CN109582798A - Automatic question-answering method, system and equipment - Google Patents
Automatic question-answering method, system and equipment Download PDFInfo
- Publication number
- CN109582798A CN109582798A CN201710909408.4A CN201710909408A CN109582798A CN 109582798 A CN109582798 A CN 109582798A CN 201710909408 A CN201710909408 A CN 201710909408A CN 109582798 A CN109582798 A CN 109582798A
- Authority
- CN
- China
- Prior art keywords
- data
- knowledge
- question
- ontology
- reply
- 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
Abstract
The embodiment of the present application provides a kind of automatic question-answering method, system and equipment.Wherein, method comprises the following steps that this for receiving user's input puts question to data;From ontology knowledge base, the corresponding at least one set of polymorphic type knowledge of acquisition at least one sample ontology relevant to this described enquirement data;Reply data is retrieved at least one set of polymorphic type knowledge;Response feedback is exported to the user according to search result.In technical solution provided by the embodiments of the present application, ontology knowledge base is that a plurality of types of knowledge provided are retrieved in response, not only improve the single problem of prior art knowledge type, and all types of knowledge can form a kind of complementary cooperation relation, a seed type knowledge is more difficult retrieve reply data when replaceable another type of knowledge retrieval, this will be helpful to reduce response without answer rate, while also ensuring higher accuracy rate, improve the automatic question answering experience of user.
Description
Technical field
This application involves field of computer technology more particularly to a kind of automatic question-answering methods, system and equipment.
Background technique
With the development of Internet application, nan-machine interrogation is by many enterprises, public institution or functional government departments
System is introduced into its website or APP, to assist or instead of being seeked advice from by the artificial enquirement for being multiplexed family back and forth.
Existing automatically request-answering system is mostly a kind of question and answer as existing " problem-answer " in set or map is known
It is found in knowledge and puts question to the answer to match to be presented to the user with user.Existing nan-machine interrogation's system is based only upon single type knowledge
(such as single question and answer to or single map knowledge knowledge) answer;Response accuracy rate is low, poor user experience.
Summary of the invention
In view of the above problems, propose the application so as to solve the above problems or at least be partially solved the above problem from
Dynamic answering method, system and equipment.
Then, in one embodiment of the application, a kind of automatic question-answering method is provided.This method comprises:
Receive this enquirement data of user's input;
From ontology knowledge base, it is corresponding at least to obtain at least one sample ontology relevant to this described enquirement data
One group of polymorphic type knowledge;
Reply data is retrieved at least one set of polymorphic type knowledge;
Response feedback is exported to the user according to search result.
In another embodiment of the application, a kind of automatic question-answering method is provided, comprising:
In response to the enquirement event of user's triggering, this enquirement data of the user are obtained;
Data are putd question to be sent to server-side described this;
The server-side is exported based on the response feedback that this puts question to data to send;
Wherein, the response feedback is generated according to search result, and the search result is using at least one response
Search modes put question to relevant at least one sample of data corresponding at least one set of multiclass in ontology knowledge base to described this
Retrieval obtains in type knowledge.
In another embodiment of the application, a kind of automatically request-answering system is provided, comprising:
Ontology knowledge base is known for storing at least one sample ontology and the corresponding one group of polymorphic type of this ontology of various kinds
Know;
Interactive unit, the enquirement event for triggering in response to user obtain this enquirement data of the user and incite somebody to action
Described this puts question to data to issue;
Response unit puts question to data for receiving described this;In the ontology knowledge base, acquisition is mentioned with described this
Ask at least one relevant sample ontology of data corresponding at least one set of polymorphic type knowledge;In at least one set of polymorphic type knowledge
Middle retrieval reply data;According to search result, response feedback is exported to the user;And
The interactive unit is also used to export the response feedback.
In another embodiment of the application, a kind of server device is provided, comprising: at first memory and first
Manage device and the first communication component;
The first memory, for storing program;
First communication component is transmitted for data;
The first processor is coupled with the first memory, for executing the institute stored in the first memory
Program is stated, to be used for:
This enquirement data of user's input are received by first communication component;
From the ontology knowledge base, it is corresponding to obtain at least one sample ontology relevant to this described enquirement data
At least one set of polymorphic type knowledge;
Reply data is retrieved at least one set of polymorphic type knowledge;
According to search result, response feedback is exported to the user by first communication component.
In another embodiment of the application, a kind of client device is provided, comprising: second memory, at second
Manage device and the second communication component;
The second memory, for storing program;
Second communication component is transmitted for data;
The second processor is coupled with the second memory, for executing the institute stored in the second memory
Program is stated, to be used for:
In response to the enquirement event of user's triggering, this enquirement data of the user are obtained;
Data are putd question to be sent to server-side described this by second communication component;
The server-side that output is received by second communication component puts question to data to send based on described this
Response feedback;
Wherein, the response feedback is generated based on search result, the search result be in ontology knowledge base with
Retrieval obtains in this described corresponding at least one set of polymorphic type knowledge of at least one relevant sample of enquirement data.
In technical solution provided by the embodiments of the present application, ontology knowledge base is a plurality of types of knowing of providing of response retrieval
Know, not only improves the single problem of prior art knowledge type, and all types of knowledge can form a kind of complementary cooperation relation,
One seed type knowledge is more difficult to retrieve replaceable another type of knowledge retrieval when reply data, this will be helpful to reduce response
It without answer rate, while also ensuring higher accuracy rate, improves the automatic question answering experience of user.
Detailed description of the invention
In order to illustrate the technical solutions in the embodiments of the present application or in the prior art more clearly, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is this Shen
Some embodiments please for those of ordinary skill in the art without creative efforts, can be with root
Other attached drawings are obtained according to these attached drawings.
Fig. 1 is the schematic illustration for the ontology knowledge base that one embodiment of the application provides;
Fig. 2 is the structural schematic diagram for the automatically request-answering system that one embodiment of the application provides;
Fig. 3 is a kind of editor circle that can be used for safeguarding all types of knowledge in ontology knowledge base that one embodiment of the application provides
Face schematic diagram;
Fig. 4 is the flow diagram for the automatic question-answering method that one embodiment of the application provides;
Fig. 5 is a kind of exemplary diagram for the automatic question answering showing interface search result that one embodiment of the application provides;
Fig. 6 is another exemplary diagram for the automatic question answering showing interface search result that one embodiment of the application provides;
Fig. 7 is the flow diagram for the automatic question-answering method that another embodiment of the application provides;
Fig. 8 is in the semi-structured question and answer of multiple groups in the automatic question-answering method that provides of one embodiment of the application to examining in knowledge
The flow diagram of rope reply data;
It is examined in the map knowledge of at least one set of structuring in the automatic question-answering method that Fig. 9 provides for one embodiment of the application
The flow diagram of rope reply data;
Figure 10 is the exemplary relational graph of map knowledge that one embodiment of the application provides;
In at least one set of non-structured text knowledge in the automatic question-answering method that Figure 11 provides for one embodiment of the application
The flow diagram of middle retrieval reply data;
Figure 12 is the exemplary diagram that the machine reading model that one embodiment of the application provides is a deep neural network model;
Figure 13 is a kind of process signal that online question and answer are carried out based on machine reading model that one embodiment of the application provides
Figure;
Figure 14 is the flow diagram for the automatic question-answering method that another embodiment of the application provides;
Figure 15 is the structural schematic diagram for the automatic call answering arrangement that one embodiment of the application provides;
Figure 16 is the structural schematic diagram for the automatic call answering arrangement that another embodiment of the application provides;
Figure 17 is the structural schematic diagram for the server device that one embodiment of the application provides;
Figure 18 is the structural schematic diagram for the client device that one embodiment of the application provides.
Specific embodiment
In order to make those skilled in the art more fully understand application scheme, below in conjunction in the embodiment of the present application
Attached drawing, the technical scheme in the embodiment of the application is clearly and completely described.
In existing automatic question answering technology, the acquisition of knowledge tends to rely on open question and answer to knowledge base.However, being based on
There are following many restrictions for automatically request-answering system of the question and answer to knowledge base:
The first, question and answer experience is bad.
In existing automatically request-answering system, due to using open question and answer to as knowledge, so when mostly wheel question and answer
System can not inherit the theme of wheel question and answer and open a dialogue with people, and each round question and answer are all primary new full dose knowledge searches,
Such setting can not form the experience of " dialogue " formula, and as the carry out precision of more wheel dialogues can be declined.Such as:
Is user puts question to 1: how Taobao's account modified? the enquirement 11 title of corresponding answer is obtained by knowledge search are as follows:
Taobao's account remedies
Does and then user put question to 2: can modify several times? the knowledge search for carrying out a new round puts question to 22 titles of corresponding answer
Are as follows: modified address does not receive goods responsibility at whom later
From content above it is found that it is correct for puing question to 1 corresponding answer 1, and 2 users is putd question to want to continue topic above
Continue to ask down, can automatically request-answering system using the prior art can not inherit wheel question and answer theme cause put question to 2 corresponding times
It is wrong for answering 2.
The second, Knowledge Source is single
Existing question and answer knowledge base, can only store question and answer pair, and richer knowledge type can not be supported to store, such as non-knot
The map knowledge of the document of structure or more structuring.The shortcoming of knowledge type, in the hair of algorithmic technique of collapse to a certain degree
Wave space.
Third, maintenance work repeat.
Existing question and answer are to knowledge base, when backstage user will establish the knowledge of a genus, require completely to repeat every time
Ground adds full text, increases maintenance cost.Such as:
Is problem 1: how Taobao's account modified? 1 title of answer: Taobao's account remedies
Can problem 2: Taobao's account be modified several times? 2 title of answer: the number that Taobao's account can be modified
Problem 3: Taobao's account, which is locked out, to be changed? 3 title of answer: whether Taobao's account is locked out can modify
Such as the knowledge of the above-mentioned related Taobao's account enumerated, backstage user needs repeatedly to add in each problem
" Taobao's account " text, to avoid there is the problem of mentioning in above-mentioned one.Automatic question answering field one of the most common type knowledge base is
Question and answer are to knowledge base.Question and answer need to require knowledge base to safeguard complete problem every time, with regard at last in description same class problem
Need to repeat all contents of typing;And the structure of knowledge and type that the knowledge base of this question and answer pair can store are all excessively single
One, it is unfavorable for the expansion and application of more forward position algorithms, and then it is difficult to cause the experience of automatic question answering to be promoted, precision decline.
