CN105095195B - Nan-machine interrogation's method and system of knowledge based collection of illustrative plates - Google Patents
Nan-machine interrogation's method and system of knowledge based collection of illustrative plates Download PDFInfo
- Publication number
- CN105095195B CN105095195B CN201510383452.7A CN201510383452A CN105095195B CN 105095195 B CN105095195 B CN 105095195B CN 201510383452 A CN201510383452 A CN 201510383452A CN 105095195 B CN105095195 B CN 105095195B
- Authority
- CN
- China
- Prior art keywords
- knowledge
- point
- entity
- knowledge point
- reasoning
- 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.)
- Active
Links
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention provides nan-machine interrogation's method and systems of knowledge based collection of illustrative plates.In nan-machine interrogation's method of the knowledge based collection of illustrative plates of the present invention, after receiving sentence input by user, by being segmented to sentence input by user；To carrying out entity extraction by the obtained word of participle；Using knowledge mapping knowledge reasoning is carried out to extracting obtained entity information by the entity；And feedback is provided to the sentence input by user according to the result of the knowledge reasoning, the accuracy for the answer that nan-machine interrogation is fed back can be improved, the problem of user is proposed targetedly is answered, to improve user satisfaction.
Description
Technical field
The present invention relates to the data processings suitable for nan-machine interrogation, more particularly to nan-machine interrogation's method of knowledge based collection of illustrative plates
And system.
Background technology
With the development of the Internet, applications, nan-machine interrogation is by many enterprises, public institution or functional government departments
System is introduced into its website or APP, with assist or instead of by manually back and forth multiplexing family consulting.
Keyword in the problem of existing nan-machine interrogation's system is mostly by extracting user is used as knowledge point, and in data
One-to-one entity mapping is carried out inside library to find out the respective items of knowledge point, then using respective items as answer feedback to use
Family.Existing nan-machine interrogation's system only establishes knowledge point on one-to-one relationship map net, between knowledge point and knowledge point
Contact is very weak, can not carry out knowledge-based reasoning, therefore it feeds back to the answer of user and is often inaccurate, or even lacks needle
To property, do not give a direct answer to a question.
Invention content
In view of above-mentioned, the knowledge reasoning of knowledge based collection of illustrative plates is introduced nan-machine interrogation by the present invention, can preferably carry out
The combing and foundation of knowledge are used so that machine understands the complicated representation of knowledge so as to targetedly accurate answer
The problem of family proposes.
According to an aspect of the invention, there is provided a kind of nan-machine interrogation's method of knowledge based collection of illustrative plates, including：It receives and uses
Family input sentence and the sentence is segmented；To carrying out entity extraction by the obtained word of participle；It utilizes
Knowledge mapping carries out knowledge reasoning to extracting obtained entity information by the entity；And according to the knowledge reasoning
As a result feedback is provided to the sentence input by user.
Nan-machine interrogation's method according to an embodiment of the invention, it is preferable that in the participle the step of, by point
Word algorithm cuts sentence input by user, retains the stop words of relation belonging to Feature Words, and removes remaining and deactivate
Word and redundancy.
Nan-machine interrogation's method according to an embodiment of the invention, it is preferable that in the step of entity extracts, profit
It is marked with the entity attribute of obtained word, wherein the entity attribute includes the part of speech of institute's predicate, institute's predicate
Product classification corresponding to dictionary paraphrase or institute's predicate.
Nan-machine interrogation's method according to an embodiment of the invention, it is preferable that the knowledge mapping include knowledge point with
And the relationship between knowledge point.
Nan-machine interrogation's method according to an embodiment of the invention, it is preferable that when including an entity information, in institute
In the step of stating knowledge reasoning, the knowledge point corresponding to the entity information is searched in the knowledge mapping；Described in acquisition
After knowledge point, the step of terminating the reasoning, and provide with corresponding to the knowledge point content information and the knowledge
Location information of the point in the knowledge mapping；And when traversing the knowledge mapping without finding the knowledge point, knot
The step of Shu Suoshu reasonings, and provide the feedback of no accordingly result.
Nan-machine interrogation's method according to an embodiment of the invention, it is preferable that when including a plurality of entity information, in institute
In the step of stating knowledge reasoning, one the first knowledge corresponded in the entity information a) is searched in the knowledge mapping
Point；B) by the out-degree point of first knowledge point, using first knowledge point with corresponding to another in the entity information
Relationship between one the second knowledge point searches second knowledge point in the knowledge mapping；C) above-mentioned b steps are repeated
Suddenly, until for lookup is completed, terminates the reasoning corresponding to whole knowledge points in the entity information the step of, and
It provides and the content information corresponding to the knowledge point；And d) repeat above-mentioned b step, when traverse the knowledge mapping without
When finding the knowledge point to be searched, the step of terminating the reasoning, and provide the feedback of no accordingly result.
According to another aspect of the present invention, a kind of nan-machine interrogation's system executing knowledge based collection of illustrative plates, institute are additionally provided
The system of stating includes：Word-dividing mode, for being segmented to sentence input by user；Entity abstraction module, for described to passing through
It segments obtained word and carries out entity extraction；Knowledge mapping module, for the relationship between stored knowledge point and knowledge point；Know
Reasoning module is known, for being pushed away using the knowledge mapping to extracting obtained entity information progress knowledge by the entity
Reason；And output module, for providing feedback to the sentence input by user according to the result of the knowledge reasoning.
Nan-machine interrogation's system according to an embodiment of the invention, it is preferable that the word-dividing mode passes through segmentation methods
Sentence input by user is cut, retain relation belonging to Feature Words stop words, and remove remaining stop words and
Redundancy.
Nan-machine interrogation's system according to an embodiment of the invention, it is preferable that the entity abstraction module utilizes gained
To the entity attribute of word be marked, wherein the entity attribute include the part of speech of institute's predicate, institute's predicate dictionary release
Product classification corresponding to justice or institute's predicate.
