CN105912527A - Method, device and system outputting answer according to natural language - Google Patents
Method, device and system outputting answer according to natural language Download PDFInfo
- Publication number
- CN105912527A CN105912527A CN201610240540.6A CN201610240540A CN105912527A CN 105912527 A CN105912527 A CN 105912527A CN 201610240540 A CN201610240540 A CN 201610240540A CN 105912527 A CN105912527 A CN 105912527A
- Authority
- CN
- China
- Prior art keywords
- answer
- natural language
- language
- man
- vocabulary
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/38—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/951—Indexing; Web crawling techniques
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Library & Information Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a method, device and system outputting answers according to natural language, and belongs to the intelligent robot technical field; the method, device and system can carry out intention understanding for the natural language from a man-machine interaction client, can query an ontology library and a knowledge atlas, thus obtaining corresponding language elements; the method also comprises the following steps: carrying out meta-search, ontology library query and knowledge atlas query according to the language elements, and obtaining backup answers based on the language elements; estimating the backup answers so as to obtain the optimal answer; compositing and improving the optimal answer; outputting the answer corresponding to the natural language to the man-machine interaction client according to the composited and improved result. The method uses meta-search to expand the corpus, can excavate the high quality answer from mass internet information when the natural language from the client is received, and can improve answer accuracy.
Description
Technical field
The present invention relates to intelligent robot technology field, particularly relate to a kind of methods, devices and systems according to natural language output answer.
Background technology
Intelligent answer in interactive process refers to that computer system (machine) can process the natural language of mankind's input, and exports the answer meeting mankind's intention.Intelligent answer has boundless application prospect.First it is to realize machine customer service, such as substitutes the FAQs (Frequent Asked Questions, FAQ) in government website, it is provided that more personalized service；Next to that Optimizing Search engine, existing search engine carries out the coupling of web page text according to the keyword that user inputs thus returns and mate entry accordingly, this mode makes user need to distinguish the answer that just can obtain oneself wanting in numerous information, intelligent answer then can be accomplished directly answer to be exported to user, reduce the workload of user's artificial cognition, optimize Consumer's Experience；3rd is to guide for emotion, and such as old Pei Liao robot solves the problem that old people is lonely.
First technical barrier to be solved of intelligent answer is the natural language of " understanding " user input.During Human communication, owing to having similar background and general knowledge, understand and be not typically a problem.And machine actually really " can not understand " that natural language, machine be good at is information search and coupling.Man-machine interaction at present is applied simpler natural language instructions response, such as vehicle-mounted voice order, can identify that such as " opening air-conditioning ", " broadcasting music " etc. are ordered.Command response need to support minority nature statement due to it, therefore just can complete by the method for rule match.And more complicated statement this for every-day language, need more complicated rule.The method processing natural language (such as automatic translation by computer) the earliest is based on the grammer of language itself, structure lays down a regulation.But due to motility and the complexity of natural language usage, this pure method by rule is proved to not be highly effective.Researcher had been invented again based on statistical method later, utilized a large amount of real corpus to find the rule of natural language, and this method has preferable effect in participle, syntax error correction.In recent years, along with the rise of big data technique, utilizing a large amount of real corpus to carry out train language model, the method excavating its inherent law has obtained bigger development.
Second problem to be solved of intelligent answer is how to obtain answer.The multiformity of language in living due to real work, the method using simple rule generation answer is the most infeasible.And the most most of intelligent answer platform is all based on corpus and the knowledge base of this locality, Utilizing question Similarity Measure finds the answer of coupling.Therefore, the quality of answer depends critically upon the scale of corpus, accuracy and organizational structure.How to expand corpus and how to give corpus accuracy be current intelligent answer platform problems faced.
Summary of the invention
In view of this, the present invention provides a kind of methods, devices and systems according to natural language output answer, it utilizes Meta Search Engine that corpus is expanded, after receiving the natural language of client, can excavate from magnanimity internet information and obtain high-quality answer, the accuracy of answer can be improved, thus more suitable for practicality.
In order to reach above-mentioned first purpose, the technical scheme of the method according to natural language output answer that the present invention provides is as follows:
The method according to natural language output answer that the present invention provides comprises the following steps:
Carry out the natural language from man-machine interaction client being intended to understand, inquire about according to ontology library inquiry and knowledge mapping, obtain corresponding language element；
According to described language element, carry out Meta Search Engine, local library inquiry and knowledge mapping inquiry, obtain alternative answer based on described language element；
Described alternative answer is estimated, obtains answer optimum in described answer；
Answer to described optimum carries out comprehensive and perfect；
According to described comprehensive and perfect result, export, to described man-machine interaction client, the answer that described natural language is corresponding.
The method according to natural language output answer that the present invention provides also can be applied to the following technical measures to achieve further.
As preferably, described carry out the natural language from man-machine interaction client is intended to understand, inquires about according to ontology library inquiry and knowledge mapping, obtains corresponding language element and include:
The described natural language from man-machine interaction client is carried out possible question sentence report, obtain the question sentence through reporting；
To the described natural language from man-machine interaction client, carry out vocabulary fractionation through the question sentence reported, obtain the vocabulary after splitting；
By local library, the vocabulary after described fractionation is carried out synonym and upper and lower Bits Expanding, the vocabulary race after being expanded；
Described vocabulary race carries out semantic disambiguation process, obtain the vocabulary race processed through disambiguation；
According to the described vocabulary race processed through disambiguation, described knowledge mapping is inquired about node and limit that described vocabulary race relates to, be the corresponding language element of the described natural language from man-machine interaction client.