In addition to it is above-mentioned based on question and answer to the automatically request-answering system of knowledge base other than, in the prior art there are also based on map knowledge
Automatically request-answering system.Automatically request-answering system based on map knowledge, it is necessary first to carry out a whole set of knowledge engineering method, such as wrap
Ontology detection, ontology link, attribute filling etc. are included, the map knowledge of structuring is constructed from text, and in map Knowledge-based
Question and answer are carried out on plinth, whole process is relatively complicated, and a large amount of map knowledge node is also difficult especially in actual commercialization scene
With maintenance, investment and output are often difficult proportional.
If a kind of new mode can be started from knowledge store, the knowledge type of more multi-source is supported, promote dialog
It tests and while precision, reduces the artificial work for repeating maintenance, this can be the primary great innovation in intelligent answer field, meaning
It is far-reaching.
Below in conjunction with the attached drawing in the embodiment of the present application, technical solutions in the embodiments of the present application carries out clear, complete
Site preparation description.Obviously, described embodiments are only a part of embodiments of the present application, instead of all the embodiments.It is based on
Embodiment in the application, those skilled in the art's every other implementation obtained without making creative work
Example, shall fall in the protection scope of this application.
In order to facilitate understanding, first to this application involves specialized vocabulary carry out it is as described below:
Ontology (Ontology) refers to a kind of formalization, for sharing the clear of concept system and being described in detail,
Object, concept, a type can be simply interpreted as.Formalization refers to: ontology be it is computer-readable (can be by computer
Reason);Shared to refer to that embody in ontology is the knowledge approved jointly, reflection is the concept set generally acknowledged in related fields.Ontology
Target be capture related fields knowledge, provide and the domain knowledge be commonly understood by, determine the word that the field is approved jointly
It converges, and provides from the formalization mode of different levels explicitly defining for correlation between these vocabulary (term) and vocabulary.
Ontology knowledge base includes at least one sample ontology, and each sample ontology is corresponding with one group of polymorphic type knowledge.For example,
The example of a sample ontology 1 shown in fig. 1, which is corresponding with 3 different types of knowledge, i.e., semi-structured
Question and answer are to knowledge 2, the map knowledge 3 of structuring, non-structured text knowledge 4 etc..Each sample ontology can also have and its word
Synset 5 similar in justice, in order to find sample corresponding with the enquirement data of user's input by way of synonym
Ontology.Technical solution provided by the embodiments of the present application is realized based on the ontology knowledge base.
Automatic question answering refers to that the problem of proposing according to the natural language of user finds a corresponding answer.Fig. 2 is automatic
The structural schematic diagram of question answering system.Now in conjunction with Fig. 2, the structure of automatically request-answering system provided by the present application is illustrated, specifically
It is as follows:
Automatically request-answering system includes: interactive unit 101, response unit 102 and ontology knowledge base 103.Wherein, ontology knowledge
Library 103 is for storing at least one sample ontology and the corresponding at least one set of polymorphic type knowledge of at least one sample ontology;It hands over
Mutual unit 101 is used for enquirement event trigger in response to user, obtain the user this put question to data and will described in this
Data are putd question to issue;Response unit 102 puts question to data for receiving described this;In the ontology knowledge base, acquisition and institute
Stating this puts question at least one relevant sample ontology of data corresponding at least one set of polymorphic type knowledge;It is more in at least one set
Reply data is retrieved in type knowledge;According to search result, response feedback is exported to the user;And the interactive unit
101 are also used to export the response feedback.
In short, interactive unit 101 is sent to response unit 102 the problem of puing question to user using natural language;
Response unit 102 is based on that data is putd question to retrieve the corresponding reply data of enquirement data in ontology knowledge base 103, and passes through interaction
Unit 101 exports search result (structural knowledge such as without answer, specific reply data or rhetorical question).
When it is implemented, interactive unit 101 may be mounted at client device (such as mobile phone, tablet computer, wearable device
Deng) on application software, such as electric business application, wechat enterprise;Response unit 102 can be a hardware ontology, such as conventional
Server, cloud device etc., a virtual center being also possible on hardware device, can also be and be mounted in server device
The application program for being able to achieve above-mentioned search function, the embodiment of the present application is not especially limited this.Ontology knowledge base 103 can be with
It is the computer readable storage medium for being stored with above-mentioned sample ontology and polymorphic type knowledge.
Sample ontology in ontology knowledge base it is to be understood that user's demand (including seeking advice from, feeding back) topic.Such as
Around " login password ", this topic has: how to modify, allow to modify number, allow mistake input number, modified where etc..
User can express its demand i.e. problem of user's reality with a variety of different form of presentation, then this topic just can be used as one
Sample ontology.Sample ontology can be obtained by excavating history Question Log, for example, will be primary by analysis of history Question Log
The more more wheels dialogue of frequency of occurrence is regarded as belonging to same topic in session (including more wheel dialogues in a session), by this
Topic is as a sample ontology.Alternatively, a business object is directly as a sample ontology, for example, a certain business object
(a such as commodity), it is all how to buy and (buy one and get one free, discount) about business object, how to use that most of user, which puts question to,
(eating), effect how etc. the problems such as;These problems are all around the business object, it is therefore necessary to as a sample ontology
To safeguard.
The corresponding semi-structured question and answer of sample ontology can be excavated to obtain to knowledge from the question and answer centering being collected into, for example,
Based on the question and answer being collected into the problems in obtain question template, question template can be for Sentence Template, semantic template etc.;Using language
Say knowledge technology to the corresponding answer of the question and answer centering problem as the corresponding response of the question template, to form one group of question and answer
It is right.Wherein, the collection process of question and answer pair can be collected automatically by system, can also be by artificially collecting, and the embodiment of the present application is to this
It is not especially limited;In addition, the process that system collects question and answer pair automatically can be found in related content in the prior art, herein no longer
It repeats.
It is semantic network on the map knowledge essence of the corresponding structuring of sample ontology, is a kind of data knot based on figure
Structure is made of node (point) and side (edge).Node A, B as shown in Figure 1, C ... etc., node A and node B it
Between between a, node B and node C while b ... etc..In map knowledge, each node indicates to deposit in real world
Information, each edge indicates " relationship " between information and information.Generally, map knowledge is exactly all variety classes
Information link together obtained from a relational network.Existing map knowledge architecture side can be used in the map knowledge of structuring
Method realizes that details are not described herein again.
But what needs to be explained here is that: map knowledge architecture method is in multiple types knowledge in the embodiment of the present application
One kind, for the more existing automatically request-answering system based entirely on map knowledge, a large amount of and perfect map knowledge section of Maintenance free
Point need to only safeguard some knowledge that solution can not can not be replied by other types knowledge, safeguard less investment, and safeguard more
Add simple.
Non-structured text knowledge can be backstage personnel human-edited generation, be also possible to using web crawlers from
What network side crawled.
In a kind of achievable technical solution, staff can be by editing interface shown in Fig. 3 come artificial maintenance
The corresponding a plurality of types of knowledge of sample ontology are manually added, deleted and/or are deleted.For example, staff can in Fig. 3
The addition of question and answer pair is triggered by clicking the "+addition related question " control on editing interface.After the completion of question and answer are to addition, work
Making personnel can also be triggered by " editor " control on touch-control editing interface to the editors of corresponding question and answer pair.Staff can also lead to
The non-structured text of editing interface editor shown in Fig. 3 is crossed as the sample ontology (the * * set meal in such as Fig. 3) one knows
Point is known, for example, staff can be by " editor sketches " control on touch-control editing interface to trigger text editing operations.
The purport of technical solution provided by the embodiments of the present application is to provide a plurality of types of knowledge for response retrieval, is improved existing
The problem for having technological know-how type single;In addition, all types of knowledge can form a kind of complementary cooperation relation, in a seed type knowledge
It is more difficult to replace another type of knowledge retrieval when retrieving reply data, this will be helpful to reduce response without answer rate, simultaneously
It also ensures higher accuracy rate, improves the automatic question answering experience of user.
Fig. 4 shows the flow diagram of the question and answer data processing method of one embodiment of the application offer.As shown in Figure 1,
It is provided in this embodiment the described method includes:
201, this enquirement data of user's input are received.
202, from the ontology knowledge base, at least one sample ontology pair relevant to this described enquirement data is obtained
At least one set of polymorphic type knowledge answered.
203, reply data is retrieved at least one set of polymorphic type knowledge.
204, response feedback is exported to the user according to search result.
In above-mentioned 201, it is defeated by human-computer interaction interface (such as interface shown in fig. 5) that this puts question to data to can be user
The question sentence text that enters or based on question sentence speech production.It wherein, can based on question sentence speech production process specifically: known using voice
Other technology identifies question sentence voice, writing text is generated according to recognition result, then using the writing text of generation as this
Secondary enquirement data.
In above-mentioned 202, data processing first can be carried out to this enquirement data of user's input before obtaining polymorphic type knowledge,
It is convenient for searching relative sample ontology in ontology knowledge base in this way.A kind of achievable mode is to put question to number to this
According to ontology extraction is carried out, to obtain at least one example ontology;Such as this is carried out to enquirement data " how Taobao's account is modified "
Body extracts, and extraction obtains " Taobao's account " and " modification " two example ontologies.
When it is implemented, ontology extraction process is as follows:
S11, it puts question to data to segment this, obtains multiple participles.
S12, obtained the multiple participle is normalized, to obtain at least one example ontology.
Wherein, natural language processing technique field is belonged to this technology for puing question to data to be segmented.For a language
Sentence, the mankind can judge which is notional word, which is function word, and then understands its meaning by the knowledge of oneself.But for machine
For device processing, then analyzed.The process of participle can be used segmentation methods in the prior art and put question to data to this
It is separated, to retain the participle with Ontology.Wherein, segmentation methods can be using existing based on string matching
Segmenting method, the segmenting method based on understanding or segmenting method based on statistics etc., the embodiment of the present application are not made this specifically
It limits.
In addition, the enquirement sentence of equivalent, different user can be expressed using different clause and different vocabulary.