Nan-machine interrogation's system according to an embodiment of the invention, it is preferable that the knowledge mapping module is by being every
Independent in-degree point and out-degree point is arranged in a knowledge point, according to identical in-degree point and out-degree point establish the knowledge point it
Between classification knowledge relation, knowledge non-directed graph is established between the knowledge point, is come between stored knowledge point and knowledge point
Relationship.
Nan-machine interrogation's system according to an embodiment of the invention, it is preferable that described when including an entity information
Knowledge reasoning module is provided when obtaining the knowledge point corresponding to the entity information by being searched in the knowledge mapping
With the location information of content information and the knowledge point in the knowledge mapping corresponding to the knowledge point；And when time
When going through the knowledge mapping without finding the knowledge point, the feedback of no accordingly result is provided.
Nan-machine interrogation's system according to an embodiment of the invention, it is preferable that described when including a plurality of entity information
Knowledge reasoning module is searched in the knowledge mapping corresponding to one the first knowledge point in the entity information；Pass through institute
The out-degree point for stating the first knowledge point is known using first knowledge point with corresponding to another second in the entity information
Know the relationship between point, second knowledge point is searched in the knowledge mapping；When repeating the above steps, until for correspondence
When lookup is completed in whole knowledge points in the entity information, provide and the content information corresponding to the knowledge point；
And when traversing the knowledge mapping without finding the knowledge point to be searched, provide the feedback of no accordingly result.
According to another aspect of the present invention, a kind of nan-machine interrogation's system of knowledge based collection of illustrative plates, the system are provided
Including：Input module, for receiving sentence input by user；Word-dividing mode, for being segmented to sentence input by user；It is real
Body abstraction module, for carrying out entity extraction by the obtained word of participle；Knowledge mapping module is used for stored knowledge
Relationship between point and knowledge point；Knowledge reasoning module, for utilizing the knowledge mapping to extract institute to passing through the entity
Obtained entity information carries out knowledge reasoning；And output module, the result according to the knowledge reasoning is used for the user
The sentence of input provides feedback.
The present invention is realized and is proposed to user by introducing knowledge reasoning in nan-machine interrogation and combining data processing
The accurate analysis of problem, and the problem of user is proposed can be targetedly answered, to reach promotion user satisfaction
Effect.
Description of the drawings
Attached drawing illustrates the embodiment of the present invention, and is used to explain the principle of the present invention together with specification.In the accompanying drawings：
Fig. 1 is the exemplary plot of the overall process process of nan-machine interrogation's method of knowledge based collection of illustrative plates according to the present invention；
Fig. 2 is the example flow diagram of nan-machine interrogation's method of knowledge based collection of illustrative plates according to the present invention；
Fig. 3 is the exemplary relational graph of knowledge mapping according to the present invention；
Fig. 4 is the example block diagram of nan-machine interrogation's system of knowledge based collection of illustrative plates according to the present invention.
Specific implementation mode
The preferred realization method of the application is described in detail below in conjunction with the accompanying drawings.
For the convenience of explanation, language is inputted as user using " what difference iPhone and Samsung mobile phone have " first herein
The example of sentence.In existing nan-machine interrogation's system, common method is：The keyword in above-mentioned sentence, " apple are extracted first
Mobile phone/Samsung mobile phone/difference "；It introduces synonymous near synonym and generates corresponding retrieval type, " iPhone and Samsung mobile phones and is (no
With or difference or difference or null) "；Then existing database is retrieved.Thus given answer is often to include
The existing entry or article of the above keyword and its near synonym, although these entries or article relate to " iPhone " simultaneously
" Samsung mobile phone ", or " difference " of the two is also given, but search for " the apple hand of user can not be directed to
Machine " and " Samsung mobile phone " provide the comparison with targetedly functionally, but need user by read entry or article come
Find targetedly answer.
In contrast, nan-machine interrogation's method as given by the present invention, then can provide more problem input by user
Targetedly answer provides explanation below with reference to attached drawing.Fig. 1 is the nan-machine interrogation of knowledge based collection of illustrative plates according to the present invention
The exemplary plot of the overall process process of method.
According to the overall process process of nan-machine interrogation's method as shown in Figure 1, machine first receives language input by user
Sentence " what difference iPhone and Samsung mobile phone have "；And then machine passes through word segmentation processing to sentence, becomes " apple hand
Machine/and/Samsung mobile phone/have/what/different "；Then entity extraction is carried out to word, obtains " iPhone：Noun, apple move
Mobile phone/Samsung mobile phone：Noun, apple, mobile phone ", and " difference " relation belonging to word correspond to " comparison ".With existing skill
The simple retrieval database of art is different, and machine can load knowledge mapping in the present invention, and after successfully loading knowledge mapping,
Using " iPhone " and " Samsung mobile phone " the two knowledge points as knowledge entrance, knowledge reasoning is carried out by recurrence reasoning, with
Carry out entity lookup.If fail to find with " iPhone " and " Samsung mobile phone " corresponding knowledge point in knowledge mapping,
Then reasoning fails.When found in knowledge mapping corresponding to the functional attributes of knowledge point " iPhone " and " Samsung mobile phone " it
Afterwards, reasoning success, and the result for taking its differentiation is supplied to user as answer by machine.It is according to the present invention man-machine to ask
Method is answered, the case where for the answer given by the above problem compared with the prior art, it is no longer necessary to which the user of enquirement is by readding
It reads a large amount of related entries or article oneself finds answer, but can more directly provide more targetedly answer.
Below with reference to attached drawing, the specific process flow of nan-machine interrogation's method according to the present invention is specifically described.Fig. 2 is root
According to the example flow diagram of nan-machine interrogation's method of the knowledge based collection of illustrative plates of one embodiment of the present of invention.As shown in Fig. 2, this is man-machine
Answering method includes step：ST10 receives sentence input by user and is segmented to sentence input by user；ST20 is right
Entity extraction is carried out by the obtained word of participle；ST30, using knowledge mapping to by obtained by entity extraction
Entity information carry out knowledge reasoning；And ST40, the sentence input by user is given according to the result of the knowledge reasoning
Go out feedback.
