CN106202476B - A kind of interactive method and device of knowledge based collection of illustrative plates - Google Patents
A kind of interactive method and device of knowledge based collection of illustrative plates Download PDFInfo
- Publication number
- CN106202476B CN106202476B CN201610565229.9A CN201610565229A CN106202476B CN 106202476 B CN106202476 B CN 106202476B CN 201610565229 A CN201610565229 A CN 201610565229A CN 106202476 B CN106202476 B CN 106202476B
- Authority
- CN
- China
- Prior art keywords
- sentence
- classification
- module
- knowledge
- threshold value
- Prior art date
Links
- 230000002452 interceptive Effects 0.000 title claims abstract description 17
- 238000000034 methods Methods 0.000 claims abstract description 6
- 230000000875 corresponding Effects 0.000 claims description 16
- 238000004458 analytical methods Methods 0.000 claims description 9
- 238000000605 extraction Methods 0.000 claims description 4
- 230000005540 biological transmission Effects 0.000 claims description 3
- 239000000284 extracts Substances 0.000 claims description 3
- 235000013399 edible fruits Nutrition 0.000 claims 1
- 241000219071 Malvaceae Species 0.000 description 5
- 229910017436 S2 Can Inorganic materials 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000000463 materials Substances 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
Abstract
Description
Technical field
The invention belongs to the data processing of nan-machine interrogation, more particularly to a kind of interactive method of knowledge based collection of illustrative plates And device.
Background technology
Intelligent Answer System in question-response form, the enquirement knowledge required for accurate locating websites user, by with Website user is interacted, and the information service of personalization is provided for website user.
Intelligent Answer System is the unordered corpus information that will build up on, and carries out the arrangement of orderly and science, and foundation is based on knowing The disaggregated model of knowledge；These disaggregated models can instruct the language material for newly increasing to seek advice from and information on services, save human resources, improve The automaticity of information processing, reduces website operating cost.Based on the basic feelings on government and enterprise accumulated for many years to website Condition FAQs and its answer, arrange the question and answer storehouse form for specification, to support the intelligent answer of various forms problem.Facilitate User, improves work efficiency, improves corporate image.
Present intelligent Answer System would generally be processed sentence by means of natural language processing and full-text search technology, Single mode generally can be than relatively low in some of accuracy rate.Conventional intelligent Answer System can only be directed to some field Question and answer, for the question answering system of open field, currently there are no a kind of general, and autgmentability relatively good handling process Occur.
Question answering system is mainly classified including sentence, information retrieval, answer extracting, semantic understanding, the aspect such as knowledge mapping Technology.
The content of the invention
In order to overcome the deficiencies in the prior art, an object of the present invention is a kind of the man-machine of knowledge based collection of illustrative plates of offer The method of dialogue, it can be answered the problem of general knowledge and Opening field.
The second object of the present invention is to provide a kind of interactive device of knowledge based collection of illustrative plates, its can to general knowledge and The problem of Opening field is answered.
An object of the present invention is realized using following technical scheme：
A kind of interactive method of knowledge based collection of illustrative plates, comprises the following steps：
S1：The sentence that user sends is received, and obtain the sentence above of the sentence, and determine the sentence and above sentence Classification；
S2：Classification according to the sentence and the above classification of sentence determine the final generic of the sentence；
S3：Feature Words in the sentence are extracted by knowledge mapping, and judges whether all Feature Words are relevant, if Have, then the association generation according to Feature Words is replied, if it is not, performing step S4；
S4：The sentence is engaged in the dialogue matching according to chat conversations storehouse, if the match is successful, generation is replied, if lost Lose, then perform step S5；
S5：Regular pattern matching is carried out to the sentence, if the match is successful, generation is corresponding to reply, if matching is lost Lose, then perform step S6；
S6：Final generic according to the sentence is analyzed and processes and generate reply.
Preferably, step S2 specifically includes following sub-step：
S21：Classification according to the sentence and the above classification of sentence determine the classification grade and language above of the sentence The classification grade of sentence；
S22：Judge whether the classification grade of the sentence is more than the classification grade of sentence above, if it is, from institute Predicate sentence classification as the sentence final generic, if it is not, then from sentence above classification as the sentence Final generic.The implementation of step S2 can be disclosed further.