For example, user 1 puts question to " how can be with reimbursement? ";User 2 puts question to " reimbursement mode is? ";User 3 puts question to " how having the money refunded ";This three
Kind enquirement is to be expressed equivalent in meaning, but is different by multiple participles that participle obtains.At this point, for unified word
Facilitate later retrieval, needs for different participles to be normalized.Wherein, the process of normalized simply understands to be exactly to incite somebody to action
Participle with identical expression is converted to same example ontology.It is corresponded to for example, the mode tabled look-up can be used to obtain each participle
Example ontology.
Sample ontology relevant to this enquirement data may is that in ontology knowledge base puts question to data pick-up to go out with from this
The identical sample ontology of example ontology, or put question to data pick-up to go out the similar sample ontology of example ontology (i.e. with from this
Example sample and sample ontology are synonym).
Certainly, it may also be searched in ontology knowledge base identical less than with the example ontology for puing question to data pick-up to go out from this
Or the case where similar sample ontology.At this point, puing question to the relevant sample ontology of data can be with to this in ontology knowledge base are as follows: with
The relevant sample data of previous enquirement data.Wherein, occur in previous enquirement data and this enquirement same session of data
It puts question to twice, which, which can be, puts question to the previous enquirement data before data earlier than this, before being also possible to
Two enquirement data.
When the example ontology that cannot be found in ontology knowledge base with put question to data pick-up to go out from this is same or similar
Sample ontology when, represent user continue saying same topic, then can continue to use related with previous enquirement data at least one
A sample ontology.Is such as the above-mentioned example enumerated: user puts question to 1: how Taobao's account modified? and then user puts question to 2: can
Is modification several times? for puing question to for 2, in ontology knowledge base be can not find it is identical as the example ontology extracted from enquirement 2 or
Similar sample ontology, therefore, put question to 2 can also continue to using with put question to 1 at least one related sample ontology, such as " Taobao
Account ".
Above content is to put question to data for non-the case where puing question to for the first time for this, if it is to mention for the first time that this, which puts question to data,
It asks, and cannot be found in ontology knowledge base and put question to the same or similar sample of example ontology of data pick-up out with from this
Ontology, then can be directly to user feedback without answer;Or data-speculative user is putd question to be intended to based on this, and be intended to according to user
At least one sample ontology is recommended to select for user for user.For example, response unit can be sent to interactive unit carries promising use
The rhetorical question response at least one sample ontology that family is recommended;Interactive unit is in response to user for providing in the rhetorical question response
At least one sample ontology that the selection event is directed toward is fed back to response list by the selection event of at least one sample ontology
User can be passed through at least one the sample ontology for selecting event confirmation to select and put question to data phase as with this by member, response unit
At least one the sample ontology closed.
When it is implemented, user can pass through interactive interface or interactive interface triggering selection event.For example, user can pass through touching
It controls the touch screen of interactive device (such as mobile phone, desktop computer, tablet computer) or clicks target sample ontology using mouse/keyboard
Carry out triggering selection event;Or the audio interface input audio letter that user is provided by interactive device (such as intelligent sound box, mobile phone)
Breath carrys out triggering selection event etc., and the present invention is not especially limit this.
To sum up, it provided in this embodiment 202 can specifically be realized using following steps:
2021, it puts question to data to carry out ontology extraction described this, obtains at least one example ontology.
2022, it in the ontology knowledge base, searches and described at least one example ontology the same or similar at least one
A sample ontology.
If 2023, finding, it is corresponding at least that at least one described sample ontology is obtained from the ontology knowledge base
One group of polymorphic type knowledge.
If 2024, not finding, the attribute information of data is putd question to determine at least one described sample according to described this
Ontology, and the corresponding at least one set of polymorphic type knowledge of at least one described sample ontology is obtained from the ontology knowledge base.
Wherein, this in above-mentioned 2024 puts question to the attribute information of data can include: this puts question to the question and answer mark of data
(ChatID) or session identification (SessionID).SessionID indicates that user enters a session.It is asked automatically when user exits
When answering interface and being again introduced into, another unique seesionID can be regenerated to indicate being a new session.ChatID
It indicates each of session and puts question to data.ChatID can be one from the number increased, and ChatID value is bigger, indicates to mention
Ask the sequence of data more rearward.If the ChatID of each enquirement data in each session is incremented by since preset initial value,
Then it can determine that this puts question to whether data are to put question to for the first time directly according to ChatID;Otherwise, it can pass through inquiry SessionID's
Whether previous enquirement data are had in session to determine that this puts question to whether data are to put question to for the first time.
Retrieving reply data in above-mentioned 203 in polymorphic type knowledge can be used the following two kinds mode and realizes:
Mode one first retrieves reply data in a seed type knowledge, if retrieving reply data, by the reply data
As search result;If not retrieving reply data, reply data is retrieved in another seed type knowledge, until all types
Reply data both participate in retrieval until.Specific step is as follows:
2031, reply data is retrieved at least one set of first kind knowledge.
If 2032, retrieving reply data, using the reply data as the search result.
If 2033, not retrieving reply data, reply data is retrieved at least one set of Second Type knowledge, until institute
It states until all types knowledge at least one set of polymorphic type knowledge both participates in retrieval.
Wherein, the selection of multiple types knowledge successively can be random, or selects successively according to preset type priority grade.
I.e. according to type priority grade, the knowledge type for participating in retrieval is sequentially chosen from least one set of polymorphic type knowledge.Assuming that
Multiple types knowledge includes three types knowledge, is respectively as follows: the map knowledge of structuring, semi-structured question and answer to knowledge and non-
The text knowledge of structuring.Preset type priority are as follows: semi-structured question and answer are to knowledge, the map knowledge of structuring, non-
The text knowledge of structuring.Correspondingly, aforesaid way can specifically:
First at least one set of semi-structured question and answer to retrieving reply data in knowledge;
If at least one set of semi-structured question and answer to retrieving reply data in knowledge, using the reply data as
The search result;Otherwise, then in the map knowledge of at least one set of structuring retrieve reply data;
If retrieving reply data in the map knowledge of at least one set of structuring, using the reply data as retrieval
As a result;Otherwise, reply data is retrieved at least one set of non-structured text knowledge;
If retrieving reply data in the map knowledge of at least one set of structuring, tied reply data as retrieval
Fruit;It otherwise, will be without answer as the search result.
What needs to be explained here is that: what is referred in the application retrieves reply data, refer to the reply data that retrieves with
This reply data for puing question to the degree of correlation of data to meet relevance threshold.Wherein, the embodiment of the present application is to relevance threshold
Value is not especially limited.
Mode two, each type knowledge all carry out reply data retrieval, if retrieving two or more answer numbers
According to then by calculating the degree of correlation of each reply data and this enquirement data, using the high reply data of the degree of correlation as retrieval
As a result.Specific step is as follows:
2031 ', reply data is retrieved at least one set of polymorphic type knowledge respectively.
If 2032 ', retrieving two or more reply datas, described two or more than two responses are calculated
Data put question to the degree of correlation of data with described this respectively.
2033 ', it according to the degree of correlation, is chosen from described two or more than two reply datas and is used as the retrieval
As a result at least one reply data.
Wherein, reply data and the respective algorithms for puing question to the degree of correlation between data that the prior art can be used are realized, this Shen
Please embodiment this is not especially limited.
There is also the need to explanations: the above-mentioned search method that reply data is retrieved in a seed type knowledge can be found in existing
There is the related content in technology, the application search method is not especially limited.
In a kind of achievable embodiment, above-mentioned 2033 ' can specifically realize with the following method: from described two or two
In a above reply data, the highest reply data of the degree of correlation is chosen as the search result.
In another achievable embodiment, above-mentioned 2033 ' can also specifically realize with the following method: from described two
Or in more than two reply datas, the highest reply data of the degree of correlation is chosen as the first response, the degree of correlation is chosen and is inferior to institute
At least one reply data of optimal response is stated as the second response, using first response and second response as described in
Search result.
In technical solution provided by the embodiments of the present application, ontology knowledge base is a plurality of types of knowing of providing of response retrieval
Know, not only improves the single problem of prior art knowledge type, and all types of knowledge can form a kind of complementary cooperation relation,
One seed type knowledge is more difficult to retrieve replaceable another type of knowledge retrieval when reply data, this will be helpful to reduce response
It without answer rate, while also ensuring higher accuracy rate, improves the automatic question answering experience of user.
Further, since technical solution provided by the embodiments of the present application is to retrieve to answer at least one set of polymorphic type knowledge
Answer evidence, it is more likely that the reply data retrieved occur is at least two, and at least two reply datas are from different groups
The case where first kind knowledge (i.e. across sample ontology).For example, the reply data retrieved is two, one of reply data
It is to be retrieved from the corresponding polymorphic type knowledge of sample ontology 1, another reply data is from the corresponding multiclass of sample ontology 2
It is retrieved in type knowledge.For this purpose, the method provided by the embodiments of the present application, may also include the steps of:
If the reply data retrieved is at least two, and at least two reply datas from different groups the
When one type knowledge, judge whether at least two reply datas are same or similar;
It is if they are the same or similar, then merge at least two reply datas;
If not same or similar, at least two reply datas are carried to user transmission and respectively correspond to sample
First rhetorical question information of ontology, according to the selection event that the user executes for the first rhetorical question information, by the selection
At least one corresponding reply data of sample ontology that event is directed toward is as search result.
Need exist for supplement: method provided by the embodiments of the present application may also include that selection event direction
At least one sample ontology is recorded as sample ontology relevant to this described enquirement data.Why the step, mesh are increased
Be in order to it is subsequent enquirement data retrieval when occur from ontology knowledge base search less than with from it is subsequent put question to data pick-up sheet
The same or similar sample ontology of body ontology is prepared.
Above-mentioned is a kind of achievable processing mode when retrieving at least two reply datas, i.e., by multiple reply datas into
Row merges or selects event therefrom to select at least one reply data as search result based on user.Another implementable
Mode in, can regard at least two reply datas retrieved as search result, can also be from least two reply datas
One, which is selected, as the first response (can be regarded as preferred response) selects at least one (can simply be interpreted as standby as the second response
Select response).
Wherein, the first response and the second response can show user simultaneously, the first response can also be used conventional display side
Formula is shown, and the second response shows the displaying of breviary around the first response.
Below for retrieving 3 reply datas, above content is illustrated.