As preferably, according to described language element, carry out Meta Search Engine, obtain alternative answer based on described language element and include:
According to described language element, determine the Question Classification of the described natural language from man-machine interaction client, obtain classification results；
According to described classification results, choose targeted website；
On described targeted website, local library and knowledge mapping, with described language element as foundation, scan for, obtain the original list of Search Results；
Entry on the original list of described Search Results and the described natural language from man-machine interaction client are carried out similarity-rough set, obtains the similarity URL higher than the entry of threshold value；
From described URL, extraction obtains alternative answer based on described language element.
As preferably, described alternative answer is estimated, obtains answer optimum in described answer and include:
Content and the described natural language from man-machine interaction client to described alternative answer carry out relevance evaluation and quality evaluation, determine that degree of association is the highest, and, the answer that alternative answer is optimum that quality is optimal.
As preferably,
The dependency of the content of described alternative answer and the described natural language from man-machine interaction client, with the quantity of described language element that relates in described alternative answer as foundation, is defined as degree of association with the most person of described language element related in described alternative answer the highest；
The quality of described alternative answer is recommended with described answer or the quantity approved of is as foundation, and to be defined as quality optimal for the recommended or most person of quantity that approves of.
In order to reach above-mentioned second purpose, the technical scheme of the device according to natural language output answer that the present invention provides is as follows:
The device according to natural language output answer that the present invention provides includes language element acquiring unit, Meta Search Engine unit, local library query unit, knowledge mapping query unit, answer assessment unit, answer is comprehensive and improves unit, answer output unit,
Described language element acquiring unit understands for the natural language from man-machine interaction client carries out intention, inquires about according to ontology library inquiry and knowledge mapping, obtains corresponding language element；
Described Meta Search Engine unit, for according to described language element, carries out Meta Search Engine, obtains first group of alternative answer based on described language element；
Described local library query unit, for according to described language element, carries out local search, obtains second group of alternative answer based on described language element；
Described knowledge mapping query unit, for according to described language element, carries out knowledge mapping inquiry, obtains the 3rd group of alternative answer based on described language element；
Described answer assessment unit, for being estimated described first group of alternative answer, second group of alternative answer and the 3rd group of alternative answer, obtains answer optimum in described answer；
Described answer is comprehensive and improves unit for carrying out comprehensive and perfect to the answer of described optimum；
Described answer output unit is for according to described comprehensive and perfect result, exporting, to described man-machine interaction client, the answer that described natural language is corresponding.
The device according to natural language output answer that the present invention provides also can be applied to the following technical measures to achieve further.
As preferably, described language element acquiring unit includes that question sentence reports module, vocabulary splits module, vocabulary extension module, vocabulary disambiguation module, language element acquisition module,
Described question sentence is reported module and is reported for the described natural language from man-machine interaction client is carried out possible question sentence, obtains the question sentence through reporting；
Described vocabulary splits module to be used for the described natural language from man-machine interaction client, carries out vocabulary fractionation through the question sentence reported, and obtains the vocabulary after splitting；
Described vocabulary extension module, for by local library, carries out synonym and upper and lower Bits Expanding, the vocabulary race after being expanded to the vocabulary after described fractionation；
Described vocabulary disambiguation module processes for described vocabulary race carries out semantic disambiguation, obtains the vocabulary race processed through disambiguation；
Language element acquisition module, for according to the described vocabulary race processed through disambiguation, inquiring about node and limit that described vocabulary race relates in described knowledge mapping, is the corresponding language element of the described natural language from man-machine interaction client.
As preferably, described Meta Search Engine unit includes that language element sort module, targeted website choose module, search module, URL acquisition module, alternative answer extracting module,
Described language element sort module is for according to described language element, determining the classification of the described natural language from man-machine interaction client, obtain classification results；
Module is chosen for according to described classification results, choosing targeted website in described targeted website；
Described search module, in described targeted website, with described language element as foundation, scans for, obtains the original list of Search Results；
Described URL acquisition module, for the entry on the original list of described Search Results and the described natural language from man-machine interaction client are carried out similarity-rough set, obtains the URL of the similarity entry higher than 80%；
Described alternative answer extracting module obtains alternative answer based on described language element for extraction from described URL.
As preferably, answer assessment unit includes correlation evaluation module, quality assessment modules,
Described correlation evaluation module is for choosing the content of described alternative answer and the described alternative answer the highest from the natural language degree of association of man-machine interaction client；
Described quality assessment modules is for choosing the alternative answer that quality in alternative answer is optimal.
As preferably, the dependency of the content of described alternative answer and the described natural language from man-machine interaction client, with the quantity of described language element that relates in described alternative answer as foundation, is defined as degree of association with the most person of described language element related in described alternative answer the highest；
The quality of described alternative answer is recommended with described answer or the quantity approved of is as foundation, and to be defined as quality optimal for the recommended or most person of quantity that approves of.
In order to reach above-mentioned 3rd purpose, the technical scheme of the system according to natural language output answer that the present invention provides is as follows:
The system according to natural language output answer that the present invention provides includes man-machine interaction client, server,
Described man-machine interaction client is used for going out out natural language to described server, and, described man-machine interaction client is for receiving the answer of described server output；
Ontology library, local library, knowledge mapping, META Search Engine it is provided with on described server,
Described ontology library is used for the relation data between storage concept and concept,
Described local library is used for storing language material and simple knowledge,
Described knowledge mapping is used for expressing the various fact；
The searching interface that described META Search Engine provides for utilizing universal search engine or specific website is to obtain information.