Preferably, step S4 specifically includes following sub-step：
S41：Full-text search is carried out to the sentence according to chat conversations storehouse；
S42：The phase of candidate's sentence and read statement in chat conversations storehouses is calculated by Jaccard Similarity algorithms Like degree；
S43：Judge that whether gained similarity, more than a threshold value, if greater than the threshold value, then selects respective statement to be returned It is multiple, if less than the threshold value, then perform step S5.Its specific implementation that can further disclose step S4.
Preferably, the threshold value is 0.7.Its setting that can further disclose threshold value.
Preferably, step S6 specifically includes following sub-step：
S61：Final generic according to the sentence extracts the attributive character related to the final generic of the sentence；
S62：According to the extraction program of gained attributive character, attribute corresponding with attributive character is extracted；
S63：Generated according to gained attribute and replied.Its implementation that can further disclose step S6.
The second object of the present invention is realized using following technical scheme：
A kind of interactive device of knowledge based collection of illustrative plates, including with lower module：
Receiver module：Sentence for receiving user's transmission, and obtains the sentence above of the sentence, and determine the sentence and The classification of sentence above；
Sort module：For the classification according to the sentence and above the classification of sentence come determine the sentence it is final belonging to class Not；
Knowledge mapping module：For extracting the Feature Words in the sentence by knowledge mapping, and judge that all Feature Words are It is no relevant, if it has, then the association generation according to Feature Words is replied, if it is not, performing chat matching module；
Chat matching module：For being engaged in the dialogue to the sentence matching according to chat conversations storehouse, if the match is successful, give birth to Into reply, if it fails, then performing canonical matching module；
Canonical matching module：For carrying out regular pattern matching to the sentence, if the match is successful, generation is corresponding to return It is multiple, if it fails to match, perform analysis and processing module；
Analysis and processing module：Reply is processed and generates for being analyzed according to the final generic of the sentence.
Preferably, sort module specifically includes following submodule：
Level determination module：Classification according to the sentence and the above classification of sentence determine the classification grade of the sentence The classification grade of sentence above；
First judge module：Judge whether the classification grade of the sentence is more than the classification grade of sentence above, if it is, The classification of the sentence is then selected as the final generic of the sentence, if it is not, then from the classification conduct of sentence above The final generic of the sentence.The specific submodule that sort module is included can be further determined that.
Preferably, chat matching module specifically includes following submodule：
Full-text search module：For carrying out full-text search to the sentence according to chat conversations storehouse；
Computing module：For by candidate's sentence and input in Jaccard Similarity algorithms calculating chat conversations storehouse The similarity of sentence；
Second judge module：For judging that whether gained similarity, more than a threshold value, if greater than the threshold value, then selects phase Answer sentence to be replied, if less than the threshold value, then perform canonical matching module.It can further disclose chat matching module institute Comprising specific module.
Preferably, the threshold value is 0.7.Its numerical value that can further disclose threshold value.
Preferably, analysis and processing module specifically includes following submodule：
Attribute extraction module：For extracting the final generic phase with the sentence according to the final generic of the sentence The attributive character of pass；
Attribute matching module：For the extraction program according to gained attributive character, attribute corresponding with attributive character is extracted；
Reply module：Replied for being generated according to gained attribute.It can further disclose what analysis and processing module was included Submodule.
Compared to existing technology, the beneficial effects of the present invention are：
The classification of control statement to a certain extent of the invention, and make the Opening field of the knowledge question of general knowledge and customization Question and answer are same in a flow, are different from existing automatically request-answering system and rely only on the carrying out in chat storehouse and disaggregated model and ask The search of answer is inscribed, and the technology of template matches and knowledge mapping search is added in invention, enable to human-computer dialogue More enrich.
Brief description of the drawings
Fig. 1 is a kind of flow chart of the interactive method of knowledge based collection of illustrative plates of the invention；
Fig. 2 is a kind of structured flowchart of the interactive device of knowledge based collection of illustrative plates of the invention；
Fig. 3 is the classification chart of statement classification of the invention.