Assuming that 3 reply datas are respectively as follows: reply data A, reply data B and reply data C, 3 reply data difference
This puts question to the similarity of data with user are as follows: a > b > c.
Firstly, selecting this with user to put question to the phase of data from three reply data A, reply data B and reply data C
It is the first response like highest reply data A is spent, similarity is inferior to the reply data B and/or reply data C of reply data A
As the second response.
Then, user is fed back to using the first response and the second response as search result.
Interactive unit receives and can show the first response and the second response according to sequencing after the search result and answering
It answers in interface, as shown in Figure 5;Alternatively, the second response is hidden as shown in fig. 6, the first reply data is presented in response interface
In combobox, in response to user, to dropdown control, (" answer is dissatisfied to Click here and can check user's click in such as Fig. 6
Touch control operation more ") shows the second response, i.e. reply data A and/or reply data C.Wherein, dropdown control can show line
Around first response display area, for example, underface or rightmost are (such as Fig. 6 institute in the lower left corner of the first response display area
Show) etc., the embodiment of the present application is not especially limited this.
Further, the structural knowledge of rhetorical question is provided in the ontology knowledge base provided in the embodiment of the present application
Point.The structural knowledge point of the rhetorical question can be used as a kind of reply data.The method i.e. provided by the embodiments of the present application can also wrap
It includes:
205, include in the search result rhetorical question structural knowledge point when, Xiang Suoshu user sends based on described anti-
The second rhetorical question information that the structural knowledge point asked generates.
206, user is received for the rhetorical question response of the second rhetorical question information feedback.
207, it according to the rhetorical question response, relocates described this and puts question to data.
208, at least one reply data recommended is sent based on this enquirement data, Xiang Suoshu user described in reorientation
Or it is retrieved in the ontology knowledge base again.
By the way that the structural knowledge point of the rhetorical question is arranged, automatically request-answering system can be based on the structural knowledge of the rhetorical question
Point actively initiates to ask in reply to user, puts question to data to collect more information with user interaction to relocate this, facilitates
Provide more accurate answer.
Assuming that user puts question to " account, which is locked out, to be changed? " based on the enquirement data retrieval to search result in wrap
Structural knowledge point containing rhetorical question, such as Taobao's account is locked, trade company is punished locking;It can be sent at this time to user and be based on being somebody's turn to do
What the structural knowledge point of rhetorical question generated " be Taobao's account is locked or trade company is punished locking " second asks in reply information.User
Rhetorical question response be " Taobao's account locked ";Then combinable " Taobao's account is locked " " account, which is locked out, to be changed? "
Relocate the enquirement of user.User based on repositioning puts question to, and directly can send at least one recommended to the user
Reply data is retrieved in the ontology knowledge base again.
Fig. 7 shows the flow diagram of the automatic question-answering method of one embodiment of the application offer.As shown in fig. 7, this reality
Applying the automatic question-answering method that example provides includes:
301, this enquirement data of user's input are received.
302, data are putd question to carry out ontology extraction this.
If the example ontology 303, extracted is one, and includes and an example sheet in the ontology knowledge base
The same or similar sample ontology A of body then retrieves answer number in the corresponding one group of polymorphic type knowledge of sample ontology A
According to, and stored sample ontology A as upper body of text.
If the example ontology 304, extracted is multiple, and only have in the ontology knowledge base sample ontology B with
An example ontology in multiple example ontology is same or similar, then in the corresponding one group of polymorphic type knowledge of sample ontology B
Middle retrieval reply data, and stored sample ontology B as upper body of text.
If the example ontology 305, extracted is multiple, and has multiple sample ontology C in the ontology knowledge base and be somebody's turn to do
The example ontology of respective numbers in multiple example ontologies is same or similar, then in multiple sample ontology C corresponding one
Reply data is retrieved in group polymorphic type knowledge, and empties stored upper body of text.
If 306, the example ontology extracted does not have same or similar sample ontology in the ontology knowledge base,
At least one sample ontology (i.e. upper body of text) relevant to previous enquirement data is obtained, and at least one sample ontology pair
Reply data is retrieved in at least one set of polymorphic type knowledge answered.
Wherein, for situation shown in 306, upper body of text is constant.
What needs to be explained here is that: the parameter of a upper body of text is increased in the present embodiment.Body of text is adjoint on this
One session whole process is generated when a session is activated, quilt when conversation end (i.e. user turns off question and answer interface)
It deletes.Parameter of body of text is exactly to store at least one sample ontology relevant to previous enquirement data on this.This reality
Apply be arranged in example upper body of text purpose be exactly in order to when subsequent enquirements data retrieval appearance from ontology knowledge base lookup less than
It prepares with from the subsequent ontology ontology the same or similar sample ontology for puing question to data pick-up.Shown in above-mentioned 305 step
In situation by the purpose that stored upper body of text is purged be because multiple sample ontology C for subsequent enquirement data retrieval
When there is above situation (having multiple example ontologies of multiple sample ontologies and corresponding number same or similar in body knowledge base)
The effect of this when is little, it is therefore desirable to user assist from multiple sample ontology C selection or to multiple sample ontology C of selection into
User's selection or spliced sample ontology is saved as upper body of text by the operations such as row splicing.Institute i.e. provided by the above embodiment
State method, further includes:
If the reply data retrieved in the corresponding one group of polymorphic type knowledge of multiple sample ontology C is at least two
A, at least two reply datas are not identical or dissimilar and are originated from different groups of polymorphic type knowledge, then send and take to the user
The first rhetorical question information that sample ontology is respectively corresponded to at least two reply data, according to the user for described the
The selection event that one rhetorical question information executes, the corresponding reply data of at least one sample ontology that the selection event is directed toward are made
For search result, and the upper body of text removed is updated at least one sample ontology C that the selection event is directed toward.
There is also the need to supplements: in the case of constant for body of text upper in above-mentioned steps 306, be based on upper body of text pair
At least one set of polymorphic type knowledge answered puts question to data to retrieve this, if retrieving less than reply data, exportable no answer,
And record primary retrieval failure;If when user puts question to data again in ontology knowledge base again without with again put question to data phase
The sample ontology of pass can then continue based on the corresponding at least one set of polymorphic type knowledge retrieval reply data of upper body of text, if still
Retrieval less than, no answer can be continued to output, and again record primary retrieval failure;If twice based on upper body of text retrieval failure,
It then needs to empty body of text.The problem of appearance unsuccessfully empties body of text twice, illustrates user has deviated from topic, if continuing to make
It is retrieved with the corresponding at least one set of polymorphic type knowledge of body of text, only increases the probability without answer.
Content in relation to above-mentioned 301~306 is no longer described in detail herein referring also to above-described embodiment.
In technical solution provided by the embodiments of the present application, ontology knowledge base is a plurality of types of knowing of providing of response retrieval
Know, not only improves the single problem of prior art knowledge type, and all types of knowledge can form a kind of complementary cooperation relation,
One seed type knowledge is more difficult to retrieve replaceable another type of knowledge retrieval when reply data, this will be helpful to reduce response
It without answer rate, while also ensuring higher accuracy rate, improves the automatic question answering experience of user.
Fig. 8 is shown in automatic question-answering method provided by the embodiments of the present application in the semi-structured question and answer of multiple groups in knowledge
Retrieve the flow diagram of reply data.As shown in Figure 8, comprising:
401, the semi-structured question and answer of at least one set are calculated to mention all problems in knowledge with described this respectively
Ask the similarity of data.
402, subject is spliced to each problem in knowledge for at least one set semi-structured question and answer, and calculates splicing
All problems after subject put question to the similarity of data with described this respectively.
403, will all problems and it is all splicing subject after problem in it is described this put question to data similarity be more than
Similarity threshold, and make for the corresponding answer of problem after at least one problem of maximum value and/or at least one splicing subject
For the reply data retrieved.
Wherein, similarity calculation can be referring specifically to related content in the prior art, and details are not described herein again.
In above-mentioned 402 for problem splicing subject can be it is pre-set, be also possible to system according to this put question to number
It it is theorized that and analyze, specific the present embodiment is not construed as limiting this.
Further, problem after all problems and all splicing subjects puts question to the similarity of data not surpass with this
When crossing similarity threshold, reply data is no answer.
Fig. 9 is shown in automatic question-answering method provided by the embodiments of the present application in the map knowledge of at least one set of structuring
Retrieve the flow diagram of reply data.As shown in Figure 9, comprising:
501, data generative semantics chain is putd question to based on described this, includes at least two examples in the semanteme chain information
Relation information between ontology and two adjacent instances ontologies.
502, the first semantic section with first example Ontology Matching is searched in the map knowledge of at least one set of structuring
Point.
503, using the relation information between the first example ontology and the second example ontology, at least one set of structuring
Map knowledge in search the second semantic node;Continue to search third semantic node using this step until all example ontologies pair
Until the semantic node answered is found, using the semantic node finally found as the reply data.
For the convenience of explanation, as shown in Figure 10, the exemplary relational graph of map knowledge is given by taking colleges and universities teacher as an example.
Wherein, " education ", " university ", " Lee * * ", " * * university ", " C institute ", " D profession " are knowledge points, these knowledge points pass through " religion
Award ", " school ", " tenure ", " department ", the relatival connection such as " profession " and form knowledge relation figure.Assuming that the enquirement of user
Data are " which institute * teacher * Lee of * * university is? ", " * * university/Lee * */is obtained after entity extraction processing
Relation information is to teach between institute ", and " * * university/Lee * * ", and relation information is department between " Lee * */institute ".It is using
When the above method carries out knowledge reasoning, first corresponding to example ontology " * * university " is found in map knowledge shown in Fig. 10
Then a semantic node finds second according to the relation information " teaching " of example ontology " * * university " and example ontology " Lee * * "
Semantic node;The semantic section of third is found further according to the relation information " department " of example ontology " Lee * * " and example ontology " institute "
Point, so far all example ontologies find corresponding semantic node, and third semantic node " C institute " is used as reply data.
Figure 11 is shown in automatic question-answering method provided by the embodiments of the present application to be known at least one set of non-structured text
The flow diagram of reply data is retrieved in knowledge.As shown in figure 11, comprising:
601, at least one set of non-structured text knowledge, destination document relevant to this enquirement data is obtained.
602, the degree of association in data in each problem participle and destination document between each document participle is putd question to according to this, from
The document snippet that can be used as reply data is extracted in destination document.