Participle technique is carried out to the sentence of input and belongs to natural language processing technique field, for a sentence, Ren Leike
Judge which is notional word with the knowledge by oneself, which is function word, and then understands its meaning, for machine processing then
It is segmented.In the step ST10 of participle, sentence input by user is cut by segmentation methods, reservation belongs to
The stop words of relationship characteristic word, and remove remaining stop words and redundancy.In the present invention, simultaneously for segmentation methods
Special limitation is not done, and the existing segmenting method based on string matching, segmenting method and base based on understanding may be used
In the various specific segmentation methods of the segmenting method of statistics.
Specifically, by taking " whom the author of master of slamdunking is " as an example, can become after participle " master of slamdunking// make
Person/be/who ".In order to save memory space and improve treatment effeciency those practical significances can be further removed after participle not
Big word or word, that is, retain the stop words of relation belonging to Feature Words, and removes remaining stop words and redundancy.For
Above-mentioned example sentence, wherein " " and "Yes" belong to stop words, and " " then belongs to redundancy as auxiliary words of mood, wherein
Although "Yes" was a stop words originally, the meaning that there is relationship to be directed toward for it herein is removing so that can be retained
After stop words and redundancy, " master of slamdunking/author/be/who " sentence become.
For stop words, on the basis of having been provided with available " deactivated vocabulary " in the prior art, the present invention is also into one
Step is provided with " relationship deactivates vocabulary ", and a series of stop words of relation belonging to Feature Words is stored in " relationship deactivates vocabulary ",
Such as, "Yes", " having ", " for " etc..In the processing for being removed stop words, by being carried out to " relationship deactivates vocabulary "
The stop words of relation belonging to Feature Words is searched and retained, can ensure the fine granularity and accuracy of word.
In the step ST20 that the entity extracts, it is marked using the entity attribute of obtained word, wherein
The entity attribute includes the product classification corresponding to the part of speech of institute's predicate, the dictionary paraphrase of institute's predicate or institute's predicate.In this hair
In bright, the algorithm extracted for entity does not do special limitation, existing various specific segmentation methods may be used, as long as its energy
Enough realize carries out entity replacement to word, judge which be name, which be commodity, which be number etc. basic handling.It is preferred that
Ground carries out entity extraction using condition random field in the present invention, to carry out including Entity recognition, part-of-speech tagging to word
Deng entity replace.
Still by taking " whom the author of master of slamdunking is " as an example, become after word segmentation processing " master of slamdunking/author/is/
Who ", when entity extracts, because " master of slamdunking " word had both corresponded to the books class product as caricature, also correspond to conduct
The audio-visual class product of animation, can " master of slamdunking " be labeled as caricature and animation, and due in the sentence of above-mentioned input simultaneously
There is no specific determiner, so the result of this 2 kinds of entities mark can be retained；" author " word as " master of slamdunking " itself
The title of attribute can't be extracted, but can be used as relative in reasoning；" who " is personal pronoun, can be extracted and with
" pronoun, personage " marks；Accordingly, for above-mentioned example, the result that entity extracts is " master of slamdunking：Caricature/animation, who：
Pronoun.For including the situation of determiner, such as " whom author of caricature master of slamdunking is " in sentence, extracted in entity
Afterwards, because in sentence including the modification of " caricature " this entity, the attribute of " master of slamdunking " as caricature can only be retained,
Therefore " master of slamdunking " can be labeled as caricature, and the result that the entity of this sentence extracts is correspondingly " master of slamdunking：Caricature, who：
Pronoun, personage ".
After completing entity and extracting, the simple mapping table being different between knowledge point in the prior art, the present invention will
Knowledge reasoning is carried out to obtained entity information using knowledge mapping.Have available knowledge mapping in the prior art, than
The Freebase of such as Google.In the present invention, the knowledge mapping of the relationship between knowledge point and knowledge point is established, it is excellent
The product or service that selection of land is provided using website or APP establish knowledge mapping as knowledge point.The knowledge mapping includes knowing
Know the relationship between point and knowledge point, wherein each knowledge point is provided with independent in-degree point and out-degree point, Yi Jisuo
It is knowing by the classification established between the knowledge point according to identical in-degree point and out-degree point to state the relationship between knowledge point
Knowledge relationship is established to establish knowledge non-directed graph between the knowledge point.
For the convenience of explanation, as shown in figure 3, giving the exemplary relational graph of knowledge mapping by taking football as an example,
Wherein " sport ", " football ", " Mei Xi ", " Barcelona ", " Real Madrid ", " Mei Xi ", " sieve C ", " Spain " are knowledge
Point, these knowledge points pass through the relatives such as " movement ", " football player ", " football club ", " effect ", " position/subordinate "
Connection and form knowledge non-directed graph.
When establishing knowledge mapping, the entity attribute of each knowledge point can be labeled, such as front example
" master of slamdunking " in son, its entity attribute will include caricature, animation, deliver time, author, producer, price etc.；Again
Such as " Mei Xi " in follow-up example, its attribute has personage, football player, gender, effect club etc., so as to machine
Device, which can recognize that, to be come, to go after searching knowledge mapping to can be derived that desired content information by entity.
It is not a kind of speech network of simple static state, it is preferable that can be to it for established knowledge mapping
The entity attribute of middle knowledge point is modified and is supplemented.For example, " Mei Xi " in above-mentioned example, when establishing knowledge mapping, meeting
Its " effect club " is labeled as " Barcelona " according to situation at that time, in order to which machine is subsequently with the feelings of the change of situation
It remains able to provide correct information under condition, the option of modification is remained with for " effect club ", and for " football
Member " and " gender " be " man " etc. these usual not attributes of malleable, then are not necessarily it and retain the option changed, with reduction
System complexity.Furthermore it is preferred that for established knowledge mapping, knowledge point can be carried out to increase newly and to wherein
The entity attribute of knowledge point is supplemented.For example, " Wuwei " word, if only based on currently provided GT grand touring product or
It services to its entity attribute labeling place name, tourism/ticketing service, can be carried when electric business enterprise increases books or animation product
When for product about " Yu Yu Hakusho ", even if user has input the sentence including " Wuwei " and " caricature ", but because know
The information about " Yu Yu Hakusho " can not be accurately obtained by knowing reasoning, and cannot provide correct result, and then may lose quotient
The conclusion of the business chance of industry dealing.In the newly-increased and supplement of entity attribute of knowledge point, in order to reduce system complexity and
Reduce storage load, be not that corresponding knowledge point or entity attribute are all increased to whole neologisms, it is preferable that based on website or
In APP increase newly product or service come increase newly knowledge point or supplement entity attribute.