Specific embodiment
Below, with reference to accompanying drawing and specific embodiment, the present invention is described further：
As shown in figure 1, the invention provides a kind of interactive method of knowledge based collection of illustrative plates, comprising the following steps：
S1：The sentence that user sends is received, and obtain the sentence above of the sentence, and determine the sentence and above sentence Classification；
S2：Classification according to the sentence and the above classification of sentence determine the final generic of the sentence；Step S2 Specifically include following sub-step：
S21：Classification according to the sentence and the above classification of sentence determine the classification grade and language above of the sentence The classification grade of sentence；
S22：Judge whether the classification grade of the sentence is more than the classification grade of sentence above, if it is, from institute Predicate sentence classification as the sentence final generic, if it is not, then from sentence above classification as the sentence Final generic.This step be in order to determine in subsequent step to the sentence with which kind of classification come proceed treatment, The determination of classification results is needed by being compared to the current class in context, then just determines, from that classification, to compare Mode mainly use and the grade of classification compares, it is preferential with high-grade classification, as shown in figure 3,1 grade is the lowest class, 46 Level is highest level, and from 1 grade to 46 grades, grade rises successively, for example, when user says " I will navigate ", this sentence is undoubtedly It is " navigation " classification, but, navigation needs a destination, so, rear end can generate " which the destination that may I ask you is " It is that such as user says " Guangzhou " when client responds to inquire the destination of user, when " Guangzhou " this phrase enters classification, it " what " this classification may be classified as, and if when the phrase that client replys is classified to other classifications, navigated This purpose cannot just be completed, so needing category preferences to correct.Due to " navigation ", this is sorted in the session of user In be not closed, so compare, it is necessary to carry out grade with " navigation " when " what " " Guangzhou " be classified into, due to " navigation " The height of grade ratio " what ", so " Guangzhou " this phrase needs to obey " navigation " this classification, in this way, many wheel dialogues are just able to Complete.
But also exception, such as it is weather typing to be in going up in short, then under be in short not divide Class, then treatment of the sentence as weather typing；
S3：Feature Words in the sentence are extracted by knowledge mapping, and judges whether all Feature Words are relevant, if Have, then the association generation according to Feature Words is replied, if it is not, performing step S4；By the Feature Words that will be drawn into as The starting point of the search of knowledge mapping, carries out the association search of knowledge, if whole Feature Words are all relevant, generates answer.
For example, including node Guangzhou in knowledge mapping, then there is attribute city flower, property value is common bombax flower, then formed Triple (Guangzhou, city flower, common bombax flower).When sentence, " what the city flower in Guangzhou is" or " common bombax flower is city flower where” When search into collection of illustrative plates, due to including Guangzhou, city flower and common bombax flower in knowledge mapping, then by the extraction to word, so After find corresponding triple, generation replys that " city flower in Guangzhou is common bombax flower.”.
S4：The sentence is engaged in the dialogue matching according to chat conversations storehouse, if the match is successful, generation is replied, if lost Lose, then perform step S5；Step S4 specifically includes following sub-step：
S41：Full-text search is carried out to the sentence according to chat conversations storehouse；
S42：The phase of candidate's sentence and read statement in chat conversations storehouses is calculated by Jaccard Similarity algorithms Like degree；
S43：Judge that whether gained similarity, more than a threshold value, if greater than the threshold value, then selects respective statement to be returned It is multiple, if less than the threshold value, then perform step S5.The threshold value is 0.7.The step is mainly the sentence of matching chat, for example When the sentence of the daily greeting chat for occurring " hello " etc in dialogue, then daily chatting can be completed by the step Its function.
S5：Regular pattern matching is carried out to the sentence, if the match is successful, generation is corresponding to reply, if matching is lost Lose, then perform step S6；The step is mainly used in systemic-function, the treatment of the special clause of sensitive word and part, for example, work as using When the sentence of " who are you " etc is asked at family, function of this question sentence equivalent to the about in app is matched and systemic-function Related information, generation is replied.
S6：Final generic according to the sentence is analyzed and processes and generate reply；Step S6 specifically includes following Sub-step：
S61：Final generic according to the sentence extracts the attributive character related to the final generic of the sentence；
S62：According to the extraction program of gained attributive character, attribute corresponding with attributive character is extracted；
S63：Generated according to gained attribute and replied.The dialogue for mainly having processed foregoing step process not in this step, example Such as：" Guangzhou is today cold for user input", the classification of this sentence is weather, mainly has four for the attribute that weather is paid close attention to Individual time, place, detailed classification and degree, three category can be extracted by the defined attribute extraction program done for weather sentence Property, the time is today, and place is Guangzhou, and degree is cold, then goes inquiry to obtain weather data according to time and city, further according to Whether the temperature record in weather data is cold to judge today, so as to complete the answer to the sentence.In this step, it will usually Need to call other modules to inquire about the information of correlation so as to generate corresponding answer.
As shown in Fig. 2 the invention provides a kind of interactive device of knowledge based collection of illustrative plates, including with lower module：
Receiver module：Sentence for receiving user's transmission, and obtains the sentence above of the sentence, and determine the sentence and The classification of sentence above；
Sort module：For the classification according to the sentence and above the classification of sentence come determine the sentence it is final belonging to class Not；Sort module specifically includes following submodule：
Level determination module：Classification according to the sentence and the above classification of sentence determine the classification grade of the sentence The classification grade of sentence above；
First judge module：Judge whether the classification grade of the sentence is more than the classification grade of sentence above, if it is, The classification of the sentence is then selected as the final generic of the sentence, if it is not, then from the classification conduct of sentence above The final generic of the sentence.
Knowledge mapping module：For extracting the Feature Words in the sentence by knowledge mapping, and judge that all Feature Words are It is no relevant, if it has, then the association generation according to Feature Words is replied, if it is not, performing chat matching module；
Chat matching module：For being engaged in the dialogue to the sentence matching according to chat conversations storehouse, if the match is successful, give birth to Into reply, if it fails, then performing canonical matching module；Chat matching module specifically includes following submodule：
Full-text search module：For carrying out full-text search to the sentence according to chat conversations storehouse；
Computing module：For by candidate's sentence and input in Jaccard Similarity algorithms calculating chat conversations storehouse The similarity of sentence；
Second judge module：For judging that whether gained similarity, more than a threshold value, if greater than the threshold value, then selects phase Answer sentence to be replied, if less than the threshold value, then perform canonical matching module.The threshold value is 0.7.
Canonical matching module：For carrying out regular pattern matching to the sentence, if the match is successful, generation is corresponding to return It is multiple, if it fails to match, perform analysis and processing module；
Analysis and processing module：Reply is processed and generates for being analyzed according to the final generic of the sentence.Analysis Processing module specifically includes following submodule：
Attribute extraction module：It is special for extracting the attribute related to the sentence classification according to the final generic of the sentence Levy；
Attribute matching module：According to the extraction program of gained attributive character, attribute corresponding with attributive character is extracted；
Reply module：Generated according to gained attribute and replied.
It will be apparent to those skilled in the art that technical scheme that can be as described above and design, make other various It is corresponding to change and deformation, and all these change and deformation should all belong to the protection domain of the claims in the present invention Within.
Claims (6)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610565229.9A CN106202476B (en) | 2016-07-14 | 2016-07-14 | A kind of interactive method and device of knowledge based collection of illustrative plates |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610565229.9A CN106202476B (en) | 2016-07-14 | 2016-07-14 | A kind of interactive method and device of knowledge based collection of illustrative plates |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106202476A CN106202476A (en) | 2016-12-07 |
| CN106202476B true CN106202476B (en) | 2017-06-06 |
Family
ID=57493787
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610565229.9A CN106202476B (en) | 2016-07-14 | 2016-07-14 | A kind of interactive method and device of knowledge based collection of illustrative plates |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106202476B (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106844732B (en) * | 2017-02-13 | 2020-05-08 | 长沙军鸽软件有限公司 | Method for automatically acquiring session scene label incapable of being directly acquired |
| CN106844734A (en) * | 2017-02-13 | 2017-06-13 | 长沙军鸽软件有限公司 | A kind of method for automatically generating session reply content |
| CN106951470B (en) * | 2017-03-03 | 2019-01-18 | 中兴耀维科技江苏有限公司 | A kind of intelligent Answer System based on the retrieval of professional knowledge figure |
| CN107133305A (en) * | 2017-04-28 | 2017-09-05 | 上海斐讯数据通信技术有限公司 | A kind of automatic construction device of chat robots knowledge base and its method |
| CN107193948B (en) * | 2017-05-22 | 2018-04-20 | 邢加和 | Human-computer dialogue data analysing method and device |