Machine reading model can be used to realize in step described in above-mentioned 601 and 602.Machine reading model can there are many
Implementation, one exemplary embodiment of the application provide a kind of implementation of machine reading model.As shown in figure 12, the machine
Reading model is a deep neural network model, such as can be Recognition with Recurrent Neural Network (Recurrent
NeuralNetwork, RNN) model includes four layers altogether.Wherein, first layer is input layer (Input Layer), and the second layer is to compile
Code layer (Encoding Layer), third layer are concerns layer (Attention Layer), and the 4th layer is output layer (Output
Layer)。
Wherein it is possible to obtain machine reading model shown in Figure 12 by model training.In model training stage, need to mention
Ask data, to put question to data it is related can provide search answer document (i.e. destination document) and destination document in corresponding answer
Content is sent into input layer as input.After obtaining machine reading model shown in Figure 12, online service stage can be entered.
In model service stage, need to put question to the document participle conduct in the problems in data participle and destination document
Input layer is sent into input.As shown in figure 12, input layer includes two parts, and a part is to need the problem of inputting participle, a part
It is the document participle for needing to input.Either problem participle or document participle, the corresponding input item of each participle, are denoted as word
Item (term).Each term is mainly by term vector (Word Embedding), the word vector of the corresponding participle of the term
(Character Embedding) and ring characteristics (Surround Features) connection composition.Term vector, word vector are
Digitized, Surround Features used herein also uses binary system (binary) feature, to realize digitlization.
Optionally, the corresponding Surround Features of each term may include whether the corresponding participle of the term is puing question to number here
According to or destination document in the feature that occurs, can also include word frequency-inverse document frequency (TermFrequency-of the term
InverseDocumentFrequency, TFIDF) weight (weighted).
As shown in figure 12, the output of input layer is by the input as coding layer.It, can be corresponding to each participle in coding layer
Term carries out coding to export the coding vector of each participle.Optionally, coding layer can use LSTM or GRU or multilayer
CNN realization, but not limited to this.
As shown in figure 12, the output of coding layer will be as the input for paying close attention to layer.In machine reading model, concern layer is one
Kind imitates human brain attention mechanism, finds the mechanism for generating biggest impact part to result in image or text and is paying close attention to layer, can
Relevant calculation is carried out with the coding vector of the coding vector for segmenting each problem and each document participle, to obtain each problem participle
With each document participle between the degree of association, in order to output layer can find in destination document to put question to the maximum text of data influence
Shelves participle.Optionally, the relevant calculation used here, which can be, directly carries out dot product for two coding vectors, and dot product result is made
For the similarity between two coding vectors;Alternatively, the Ming Shi distance of two coding vectors can also be calculated, Ming Shi distance is made
For the similarity between two coding vectors;Alternatively, the difference between the maximum value in two coding vectors can also be calculated, it will
The difference is the similarity being used as between two coding vectors;Alternatively, can also calculate minimum value in two coding vectors it
Between difference, be to be used as similarity between two coding vectors by the difference.
As shown in figure 12, the output of layer is paid close attention to by the input as output layer.As shown in figure 12, Zong Xiangshang, output layer packet
Two sublayers are included, are to calculate sublayer and selection sublayer respectively;Horizontally, output layer includes two-way calculate node, is needle all the way
It is the calculate node for answer end point all the way to the calculate node of answer starting point.Sublayer is being calculated, will opened for answer
The calculate node of initial point is denoted as Start_Score, will be denoted as End_Score for the calculate node of answer end point;It is selecting
Sublayer will be denoted as Start_Mask for the calculate node of answer starting point, will be denoted as the calculate node of answer end point
End_Mask.As shown in figure 12, sublayer is being calculated, the output result of calculate node Start_Score is in addition to being sent into calculate node
Except Start_Mask, it is also necessary to output it result and be sent into calculate node End_Score.
Calculate node Start_Score is mainly according to the pass between each problem participle and each document participle of concern layer output
Connection degree calculates probability of each document participle as answer starting point.Calculate node End_Score is mainly according to calculate node
Start_Score output each document participle as answer starting point probability and concern layer output each problem participle with respectively
The degree of association between document participle calculates probability of each document participle as answer end point.Calculate node Start_Mask and
Calculate node End_Mask mainly carries out the various processing such as output weighting, normalization to each term, it is therefore an objective to control output
Mode, in order to support structuring to export or support output on demand.
Wherein, a kind of process for carrying out online question and answer based on machine reading model shown in Figure 12 is as shown in figure 13, including with
Lower operation:
Problem receives: the mainly enquirement data of reception user transmission.
Document positioning: mainly according to data are putd question to, determination can furnish an answer from business scenario relevant document
Destination document.It is alternatively possible to using but be not limited to: text classification, text retrieval or question template intercept positioning target text
Shelves.In document position fixing process, if necessary, can be cached to document.
Pretreatment: mainly destination document and enquirement data are pre-processed, machine shown in Figure 12 is met with acquisition and is read
Read the vectorization information of the input layer demand of model.Here pretreatment includes for enquirement data and for the pre- of destination document
Treatment process is similar, and main includes participle, extraction language feature and vectorization.Optionally, destination document is pre-processed
In the process, it is contemplated that document may can be formatted processing with structural informations such as multistage paragraph, special tags.
For example, word segmentation result may include three problems participle so as to put question to data be " Huawei's mobile phone is how " as an example, point
Be not " Huawei ", " mobile phone ", " how ".To these three problems segment, can determine respectively its corresponding term vector and word to
Amount.By taking " mobile phone " as an example, " mobile phone " corresponding vector expression way, and " hand " and " machine " corresponding vector expression can be determined
Mode.In addition, these three participles can be counted its language feature respectively and be gone forward side by side row vector.By taking " mobile phone " as an example, part of speech
It is noun, occurs in puing question to data, belong to subject in puing question to data, it is assumed that counting its TFIDF is 0.05.To " Huawei ",
" mobile phone " and " how " for three problem participles, its term vector, word vector and feature vector can be connected into shape
At corresponding term.Document participle can be processed similarly.
Model calls: mainly machine reading model shown in calling figure 12.By the vectorization letter of above-mentioned formation respectively segmented
Breath is sent into input layer respectively as input, and each layer is successively handled, until output layer exports each document participle and rises as answer
The probability of initial point and answer end point.
Answer extracting: each document participle is mainly exported as the general of answer starting point and answer end point according to output layer
Rate extracts the document snippet that can be used as answer from destination document, and enters post-processing stages.
Post-processing: mainly extracted document snippet is post-processed, to improve readability.
Figure 14 shows the flow diagram for the automatic question-answering method that another embodiment of the application provides.As shown in figure 14,
Include:
701, in response to the enquirement event of user's triggering, this enquirement data of the user are obtained.
702, data are putd question to be sent to server-side described this.
703, the server-side is exported based on the response feedback that this puts question to data to send.
Wherein, the response feedback is generated according to search result, and the search result is using at least one response
Search modes put question to relevant at least one sample of data corresponding at least one set of multiclass in ontology knowledge base to described this
Retrieval obtains in type knowledge.
What needs to be explained here is that: the generation of search result described in above content can be found in the phase in the various embodiments described above
Hold inside the Pass, is no longer described in detail herein.
User can trigger enquirement event by automatic question answering interface or question and answer interface in above-mentioned 701.For example, user can pass through
Input puts question to data and triggers the enquirement event after clicking transmission on automatic question answering interface, alternatively, user passes through touch-control
Voice control on automatic question answering interface and say put question to sentence after trigger the enquirement event;Or user passes through touch-control
Intelligent answer equipment (such as intelligent sound box) provide audio collection control and say put question to sentence after trigger the enquirement event
Deng the present invention is not especially limit this.
Exporting response feedback in above-mentioned 703 can specifically: is answered described in display in the question and answer interface in a manner of text
Answer feedback;Or the response feedback is exported in a manner of voice, etc., the embodiment of the present application is not especially limited this.
In technical solution provided by the embodiments of the present application, ontology knowledge base is a plurality of types of knowing of providing of response retrieval
Know, not only improves the single problem of prior art knowledge type, and all types of knowledge can form a kind of complementary cooperation relation,
One seed type knowledge is more difficult to retrieve replaceable another type of knowledge retrieval when reply data, this will be helpful to reduce response
It without answer rate, while also ensuring higher accuracy rate, improves the automatic question answering experience of user.
In some processes described in the description of the present application, claims and above-mentioned attached drawing, contain according to spy
Multiple operations that fixed sequence occurs, these operations can not be executed according to its sequence what appears in this article or be executed parallel.
Serial number of operation such as 201,202 etc. is only used for distinguishing each different operation, and it is suitable that serial number itself does not represent any execution
Sequence.In addition, these processes may include more or fewer operations, and these operations can be executed in order or be held parallel
Row.It should be noted that the description such as herein " first ", " second ", be for distinguishing different message, equipment, module etc.,
Sequencing is not represented, " first " and " second " is not also limited and is different type.
Figure 15 shows the structural schematic diagram of the automatic call answering arrangement of one embodiment of the application offer.As shown in figure 15, originally
Embodiment provide described device include:
Receiving module 801 is used to receive this enquirement data of user's input;
First, which obtains module 802, is used for from the ontology knowledge base, obtains and puts question to data relevant extremely to described this
The corresponding at least one set of polymorphic type knowledge of a sample ontology less;
Retrieval module 803 is used to retrieve reply data at least one set of polymorphic type knowledge;
Feedback module 804 is used to export response feedback to the user according to search result.
Further, the first acquisition module 802 is also used to put question to data to carry out ontology extraction described this, obtains
At least one example ontology;In the ontology knowledge base, search the same or similar extremely at least one described example ontology
A few sample ontology;If finding, obtained from the ontology knowledge base at least one described sample ontology it is corresponding to
Few one group of polymorphic type knowledge;If not finding, according to described this put question to the attribute information of data determine it is described at least one
Sample ontology, and obtain the corresponding at least one set of polymorphic type of at least one described sample ontology from the ontology knowledge base and know
Know.
Further, the first acquisition module 802 is also used to put question to the attribute information of data according to described this, determines
It is described this put question to data whether headed by time this enquirement;It puts question to, then obtains relevant to previous enquirement data for the first time if non-
At least one described sample ontology.
Further, the retrieval module 803 is also used to retrieve reply data at least one set of first kind knowledge;If
Reply data is retrieved, then using the reply data as the search result;If not retrieving reply data, at least one
Reply data is retrieved in group Second Type knowledge, until all types knowledge at least one set of polymorphic type knowledge both participates in
Until retrieval.
Further, described device further include: choose module be used for according to type priority grade, sequentially from it is described at least
The knowledge type for participating in retrieval is chosen in one group of polymorphic type knowledge.
Further, the first kind knowledge is semi-structured question and answer to knowledge.Above-mentioned retrieval module 803 is also used
The phase of data is putd question to described this respectively to all problems in knowledge in calculating the semi-structured question and answer of at least one set
Like degree;Be the semi-structured question and answer of at least one set in knowledge each problem splice subject, and calculate splicing subject after
All problems respectively with it is described this put question to data similarity;By all problems and it is all splicing subject after problem in
It is described this put question to data similarity be more than similarity threshold, and for maximum value at least one problem and/or at least one
The corresponding answer of problem after splicing subject is as the reply data retrieved.
Further, described device may also include that
Judgment module, if the reply data for retrieving is at least two, and at least two reply datas
When from different groups of first kind knowledge, judge whether at least two reply datas are same or similar;
Merging module then merges at least two reply datas for if they are the same or similar;
The feedback module, if be also used to it is not same or similar, to the user send carry described at least two
Reply data respectively corresponds to the first rhetorical question information of sample ontology, is executed according to the user for the first rhetorical question information
Selection event, using described at least one corresponding reply data of sample ontology for selecting event to be directed toward as search result.
Further, described device further include: logging modle is used at least one sample for being directed toward the selection event
This ontology is recorded as sample ontology relevant to this described enquirement data.
Further, the retrieval module 803 is also used to retrieve response at least one set of polymorphic type knowledge respectively
Data;If retrieving two or more reply datas, described two or more than two reply data difference are calculated
The degree of correlation of data is putd question to described this;According to the degree of correlation, selected from described two or more than two reply datas
It is taken as at least one reply data for the search result.
Further, above-mentioned retrieval module 803 can also be used in:
From described two or more than two reply datas, the highest reply data of the degree of correlation is chosen as the retrieval
As a result;Or
From described two or more than two reply datas, chooses the highest reply data of the degree of correlation and answered as first
Answer, choose at least one reply data of the degree of correlation inferior to the optimal response as the second response, will first response with
Second response is as the search result.
Further, described device may also include that
When the feedback module 804 is also used in the search result include the structural knowledge point of rhetorical question, Xiang Suoshu
User sends the second rhetorical question information that the structural knowledge point based on the rhetorical question generates;
The receiving module 801 is also used to receive user for the rhetorical question response of the second rhetorical question information feedback;
Module is relocated, for described this being relocated and puing question to data according to the rhetorical question response;
The feedback module 804 is also used to based on this enquirement data described in reorientation, and Xiang Suoshu user, which sends, to be recommended
At least one reply data;Or the retrieval module is also used to exist again based on this enquirement data described in reorientation
It is retrieved in the ontology knowledge base.
Further, the multiple types knowledge includes: the map knowledge of structuring, semi-structured question and answer to knowledge
And/or non-structured text knowledge.
What needs to be explained here is that: automatic call answering arrangement provided by the above embodiment can be realized in above-mentioned each method embodiment
The principle of the technical solution of description, above-mentioned each module or unit specific implementation can be found in corresponding interior in above-mentioned each method embodiment
Hold, details are not described herein again.
In technical solution provided by the embodiments of the present application, ontology knowledge base is a plurality of types of knowing of providing of response retrieval
Know, not only improves the single problem of prior art knowledge type, and all types of knowledge can form a kind of complementary cooperation relation,
One seed type knowledge is more difficult to retrieve replaceable another type of knowledge retrieval when reply data, this will be helpful to reduce response
It without answer rate, while also ensuring higher accuracy rate, improves the automatic question answering experience of user.
Figure 16 shows the structural schematic diagram for the automatic call answering arrangement that another embodiment of the application provides.As shown in figure 16,
Described device provided in this embodiment includes:
The enquirement event that second acquisition module 901 is used to trigger in response to user obtains this enquirement number of the user
According to;
Sending module 902 is used to put question to data to be sent to server-side described this;
Output module 903 is used to export the server-side based on the response feedback that this puts question to data to send;
Wherein, the response feedback is generated according to search result, and the search result is using at least one response
Search modes put question to relevant at least one sample of data corresponding at least one set of multiclass in ontology knowledge base to described this
Retrieval obtains in type knowledge.
Further, the output module 903 is also used to answer described in display in the question and answer interface in a manner of text
Answer feedback;Or the response feedback is exported in a manner of voice.
What needs to be explained here is that: automatic call answering arrangement provided by the above embodiment can be realized in above-mentioned each method embodiment
The principle of the technical solution of description, above-mentioned each module or unit specific implementation can be found in corresponding interior in above-mentioned each method embodiment
Hold, details are not described herein again.
In technical solution provided by the embodiments of the present application, ontology knowledge base is a plurality of types of knowing of providing of response retrieval
Know, not only improves the single problem of prior art knowledge type, and all types of knowledge can form a kind of complementary cooperation relation,
One seed type knowledge is more difficult to retrieve replaceable another type of knowledge retrieval when reply data, this will be helpful to reduce response
It without answer rate, while also ensuring higher accuracy rate, improves the automatic question answering experience of user.
Figure 17 shows the structural schematic diagrams for the server device that one embodiment of the application provides.As shown in figure 17, described
Server device includes: first memory 1001, first processor 1002 and the first communication component 1003;
First memory 1001, for storing computer program.In addition to this, first memory 1001 are additionally configured to
Various other data are stored to support the operation in server device.The example of these data includes in server device
The instruction of any application or method of upper operation, message, picture, video etc..
First memory 1001 can be by any kind of volatibility or non-volatile memory device or their combination
It realizes, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable
Read-only memory (EPROM), programmable read only memory (PROM), read-only memory (ROM), magnetic memory, flash memory,
Disk or CD.
First communication component 1003 is transmitted for data.
First processor 1002 is coupled with memory 1001, for executing the institute stored in the first memory 1001
Program is stated, to be used for:
This enquirement data of user's input are received by first communication component 1001;
From the ontology knowledge base, it is corresponding to obtain at least one sample ontology relevant to this described enquirement data
At least one set of polymorphic type knowledge;
Reply data is retrieved at least one set of polymorphic type knowledge;
According to search result, response feedback is exported to the user by first communication component 1001.
Further, first processor 1002 can also be used in: to it is described this put question to data carry out ontology extraction, obtain to
A few example ontology;In the ontology knowledge base, search the same or similar at least at least one described example ontology
One sample ontology;If finding, it is corresponding at least that at least one described sample ontology is obtained from the ontology knowledge base
One group of polymorphic type knowledge;If not finding, the attribute information of data is putd question to determine at least one described sample according to described this
This ontology, and the corresponding at least one set of polymorphic type knowledge of at least one described sample ontology is obtained from the ontology knowledge base.
Further, first processor 1002 can also be used in: puing question to the attribute information of data according to described this, determines institute
State this put question to data whether headed by time this enquirement;It is putd question to for the first time if non-, then obtains institute relevant to previous enquirement data
State at least one sample ontology.
Further, first processor 1002 can also be used in: retrieve reply data at least one set of first kind knowledge;
If retrieving reply data, using the reply data as the search result;If not retrieving reply data, at least
Reply data is retrieved in one group of Second Type knowledge, until all types knowledge at least one set of polymorphic type knowledge is joined
Until retrieval.
Further, first processor 1002 can also be used in: sequentially more from at least one set according to type priority grade
The knowledge type for participating in retrieval is chosen in type knowledge.
Further, when first kind knowledge be semi-structured question and answer to knowledge when, first processor 1002 also can be used
In: it calculates the semi-structured question and answer of at least one set and puts question to the phase of data with described this respectively to all problems in knowledge
Like degree;Be the semi-structured question and answer of at least one set in knowledge each problem splice subject, and calculate splicing subject after
All problems respectively with it is described this put question to data similarity;By all problems and it is all splicing subject after problem in
It is described this put question to data similarity be more than similarity threshold, and for maximum value at least one problem and/or at least one
The corresponding answer of problem after splicing subject is as the reply data retrieved.
Further, first processor 1002 can also be used in: if the reply data retrieved is at least two, and extremely
When few two reply datas are originated from different groups of first kind knowledge, judge whether at least two reply datas are identical
Or it is similar;It is if they are the same or similar, then merge at least two reply datas;If not same or similar, sent out to the user
It send and carries the first rhetorical question information that at least two reply datas respectively correspond to sample ontology, institute is directed to according to the user
The selection event for stating the execution of the first rhetorical question information, at least one corresponding answer number of sample ontology that the selection event is directed toward
According to as search result.
Further, first processor 1002 can also be used in: at least one sample ontology that the selection event is directed toward
It is recorded as sample ontology relevant to this described enquirement data.
Further, first processor 1002 can also be used in: retrieving answer at least one set of polymorphic type knowledge respectively
Answer evidence;If retrieving two or more reply datas, described two or more than two reply datas point are calculated
The degree of correlation of data is not putd question to described this;According to the degree of correlation, from described two or more than two reply datas
Choose at least one reply data as the search result.
Further, first processor 1002 can also be used in: from described two or more than two reply datas, choose
The highest reply data of the degree of correlation is as the search result;Or
From described two or more than two reply datas, chooses the highest reply data of the degree of correlation and answered as first
Answer, choose at least one reply data of the degree of correlation inferior to the optimal response as the second response, will first response with
Second response is as the search result.
Further, first processor 1002 can also be used in: include the structural knowledge of rhetorical question in the search result
When point, Xiang Suoshu user sends the second rhetorical question information that the structural knowledge point based on the rhetorical question generates;User is received to be directed to
The rhetorical question response of the second rhetorical question information feedback;According to the rhetorical question response, relocates described this and put question to data;It is based on
Described this of reorientation puts question to data, and Xiang Suoshu user sends at least one reply data recommended or again in the ontology
It is retrieved in knowledge base.
Further, as shown in figure 17, server device may also include that the first display 1004, the first power supply module 1005,
Other components such as the first audio component 1006.Members are only schematically provided in Figure 17, are not meant to server device only
Including component shown in Figure 17.
Correspondingly, the embodiment of the present application also provides a kind of computer readable storage medium for being stored with computer program, institute
It states and can be realized method and step relevant to server or function in the various embodiments described above when computer program is computer-executed.
In technical solution provided by the embodiments of the present application, ontology knowledge base is a plurality of types of knowing of providing of response retrieval
Know, not only improves the single problem of prior art knowledge type, and all types of knowledge can form a kind of complementary cooperation relation,
One seed type knowledge is more difficult to retrieve replaceable another type of knowledge retrieval when reply data, this will be helpful to reduce response
It without answer rate, while also ensuring higher accuracy rate, improves the automatic question answering experience of user.
Figure 18 shows the structural schematic diagram of the client device of one embodiment of the application offer.As shown in figure 18, described
Client device includes: second memory 1101, second processor 1102 and the second communication component 1103;
The second memory 1101, for storing program.In addition to this, memory 1101 are additionally configured to store it
Its various data is to support operation on the terminal device.The example of these data includes appointing for what is operated on the terminal device
The instruction of what application program or method, contact data, telephone book data, message, picture, video etc..
Memory 1101 can realize by any kind of volatibility or non-volatile memory device or their combination,
Such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable is read-only
Memory (EPROM), programmable read only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk
Or CD.
Second communication component 1103 is transmitted for data.
The second processor 1102, with the second memory, 101 couplings, for executing the second memory
The described program stored in 1101, to be used for:
In response to the enquirement event of user's triggering, this enquirement data of the user are obtained;
Data are putd question to be sent to server-side described this by second communication component;
Output is based on described this by the server-side that second communication component 1103 receives and puts question to data hair
The response feedback sent;
Wherein, the response feedback is generated based on search result, the search result be in ontology knowledge base with
Retrieval obtains in this described corresponding at least one set of polymorphic type knowledge of at least one relevant sample of enquirement data.
Further, the second processor 1102 can also be used in: be shown in the question and answer interface in a manner of text
The response feedback;Or the response feedback is exported in a manner of voice.
Further, as shown in figure 18, client device may also include that second display 1104, second source component 1105,
Other components such as the second audio component 1106.
Correspondingly, the embodiment of the present application also provides a kind of computer readable storage medium for being stored with computer program, institute
It states and can be realized method and step relevant to server or function in the various embodiments described above when computer program is computer-executed.
Communication component in Figure 17 and Figure 18 can be configured to convenient between communication component corresponding device and other equipment
The communication of wired or wireless way.Communication component corresponding device can access the wireless network based on communication standard, such as WiFi, 2G
Or 3G or their combination.In one exemplary embodiment, communication component receives via broadcast channel and comes from external broadcasting pipe
The broadcast singal or broadcast related information of reason system.In one exemplary embodiment, the communication component further includes that near field is logical
(NFC) module is believed, to promote short range communication.For example, radio frequency identification (RFID) technology, infrared data association can be based in NFC module
Meeting (IrDA) technology, ultra wide band (UWB) technology, bluetooth (BT) technology and other technologies are realized.
Display in Figure 17 and Figure 18, may include screen, and screen may include liquid crystal display (LCD) and touching
Touch panel (TP).If screen includes touch panel, screen may be implemented as touch screen, to receive input letter from the user
Number.Touch panel includes one or more touch sensors to sense the gesture on touch, slide, and touch panel.The touch
Sensor can not only sense the boundary of a touch or slide action, but also detect associated with the touch or slide operation hold
Continuous time and pressure.
Power supply module in Figure 17 and Figure 18, the various assemblies for power supply module corresponding device provide electric power.Power supply group
Part may include power-supply management system, one or more power supplys and other with for power supply module corresponding device generate, management and point
With the associated component of electric power.
Audio component in Figure 17 and Figure 18, is configured as output and/or input audio signal.For example, audio component
Including a microphone (MIC), when audio component corresponding device is in operation mode, such as call model, logging mode and voice
When recognition mode, microphone is configured as receiving external audio signal.The received audio signal can be further stored in
Memory is sent via communication component.In some embodiments, audio component further includes a loudspeaker, for exporting audio
Signal.
The apparatus embodiments described above are merely exemplary, wherein described, unit can as illustrated by the separation member
It is physically separated with being or may not be, component shown as a unit may or may not be physics list
Member, it can it is in one place, or may be distributed over multiple network units.It can be selected according to the actual needs
In some or all of the modules achieve the purpose of the solution of this embodiment.Those of ordinary skill in the art are not paying creativeness
Labour in the case where, it can understand and implement.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment can
It realizes by means of software and necessary general hardware platform, naturally it is also possible to pass through hardware.Based on this understanding, on
Stating technical solution, substantially the part that contributes to existing technology can be embodied in the form of software products in other words, should
Computer software product may be stored in a computer readable storage medium, such as ROM/RAM, magnetic disk, CD, including several fingers
It enables and using so that a computer equipment (can be personal computer, server or the network equipment etc.) executes each implementation
Method described in certain parts of example or embodiment.
Finally, it should be noted that above embodiments are only to illustrate the technical solution of the application, rather than its limitations;Although
The application is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it still may be used
To modify the technical solutions described in the foregoing embodiments or equivalent replacement of some of the technical features;
And these are modified or replaceed, each embodiment technical solution of the application that it does not separate the essence of the corresponding technical solution spirit and
Range.
Claims (17)
1. a kind of automatic question-answering method characterized by comprising
Receive this enquirement data of user's input;
From ontology knowledge base, at least one corresponding at least one set of sample ontology relevant to this described enquirement data is obtained
Polymorphic type knowledge;
Reply data is retrieved at least one set of polymorphic type knowledge;
Response feedback is exported to the user according to search result.
2. acquisition is mentioned with described this method according to claim 1, wherein described from ontology knowledge base
Ask at least one relevant sample ontology of data corresponding at least one set of polymorphic type knowledge, comprising:
It puts question to data to carry out ontology extraction described this, obtains at least one example ontology;
In the ontology knowledge base, search and at least one the same or similar sample sheet of at least one described example ontology
Body;
If finding, the corresponding at least one set of polymorphic type of at least one described sample ontology is obtained from the ontology knowledge base
Knowledge;
If not finding, put question to the attribute information of data at least one determining described sample ontology according to described this, and from
The corresponding at least one set of polymorphic type knowledge of at least one described sample ontology is obtained in the ontology knowledge base.
3. according to the method described in claim 2, it is characterized in that, puing question to the attribute information of data to determine institute according to described this
State at least one sample ontology, comprising:
Put question to the attribute information of data according to described this, determine described this put question to data whether headed by time this enquirement;
It is putd question to for the first time if non-, then obtains at least one described sample ontology relevant to previous enquirement data.
4. according to the method in any one of claims 1 to 3, which is characterized in that at least one set of polymorphic type knowledge
Middle retrieval reply data, comprising:
Reply data is retrieved at least one set of first kind knowledge;
If retrieving reply data, using the reply data as the search result;
If not retrieving reply data, reply data is retrieved at least one set of Second Type knowledge, until described at least one
Until all types knowledge in group polymorphic type knowledge both participates in retrieval.
5. method according to claim 4, which is characterized in that further include:
According to type priority grade, the knowledge type for participating in retrieval is sequentially chosen from least one set of polymorphic type knowledge.
6. according to the method described in claim 4, it is characterized in that, the first kind knowledge is semi-structured question and answer to knowing
Know;And
It is described to retrieve reply data at least one set of first kind knowledge, comprising:
It calculates the semi-structured question and answer of at least one set and puts question to data with described this respectively to all problems in knowledge
Similarity;
Be the semi-structured question and answer of at least one set in knowledge each problem splice subject, and calculate splicing subject after
All problems put question to the similarity of data with described this respectively;
It is more than similarity threshold by the similarity of data is putd question to described this in the problem after all problems and all splicing subjects
Value, and be used as and retrieve for the corresponding answer of problem after at least one problem of maximum value and/or at least one splicing subject
The reply data.
7. according to the method in any one of claims 1 to 3, which is characterized in that further include:
If the reply data retrieved is at least two, and at least two reply datas are originated from different groups of the first kind
When type knowledge, judge whether at least two reply datas are same or similar;
It is if they are the same or similar, then merge at least two reply datas;
If not same or similar, at least two reply datas are carried to user transmission and respectively correspond to sample ontology
First rhetorical question information, according to the user for it is described first rhetorical question information execute selection event, by the selection event
At least one the corresponding reply data of sample ontology being directed toward is as search result.
8. the method according to the description of claim 7 is characterized in that further include:
At least one sample ontology that the selection event is directed toward is recorded as sample sheet relevant to this described enquirement data
Body.
9. according to the method in any one of claims 1 to 3, which is characterized in that at least one set of polymorphic type knowledge
Middle retrieval reply data, comprising:
Reply data is retrieved at least one set of polymorphic type knowledge respectively;
If retrieving two or more reply datas, calculate described two or more than two reply datas respectively with
Described this puts question to the degree of correlation of data;
According to the degree of correlation, chosen as the search result at least from described two or more than two reply datas
One reply data.
10. according to the method described in claim 9, it is characterized in that, according to the degree of correlation, from described two or more than two
Reply data in choose at least one reply data as the search result, comprising:
From described two or more than two reply datas, chooses the highest reply data of the degree of correlation and tied as the retrieval
Fruit;Or
From described two or more than two reply datas, the highest reply data of the degree of correlation is chosen as the first response, choosing
Take the degree of correlation inferior at least one reply data of the optimal response as the second response, by first response and described the
Two responses are as the search result.
11. according to the method in any one of claims 1 to 3, which is characterized in that further include:
When in the search result including the structural knowledge point of rhetorical question, Xiang Suoshu user sends the structure based on the rhetorical question
Change the second rhetorical question information that knowledge point generates;
User is received for the rhetorical question response of the second rhetorical question information feedback;
According to the rhetorical question response, relocates described this and put question to data;
It sends at least one reply data recommended based on this enquirement data, Xiang Suoshu user described in reorientation or exists again
It is retrieved in the ontology knowledge base.