The present invention knowledge mapping in, there are independent in-degree point and out-degree point in each knowledge point, according to it is identical enter
Degree point and out-degree point set up the knowledge relation of level-one, two level, three-level, finally set up a knowledge non-directed graph.Referring again to
For knowledge mapping illustrated in fig. 3, wherein " sport " is the knowledge point of level-one, " football " is the knowledge point of two level, " Mei Xi ",
" Barcelona ", " sieve C ", " Real Madrid " are the knowledge of three-level, and the knowledge of three-level allows have other high level knowledge
In-degree point is done, for example, " Mei Xi " is personage, " Barcelona " is club etc..Based on exemplary knowledge mapping, Ke Yiyou
" Mei Xi " infer he be " sport " field personage either " football " field personage, can be with by the relationship between peer
It infers " Mei Xi " to serve " Barcelona ", and " Mei Xi " and " Real Madrid " currently without intersection etc..
In the knowledge mapping of the present invention, which is accessed by the in-degree point of the different stage of each knowledge point,
Knowledge point at the same level is accessed in the out-degree point by the knowledge point.Line between knowledge point embodies between each knowledge point
Relationship, for example " Li Yuan " can be connected with " Li Shih-min " by " father and son " relationship, and " master of slamdunking " and " Takehiko Inoue " meeting
It is connected by " author " relationship.
In the step of using knowledge mapping to extracting obtained entity information progress knowledge reasoning by the entity
In ST30, it is preferable that when including an entity information, knowing corresponding to the entity information is searched in the knowledge mapping
Know point；After obtaining the knowledge point, the step of terminating the reasoning, and provide and the content corresponding to the knowledge point is believed
Breath, and in order to quick-searching to the knowledge point, provide location information of the knowledge point in the knowledge mapping；
And when traversing the knowledge mapping without finding the knowledge point, the step of terminating the reasoning, and provide without corresponding
As a result feedback.
Still by taking knowledge mapping illustrated in fig. 3 as an example, such as sentence input by user is " whom Mei Xi is ", at participle
Can become after reason " Mei Xi/be/who ", it can become " Mei Xi after entity extraction processing：Personage, football player, who：Pronoun, people
Object ".In knowledge reasoning, the knowledge point corresponding to entity information " Mei Xi " is found in knowledge mapping, and is provided and corresponded to " people
The information " football player " of object ", reasoning terminates.Sentence for example input by user is " whom Marcos Soares are " again, by participle
Can become after processing " Marcos Soares/be/who ", it can not be found in current knowledge collection of illustrative plates corresponding to entity information " Marcos Soares "
Knowledge point, therefore reasoning terminates, and provides the feedback of " no accordingly result ".
In the step of using knowledge mapping to extracting obtained entity information progress knowledge reasoning by the entity
In ST30, it is preferable that when including a plurality of entity information, a) search and correspond in the entity information in the knowledge mapping
One the first knowledge point；B) by the out-degree point of first knowledge point, using first knowledge point and correspond to institute
The relationship between the second knowledge point of another in entity information is stated, second knowledge is searched in the knowledge mapping
Point；C) above-mentioned b step is repeated, until for lookup is completed corresponding to whole knowledge points in the entity information, terminating
The step of reasoning, and provide and the content information corresponding to the knowledge point；And above-mentioned b step d) is repeated, work as traversal
When the knowledge mapping is without finding the knowledge point to be searched, the step of terminating the reasoning, and provide no accordingly result
Feedback.
Still by taking knowledge mapping illustrated in fig. 3 as an example, such as sentence input by user is " club where plum west is ",
It can become after word segmentation processing " Mei Xi/place/club/is ", can become " Mei Xi after entity extraction processing：Personage, football
Sportsman, club：Football, group, place ".In knowledge reasoning, found in knowledge mapping corresponding to entity information " plum
West " knowledge point, by the point using " Mei Xi " as out-degree, using " effect " as relative, can find " Barcelona " this
Thus one knowledge point obtains " club where plum west is Barcelona ", thus reasoning terminates as in-degree point.For aforementioned
" whom the author of master of slamdunking is " example, in knowledge reasoning, found in knowledge mapping corresponding to entity information " fill
It finds by the point using " master of slamdunking " as out-degree and is marked as closing therewith in the knowledge point of " personage " in the knowledge point of basket master-hand "
System be " author " correspondence knowledge point as in-degree point, thus reasoning terminates.
For problem " club where plum west is ", inventor does not use the search of knowledge mapping technology to draw existing
The page input above problem is held up to make comparisons with the method for the present invention.The result that existing search engine provides includes by complete
Character match and the sentence found in existing database, which includes " which the club where plum west is " etc. is similar
Problem, and answer still need to user by consult corresponding web page come find and result in also include about " Mei Xi " this word
Information corresponding to item, and specific specific aim answer corresponds to word by reading there is still a need for user and finds to obtain.In contrast,
It is then to give specific answer to enquirement according to the method for the present invention, this for a user can be more intuitive.
Sentence for example input by user is " club where Marcos Soares is " again, can be become after word segmentation processing
" Marcos Soares/place/club/are " can not be found corresponding to entity information in the knowledge mapping shown in current Fig. 3
The knowledge point of " Marcos Soares ", therefore reasoning terminates, and provide the feedback of " no accordingly result ".