| CN109726387A (en) * | 2017-10-31 | 2019-05-07 | 科沃斯商用机器人有限公司 | Man-machine interaction method and system |
| CN107943998A (en) * | 2017-12-05 | 2018-04-20 | 竹间智能科技（上海）有限公司 | A kind of human-machine conversation control system and method for knowledge based collection of illustrative plates |
| CN109002498B (en) * | 2018-06-29 | 2020-05-05 | 北京百度网讯科技有限公司 | Man-machine conversation method, device, equipment and storage medium |
| CN109033223A (en) * | 2018-06-29 | 2018-12-18 | 北京百度网讯科技有限公司 | For method, apparatus, equipment and computer readable storage medium across type session |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104471568A (en) * | 2012-07-02 | 2015-03-25 | 微软公司 | Learning-based processing of natural language questions |
| CN105677795A (en) * | 2015-12-31 | 2016-06-15 | 上海智臻智能网络科技股份有限公司 | Abstract semantics recommendation method, recommendation device and recommendation system |
| CN105677783A (en) * | 2015-12-31 | 2016-06-15 | 上海智臻智能网络科技股份有限公司 | Information processing method and device for intelligent question-answering system |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9092988B2 (en) * | 2012-11-16 | 2015-07-28 | International Business Machines Corporation | Multi-dimensional feature merging for open domain question answering |
| CN103853842B (en) * | 2014-03-20 | 2017-07-18 | 百度在线网络技术（北京）有限公司 | A kind of automatic question-answering method and system |
| CN105183848A (en) * | 2015-09-07 | 2015-12-23 | 百度在线网络技术（北京）有限公司 | Human-computer chatting method and device based on artificial intelligence |
-
2016
- 2016-07-14 CN CN201610565229.9A patent/CN106202476B/en active IP Right Grant
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104471568A (en) * | 2012-07-02 | 2015-03-25 | 微软公司 | Learning-based processing of natural language questions |
| CN105677795A (en) * | 2015-12-31 | 2016-06-15 | 上海智臻智能网络科技股份有限公司 | Abstract semantics recommendation method, recommendation device and recommendation system |
| CN105677783A (en) * | 2015-12-31 | 2016-06-15 | 上海智臻智能网络科技股份有限公司 | Information processing method and device for intelligent question-answering system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106202476A (en) | 2016-12-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10013977B2 (en) | Smart home control method based on emotion recognition and the system thereof | |
| US10115056B2 (en) | Method and apparatus for responding to an inquiry | |
| EP3183728B1 (en) | Orphaned utterance detection system and method | |
| US10679624B2 (en) | Personal directory service | |
| US20190102381A1 (en) | Exemplar-based natural language processing | |
| WO2017041372A1 (en) | Man-machine interaction method and system based on artificial intelligence | |
| US10133729B2 (en) | Semantically-relevant discovery of solutions | |
| JP5936698B2 (en) | Word semantic relation extraction device | |
| CN106503805B (en) | A kind of bimodal based on machine learning everybody talk with sentiment analysis method | |
| US9582757B1 (en) | Scalable curation system | |
| US20190057697A1 (en) | Better resolution when referencing to concepts | |
| US6574624B1 (en) | Automatic topic identification and switch for natural language search of textual document collections | |
| US7974835B2 (en) | Method, system, and apparatus for natural language mixed-initiative dialogue processing | |
| US8073877B2 (en) | Scalable semi-structured named entity detection | |
| JP6305389B2 (en) | Method and apparatus for intelligent chat between human and machine using artificial intelligence | |
| US20150074112A1 (en) | Multimedia Question Answering System and Method | |
| JP2017505964A (en) | Automatic task classification based on machine learning | |
| DE112016004863T5 (en) | Parameter collection and automatic dialog generation in dialog systems | |
| US8041568B2 (en) | Business listing search | |
| JP2018077858A (en) | System and method for conversation-based information search | |
| CN102262634B (en) | Automatic questioning and answering method and system | |
| CN105408890A (en) | Performing an operation relative to tabular data based upon voice input | |
| US10431214B2 (en) | System and method of determining a domain and/or an action related to a natural language input | |
| US20080189257A1 (en) | World-wide classified listing search with translation | |
| CN101599071B (en) | Automatic extraction method of conversation text topic |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| C06 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| C10 | Entry into substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20191223 Address after: 518019 2-24 / F, building a, Zhongshe Plaza, No. 1028, Buji Road, Dongxiao street, Luohu District, Shenzhen City, Guangdong Province Patentee after: Shenzhen Softcom Power Information Technology Co., Ltd Address before: 510000. Room two, No. 122, sports east road, Guangzhou, Guangdong, Tianhe District 3304, China Patentee before: Guangzhou hope Mdt InfoTech Ltd |
|
| TR01 | Transfer of patent right |