12. according to the method in any one of claims 1 to 3, which is characterized in that the multiple types knowledge includes: knot
The map knowledge of structure, semi-structured question and answer are to knowledge and/or non-structured text knowledge.
13. a kind of automatic question-answering method characterized by comprising
In response to the enquirement event of user's triggering, this enquirement data of the user are obtained;
Data are putd question to be sent to server-side described this;
The server-side is exported based on the response feedback that this puts question to data to send;
Wherein, the response feedback is generated according to search result, and the search result is using at least one response retrieval
Mode puts question to the corresponding at least one set of polymorphic type of relevant at least one sample of data to know in ontology knowledge base to described this
Retrieval obtains in knowledge.
14. according to the method for claim 13, which is characterized in that output response feedback, comprising:
The response feedback is shown in the question and answer interface in a manner of text;Or
The response feedback is exported in a manner of voice.
15. a kind of automatically request-answering system characterized by comprising
Ontology knowledge base, for storing at least one sample ontology and the corresponding one group of polymorphic type knowledge of this ontology of various kinds;
Interactive unit, the enquirement event for triggering in response to user, this for obtaining the user put question to data and will be described
This puts question to data to issue;
Response unit puts question to data for receiving described this;In the ontology knowledge base, obtains and put question to number with described this
According to the corresponding at least one set of polymorphic type knowledge of at least one relevant sample ontology;It is examined at least one set of polymorphic type knowledge
Rope reply data;According to search result, response feedback is exported to the user;And
The interactive unit is also used to export the response feedback.
16. a kind of server device characterized by comprising first memory and first processor and the first communication component;
The first memory, for storing program;
First communication component is transmitted for data;
The first processor is coupled with the first memory, for executing the journey stored in the first memory
Sequence, to be used for:
This enquirement data of user's input are received by first communication component;
From ontology knowledge base, at least one corresponding at least one set of sample ontology relevant to this described enquirement data is obtained
Polymorphic type knowledge;
Reply data is retrieved at least one set of polymorphic type knowledge;
According to search result, response feedback is exported to the user by first communication component.
17. a kind of client device characterized by comprising second memory, second processor and the second communication component;
The second memory, for storing program;
Second communication component is transmitted for data;
The second processor is coupled with the second memory, for executing the journey stored in the second memory
Sequence, to be used for:
In response to the enquirement event of user's triggering, this enquirement data of the user are obtained;
Data are putd question to be sent to server-side described this by second communication component;
The server-side that output is received by second communication component is based on the response that this puts question to data to send
Feedback;
Wherein, the response feedback is generated based on search result, the search result be in ontology knowledge base with it is described
Retrieval obtains in this corresponding at least one set of polymorphic type knowledge of at least one relevant sample of enquirement data.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710909408.4A CN109582798A (en) | 2017-09-29 | 2017-09-29 | Automatic question-answering method, system and equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710909408.4A CN109582798A (en) | 2017-09-29 | 2017-09-29 | Automatic question-answering method, system and equipment |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109582798A true CN109582798A (en) | 2019-04-05 |
Family
ID=65918887
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710909408.4A Pending CN109582798A (en) | 2017-09-29 | 2017-09-29 | Automatic question-answering method, system and equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109582798A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110516027A (en) * | 2019-07-22 | 2019-11-29 | 北京达佳互联信息技术有限公司 | Update method, device, electronic equipment and the storage medium of information aggregate |
CN110728356A (en) * | 2019-09-17 | 2020-01-24 | 阿里巴巴集团控股有限公司 | Dialogue method and system based on recurrent neural network and electronic equipment |
CN111506827A (en) * | 2020-03-18 | 2020-08-07 | 威比网络科技(上海)有限公司 | Salesman question and answer interaction method, platform, equipment and storage medium |
CN111597313A (en) * | 2020-04-07 | 2020-08-28 | 深圳追一科技有限公司 | Question answering method, device, computer equipment and storage medium |
CN112084319A (en) * | 2020-09-29 | 2020-12-15 | 四川省人工智能研究院(宜宾) | Relational network video question-answering system and method based on actions |
CN112507706A (en) * | 2020-12-21 | 2021-03-16 | 北京百度网讯科技有限公司 | Training method and device of knowledge pre-training model and electronic equipment |
CN113411618A (en) * | 2020-11-26 | 2021-09-17 | 腾讯科技(深圳)有限公司 | Data processing method and device based on social application and computer storage medium |
EP4152176A1 (en) * | 2021-09-18 | 2023-03-22 | Beijing Baidu Netcom Science And Technology Co. Ltd. | Intelligent question-answering processing method and system, electronic device and storage medium |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103049433A (en) * | 2012-12-11 | 2013-04-17 | 微梦创科网络科技(中国)有限公司 | Automatic question answering method, automatic question answering system and method for constructing question answering case base |
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 |
US20160180437A1 (en) * | 2014-12-18 | 2016-06-23 | Nuance Communications, Inc. | Product recommendation with ontology-linked product review |
US20170091313A1 (en) * | 2015-09-28 | 2017-03-30 | Microsoft Technology Licensing, Llc | Domain-specific unstructured text retrieval |
CN106649394A (en) * | 2015-11-03 | 2017-05-10 | 中兴通讯股份有限公司 | Fusion knowledge base processing method and device and knowledge base management system |
CN106919655A (en) * | 2017-01-24 | 2017-07-04 | 网易(杭州)网络有限公司 | A kind of answer provides method and apparatus |
-
2017
- 2017-09-29 CN CN201710909408.4A patent/CN109582798A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103049433A (en) * | 2012-12-11 | 2013-04-17 | 微梦创科网络科技(中国)有限公司 | Automatic question answering method, automatic question answering system and method for constructing question answering case base |
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 |
US20160180437A1 (en) * | 2014-12-18 | 2016-06-23 | Nuance Communications, Inc. | Product recommendation with ontology-linked product review |
US20170091313A1 (en) * | 2015-09-28 | 2017-03-30 | Microsoft Technology Licensing, Llc | Domain-specific unstructured text retrieval |
CN106649394A (en) * | 2015-11-03 | 2017-05-10 | 中兴通讯股份有限公司 | Fusion knowledge base processing method and device and knowledge base management system |
CN106919655A (en) * | 2017-01-24 | 2017-07-04 | 网易(杭州)网络有限公司 | A kind of answer provides method and apparatus |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110516027A (en) * | 2019-07-22 | 2019-11-29 | 北京达佳互联信息技术有限公司 | Update method, device, electronic equipment and the storage medium of information aggregate |
CN110516027B (en) * | 2019-07-22 | 2022-04-22 | 北京达佳互联信息技术有限公司 | Information set updating method and device, electronic equipment and storage medium |
CN110728356A (en) * | 2019-09-17 | 2020-01-24 | 阿里巴巴集团控股有限公司 | Dialogue method and system based on recurrent neural network and electronic equipment |
CN110728356B (en) * | 2019-09-17 | 2023-08-04 | 创新先进技术有限公司 | Dialogue method and system based on cyclic neural network and electronic equipment |
CN111506827A (en) * | 2020-03-18 | 2020-08-07 | 威比网络科技(上海)有限公司 | Salesman question and answer interaction method, platform, equipment and storage medium |
CN111597313A (en) * | 2020-04-07 | 2020-08-28 | 深圳追一科技有限公司 | Question answering method, device, computer equipment and storage medium |
CN112084319A (en) * | 2020-09-29 | 2020-12-15 | 四川省人工智能研究院(宜宾) | Relational network video question-answering system and method based on actions |
CN113411618A (en) * | 2020-11-26 | 2021-09-17 | 腾讯科技(深圳)有限公司 | Data processing method and device based on social application and computer storage medium |
CN113411618B (en) * | 2020-11-26 | 2024-03-22 | 腾讯科技(深圳)有限公司 | Data processing method and device based on social application and computer storage medium |
CN112507706A (en) * | 2020-12-21 | 2021-03-16 | 北京百度网讯科技有限公司 | Training method and device of knowledge pre-training model and electronic equipment |
EP4152176A1 (en) * | 2021-09-18 | 2023-03-22 | Beijing Baidu Netcom Science And Technology Co. Ltd. | Intelligent question-answering processing method and system, electronic device and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109582798A (en) | Automatic question-answering method, system and equipment | |
US11334635B2 (en) | Domain specific natural language understanding of customer intent in self-help | |
CN111177569B (en) | Recommendation processing method, device and equipment based on artificial intelligence | |
US11082554B2 (en) | Method for conversion and classification of data based on context | |
US20150286943A1 (en) | Decision Making and Planning/Prediction System for Human Intention Resolution | |
CN108228559A (en) | A kind of human-computer interaction realization method and system for customer service | |
CN108139918A (en) | Using every user as basic custom program feature | |
CN110073349A (en) | Consider the word order suggestion of frequency and formatted message | |
US20220138770A1 (en) | Method and apparatus for analyzing sales conversation based on voice recognition | |
US11809419B2 (en) | System to convert natural-language financial questions into database queries | |
US11635883B2 (en) | Indication of content linked to text | |
CN103534697A (en) | Training statistical dialog managers in spoken dialog systems with web data | |
WO2024032652A1 (en) | Task assistance method and apparatus, electronic device, and computer readable storage medium | |
Kunte et al. | Analysis of machine learning algorithms for predicting personality: brief survey and experimentation | |
WO2020263182A1 (en) | Method and system for conducting a brainstorming session with a virtual expert | |
Ye et al. | Tooleyes: Fine-grained evaluation for tool learning capabilities of large language models in real-world scenarios | |
WO2023040516A1 (en) | Event integration method and apparatus, and electronic device, computer-readable storage medium and computer program product | |
Hoque et al. | An interactive system for exploring community question answering forums | |
Yarlagadda et al. | DocTable: Table-Oriented Interactive Machine Learning for Text Corpora | |
US11966570B2 (en) | Automated processing and dynamic filtering of content for display | |
US11914844B2 (en) | Automated processing and dynamic filtering of content for display | |
CN116775813B (en) | Service searching method, device, electronic equipment and readable storage medium | |
CN109165277A (en) | A kind of output method and facility for study of composition | |
Dikshit et al. | Automating Questions and Answers of Good and Services Tax system using clustering and embeddings of queries | |
US20230260533A1 (en) | Automated segmentation of digital presentation data |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190405 |
|
RJ01 | Rejection of invention patent application after publication |