For another example, by taking " why more expensive than common keyboard mechanical keyboard is " as an example, inventor does not use knowledge mapping existing
The interaction page of nan-machine interrogation's system of technology has input the above problem to make comparisons with the method for the present invention.It is existing man-machine to ask
The result that the system of answering provides is, for example, that " because the reasons such as supplier's difference, and different businessman's promotion influence, commodity price may deposit
In difference ", this gives general pervasive explanation only in " expensive ", and there is no compare for two input by user
Object is provided and any is targetedly compared.
For the above problem input by user " why more expensive than common keyboard mechanical keyboard is ", according to the method for the present invention can
Word segmentation processing is carried out to it first, can become after word segmentation processing " mechanical keyboard/why/ratio/common keyboard/expensive ", it is real
It can become " mechanical keyboard after body extraction processing：Noun, input product, common keyboard：Noun, input product ".Have herein
The relatival keyword of conduct of " expensive " and " than " can be found in knowledge mapping corresponding to entity information in knowledge reasoning
The knowledge point of " mechanical keyboard " and " common keyboard/membrane keyboard " carries out difference ratio by the cost of the attribute to belonging to the two
The comparison result for relatively comparing to do attribute, and obtaining the materials about such as product, service life, user experience etc. is used as
Answer.
Fig. 4 is the example block diagram of nan-machine interrogation's system of knowledge based collection of illustrative plates according to the present invention, shown in people
Machine question answering system 100 includes：Input module 10, for receiving sentence input by user；Word-dividing mode 20, for being inputted to user
Sentence segmented；Entity abstraction module 30, for carrying out entity extraction by the obtained word of participle；Knowledge graph
Compose module 40, for store include relationship between knowledge point and knowledge point knowledge mapping；Knowledge reasoning module 50, is used for
Using the knowledge mapping knowledge reasoning is carried out to extracting obtained entity information by the entity；And output module
60, for providing feedback to the sentence input by user according to the result of the knowledge reasoning.
Preferably, in one embodiment, input module 10 can be received with words input, voice input, and/or ability
The sentence input by user of other methods typing known to domain.
Preferably, in one embodiment, word-dividing mode 20 cuts sentence input by user by segmentation methods,
Retain the stop words of relation belonging to Feature Words, and removes remaining stop words and redundancy.
Preferably, in one embodiment, entity abstraction module 30 carries out it using the entity attribute of obtained word
Label, wherein the entity attribute includes the product corresponding to the part of speech of institute's predicate, the dictionary paraphrase of institute's predicate or institute's predicate
Classification.
Preferably, in one embodiment, knowledge mapping module 40 by for each knowledge point be arranged it is independent enter
Degree point and out-degree point, the knowledge relation of the classification between the knowledge point are established according to identical in-degree point and out-degree point, in institute
It states and establishes knowledge non-directed graph between knowledge point, the relationship between knowledge point and knowledge point to store the knowledge mapping.
Preferably, in one embodiment, knowledge reasoning module 50 is worked as and is obtained by being searched in the knowledge mapping
Corresponding to the entity information knowledge point when, provide with corresponding to the knowledge point content information and the knowledge point exist
Location information in the knowledge mapping；And when traversing the knowledge mapping without finding the knowledge point, provide nothing
The feedback of accordingly result.
Preferably, in one embodiment, knowledge reasoning module 50 is searched in the knowledge mapping corresponds to the reality
One the first knowledge point in body information；By the out-degree point of first knowledge point, using first knowledge point with it is right
The relationship between another the second knowledge point in entity information described in Ying Yu, searches described second in the knowledge mapping
Knowledge point；When repeating the above steps, when for lookup is completed corresponding to whole knowledge points in the entity information,
It provides and the content information corresponding to the knowledge point；And when the traversal knowledge mapping is without finding the knowledge to be searched
When point, the feedback of no accordingly result is provided.
Preferably, in one embodiment, output module 60 can in a visual manner, audible mode, and/or this field
The other modes known provide the respective feedback to user's read statement.
Nan-machine interrogation's method and system present invention as described above, can be applied to the automatic answering system such as JIMI,
It can be embedded in search engine, to carry out the knowledge reasoning of knowledge based collection of illustrative plates to problem input by user, make knowledge reasoning
Depth and range greatly improve, and then provide more targeted answer as feedback.
The basic principle that the present invention is described above in association with specific embodiment, however, it is desirable to, it is noted that this field
For those of ordinary skill, it is to be understood that the whole either any steps or component of the process and apparatus of the present invention, Ke Yi
Any computing device (including processor, storage medium etc.) either in the network of computing device with hardware, firmware, software or
Combination thereof is realized that this is that those of ordinary skill in the art use them in the case where having read the explanation of the present invention
Basic programming skill can be achieved with.
Therefore, the purpose of the present invention can also by run on any computing device a program or batch processing come
It realizes.The computing device can be well known fexible unit.Therefore, the purpose of the present invention can also include only by offer
The program product of the program code of the method or device is realized to realize.That is, such program product is also constituted
The present invention, and the storage medium for being stored with such program product also constitutes the present invention.Obviously, the storage medium can be
Any well known storage medium or any storage medium developed in the future.
It may also be noted that in apparatus and method of the present invention, it is clear that each component or each step are can to decompose
And/or reconfigure.These decompose and/or reconfigure the equivalent scheme that should be regarded as the present invention.Also, execute above-mentioned series
The step of processing, can execute according to the sequence of explanation in chronological order naturally, but not need to centainly sequentially in time
It executes.Certain steps can execute parallel or independently of one another.
Above-mentioned specific implementation mode, does not constitute limiting the scope of the invention.Those skilled in the art should be bright
It is white, design requirement and other factors are depended on, various modifications, combination, sub-portfolio and replacement can occur.It is any
Modifications, equivalent substitutions and improvements made by within the spirit and principles in the present invention etc., should be included in the scope of the present invention
Within.
Claims (12)
1. a kind of nan-machine interrogation's method of knowledge based collection of illustrative plates, the method includes：
It receives sentence input by user and the sentence is segmented, including：By segmentation methods to language input by user
Sentence is cut, and deactivates the stop words of vocabulary reservation relation belonging to Feature Words by inquiring relationship, and remove remaining and deactivate
Word and redundancy；
According to condition random field to carrying out entity extraction by the obtained word of participle；
Using knowledge mapping knowledge reasoning is carried out to extracting obtained entity information by the entity；And
Feedback is provided to the sentence input by user according to the result of the knowledge reasoning.