The system according to natural language output answer that the present invention provides also can be applied to the following technical measures to achieve further.
As preferably,
Relation between described concept and concept includes synonymy and/or hyponymy；
The described various fact includes entity-attribute-value, entity-relationship-entity.
The methods, devices and systems according to natural language output answer that the present invention provides understand by the natural language from man-machine interaction client carries out intention, inquire about according to ontology library inquiry and knowledge mapping, obtain corresponding language element；According to language element, carry out Meta Search Engine, local library inquiry and knowledge mapping inquiry, obtain alternative answer based on language element；Alternative answer is estimated, obtains answer optimum in answer；Optimum answer is carried out comprehensive and perfect；According to comprehensive and perfect result, to the answer that man-machine interaction client output natural language is corresponding.It utilizes Meta Search Engine that corpus is expanded, after receiving the natural language of client, it is possible to excavates from magnanimity internet information and obtains high-quality answer, it is possible to increase the accuracy of answer.
Accompanying drawing explanation
By reading the detailed description of hereafter preferred implementation, various other advantage and benefit those of ordinary skill in the art be will be clear from understanding.Accompanying drawing is only used for illustrating the purpose of preferred implementation, and is not considered as limitation of the present invention.And in whole accompanying drawing, it is denoted by the same reference numerals identical parts.In the accompanying drawings:
The general steps flow chart of the method according to natural language output answer that Fig. 1 provides for the embodiment of the present invention；
The concrete steps flow chart of the method according to natural language output answer that Fig. 2 provides for the embodiment of the present invention；
Fig. 3 flows to relation generalization schematic diagram for the signal of the device according to natural language output answer that the embodiment of the present invention provides；
Fig. 4 flows to relation generalization schematic diagram for the signal of the system according to natural language output answer that the embodiment of the present invention provides；
After man-machine interaction client input " birthday of doctor Zhang " that Fig. 5 provides for the embodiment of the present invention, obtain the logical relation schematic diagram of knowledge mapping during correct option.
Detailed description of the invention
The present invention solves the problem that prior art exists, a kind of methods, devices and systems according to natural language output answer are provided, it utilizes Meta Search Engine that corpus is expanded, after receiving the natural language of client, can excavate from magnanimity internet information and obtain high-quality answer, the accuracy of answer can be improved, thus more suitable for practicality.
By further illustrating the technological means and effect that the present invention taked by reaching predetermined goal of the invention, below in conjunction with accompanying drawing and preferred embodiment, to the methods, devices and systems according to natural language output answer proposed according to the present invention, its detailed description of the invention, structure, feature and effect thereof, after describing in detail such as.In the following description, what different " embodiments " or " embodiment " referred to is not necessarily same embodiment.Additionally, special characteristic, structure or feature in one or more embodiment can be combined by any suitable form.
The terms "and/or", a kind of incidence relation describing affiliated partner, three kinds of relations can be there are in expression, such as, A and/or B, concrete is interpreted as: can include A and B simultaneously, can be with individualism A, it is also possible to individualism B, it is possible to possess above-mentioned three kinds of any one situations.
Seeing accompanying drawing 1 and accompanying drawing 2, the method according to natural language output answer that the present invention provides comprises the following steps:
Step S1: carry out the natural language from man-machine interaction client being intended to understand, inquire about according to ontology library inquiry and knowledge mapping, obtain corresponding language element；
Step S2: according to language element, carries out Meta Search Engine, local library inquiry, knowledge mapping inquiry, obtains alternative answer based on language element；
Step S3: be estimated alternative answer, obtains answer optimum in answer；
Step S4: optimum answer is carried out comprehensive and perfect；
Step S5: according to comprehensive and perfect result, to the answer that man-machine interaction client output natural language is corresponding.
The method according to natural language output answer that the present invention provides understands by the natural language from man-machine interaction client carries out intention, inquires about according to ontology library inquiry and knowledge mapping, obtains corresponding language element；According to language element, carry out Meta Search Engine, local library inquiry and knowledge mapping inquiry, obtain alternative answer based on language element；Alternative answer is estimated, obtains answer optimum in answer；Optimum answer is carried out comprehensive and perfect；According to comprehensive and perfect result, to the answer that man-machine interaction client output natural language is corresponding.It utilizes Meta Search Engine that corpus is expanded, after receiving the natural language of client, it is possible to excavates from magnanimity internet information and obtains high-quality answer, it is possible to increase the accuracy of answer.
Wherein, carry out the natural language from man-machine interaction client being intended to understand, inquire about according to ontology library inquiry and knowledge mapping, obtain corresponding language element and include:
Step S11: the natural language from man-machine interaction client is carried out possible question sentence and reports, obtain the question sentence through reporting；
Step S12: to from man-machine interaction client natural language, carry out vocabulary fractionation through the question sentence reported, obtain the vocabulary after splitting；
Step S13: by local library, the vocabulary after splitting is carried out synonym and upper and lower Bits Expanding, the vocabulary race after being expanded；
Step S14: vocabulary race is carried out semantic disambiguation and processes, obtain the vocabulary race processed through disambiguation；
Step S15: according to the vocabulary race processed through disambiguation, inquire about node and limit that vocabulary race relates in knowledge mapping, be the corresponding language element of the natural language from man-machine interaction client.
