prestore in conversation history
Agent, session prestore Agent as the 2nd Agent, if the 2nd Agent and above-mentioned first Agent matching steps institute are matched
Agent is identical, then only there are one the first reply answer outputs for the secondary session；If there are sessions to prestore in conversation history
Agent, then it is the 2nd Agent to name the session Agent that prestores, and the 2nd Agent and the first Agent are namely walked with matching
The rapid matched Agent of institute is different, then the secondary session can provide the second reply, comparison step compares the first reply and second and replys
Weight, each reply will carry the scoring of weight, and weight scores ranging from 0-1, and score is bigger, and expression weight is higher.
Analysis by algorithm to input instruction vocabulary and semanteme, the two answers have corresponding scoring, and it is larger to export scoring
Answer, and record the corresponding Agent of the larger reply of weighted value.
Above-mentioned user instruction treatment method, the grader include segmenter as follows to the user instruction into
Row participle：Word-phrase separation step is based on preset vocabulary and vocabulary division rule, and word-phrase separation is carried out to the user instruction；
Screening and filtration step, it is based on word-phrase separation as a result, being screened and being filtered；Semantic analysis step, based on screening and filtering
As a result, carry out user instruction semantic analysis.Word-phrase separation step, preset vocabulary and vocabulary division rule are by general in industry
It is foundation all over generally acknowledged vocabulary；For example, using the equipment application of this system medically, preset vocabulary and vocabulary division rule
It is just divided according to medical industry vocabulary, using the equipment application of this system on cuisines, preset vocabulary and vocabulary divide
Rule is just according to just more cuisines trade division.Screening and filtration step, according to words classification results, by nonsensical or interference
Words does not have influential words to filter out instruction, then carries out semantic analysis to the rear instruction screened and filtered, makes analysis
As a result more accurate, keep the result of matching Agent more accurate.
Above-mentioned user instruction treatment method, in the word-phrase separation step, the word-phrase separation be based on acquiescence dictionary and/or
Customize dictionary；In the screening and filtration step, the screening process is carried out with reference to preset words blacklist.The acquiescence word
Library is generally acknowledged in the art common words, and the customization dictionary is that this field is of little use or no vocabulary；The black name
Refer to singly those it is nonsensical or interference words or on instruction do not have influential words.
According to further aspect of the application, a kind of user instruction processing unit based on AI systems is provided, including with
In the user instruction module for receiving input；Described device further includes the first responder module, and first responder module includes：Classification
Unit carries out semantic analysis for being based on grader to the user instruction, determines that user instruction is classified；First Agent is matched
Unit is classified according to user instruction, in multiple Agent of AI systems, is determined highest with the user instruction matching degree
First Agent；Be intended to matching unit, by the semantic analysis to user instruction, determine under the first Agent with the user
The corresponding intention (Intent) of instruction；First replys unit, and by closed domain AI systems, institute is based on by the first Agent
The intention corresponding to user instruction is stated, carrying out first to the user instruction replys.Each Agent includes multiple Intent
(mapping relations of language and behavior), Entity (set of parameter in input content) and Context (context environmental).It receives
The instruction sent out to user, the instruction that user sends out can be natural language, behavior or word etc.；Analysis user refers to first
It enables, the semanteme of instruction is analyzed, the intention of analysis instruction, to this instruction classification, instruction is subdivided into different functions
Module or function classification, as Agent, the first Agent matching units, according to user instruction classification, the intention instructed multiple
Found in Agent with the highest Agent of the instructions match degree, be the first Agent, it is intended that matching unit is found in the first Agent
Intent corresponding with instruction intention passes through the Intent and the instruction according to Intent is corresponded in this first Agent
Including Entity match corresponding reply content, closed domain AI systems provide the according to corresponding Intent under the first Agent
One replys.This completes primary exchanges, and in this process, the addition of classifying step makes before replying each instruction, right
The analysis of the intention of the instruction received is found in existing multiple Agent, is matched, find matching degree most in systems
High Agent, and Intent corresponding with the intention in the Agent is found, each Agent can have the machine of recycling
Meeting completes do not have new Agent to generate while exchange, while will not store the content that other Agent were defined, and docks
The Accurate classification of command content is received, Rapid matching corresponding function module promotes AI efficiency and accuracy rate.