2. nan-machine interrogation's method according to claim 1, wherein the step of entity extracts include：Using acquired
The entity attribute of word be marked, wherein the entity attribute include the part of speech of institute's predicate, institute's predicate dictionary release
Product classification corresponding to justice or institute's predicate.
3. nan-machine interrogation's method according to claim 1, wherein the knowledge mapping include knowledge point and knowledge point it
Between relationship, wherein each knowledge point is provided with the pass between independent in-degree point and out-degree point and the knowledge point
System is the knowledge relation by establishing the classification between the knowledge point according to identical in-degree point and out-degree point, to know described
It establishes knowledge non-directed graph between knowledge point and establishes.
4. nan-machine interrogation's method according to claim 1, wherein the step of knowledge reasoning includes：
The knowledge point corresponding to the entity information is searched in the knowledge mapping；
After obtaining the knowledge point, the step of terminating the reasoning, and provide and the content corresponding to the knowledge point is believed
The location information of breath and the knowledge point in the knowledge mapping；And
When traversing the knowledge mapping without finding the knowledge point, the step of terminating the reasoning, and provide without corresponding
As a result feedback.
5. nan-machine interrogation's method according to claim 1, wherein the step of knowledge reasoning includes：
A) one the first knowledge point corresponded in the entity information is searched in the knowledge mapping；
B) by the out-degree point of first knowledge point, using first knowledge point with corresponding to another in the entity information
Relationship between one the second knowledge point searches second knowledge point in the knowledge mapping；
C) above-mentioned b step is repeated, until for lookup is completed corresponding to whole knowledge points in the entity information, terminating
The step of reasoning, and provide and the content information corresponding to the knowledge point；And
D) above-mentioned b step is repeated, when traversing the knowledge mapping without finding the knowledge point to be searched, terminates the reasoning
The step of, and provide the feedback of no accordingly result.
6. a kind of nan-machine interrogation's system of knowledge based collection of illustrative plates, the system comprises：
Input module, for receiving sentence input by user；
Word-dividing mode, for being segmented to sentence input by user, including：By segmentation methods to sentence input by user into
Row cutting, by inquire relationship deactivate vocabulary retain relation belonging to Feature Words stop words, and remove remaining stop words with
And redundancy；
Entity abstraction module is used to carry out entity extraction to passing through the obtained word of participle according to condition random field；
Knowledge mapping module, for store include relationship between knowledge point and knowledge point knowledge mapping；
Knowledge reasoning module, for being known extracting obtained entity information by the entity using the knowledge mapping
Know reasoning；And
Output module, for providing feedback to the sentence input by user according to the result of the knowledge reasoning.
7. nan-machine interrogation's system according to claim 6, wherein the entity abstraction module utilizes the reality of obtained word
Body attribute is marked, wherein the entity attribute includes the dictionary paraphrase or described of the part of speech of institute's predicate, institute's predicate
Product classification corresponding to word.
8. nan-machine interrogation's system according to claim 6, wherein the knowledge mapping module is by for each knowledge
Independent in-degree point and out-degree point is arranged in point, and the classification between the knowledge point is established according to identical in-degree point and out-degree point
Knowledge relation establishes knowledge non-directed graph between the knowledge point, to store the knowledge point and knowledge point of the knowledge mapping
Between relationship.
9. nan-machine interrogation's system according to claim 6, wherein the knowledge reasoning module is when by the knowledge graph
When being searched in spectrum and obtaining the knowledge point corresponding to the entity information, provide with the content information corresponding to the knowledge point with
And location information of the knowledge point in the knowledge mapping；And described know without finding when traversing the knowledge mapping
When knowing point, the feedback of no accordingly result is provided.
10. nan-machine interrogation's system according to claim 6, wherein the knowledge reasoning module is in the knowledge mapping
Search one the first knowledge point corresponded in the entity information；By the out-degree point of first knowledge point, institute is utilized
The first knowledge point is stated and corresponding to the relationship between another the second knowledge point in the entity information, in the knowledge graph
Second knowledge point is searched in spectrum；When repeating the above steps, until for knowing corresponding to the whole in the entity information
When lookup is completed in knowledge point, provide and the content information corresponding to the knowledge point；And when the traversal knowledge mapping does not have
Have when finding the knowledge point to be searched, provides the feedback of no accordingly result.
11. a kind of electronic equipment, which is characterized in that including：
One or more processors；
Storage device, for storing 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 to 5.