Wherein, according to language element, carry out Meta Search Engine, obtain alternative answer based on language element and include:
Step S21: according to language element, determines the classification of natural language from man-machine interaction client, obtains classification results；
Step S22: according to classification results, chooses targeted website；
Step S23: on targeted website, with language element as foundation, scans for, and obtains the original list of Search Results；In the present embodiment, the original list searched by crawler capturing；
Step S24: with the natural language from man-machine interaction client, the entry on the original list of Search Results is carried out similarity-rough set, obtains the similarity URL higher than the entry of threshold value, and in the present embodiment, threshold value is 80%；
Step S25: extraction obtains alternative answer based on language element from URL；In the present embodiment, by crawler capturing, from URL, extraction obtains alternative answer based on language element.
Wherein, alternative answer is estimated, obtains answer optimum in answer and include:
Content to alternative answer carries out relevance evaluation and quality evaluation with the natural language from man-machine interaction client, determines that degree of association is the highest, and, the answer that alternative answer is optimum that quality is optimal.
Wherein, the dependency of the content of alternative answer and the natural language from man-machine interaction client, with the quantity of language element that relates in alternative answer as foundation, is defined as degree of association with the most person of language element related in alternative answer the highest；
The quality of alternative answer is recommended with answer or the quantity approved of is as foundation, and to be defined as quality optimal for the recommended or most person of quantity that approves of.
Seeing accompanying drawing 3, the device according to natural language output answer that the present invention provides includes language element acquiring unit, Meta Search Engine unit, local library query unit, knowledge mapping query unit, answer assessment unit, answer is comprehensive and improves unit, answer output unit.Language element acquiring unit understands for the natural language from man-machine interaction client carries out intention, inquires about according to ontology library inquiry and knowledge mapping, obtains corresponding language element；Meta Search Engine unit is for carrying out Meta Search Engine according to language element, local library query unit is for carrying out local search according to language element, knowledge mapping query unit is for carrying out knowledge mapping inquiry according to language element, comprehensive Meta Search Engine, local search and knowledge mapping inquiry as a result, it is possible to obtain alternative answer based on language element；Answer assessment unit, for being estimated alternative answer, obtains answer optimum in answer；Answer is comprehensive and improves unit for carrying out comprehensive and perfect to optimum answer；Answer output unit is for according to comprehensive and perfect result, to the answer that man-machine interaction client output natural language is corresponding.
Natural language from man-machine interaction client, by language element acquiring unit, is carried out being intended to understand, inquires about according to ontology library inquiry and knowledge mapping, obtain corresponding language element by the device according to natural language output answer that the present invention provides；By Meta Search Engine unit, local library query unit, knowledge mapping query unit, according to language element, carry out Meta Search Engine, local library inquiry and knowledge mapping inquiry, obtain alternative answer based on language element；Assess unit by answer, alternative answer is estimated, obtain answer optimum in answer；Comprehensively and improve unit by answer, optimum answer is carried out comprehensive and perfect；By answer output unit, according to comprehensive and perfect result, to the answer that man-machine interaction client output natural language is corresponding.It utilizes Meta Search Engine that corpus is expanded, after receiving the natural language of client, it is possible to excavates from magnanimity internet information and obtains high-quality answer, it is possible to increase the accuracy of answer.
Wherein, language element acquiring unit includes that question sentence reports module, vocabulary splits module, vocabulary extension module, vocabulary disambiguation module, language element acquisition module.Question sentence is reported module and is reported for the natural language from man-machine interaction client is carried out possible question sentence, obtains the question sentence through reporting；Vocabulary split module for from man-machine interaction client natural language, carry out vocabulary fractionation through the question sentence reported, obtain the vocabulary after splitting；Vocabulary extension module is for by local library, carrying out synonym and upper and lower Bits Expanding to the vocabulary after splitting, the vocabulary race after being expanded；Vocabulary disambiguation module processes for vocabulary race carries out semantic disambiguation, obtains the vocabulary race processed through disambiguation；Language element acquisition module is for according to the vocabulary race processed through disambiguation, inquiring about node and limit that vocabulary race relates to, be the corresponding language element of the natural language from man-machine interaction client in knowledge mapping.
Wherein, Meta Search Engine unit includes that language element sort module, targeted website choose module, search module, URL acquisition module, alternative answer extracting module.Language element sort module for according to language element, determines the classification of natural language from man-machine interaction client, obtains classification results；Module is chosen for according to classification results in targeted website, chooses targeted website；Search module, in targeted website, with language element as foundation, scans for, obtains the original list of Search Results；URL acquisition module, for the entry on the original list of Search Results is carried out similarity-rough set with the natural language from man-machine interaction client, obtains the URL of the similarity entry higher than 80%；Alternative answer extracting module obtains alternative answer based on language element for extraction from URL.
Wherein, answer assessment unit includes correlation evaluation module, quality assessment modules.Correlation evaluation module is for choosing the content of alternative answer and the alternative answer the highest from the natural language degree of association of man-machine interaction client；Quality assessment modules is for choosing the alternative answer that quality in alternative answer is optimal.
Wherein, the dependency of the content of alternative answer and the natural language from man-machine interaction client, with the quantity of language element that relates in alternative answer as foundation, is defined as degree of association with the most person of language element related in alternative answer the highest；The quality of alternative answer is recommended with answer or the quantity approved of is as foundation, and to be defined as quality optimal for the recommended or most person of quantity that approves of.
Seeing accompanying drawing 4, the system according to natural language output answer that the present invention provides includes man-machine interaction client, server.Man-machine interaction client is used for going out out natural language to server, and, man-machine interaction client is for receiving the answer of server output；Being provided with ontology library, local library, knowledge mapping, META Search Engine on server, ontology library is for the relation data between storage concept and concept, and local library is used for storing various language material and simple knowledge, and knowledge mapping is used for expressing the various fact；The searching interface that META Search Engine provides for utilizing universal search engine or specific website is to obtain information.
Wherein, the relation between concept and concept includes synonymy and/or hyponymy；The various facts include entity-attribute-value, entity-relationship-entity.
Embodiment
As a example by " Zhang San ", wherein, owing to the educational background of Zhang San is doctor, so sometimes, being also designated as " doctor Zhang ".
See accompanying drawing 5, in the present embodiment, the problem of user's input " birthday of doctor Zhang ", it is desirable to obtain a correct answer.
The first step, it is intended that understand.First pretreatment, carries out participle and part-of-speech tagging to this question sentence, and remove stop words " ", obtain data structure {<doctor Zhang, noun>,<birthday, noun>}.Secondly word is carried out query expansion and semantic disambiguation.By inquiry ontology library, obtain the synonym of " doctor Zhang " for " Zhang San ", and the synonym of " birthday " is " date of birth ", and obtain they nodes in knowledge mapping and attribute limit according to " Zhang San " and " date of birth ".
Second step, carries out Meta Search Engine, local library inquiry and knowledge mapping inquiry parallel according to the language element being intended to understand.
Step 2.1 Meta Search Engine.
The classification of step 2.1.1 problem is judged to " society the people's livelihood ", finds maximally related Liang Ge website Baidu know (zhidao.baidu.com) and search and ask (wenwen.sogou.com).
Step 2.1.2 carries out the query composition of following four set of keyword to the two website:<doctor Zhang, birthday>,<Zhang San, birthday>,<doctor Zhang, date of birth>,<Zhang San, date of birth>.
The step 2.1.3 original list to obtaining in 2.1.2 carries out similarity-rough set, in finding Baidu to know, the problem of an entry is " doctor's Zhang birthday " for most mating entry, and its URL is " http://zhidao.baidu.com/question/********.html？Loc_ans=******** "
Network address in step 2.1.4 crawl step 2.1.3, obtains 5 answers.
Step 2.1.5 carries out answer assessment to five answers.Similarity assessment finds have " Zhang San ", " being born in " printed words in the 1st and the 2nd article of answer, higher with problem similarity；Finding in quality evaluation, the Article 1 answer person of being asked adopts and by favorable comment more than 700 time, therefore quality is higher.Therefore Meta Search Engine returns answer is " solar calendar December 26, lunar calendar November 19 ... ".
Step 2.2 inquires about local library.
Step 2.3 inquires about knowledge mapping, according to the node arrived in the first step and attribute limit, and structure collection of illustrative plates inquiry, and to obtain answer be " on December 26th, 1898 " (with reference to accompanying drawing 5).
3rd step, answer is comprehensive and perfect.In this example, knowledge mapping inquiry returns result, the most comprehensively selects the result of knowledge mapping, and this result was improved as " birthday of doctor Zhang be 1898 on December 26, ".
In addition, the methods, devices and systems according to natural language output answer that the present invention provides can also capture various question and answer storehouses, knowledge base and other natural language texts on the Internet by off-line, extract useful knowledge, and store this locality with certain type of organization, carry out local search during search answer and substitute Meta Search Engine.
Although preferred embodiments of the present invention have been described, but those skilled in the art once know basic creative concept, then these embodiments can be made other change and amendment.So, claims are intended to be construed to include preferred embodiment and fall into all changes and the amendment of the scope of the invention.
Obviously, those skilled in the art can carry out various change and modification without departing from the spirit and scope of the present invention to the present invention.So, if these amendments of the present invention and modification belong within the scope of the claims in the present invention and equivalent technologies thereof, then the present invention is also intended to comprise these change and modification.
Claims (10)
1. the method according to natural language output answer, it is characterised in that comprise the following steps:
Carry out the natural language from man-machine interaction client being intended to understand, according to ontology library inquiry and knowledge
Collection of illustrative plates is inquired about, and obtains corresponding language element；
According to described language element, carry out Meta Search Engine, local library inquiry and knowledge mapping inquiry, obtain based on
The alternative answer of described language element；
Described alternative answer is estimated, obtains answer optimum in described answer；
Answer to described optimum carries out comprehensive and perfect；
According to described comprehensive and perfect result, export described natural language pair to described man-machine interaction client
The answer answered.
Method according to natural language output answer the most according to claim 1, it is characterised in that institute
State and carry out the natural language from man-machine interaction client being intended to understand, according to ontology library inquiry and knowledge graph
Spectrum inquiry, obtains corresponding language element and includes:
The described natural language from man-machine interaction client is carried out possible question sentence report, obtain through turning
The question sentence stated；
To the described natural language from man-machine interaction client, carry out vocabulary fractionation through the question sentence reported,
Obtain the vocabulary after splitting；
By ontology library, the vocabulary after described fractionation is carried out synonym and upper and lower Bits Expanding, after being expanded
Vocabulary race；
Described vocabulary race carries out semantic disambiguation process, obtain the vocabulary race processed through disambiguation；
According to the described vocabulary race processed through disambiguation, described knowledge mapping is inquired about described vocabulary race and relates to
Node and limit, be the corresponding language element of the described natural language from man-machine interaction client.
Method according to natural language output answer the most according to claim 1, it is characterised in that root
According to described language element, carry out Meta Search Engine, local library inquiry and knowledge mapping inquiry, obtain based on institute's predicate
The alternative answer of speech element includes:
According to described language element, determine that the described problem from the natural language of man-machine interaction client is divided
Class, obtains classification results；
According to described classification results, choose targeted website；
On described targeted website, local library and knowledge mapping, with described language element as foundation, search
Rope, obtains the original list of Search Results；
By the entry on the original list of described Search Results and the described nature from man-machine interaction client
Language carries out similarity-rough set, obtains the similarity URL higher than the entry of threshold value；
From described URL, extraction obtains alternative answer based on described language element.
Method according to natural language output answer the most according to claim 1, it is characterised in that right
Described alternative answer is estimated, and obtains answer optimum in described answer and includes:
The content of described alternative answer is carried out relevant to the described natural language from man-machine interaction client
Property assessment and quality evaluation, determine that degree of association is the highest, and, the optimal alternative answer of quality is optimum answering
Case；
As preferably,
The dependency of the content of described alternative answer and the described natural language from man-machine interaction client with
The quantity of the described language element related in described alternative answer is foundation, with relate in described alternative answer
It is the highest that the most person of described language element is defined as degree of association；
The quality of described alternative answer recommended with described answer or approve of quantity as foundation, recommended or
It is optimal that the most person of quantity that person approves of is defined as quality.
5. the device according to natural language output answer, it is characterised in that include that language element obtains single
Unit, Meta Search Engine unit, local library query unit, knowledge mapping query unit, answer assessment unit, answer
It is comprehensive and improve unit, answer output unit,
Described language element acquiring unit is for being intended to the natural language from man-machine interaction client
Understand, inquire about according to ontology library inquiry and knowledge mapping, obtain corresponding language element；
Described Meta Search Engine unit, for according to described language element, carries out Meta Search Engine, obtains based on described language
First group of alternative answer of element；
Described local library query unit, for according to described language element, carries out local search, obtains based on institute
State second group of alternative answer of language element；
Described knowledge mapping query unit, for according to described language element, carries out knowledge mapping inquiry, obtains
The 3rd group of alternative answer based on described language element；
Described answer assessment unit is for described first group of alternative answer, second group of alternative answer and the 3rd group
Alternative answer is estimated, and obtains answer optimum in described answer；
Described answer is comprehensive and improves unit for carrying out comprehensive and perfect to the answer of described optimum；
Described answer output unit is for according to described comprehensive and perfect result, to described man-machine interaction client
End exports the answer that described natural language is corresponding.
Device according to natural language output answer the most according to claim 5, it is characterised in that institute
State language element acquiring unit and include that question sentence reports module, vocabulary splits module, vocabulary extension module, vocabulary
Disambiguation module, language element acquisition module,
Described question sentence reports module for the described natural language from man-machine interaction client is carried out possibility
Question sentence report, obtain the question sentence through reporting；
Described vocabulary splits module for the described natural language from man-machine interaction client, through reporting
Question sentence carry out vocabulary fractionation, obtain through fractionation after vocabulary；
Described vocabulary extension module, for by local library, carries out synonym with upper and lower to the vocabulary after described fractionation
Bits Expanding, the vocabulary race after being expanded；
Described vocabulary disambiguation module processes for described vocabulary race carries out semantic disambiguation, obtains at disambiguation
The vocabulary race of reason；
Language element acquisition module is for according to the described vocabulary race processed through disambiguation, at described knowledge mapping
Node that middle inquiry described vocabulary race relates to and limit, be the described natural language from man-machine interaction client
Corresponding language element.
Device according to natural language output answer the most according to claim 5, it is characterised in that institute
State Meta Search Engine unit and include that language element sort module, targeted website choose module, search module, URL
Acquisition module, alternative answer extracting module,
Described language element sort module, for according to described language element, determines described from man-machine interaction visitor
The classification of the natural language of family end, obtains classification results；
Module is chosen for according to described classification results, choosing targeted website in described targeted website；
Described search module, in described targeted website, with described language element as foundation, scans for,
Obtain the original list of Search Results；
Described URL acquisition module for by the entry on the original list of described Search Results with described from
The natural language of man-machine interaction client carries out similarity-rough set, obtains the similarity entry higher than 80%
URL；
Described alternative answer extracting module obtains based on described language element for extraction from described URL
Alternative answer.
Device according to natural language output answer the most according to claim 7, it is characterised in that answer
Case assessment unit includes correlation evaluation module, quality assessment modules,
Described correlation evaluation module is for choosing the content of described alternative answer with described from man-machine interaction
The alternative answer that the natural language degree of association of client is the highest；
Described quality assessment modules is for choosing the alternative answer that quality in alternative answer is optimal；
As preferably, the content of described alternative answer and the described natural language from man-machine interaction client
Dependency is with the quantity of described language element that relates in described alternative answer as foundation, with described alternative answer
In the most person of described language element that relates to be defined as degree of association the highest；
The quality of described alternative answer recommended with described answer or approve of quantity as foundation, recommended or
It is optimal that the most person of quantity that person approves of is defined as quality.
9. one kind according to natural language output answer system, it is characterised in that include man-machine interaction client,
Server,
Described man-machine interaction client is used for going out out natural language to described server, and, described man-machine friendship
Client is for receiving the answer of described server output mutually；
Ontology library, local library, knowledge mapping, META Search Engine it is provided with on described server,
Described ontology library is used for the relation data between storage concept and concept,
Described local library is used for storing various language material and simple knowledge,
Described knowledge mapping is used for expressing the various fact；
The searching interface that described META Search Engine provides for utilizing universal search engine or specific website is come
Acquisition information.
System according to natural language output answer the most according to claim 9, it is characterised in that
Relation between described concept and concept includes synonymy and/or hyponymy；
The described various fact includes entity-attribute-value, entity-relationship-entity.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610240540.6A CN105912527A (en) | 2016-04-19 | 2016-04-19 | Method, device and system outputting answer according to natural language |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610240540.6A CN105912527A (en) | 2016-04-19 | 2016-04-19 | Method, device and system outputting answer according to natural language |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105912527A true CN105912527A (en) | 2016-08-31 |
Family
ID=56747271
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610240540.6A Pending CN105912527A (en) | 2016-04-19 | 2016-04-19 | Method, device and system outputting answer according to natural language |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105912527A (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106844506A (en) * | 2016-12-27 | 2017-06-13 | 竹间智能科技（上海）有限公司 | The knowledge retrieval method and the automatic improving method of knowledge base of a kind of artificial intelligence dialogue |
| CN107622052A (en) * | 2017-09-20 | 2018-01-23 | 广东欧珀移动通信有限公司 | Natural language processing method, apparatus, storage medium and terminal device |
| CN107656997A (en) * | 2017-09-20 | 2018-02-02 | 广东欧珀移动通信有限公司 | Natural language processing method, apparatus, storage medium and terminal device |
| CN107679039A (en) * | 2017-10-17 | 2018-02-09 | 北京百度网讯科技有限公司 | The method and apparatus being intended to for determining sentence |
| CN108170704A (en) * | 2017-11-21 | 2018-06-15 | 北京明略软件系统有限公司 | A kind of method and device of atlas analysis |
| CN108920530A (en) * | 2018-06-08 | 2018-11-30 | 泰康保险集团股份有限公司 | A kind of information processing method, device, storage medium and electronic equipment |
| CN109033223A (en) * | 2018-06-29 | 2018-12-18 | 北京百度网讯科技有限公司 | For method, apparatus, equipment and computer readable storage medium across type session |
| CN109213847A (en) * | 2018-09-14 | 2019-01-15 | 广州神马移动信息科技有限公司 | Layered approach and its device, electronic equipment, the computer-readable medium of answer |
| CN109844743A (en) * | 2017-06-26 | 2019-06-04 | 微软技术许可有限责任公司 | Response is generated in automatic chatting |
| CN109933707A (en) * | 2018-10-31 | 2019-06-25 | 中国科学院信息工程研究所 | A kind of theme corpus construction method and system based on search engine |
| CN109933653A (en) * | 2019-01-24 | 2019-06-25 | 平安科技（深圳）有限公司 | Question and answer querying method, system and the computer equipment of question answering system |
| CN109947916A (en) * | 2019-03-01 | 2019-06-28 | 河北尚云信息科技有限公司 | Question answering system device and answering method based on meteorological field knowledge mapping |
| CN110543951A (en) * | 2018-05-28 | 2019-12-06 | 中国铁道科学研究院铁道建筑研究所 | Virtual assistant system for maintenance of railway bridge |
| WO2022012234A1 (en) * | 2020-07-17 | 2022-01-20 | 海信视像科技股份有限公司 | Associated recommendation method, smart device and service device |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103279528A (en) * | 2013-05-31 | 2013-09-04 | 俞志晨 | Question-answering system and question-answering method based on man-machine integration |
| 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 |
| CN104050256A (en) * | 2014-06-13 | 2014-09-17 | 西安蒜泥电子科技有限责任公司 | Initiative study-based questioning and answering method and questioning and answering system adopting initiative study-based questioning and answering method |
| CN104361127A (en) * | 2014-12-05 | 2015-02-18 | 广西师范大学 | Multilanguage question and answer interface fast constituting method based on domain ontology and template logics |
| CN104471568A (en) * | 2012-07-02 | 2015-03-25 | 微软公司 | Learning-based processing of natural language questions |
| CN104915340A (en) * | 2014-03-10 | 2015-09-16 | 北京大学 | Natural language question-answering method and device |
| US20160055234A1 (en) * | 2014-08-19 | 2016-02-25 | International Business Machines Corporation | Retrieving Text from a Corpus of Documents in an Information Handling System |
-
2016
- 2016-04-19 CN CN201610240540.6A patent/CN105912527A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104471568A (en) * | 2012-07-02 | 2015-03-25 | 微软公司 | Learning-based processing of natural language questions |
| CN103279528A (en) * | 2013-05-31 | 2013-09-04 | 俞志晨 | Question-answering system and question-answering method based on man-machine integration |
| 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 |
| CN104915340A (en) * | 2014-03-10 | 2015-09-16 | 北京大学 | Natural language question-answering method and device |
| CN104050256A (en) * | 2014-06-13 | 2014-09-17 | 西安蒜泥电子科技有限责任公司 | Initiative study-based questioning and answering method and questioning and answering system adopting initiative study-based questioning and answering method |
| US20160055234A1 (en) * | 2014-08-19 | 2016-02-25 | International Business Machines Corporation | Retrieving Text from a Corpus of Documents in an Information Handling System |
| CN104361127A (en) * | 2014-12-05 | 2015-02-18 | 广西师范大学 | Multilanguage question and answer interface fast constituting method based on domain ontology and template logics |
Non-Patent Citations (1)
| Title |
|---|
| 刘峤, 李杨, 段宏, 刘瑶, 秦志光: "知识图谱构建技术综述", 《计算机研究与发展》 * |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106844506A (en) * | 2016-12-27 | 2017-06-13 | 竹间智能科技（上海）有限公司 | The knowledge retrieval method and the automatic improving method of knowledge base of a kind of artificial intelligence dialogue |
| 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 |
| CN109844743A (en) * | 2017-06-26 | 2019-06-04 | 微软技术许可有限责任公司 | Response is generated in automatic chatting |
| CN107622052A (en) * | 2017-09-20 | 2018-01-23 | 广东欧珀移动通信有限公司 | Natural language processing method, apparatus, storage medium and terminal device |
| CN107656997A (en) * | 2017-09-20 | 2018-02-02 | 广东欧珀移动通信有限公司 | Natural language processing method, apparatus, storage medium and terminal device |
| CN107622052B (en) * | 2017-09-20 | 2021-01-22 | Oppo广东移动通信有限公司 | Natural language processing method and device, storage medium and terminal equipment |
| CN107656997B (en) * | 2017-09-20 | 2021-01-15 | Oppo广东移动通信有限公司 | Natural language processing method and device, storage medium and terminal equipment |
| CN107679039A (en) * | 2017-10-17 | 2018-02-09 | 北京百度网讯科技有限公司 | The method and apparatus being intended to for determining sentence |
| 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 |
| CN108170704A (en) * | 2017-11-21 | 2018-06-15 | 北京明略软件系统有限公司 | A kind of method and device of atlas analysis |
| CN110543951A (en) * | 2018-05-28 | 2019-12-06 | 中国铁道科学研究院铁道建筑研究所 | Virtual assistant system for maintenance of railway bridge |
| CN110543951B (en) * | 2018-05-28 | 2022-05-17 | 中国铁道科学研究院铁道建筑研究所 | Virtual assistant system for maintenance of railway bridge |
| CN108920530A (en) * | 2018-06-08 | 2018-11-30 | 泰康保险集团股份有限公司 | A kind of information processing method, device, storage medium and electronic equipment |
| CN109033223A (en) * | 2018-06-29 | 2018-12-18 | 北京百度网讯科技有限公司 | For method, apparatus, equipment and computer readable storage medium across type session |
| CN109213847A (en) * | 2018-09-14 | 2019-01-15 | 广州神马移动信息科技有限公司 | Layered approach and its device, electronic equipment, the computer-readable medium of answer |
| CN109933707A (en) * | 2018-10-31 | 2019-06-25 | 中国科学院信息工程研究所 | A kind of theme corpus construction method and system based on search engine |
| CN109933707B (en) * | 2018-10-31 | 2022-10-14 | 中国科学院信息工程研究所 | Topic corpus construction method and system based on search engine |
| CN109933653A (en) * | 2019-01-24 | 2019-06-25 | 平安科技（深圳）有限公司 | Question and answer querying method, system and the computer equipment of question answering system |
| CN109947916A (en) * | 2019-03-01 | 2019-06-28 | 河北尚云信息科技有限公司 | Question answering system device and answering method based on meteorological field knowledge mapping |
| CN109947916B (en) * | 2019-03-01 | 2023-08-08 | 河北尚云信息科技有限公司 | Question-answering system device and question-answering method based on knowledge graph of meteorological field |
| WO2022012234A1 (en) * | 2020-07-17 | 2022-01-20 | 海信视像科技股份有限公司 | Associated recommendation method, smart device and service device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105912527A (en) | Method, device and system outputting answer according to natural language | |
| CN106919646B (en) | Chinese text abstract generating system and method | |
| CN109726274B (en) | Question generation method, device and storage medium | |
| CN103440243B (en) | A kind of teaching resource recommendation method and device thereof | |
| CN105844424A (en) | Product quality problem discovery and risk assessment method based on network comments | |
| Delen et al. | A holistic framework for knowledge discovery and management | |
| CN107918644B (en) | News topic analysis method and implementation system in reputation management framework | |
| WO2007008798A3 (en) | System and method for searching for network-based content in a multi-modal system using spoken keywords | |
| JP2009048441A (en) | Information retrieval system and method and program, and information retrieval service provision method | |
| CN102262634A (en) | Automatic questioning and answering method and system | |
| CN108021660B (en) | Topic self-adaptive microblog emotion analysis method based on transfer learning | |
| CN105718585B (en) | Document and label word justice correlating method and its device | |
| CN103106287A (en) | Processing method and processing system for retrieving sentences by user | |
| CN105243149B (en) | A kind of semantic-based web query recommended method and system | |
| CN105930490A (en) | Intelligent selecting system for teaching resources | |
| CN105653673A (en) | Information searching method and apparatus | |
| CN108710653B (en) | On-demand method, device and system for reading book | |
| Hong et al. | Automatically extracting word relationships as templates for pun generation | |
| CN105740310A (en) | Automatic answer summarizing method and system for question answering system | |
| CN107977395B (en) | Method for helping user read and understand electronic article and intelligent voice assistant | |
| CN104714940A (en) | Method and device for identifying unregistered word in intelligent interaction system | |
| Johnson et al. | More effective web search using bigrams and trigrams | |
| CN110209804B (en) | Target corpus determining method and device, storage medium and electronic device | |
| CN103020311A (en) | Method and system for processing user search terms | |
| CN114238735B (en) | Intelligent internet data acquisition 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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20160831 |
|
| WD01 | Invention patent application deemed withdrawn after publication |