Above-mentioned user instruction processing unit further includes after first responder module：First logging modle, for storing
State the first reply and corresponding first Agent of first reply.In this way in next time talks with, store herein
Agent is that session prestores Agent.
Above-mentioned user instruction processing unit, further include second responder module parallel with first responder module, compared with
Module and the second logging modle, second responder module include：Second matching unit, in the user instruction, that there are sessions is pre-
In the case of depositing Agent, the session is enabled to prestore Agent as the 2nd Agent；Second replys unit, by closed domain AI systems,
If the 2nd Agent is identical as the first Agent, the first reply is provided；If the 2nd Agent is different from the first Agent, by described
2nd Agent carries out second to the user instruction and replys.In above-mentioned second matching step unit, prestore if there is no session
Agent, then the second matching step terminate, system according to the first Agent provide it is described first reply；The comparison module is used
In to it is described first reply and it is described second reply compared based on weight, select weight it is larger reply as the user
Instruct corresponding reply；If replied without second, the first reply is directly given.Second logging modle is described for storing
User instruction and the larger reply of the weight and described reply corresponding Agent.
According to further aspect of the application, a kind of computer equipment is provided, including memory, processor and be stored in
In the memory and the computer program that can be run by the processor, wherein the processor execution computer journey
Method as described in any one of the above embodiments is realized when sequence.
According to further aspect of the application, provide a kind of computer readable storage medium, it is preferably non-volatile can
Storage medium is read, computer program is stored with, the computer program is realized when executed by the processor as any of the above-described
Method described in.
According to the accompanying drawings to the detailed description of the specific embodiment of the application, those skilled in the art will be more
Above-mentioned and other purposes, the advantages and features of the application are illustrated.
Description of the drawings
Some specific embodiments of the application are described in detail by way of example rather than limitation with reference to the accompanying drawings hereinafter.
Identical reference numeral denotes same or similar component or part in attached drawing.It should be appreciated by those skilled in the art that these
What attached drawing was not necessarily drawn to scale.In attached drawing：
Fig. 1 is the flow chart according to the user instruction treatment method of the AI systems of the application one embodiment；
Fig. 2 is the flow chart according to the user instruction treatment method of the AI systems of the application another embodiment；
Fig. 3 is the flow chart of sorting technique in Fig. 2；
Fig. 4 is the flow chart according to the user instruction treatment method of AI systems in the application another embodiment；
Fig. 5 is the schematic diagram according to the application user instruction processing unit of AI systems in a specific embodiment；
Fig. 6 is the schematic diagram of the first responder module shown in Fig. 5；
Fig. 7 is the schematic diagram of the second responder module shown in Fig. 5；
Fig. 8 is each component signal of the user instruction processing unit of AI systems in another specific embodiment according to the application
Figure；
Fig. 9 is the schematic diagram according to the computer equipment of the application one embodiment；
Figure 10 is the schematic diagram according to the computer readable storage medium of the application one embodiment.
Specific implementation mode
Fig. 1 is the flow chart according to the user instruction treatment method of the AI systems of the application one embodiment；Have at one
In the embodiment of body, the user instruction treatment method of the AI systems of the application includes the user instruction 100 for receiving input；Further include
First response steps 200, first response steps 200 include：
Classifying step 210 is based on grader, and semantic analysis is carried out to the user instruction, determines that user instruction is classified.
First Agent matching steps 220, classify according to user instruction, in multiple Agent of AI systems, determining and institute
State the first highest Agent of user instruction matching degree.
Be intended to matching step 230, by the semantic analysis to user instruction, determine under the first Agent with the use
The corresponding intention of family instruction.
It is right to be based on user instruction institute by closed domain AI systems by the first Agent for first return phase 240
The intention answered carries out first to the user instruction and replys.
The instruction that user sends out is received, the instruction that user sends out can be natural language, behavior or word etc.；First
User instruction is analyzed, the semanteme of instruction is analyzed, highest first Agent of intention matching degree corresponding with instruction is changed, point
Analysis instruction intention corresponding intention (Intent) in the first Agent is found to this instruction classification, according to this first
The intention of Agent, closed domain AI systems provide the first reply.This completes primary exchanges, in this process, classification step
Rapid addition makes when replying each instruction, reception each of instruct can existing Agent in searching system, matched,
Each Agent can have the chance of recycling, complete do not have new Agent to generate while exchange.
The flow chart of the user instruction treatment method of the AI systems of another embodiment of Fig. 2；
Classifying step 210 is based on grader, and semantic analysis is carried out to the user instruction, determines that user instruction is classified.
First Agent matching steps 220, classify according to user instruction, in multiple Agent of AI systems, determining and institute
State the first highest Agent of user instruction matching degree.
Be intended to matching step 230, by the semantic analysis to user instruction, determine under the first Agent with the use
The corresponding intention of family instruction.
It is right to be based on user instruction institute by closed domain AI systems by the first Agent for first return phase 240
The intention answered carries out first to the user instruction and replys.
First recording step 250, storage described first is replied and described first replys corresponding first Agent.
Fig. 3 is the flow chart of classifying step, and grader includes that segmenter as follows carries out the user instruction
Participle：
Word-phrase separation step 211 is based on preset vocabulary and vocabulary division rule, and words point is carried out to the user instruction
From.
Screening and filtration step 212, it is based on word-phrase separation as a result, being screened and being filtered.
Semantic analysis step 213, based on screening and filter as a result, carrying out the semantic analysis of user instruction.Word-phrase separation
It is foundation that step, preset vocabulary and vocabulary division rule, which have universally recognized vocabulary in industry,；In word-phrase separation step, words
Separation is based on acquiescence dictionary and/or customization dictionary；Acquiescence dictionary and customization dictionary are the existing dictionary of system in fact, give tacit consent to dictionary
It is divided according to the vocabulary that the industry is generally acknowledged with the classification for customizing dictionary, screening and filtration step, screening are filtered with filtered to language
Justice expression does not have effective words, screening process to be carried out with reference to preset words blacklist.For example, the equipment using this system is answered
Used in medically, preset vocabulary and vocabulary division rule are just divided according to medical industry vocabulary, using setting for this system
Standby to apply on cuisines, preset vocabulary and vocabulary division rule are just according to just more cuisines trade division.Screening and filtering step
Suddenly, according to words classification results, by nonsensical or interference words or influential words is not had to filter out instruction, then to sieve
The instruction of choosing and filtering carries out semantic analysis, is that analysis result is more accurate, keeps the result of matching Agent more accurate.
Fig. 4 is the flow chart according to the user instruction treatment method of AI systems in the application another embodiment；Another
In a specific embodiment, the user instruction treatment method of the AI systems of the application includes the user instruction 100 for receiving input；Also
Including the first response steps 200 and the second response steps 300, the first response steps 200 include：
Classifying step 210 is based on grader, and semantic analysis is carried out to the user instruction, determines that user instruction is classified.
First Agent matching steps 220, classify according to user instruction, in multiple Agent of AI systems, determining and institute
State the first highest Agent of user instruction matching degree.
Be intended to matching step 230, by the semantic analysis to user instruction, determine under the first Agent with the use
The corresponding intention of family instruction.
It is right to be based on user instruction institute by closed domain AI systems by the first Agent for first return phase 240
The intention answered carries out first to the user instruction and replys.Second response steps 300 include：
Second matching step 310 enables the session pre- in the case where the user instruction prestores Agent there are session
It is the 2nd Agent to deposit Agent；It prestores Agent if there is no session, then the second matching step terminates, and system is according to described
One Agent provides described first and replys.
Second return phase 320, if the 2nd Agent is identical as the first Agent, provides by closed domain AI systems
One replys；If the 2nd Agent is different from the first Agent, the user instruction is carried out second time by the 2nd Agent
It is multiple.
First response steps 200 obtain the first reply, if the second response steps 300 obtain the second reply, carry out
In next step, comparison step 400；If replied without second, the first reply is directly given, 400 comparison steps, to described the
One reply and it is described second reply compared based on weight, as it was noted above, selection weight it is larger reply as the use
Family instructs corresponding reply；If replied without second, directly replied with first to reply.
Second recording step 500, store the user instruction and the larger reply of the weight and it is described reply for
Agent.
It should be noted that the first Agent herein is used only to distinguish with " first " and " second " in the 2nd Agent
It names, not substantive meaning.
Fig. 5 is the schematic diagram according to the application user instruction processing unit of AI systems in a specific embodiment；
The user instruction processing unit 600 of AI systems include for receive input user instruction module 601, the first responder module 602,
Second responder module 603, comparison module 604 and the second logging modle 605 receive the user of input under specific works state
Instruction module 601 receives the instruction of user, which can be natural language, behavior or word etc., after receiving instruction
There are the first responder module 602, the analysis of the second responder module 603 to provide reply respectively, comparison module 604 is compared by weight, power
It weighs two and replys, two are replied and is scored, a big reply of fractional value is most total reply, and the second logging modle 605 records this
It is a to reply corresponding Agent.
Specific implementation process such as Fig. 6 of first responder module 700, the first responder module include taxon 701, first
Agent matching units 702 are intended to the reply unit 704 of matching unit 703 and first；The course of work of taxon 701 is as follows,
To the instruction word-phrase separation step first received, it is based on preset vocabulary and vocabulary division rule, in taxon 701
The participle principle that dictionary and customization dictionary are given tacit consent to there are two dictionary, respectively acquiescence dictionary and customization dictionary is prestored with this
The common recognition of industry technology personnel is classified, and words is screened after carrying out word-phrase separation to user instruction；Screening and filtering are to instruction
Expression does not have effective words, blacklist is added in these words, the frequent processed instruction of taxon is that device is recognizable
State.First Agent matching units 702, classify according to user instruction, and matching degree is found in multiple Agent of AI systems
That highest Agent, it is intended that matching unit 703 finds corresponding intention in the first Agent according to the intention of instruction, and first
The intention that unit 703 is closed the first Agent of AI network analyses is replied, the first reply is provided；First responder module completes one
Task.
Specific implementation process such as Fig. 7 of second responder module 800, the second responder module and the first responder module simultaneously into
Row, the second responder module include that the second matching unit 801 and second reply unit 802 receives under specific working condition
User instruction simultaneously in the second responder module 800, user instruction is matched by the second matching unit 801 first,
If user instruction is prestored there are session in the case of Agent, session is enabled to prestore Agent as the 2nd Agent；Above-mentioned second matching
It in step, prestores Agent if there is no session, then the second matching step terminates, and system provides institute according to the first Agent
State the first reply；First replys the reply for the dialogue；Second replys unit 802 as closed AI systems, passes through this system
Provide the second reply；If the 2nd Agent is identical as the first Agent, the first reply is provided by closed domain AI systems；Second
If Agent is different from the first Agent, the 2nd Agent is corresponding to be intended to carry out the second reply to the user instruction.
Fig. 8 is that the user instruction processing unit 900 of AI systems in one embodiment connects in specific implementation process
The user instruction module 901 for receiving input is responsible for receiving the instruction of user, and instruction can be language, word or behavior etc., receive
To instruction be carried out at the same time two processing procedures, the first responder module 910 and the second responder module 920, wherein the first response mould
Block 910 includes four units, and taxon 911 is responsible for classification, and assorting process is：User instruction is analyzed first, to instruction
Semanteme is analyzed, and to this instruction classification, has been prestored in taxon 911 there are two dictionary, has respectively been given tacit consent to word
The participle principle of library and customization dictionary, acquiescence dictionary and customization dictionary is using the common recognition of industry technical staff as foundation, to receiving
The user instruction module 901 of input is classified into every trade, and words is screened after carrying out word-phrase separation to user instruction；Screening and filtering pair
Blacklist is added in these words by the words that instruction expression does not act on, and the frequent processed instruction of taxon is that device can
The state of identification.First Agent matching units 912 are classified according to user instruction, in multiple Agent of AI systems, determine with
The first highest Agent of the user instruction matching degree；It is intended to matching step 913, by the semantic analysis to user instruction,
Determine under the first Agent with the intention corresponding to the user instruction；First replys unit 914 as closed AI systems point
The intention for analysing the first Agent provides the first reply；First responder module completes one action.At the same time, the second responder module
921 also match the instruction for receiving the user instruction module 901 inputted, the second matching unit 931, if user instruction is deposited
In the case where session prestores Agent, the session is enabled to prestore Agent as the 2nd Agent；If session is not present in user instruction
It prestores in the case of Agent, 931 process of the second matching unit terminates, and replys and is replied for the first Agent corresponding first；Second
The AI systems that unit 932 is closed domain are replied to provide by closed domain AI systems if the 2nd Agent is identical as the first Agent
Direct first replys；If the 2nd Agent is different from the first Agent, the is carried out to the user instruction by the 2nd Agent
Two reply；Comparison module 930, carries out weight comparison, the big reply of output weight fiducial value, each reply will carry one
Weight scores, and weight scoring ranging from 0-1, score is bigger, and expression weight is higher.The two answers have corresponding scoring, can be defeated
Go out higher answer of scoring, and record the larger reply of weighted value for Agent.It is replied and second time if there is first
It is multiple, weight comparison is carried out, if replied without second, directly gives the first reply, the second logging modle 940 records this time
Multiple corresponding Agent.The system of the application keeps the instruction received accurate when completing a dialogue by sort module
Corresponding Agent is assigned to, without scheduling the content of other Agent defined mistakes again, and additional set will not be led to
Meter and training, save the space of system, improve effect.
According to the accompanying drawings to the detailed description of the specific embodiment of the application, those skilled in the art will be more
Above-mentioned and other purposes, the advantages and features of the application are illustrated.
Application embodiment additionally provides a kind of computing device, and with reference to Fig. 9, which includes memory 1120, processing
Device 1110 and it is stored in the computer program that can be run in the memory 1120 and by the processor 1110, the computer journey
Sequence is stored in the space 1130 for program code in memory 1120, and the computer program by processor 1110 when being executed
It realizes for executing any one steps of a method in accordance with the invention 1131.
The embodiment of the present application also provides a kind of computer readable storage mediums.Referring to Fig.1 0, the computer-readable storage
Medium includes the storage unit for program code, which is provided with for executing steps of a method in accordance with the invention
Program 1131 ', the program are executed by processor.
The embodiment of the present application also provides a kind of computer program products including instruction.When the computer program product exists
When being run on computer so that computer executes steps of a method in accordance with the invention.
In the above-described embodiments, can come wholly or partly by software, hardware, firmware or its arbitrary combination real
It is existing.When implemented in software, it can entirely or partly realize in the form of a computer program product.The computer program
Product includes one or more computer instructions.When computer loads and executes the computer program instructions, whole or portion
Ground is divided to generate according to the flow or function described in the embodiment of the present application.The computer can be all-purpose computer, dedicated computing
Machine, computer network obtain other programmable devices.The computer instruction can be stored in computer readable storage medium
In, or from a computer readable storage medium to the transmission of another computer readable storage medium, for example, the computer
Instruction can pass through wired (such as coaxial cable, optical fiber, number from a web-site, computer, server or data center
User's line (DSL)) or wireless (such as infrared, wireless, microwave etc.) mode to another web-site, computer, server or
Data center is transmitted.The computer readable storage medium can be any usable medium that computer can access or
It is comprising data storage devices such as one or more usable mediums integrated server, data centers.The usable medium can be with
It is magnetic medium, (for example, floppy disk, hard disk, tape), optical medium (for example, DVD) or semiconductor medium (such as solid state disk
Solid State Disk (SSD)) etc..
Professional should further appreciate that, described in conjunction with the examples disclosed in the embodiments of the present disclosure
Unit and algorithm steps, can be realized with electronic hardware, computer software, or a combination of the two, hard in order to clearly demonstrate
The interchangeability of part and software generally describes each exemplary composition and step according to function in the above description.
These functions are implemented in hardware or software actually, depend on the specific application and design constraint of technical solution.
Professional technician can use different methods to achieve the described function each specific application, but this realization
It is not considered that exceeding scope of the present application.
One of ordinary skill in the art will appreciate that implement the method for the above embodiments be can be with
It is completed come instruction processing unit by program, the program can be stored in computer readable storage medium, and the storage is situated between
Matter is non-transitory (English：Non-transitory) medium, such as random access memory, read-only memory, flash
Device, hard disk, solid state disk, tape (English：Magnetic tape), floppy disk (English：Floppy disk), CD (English：
Optical disc) and its arbitrary combination.
The above, serial number involved in present specification and the content of sequence do not represent it is actual execute sequence, and do not have
The difference of priority and priority is used only to distinguish the description of different step, different original papers.
The preferable specific implementation mode of the above, only the application, but the protection domain of the application is not limited thereto,
Any one skilled in the art is in the technical scope that the application discloses, the change or replacement that can be readily occurred in,
It should all cover within the protection domain of the application.Therefore, the protection domain of the application should be with scope of the claims
Subject to.
Claims (10)
1. a kind of user instruction treatment method based on AI systems, including receive the user instruction of input；It is characterized in that,
The method further includes the first response steps, and first response steps include：
Classifying step is based on grader, and semantic analysis is carried out to the user instruction, determines that user instruction is classified；
First Agent matching steps, classify according to user instruction, and in multiple Agent of AI systems, determination refers to the user
Enable the first Agent that matching degree is highest；
Be intended to matching step, by the semantic analysis to user instruction, determine under the first Agent with the user instruction institute
Corresponding intention；
First return phase, by closed domain AI systems, by the first Agent based on the meaning corresponding to the user instruction
Figure carries out first to the user instruction and replys.
2. according to user instruction treatment method of the claim 1 based on AI systems, which is characterized in that after first response steps
Further include：
First recording step, storage described first is replied and described first replys corresponding first Agent.
3. the user instruction treatment method according to claim 1 based on AI systems, which is characterized in that further include with it is described
First response steps parallel the second response steps, comparison step and the second recording step, second response steps include：
Second matching step enables the session prestore Agent in the case where the user instruction prestores Agent there are session
For the 2nd Agent；
Second return phase provides if the 2nd Agent is identical as the first Agent by closed domain AI systems
One replys；If the 2nd Agent is different from the first Agent, the user instruction is carried out by the 2nd Agent
Second replys；
The comparison step is to reply described first to reply with described second based on weight compare, and selects weight larger
Reply as the corresponding reply of the user instruction；If replied without described second, directly gives described first and reply；
Second recording step, stores the user instruction and the larger reply of the weight and the reply is corresponding
Agent。
4. the user instruction treatment method according to claim 3 based on AI systems, which is characterized in that the grader packet
Segmenter is included as follows to segment the user instruction：
Word-phrase separation step is based on preset vocabulary and vocabulary division rule, and word-phrase separation is carried out to the user instruction；
Screening and filtration step, it is based on word-phrase separation as a result, being screened and being filtered；
Semantic analysis step, based on screening and filter as a result, carrying out the semantic analysis of user instruction.
5. the user instruction treatment method according to claim 4 based on AI systems, which is characterized in that
In the word-phrase separation step, the word-phrase separation is based on acquiescence dictionary and/or customization dictionary；
In the screening and filtration step, the screening process is carried out with reference to preset words blacklist.
6. a kind of user instruction processing unit based on AI systems includes the user instruction module for receiving input；Its feature
It is, described device further includes the first responder module, and first responder module includes：
Taxon carries out semantic analysis for being based on grader to the user instruction, determines that user instruction is classified；
First Agent matching units, classify according to user instruction, and in multiple Agent of AI systems, determination refers to the user
Enable the first Agent that matching degree is highest；
Be intended to matching unit, by the semantic analysis to user instruction, determine under the first Agent with the user instruction institute
Corresponding intention；
First replys unit, by closed domain AI systems, by the first Agent based on the meaning corresponding to the user instruction
Figure carries out first to the user instruction and replys.
7. according to user instruction processing unit of the claim 6 based on AI systems, which is characterized in that first responder module is also
Including：
First recording unit, for storing first reply and corresponding first Agent of first reply.
8. according to user instruction processing unit of the claim 6 based on AI systems, which is characterized in that further include being answered with described first
The second responder module, comparison module and the second logging modle of modular concurrent are answered, second responder module includes：
Second matching unit enables the session prestore Agent in the case where the user instruction prestores Agent there are session
For the 2nd Agent；
Second reply unit provides institute if the 2nd Agent is identical as the first Agent by closed domain AI systems
State the first reply；If the 2nd Agent is different from the first Agent, by the 2nd Agent to the user instruction
Second is carried out to reply；
The comparison module is used to reply described first and second reply compared based on weight, selects weight larger
Reply as the corresponding reply of the user instruction；If replied without described second, directly gives described first and reply；
Second logging modle is used to store the user instruction and the larger reply of the weight and the reply corresponds to
Agent.
9. a kind of computer equipment, including memory, processor and storage can be transported in the memory and by the processor
Capable computer program, wherein the processor is realized when executing the computer program such as any one of claim 1-5 institutes
The method stated.
10. a kind of computer readable storage medium, preferably non-volatile readable storage medium, are stored with computer journey
Sequence, the computer program realize the method as described in any one of claim 1-5 when executed by the processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810349754.6A CN108573046B (en) | 2018-04-18 | 2018-04-18 | User instruction processing method and device based on AI system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810349754.6A CN108573046B (en) | 2018-04-18 | 2018-04-18 | User instruction processing method and device based on AI system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108573046A true CN108573046A (en) | 2018-09-25 |
| CN108573046B CN108573046B (en) | 2021-06-29 |
Family
ID=63575075
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810349754.6A Expired - Fee Related CN108573046B (en) | 2018-04-18 | 2018-04-18 | User instruction processing method and device based on AI system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108573046B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101178705A (en) * | 2007-12-13 | 2008-05-14 | 中国电信股份有限公司 | Free-running speech comprehend method and man-machine interactive intelligent system |
| US20100257022A1 (en) * | 2009-04-07 | 2010-10-07 | Yahoo! Inc. | Finding Similar Campaigns for Internet Advertisement Targeting |
| CN103020047A (en) * | 2012-12-31 | 2013-04-03 | 威盛电子股份有限公司 | Method for revising voice response and natural language dialogue system |
| CN104598445A (en) * | 2013-11-01 | 2015-05-06 | 腾讯科技（深圳）有限公司 | Automatic question-answering system and method |
| US20150227579A1 (en) * | 2014-02-12 | 2015-08-13 | Tll, Llc | System and method for determining intents using social media data |
| CN106844400A (en) * | 2015-12-07 | 2017-06-13 | 南京中兴新软件有限责任公司 | Intelligent response method and device |
| CN107370649A (en) * | 2017-08-31 | 2017-11-21 | 广东美的制冷设备有限公司 | Household electric appliance control method, system, control terminal and storage medium |
| CN107741976A (en) * | 2017-10-16 | 2018-02-27 | 泰康保险集团股份有限公司 | Intelligent response method, apparatus, medium and electronic equipment |
-
2018
- 2018-04-18 CN CN201810349754.6A patent/CN108573046B/en not_active Expired - Fee Related
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101178705A (en) * | 2007-12-13 | 2008-05-14 | 中国电信股份有限公司 | Free-running speech comprehend method and man-machine interactive intelligent system |
| US20100257022A1 (en) * | 2009-04-07 | 2010-10-07 | Yahoo! Inc. | Finding Similar Campaigns for Internet Advertisement Targeting |
| CN103020047A (en) * | 2012-12-31 | 2013-04-03 | 威盛电子股份有限公司 | Method for revising voice response and natural language dialogue system |
| CN104598445A (en) * | 2013-11-01 | 2015-05-06 | 腾讯科技（深圳）有限公司 | Automatic question-answering system and method |
| US20150227579A1 (en) * | 2014-02-12 | 2015-08-13 | Tll, Llc | System and method for determining intents using social media data |
| CN106844400A (en) * | 2015-12-07 | 2017-06-13 | 南京中兴新软件有限责任公司 | Intelligent response method and device |
| CN107370649A (en) * | 2017-08-31 | 2017-11-21 | 广东美的制冷设备有限公司 | Household electric appliance control method, system, control terminal and storage medium |
| CN107741976A (en) * | 2017-10-16 | 2018-02-27 | 泰康保险集团股份有限公司 | Intelligent response method, apparatus, medium and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108573046B (en) | 2021-06-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20210049512A1 (en) | Explainers for machine learning classifiers | |
| ȚĂRANU | Data mining in healthcare: decision making and precision | |
| CN105701120B (en) | The method and apparatus for determining semantic matching degree | |
| Amin et al. | Comparison of different classification techniques using WEKA for hematological data | |
| JP2021510429A (en) | Machine learning to integrate knowledge and natural language processing | |
| Gupta et al. | Performance analysis of various data mining classification techniques on healthcare data | |
| CN108664897A (en) | Bank slip recognition method, apparatus and storage medium | |
| Alweshah | Firefly algorithm with artificial neural network for time series problems | |
| CN108960264A (en) | The training method and device of disaggregated model | |
| CN106485146B (en) | A kind of information processing method and server | |
| CN110929752B (en) | Grouping method based on knowledge driving and data driving and related equipment | |
| Rasheed et al. | Metagenomic taxonomic classification using extreme learning machines | |
| CN106164896A (en) | For finding multidimensional recursive learning process and the system of complicated two analogues or many analogues relation | |
| CN109983459A (en) | Method and apparatus for identifying the counting of the N-GRAM occurred in corpus | |
| CN107368526A (en) | A kind of data processing method and device | |
| Cerquitelli et al. | Data mining for better healthcare: A path towards automated data analysis? | |
| Ashoori et al. | Using clustering methods for identifying blood donors behavior | |
| Labib et al. | Data mining for cancer management in Egypt case study: childhood acute lymphoblastic leukemia | |
| CN108805152A (en) | A kind of scene classification method and device | |
| CN108573046A (en) | A kind of user instruction treatment method and device based on AI systems | |
| CN108009157A (en) | A kind of sentence classifying method and device | |
| Utomo et al. | Best-parameterized sigmoid elm for benign and malignant breast cancer detection | |
| CN112949685B (en) | Aortic dissection surgery risk prediction method based on lifting tree model | |
| CN111984762B (en) | Text classification method sensitive to attack resistance | |
| CN112487406B (en) | Network behavior analysis method based on machine learning |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210629 |