12. a kind of computer-readable medium, is stored thereon with computer program, which is characterized in that described program is held by processor
The method as described in any in claim 1 to 5 is realized when row.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510383452.7A CN105095195B (en) | 2015-07-03 | 2015-07-03 | Nan-machine interrogation's method and system of knowledge based collection of illustrative plates |
| HK16105720.8A HK1217789A1 (en) | 2015-07-03 | 2016-05-18 | Method and system for human-machine questioning and answering based on knowledge graph |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510383452.7A CN105095195B (en) | 2015-07-03 | 2015-07-03 | Nan-machine interrogation's method and system of knowledge based collection of illustrative plates |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105095195A CN105095195A (en) | 2015-11-25 |
| CN105095195B true CN105095195B (en) | 2018-09-18 |
Family
ID=54575666
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510383452.7A Active CN105095195B (en) | 2015-07-03 | 2015-07-03 | Nan-machine interrogation's method and system of knowledge based collection of illustrative plates |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN105095195B (en) |
| HK (1) | HK1217789A1 (en) |
Families Citing this family (52)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105677822A (en) * | 2016-01-05 | 2016-06-15 | 首都师范大学 | Enrollment automatic question-answering method and system based on conversation robot |
| CN107180059A (en) * | 2016-03-11 | 2017-09-19 | 北大方正集团有限公司 | Data retrieval method and data retrieval system |
| CN105843875B (en) * | 2016-03-18 | 2019-09-13 | 北京光年无限科技有限公司 | A kind of question and answer data processing method and device towards intelligent robot |
| CN107391512B (en) * | 2016-05-17 | 2021-05-11 | 北京邮电大学 | Method and device for predicting knowledge graph |
| CN106095932B (en) * | 2016-06-13 | 2019-12-20 | 竹间智能科技（上海）有限公司 | Encyclopedic knowledge question recognition method and device |
| CN107545000A (en) * | 2016-06-28 | 2018-01-05 | 百度在线网络技术（北京）有限公司 | The information-pushing method and device of knowledge based collection of illustrative plates |
| CN106294325B (en) * | 2016-08-11 | 2019-01-04 | 海信集团有限公司 | The optimization method and device of spatial term sentence |
| CN108108982A (en) * | 2016-11-25 | 2018-06-01 | 百度在线网络技术（北京）有限公司 | A kind of recognition methods of authorization message and device |
| CN106599179B (en) * | 2016-12-13 | 2020-02-14 | 竹间智能科技（上海）有限公司 | Man-machine conversation control method and device integrating knowledge graph and memory graph |
| CN106934012B (en) * | 2017-03-10 | 2020-05-08 | 上海数眼科技发展有限公司 | Natural language question-answering implementation method and system based on knowledge graph |
| CN109844743B (en) * | 2017-06-26 | 2023-10-17 | 微软技术许可有限责任公司 | Generating responses in automated chat |
| WO2019000326A1 (en) * | 2017-06-29 | 2019-01-03 | Microsoft Technology Licensing, Llc | Generating responses in automated chatting |
| CN109388793B (en) * | 2017-08-03 | 2023-04-07 | 阿里巴巴集团控股有限公司 | Entity marking method, intention identification method, corresponding device and computer storage medium |
| CN107748757B (en) * | 2017-09-21 | 2021-05-07 | 北京航空航天大学 | Question-answering method based on knowledge graph |
| CN107766483A (en) * | 2017-10-13 | 2018-03-06 | 华中科技大学 | The interactive answering method and system of a kind of knowledge based collection of illustrative plates |
| CN107679039B (en) * | 2017-10-17 | 2020-12-29 | 北京百度网讯科技有限公司 | Method and device for determining statement intention |
| EP3625699A1 (en) | 2017-10-25 | 2020-03-25 | Google LLC | Natural language processing with an n-gram machine |
| CN107885842B (en) * | 2017-11-10 | 2021-01-08 | 上海智臻智能网络科技股份有限公司 | Intelligent question and answer method, device, server and storage medium |
| CN107992528B (en) * | 2017-11-13 | 2022-07-05 | 清华大学 | Multi-relational question-answering system using interpretable reasoning network |
| CN110019710A (en) * | 2017-11-27 | 2019-07-16 | 厦门快商通信息技术有限公司 | A kind of topic forest formula interactive method and system |
| CN108452526B (en) * | 2017-11-28 | 2020-12-25 | 腾讯科技（上海）有限公司 | Game fault reason query method and device, storage medium and electronic device |
| CN107943998B (en) * | 2017-12-05 | 2021-05-11 | 竹间智能科技（上海）有限公司 | Man-machine conversation control system and method based on knowledge graph |
| US10776586B2 (en) | 2018-01-10 | 2020-09-15 | International Business Machines Corporation | Machine learning to integrate knowledge and augment natural language processing |
| US10606958B2 (en) | 2018-01-10 | 2020-03-31 | International Business Machines Corporation | Machine learning modification and natural language processing |
| US10423726B2 (en) | 2018-01-10 | 2019-09-24 | International Business Machines Corporation | Machine learning to integrate knowledge and natural language processing |
| CN108519998B (en) * | 2018-03-07 | 2021-05-14 | 云知声智能科技股份有限公司 | Problem guiding method and device based on knowledge graph |
| CN108509563A (en) * | 2018-03-23 | 2018-09-07 | 深圳狗尾草智能科技有限公司 | Robot reasoning association method, device, equipment and the medium of knowledge based collection of illustrative plates |
| CN108733654A (en) * | 2018-05-21 | 2018-11-02 | 宁波薄言信息技术有限公司 | A kind of information processing method |
| CN109062896A (en) * | 2018-07-25 | 2018-12-21 | 南京瓦尔基里网络科技有限公司 | A kind of matching process and system based on artificial intelligence words art model |
| CN109145102B (en) * | 2018-09-06 | 2021-02-09 | 杭州安恒信息技术股份有限公司 | Intelligent question answering method and knowledge graph system construction method, device and equipment thereof |
| CN109492077B (en) * | 2018-09-29 | 2020-09-29 | 北京智通云联科技有限公司 | Knowledge graph-based petrochemical field question-answering method and system |
| CN109597894B (en) * | 2018-09-30 | 2023-10-03 | 创新先进技术有限公司 | Correlation model generation method and device, and data correlation method and device |
| CN109446387A (en) * | 2018-10-09 | 2019-03-08 | 众蚁(上海)信息技术有限公司 | A kind of Owners Committee's intelligent Answer System based on artificial intelligence |
| CN109543007A (en) * | 2018-10-16 | 2019-03-29 | 深圳壹账通智能科技有限公司 | Put question to data creation method, device, computer equipment and storage medium |
| CN109522465A (en) * | 2018-10-22 | 2019-03-26 | 国家电网公司 | The semantic searching method and device of knowledge based map |
| CN109271504B (en) * | 2018-11-07 | 2021-06-25 | 爱因互动科技发展（北京）有限公司 | Inference dialogue method based on knowledge graph |
| CN111291168A (en) * | 2018-12-07 | 2020-06-16 | 北大方正集团有限公司 | Book retrieval method and device and readable storage medium |
| CN109300472A (en) * | 2018-12-21 | 2019-02-01 | 深圳创维－Rgb电子有限公司 | A kind of audio recognition method, device, equipment and medium |
| CN109933671A (en) * | 2019-01-31 | 2019-06-25 | 平安科技（深圳）有限公司 | Construct method, apparatus, computer equipment and the storage medium of personal knowledge map |
| CN109902165B (en) * | 2019-03-08 | 2021-02-23 | 中国科学院自动化研究所 | Intelligent interactive question-answering method, system and device based on Markov logic network |
| CN111859974A (en) * | 2019-04-22 | 2020-10-30 | 广东小天才科技有限公司 | Semantic disambiguation method and device combined with knowledge graph and intelligent learning equipment |
| CN110175227B (en) * | 2019-05-10 | 2021-03-02 | 神思电子技术股份有限公司 | Dialogue auxiliary system based on team learning and hierarchical reasoning |
| CN112231445A (en) * | 2020-03-27 | 2021-01-15 | 北京来也网络科技有限公司 | Searching method, device, equipment and storage medium combining RPA and AI |
| CN111460172A (en) * | 2020-03-31 | 2020-07-28 | 北京小米移动软件有限公司 | Method and device for determining answers to product questions and electronic equipment |
| CN111488741A (en) * | 2020-04-14 | 2020-08-04 | 税友软件集团股份有限公司 | Tax knowledge data semantic annotation method and related device |
| CN111552880B (en) * | 2020-04-30 | 2023-06-30 | 杭州网易再顾科技有限公司 | Knowledge graph-based data processing method and device, medium and electronic equipment |
| CN111966834A (en) * | 2020-07-29 | 2020-11-20 | 深圳市元征科技股份有限公司 | File generation method, file generation device and server |
| CN111949855A (en) * | 2020-07-31 | 2020-11-17 | 国网上海市电力公司 | Knowledge map-based engineering technology knowledge retrieval platform and method thereof |
| CN111930916B (en) * | 2020-09-18 | 2021-02-05 | 北京百度网讯科技有限公司 | Dialog generation method and device, electronic equipment and storage medium |
| CN113434658A (en) * | 2021-08-25 | 2021-09-24 | 西安热工研究院有限公司 | Thermal power generating unit operation question-answer generation method, system, equipment and readable storage medium |
| CN117573849B (en) * | 2024-01-16 | 2024-04-19 | 之江实验室 | Knowledge graph multi-hop question-answering method, device, equipment and storage medium |
| CN118210983B (en) * | 2024-05-22 | 2024-07-30 | 山东浪潮科学研究院有限公司 | Intelligent self-adaptive retrieval enhancement system, method and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103902652A (en) * | 2014-02-27 | 2014-07-02 | 深圳市智搜信息技术有限公司 | Automatic question-answering system |
| CN104252533A (en) * | 2014-09-12 | 2014-12-31 | 百度在线网络技术（北京）有限公司 | Search method and search device |
| CN104462504A (en) * | 2014-12-19 | 2015-03-25 | 北京奇虎科技有限公司 | Method and device for providing reasoning process data in search |
-
2015
- 2015-07-03 CN CN201510383452.7A patent/CN105095195B/en active Active
-
2016
- 2016-05-18 HK HK16105720.8A patent/HK1217789A1/en unknown
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103902652A (en) * | 2014-02-27 | 2014-07-02 | 深圳市智搜信息技术有限公司 | Automatic question-answering system |
| CN104252533A (en) * | 2014-09-12 | 2014-12-31 | 百度在线网络技术（北京）有限公司 | Search method and search device |
| CN104462504A (en) * | 2014-12-19 | 2015-03-25 | 北京奇虎科技有限公司 | Method and device for providing reasoning process data in search |
Non-Patent Citations (2)
| Title |
|---|
| Discovering Authorities in Question Answer Communities by Using Link Analysis;Pawel Jurczyk 等;《ACM》;20071130;第1-4页 * |
| 网络教学平台中问答系统的关键技术研究;曾庆鹏 等;《计算机与现代化》;20101231;第23-26页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105095195A (en) | 2015-11-25 |
| HK1217789A1 (en) | 2017-01-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105095195B (en) | Nan-machine interrogation's method and system of knowledge based collection of illustrative plates | |
| CN106776711B (en) | Chinese medical knowledge map construction method based on deep learning | |
| CN108804512B (en) | Text classification model generation device and method and computer readable storage medium | |
| CN107436875B (en) | Text classification method and device | |
| CN106649818B (en) | Application search intention identification method and device, application search method and server | |
| US12039447B2 (en) | Information processing method and terminal, and computer storage medium | |
| US9971967B2 (en) | Generating a superset of question/answer action paths based on dynamically generated type sets | |
| CN111191022B (en) | Commodity short header generation method and device | |
| US9348900B2 (en) | Generating an answer from multiple pipelines using clustering | |
| US9104979B2 (en) | Entity recognition using probabilities for out-of-collection data | |
| CN110134792B (en) | Text recognition method and device, electronic equipment and storage medium | |
| CN106033416A (en) | A string processing method and device | |
| CN106815252A (en) | A kind of searching method and equipment | |
| CN111581990A (en) | Cross-border transaction matching method and device | |
| CN112507160A (en) | Automatic judgment method and device for trademark infringement, electronic equipment and storage medium | |
| US11030533B2 (en) | Method and system for generating a transitory sentiment community | |
| CN112506864A (en) | File retrieval method and device, electronic equipment and readable storage medium | |
| Rachman et al. | Sentiment analysis of Madura tourism in new normal era using text blob and KNN with hyperparameter tuning | |
| Blanco et al. | Overview of NTCIR-13 Actionable Knowledge Graph (AKG) Task. | |
| CN115248890B (en) | User interest portrait generation method and device, electronic equipment and storage medium | |
| CN113505190B (en) | Address information correction method, device, computer equipment and storage medium | |
| CN110110218A (en) | A kind of Identity Association method and terminal | |
| CN112597768B (en) | Text auditing method, device, electronic equipment, storage medium and program product | |
| CN112800179B (en) | Associated database query method and device, storage medium and electronic equipment | |
| CN104572628B (en) | A kind of science based on syntactic feature defines automatic extraction system and method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 1217789 Country of ref document: HK |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |