CN107679224A - It is a kind of towards the method and system without structure text intelligent answer - Google Patents
It is a kind of towards the method and system without structure text intelligent answer Download PDFInfo
- Publication number
- CN107679224A CN107679224A CN201710985745.1A CN201710985745A CN107679224A CN 107679224 A CN107679224 A CN 107679224A CN 201710985745 A CN201710985745 A CN 201710985745A CN 107679224 A CN107679224 A CN 107679224A
- Authority
- CN
- China
- Prior art keywords
- vector
- text
- answer
- phrase
- hidden
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- 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
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention belongs to computer intelligence dialogue technoloyg field, there is provided and it is a kind of towards the method and system without structure text intelligent answer, including:S1, coding layer are encoded the text got and problem respectively, are obtained text hidden vector sum problem and are hidden vector;S2, the text hidden vector sum described problem is hidden Vector Fusion and got up by information fusion layer, and the interconnection vector group after being merged;S3, decoding layer decode according to the interconnection vector group to the text, obtain the answer of described problem, and export the answer.The present invention can directly give answer for the enquirement without structure text, it is not necessary to establish question and answer storehouse in advance;The type of enquirement is not limited;The answer of return is more accurate;Data-driven, effectively utilize big data.
Description
Technical field
The invention belongs to computer intelligence dialogue technoloyg field, and in particular to a kind of towards without structure text intelligent answer
Method and system.
Background technology
Refer to any given one section of structureless text without structure text intelligent answer, and any one is directed to the text
The enquirement for meeting following condition, i.e. the answer of the enquirement appeared in given no structure text.Under such conditions, intelligence
Energy question answering system will can find out corresponding answer to answer the enquirement.
Technology at present without structure text intelligent answer mainly has four kinds, but the shortcomings that have its each:
Method based on question and answer storehouse, it is difficult to build question and answer storehouse, especially can not know the situation of no structure text in advance
Under.Simultaneously, it is contemplated that the opening that user puts question to, it is difficult to list in advance for all problems without structure text and answer.
Method based on retrieval especially birth defect.The similarity putd question to first only according to the sentence being syncopated as with user
To be answered, it is possible to give an irrelevant answer.Meanwhile whole sentence is returned as answer, granularity is too big, does not find most accurate
Answer.
Method based on name Entity recognition, it is necessary first to the intention putd question to is judged, only when enquirement is named in inquiry
Just it is adapted to answer when entity.Therefore the enquirement that this method can be answered is limited, and the enquirement for non-name entity can not return
Answer.Meanwhile when occurring the name entity of multiple same types in without structure text, how this selects, and judge to put question to and be intended to
It is possible to inaccurate, these can all influence the validity of this method.
Method based on structure atlas analysis, it is necessary first to which analysis takes out key element therein entirely without structure text
Take out structure collection of illustrative plates.How to analyze collection of illustrative plates to go forward side by side and find out answer, there is presently no very perfect method, be more based on
Various rules obtain answer, and accuracy is not too high.The identical element meeting occurred for different places in long article and article
Increase the difficulty of atlas analysis.
In summary, the technology at present without structure text intelligent answer is primarily present following defect:Need to build in advance and ask
Answer storehouse;It is excessively thick or meticulous that the answer of return is possible to granularity, is not especially accurate;The enquirement type that can accurately answer relatively has
Limit;Big data can not effectively be utilized.
The content of the invention
For the deficiency of problem above, the invention provides a kind of method towards without structure text intelligent answer and it is
System, the present invention can directly give answer for the enquirement without structure text, it is not necessary to establish question and answer storehouse in advance;To the type of enquirement
Do not limit;The answer of return is more accurate;Data-driven, effectively utilize big data.
To achieve the above object, it is provided by the invention a kind of towards the method without structure text intelligent answer, including:
S1, coding layer are encoded the text got and problem respectively, are obtained text hidden vector sum problem and are hidden
Vector;
S2, the text hidden vector sum described problem is hidden Vector Fusion and got up by information fusion layer, and is merged
Interconnection vector group afterwards;
S3, decoding layer decode according to the interconnection vector group to the text, obtain the answer of described problem, and defeated
Go out the answer.
Preferably, the S1 specific methods are:
S11, obtain the text and problem of input;
S12:The text and described problem are segmented, obtain textual phrase and problem phrase;
S13:The textual phrase and described problem phrase are respectively mapped to corresponding term vector, obtain textual phrase to
Amount and problem phrase vector;
S14:The textual phrase vector sum described problem phrase vector is encoded using bidirectional circulating neutral net,
Obtain text hidden vector sum problem and hide vector.
Preferably, the S12 specific methods are:
Text C and problem Q are segmented respectively, obtain textual phrase C1:With problem phrase Q1:
WhereinFor i-th of word in textual phrase,For j-th of word in problem phrase, n is the total number of word in textual phrase, and m is
The total number of word in problem phrase.
Preferably, the S13 specific methods are:
By textual phrase C1:With problem phrase Q1:Corresponding term vector is respectively mapped to, obtains text
Phrase vector C2:With problem phrase vector Q2:WhereinForCorresponding term vector,ForIt is right
The term vector answered.
Preferably, the S14 is specially:
Using bidirectional circulating neutral net to textual phrase vector C2:With problem phrase vector Q2:Point
Do not encoded, obtain text hidden vector C3:Vectorial Q is hidden with problem3:WhereinForIt is corresponding
Hide vector,ForCorresponding hiding vector, wherein
Preferably, the S2 is specially:
Problem is hidden to last hiding vector of vectorWith text hidden vectorIn each hide
Vector is weighted operation, calculates similar value
Wherein,Represent to hide vectorTransposition, wsFor parameter matrix;
By similar value aiWith text hidden vectorIn each hiding multiplication of vectors, calculate the association after fusion
Vector Groups
Preferably, the S3 is specially:
Interconnection vector group after fusion is merged, g=concat (Hi), using the vectorial g after merging as two
The input of different fully-connected networks, two different fully-connected networks include the first fully-connected network and the second fully connected network
Network, the output valve of the first fully-connected network are the probability distribution p of prediction answer starting position1, second fully-connected network it is defeated
Go out probability distribution p of the value for prediction answer end position2,
p1=softmax (w1g)
p2=softmax (w2g)
Wherein w1And w2For parameter, the starting position p of answer is calculatedsWith end position pe,
ps=argmax (p1)
pe=argmax (p2)
In the text, starting position p is extractedsWith end position peBetween content of text as problem answer simultaneously
Export answer.
It is a kind of towards the system without structure text intelligent answer, including:
Coding layer module, for the text got and problem to be encoded respectively, obtain text hidden vector sum and ask
Topic hides vector;
Information fusion layer module, get up for the text hidden vector sum described problem to be hidden into Vector Fusion, and
Interconnection vector group after to fusion;
Decoding layer module, for being decoded according to the interconnection vector group to the text, obtain answering for described problem
Case, and export the answer.
From such scheme, the present invention is a kind of towards the method and system without structure text intelligent answer, can be directed to nothing
The enquirement of structure text directly gives answer, it is not necessary to establishes question and answer storehouse in advance;The type of enquirement is not limited;What is returned answers
Case is more accurate;Data-driven, effectively utilize big data.
Brief description of the drawings
, below will be to specific in order to illustrate more clearly of the specific embodiment of the invention or technical scheme of the prior art
The required accompanying drawing used is briefly described in embodiment or description of the prior art.In all of the figs, similar element
Or part is typically identified by similar reference.In accompanying drawing, each element or part might not be drawn according to the ratio of reality.
Fig. 1 is towards the method flow diagram without structure text intelligent answer in the present embodiment;
Fig. 2 is towards the system architecture diagram without structure text intelligent answer in the present embodiment.
Embodiment
Embodiments of the invention are described in detail below in conjunction with accompanying drawing.Following examples are only used for clearer
Ground illustrates the product of the present invention, therefore be intended only as example, and can not be limited the scope of the invention with this.
Embodiment:
The embodiment provides a kind of towards the method without structure text intelligent answer, as shown in figure 1, including:
S1, coding layer are encoded the text got and problem respectively, are obtained text hidden vector sum problem and are hidden
Vector;
S2, the text hidden vector sum described problem is hidden Vector Fusion and got up by information fusion layer, and is merged
Interconnection vector group afterwards;
S3, decoding layer decode according to the interconnection vector group to the text, obtain the answer of described problem, and defeated
Go out the answer.
Preferably, the S1 specific methods are:
S11, obtain the text and problem of input;
S12:The text and described problem are segmented, obtain textual phrase and problem phrase;
S13:Each word in the textual phrase and described problem phrase is respectively mapped to corresponding term vector, obtained
Text term vector and problem term vector;
S14:The textual phrase vector sum described problem phrase vector is encoded using bidirectional circulating neutral net,
Obtain text hidden vector sum problem and hide vector.
The S12 specific methods are:
Text C and problem Q are segmented respectively, obtain textual phrase C1:With problem phrase Q1:
WhereinFor i-th of word in textual phrase,For j-th of word in problem phrase, n is the total number of word in textual phrase, and m is
The total number of word in problem phrase.
The S13 specific methods are:
By textual phrase C1:With problem phrase Q1:Corresponding term vector is respectively mapped to, obtains text
This phrase vector C2:With problem phrase vector Q2:WhereinForCorresponding term vector,ForIt is right
The term vector answered.
The S14 is specially:
Using bidirectional circulating neutral net to textual phrase vector C2:With problem phrase vector Q2:Point
Do not encoded, obtain text hidden vector C3:Vectorial Q is hidden with problem3:WhereinForIt is corresponding
Hide vector,ForCorresponding hiding vector, wherein
The S2 is specially:
Problem is hidden to last hiding vector of vectorWith text hidden vectorIn each hide
Vector is weighted operation, calculates similar value
Wherein,Represent to hide vectorTransposition, wsFor parameter matrix;
By similar value aiWith text hidden vectorIn each hiding multiplication of vectors, calculate the association after fusion to
Amount group
The S3 is specially:
Interconnection vector group after fusion is merged, g=concat (Hi), using the vectorial g after merging as two
The input of different fully-connected networks, two different fully-connected networks include the first fully-connected network and the second fully connected network
Network, the output valve of the first fully-connected network are the probability distribution p of prediction answer starting position1, second fully-connected network it is defeated
Go out probability distribution p of the value for prediction answer end position2,
p1=softmax (w1g)
p2=softmax (w2g)
Wherein w1And w2For parameter, the starting position p of answer is calculatedsWith end position pe,
ps=argmax (p1)
pe=argmax (p2)
In the text, starting position p is extractedsWith end position peBetween content of text as problem answer simultaneously
Export answer.
It is a kind of towards the system without structure text intelligent answer, as shown in Fig. 2 including:
Coding layer module, for the text got and problem to be encoded respectively, obtain text hidden vector sum and ask
Topic hides vector;
Information fusion layer module, get up for the text hidden vector sum described problem to be hidden into Vector Fusion, and
Interconnection vector group after to fusion;
Decoding layer module, for being decoded according to the interconnection vector group to the text, obtain answering for described problem
Case, and export the answer.
BiLSTM and softmax is existing machine learning algorithm in the present embodiment, and concat () method is used to connect
Two or more arrays, but do not change existing array, argmax represents to find the parameter with maximum scores.The present embodiment can be with
The various aspects in life are applied, for example can be applied in chat system, user is answered and is carried on the correlation of dialog history
Ask, to promote the understanding to dialogue;It can also apply in the system with user mutual, for example story is interactive, system is to user
A story is told, user can be putd question to story, and system is answered.The technology also has otherwise application, such as
It can be used for understanding tediously long specification, answer the query of user;The technology may also be used for understanding legal provision, answer relevant
Legal problem;Even the document of leaving of mechanism can be converted into question answering system etc..
Embodiment one:Story comprehension
Text is:On the tree in spring, tender bud valve is grown;On the tree in summer, fat blade is overworked;The tree in autumn
On, leaf fills scarlet and golden yellow;Under the tree in winter, leaf landing chemical conversion soil.Fallen leaves were the stamps of the Nature, 1 year four
Season sends you, sends me, sends everybody.
Problem is:What has grown on the tree in spring
After the present embodiment is analyzed text and problem, starting position and the end position of answer are calculated, and is extracted
Answer of the content of text as described problem between starting position and end position, answer are:Grow tender bud valve
Embodiment two:Legal provision understands
Text is:Chapter 4, museum's community service
Article 28 museum should open to the public within its 6 months day for obtaining the certificate of registration.
Article 29 museum should announce the specific open hour to the public.National legal festivals and holidays and school's cold and heat again
During the vacation, museum should open.
Article 30 museum holds display and exhibition, should observe following regulation:
Theme and content should basic principle determined by constitutionality and safeguard national security with national unity, carry forward love
State's doctrine ...
Problem is:How long museum interior should open to the public
After the present embodiment is analyzed text and problem, starting position and the end position of answer are calculated, and is extracted
Answer of the content of text as described problem between starting position and end position, answer are:Article 28 museum should
6 months from from the acquirement certificate of registration
Implement three:News understands
Text is:Thunderclap is met head in rocket today home court, and prologue rocket obtains 9-0 beginning, and this Brooker of subsequent prestige is led the team
Cling to score on one's trail, WILLIAMS-DARLING Ton performance is active after rotation, and partial node rocket enters under his leading attacks and must turn out a prestigious institution, point difference by
20 points are gradually approached, it is leading with 79-59 that half-court terminates rocket.The second half just returned great of thunderclap will definitely a point difference be contracted to 12 points,
But hereafter rocket three divides such as rain, and 7 three points of note in hurricane, rocket is still leading 25 points at the end of half-court, get to after minor details 4 minutes prestige this
Brooker is led group of people and a point difference is narrowed down into 12 points again, but hereafter rocket firmly controls situation, and final rocket is won victory.
Thunderclap data are that Russell-Wei Si Brookers 39 divide 11 backboards 13 to assist;Victor-Aura enlightening ripple 15 divides 4 secondary attacks;
Shi Diwen-Adams 11 divide 4 backboards;Ai Neisi-Kan Te 23 divide 4 backboards;Tai-Ji Busen 12 divide 4 backboards.
Problem is:Ali pricks how many points
After the present embodiment is analyzed text and problem, starting position and the end position of answer are calculated, and is extracted
Answer of the content of text as described problem between starting position and end position, answer are:24 point of 5 backboard
Finally it should be noted that:The above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations;Although
The present invention is described in detail with reference to foregoing embodiments, it will be understood by those within the art that:It is still
Technical scheme described in foregoing embodiments can be modified, either which part or all technical characteristic are carried out
Equivalent substitution;And these modifications or replacement, the essence of appropriate technical solution is departed from various embodiments of the present invention technical side
The scope of case, it all should cover among the claim of the present invention and the scope of specification.
Claims (8)
- It is 1. a kind of towards the method without structure text intelligent answer, it is characterised in that including:S1, coding layer are encoded the text got and problem respectively, are obtained text hidden vector sum problem and are hidden vector;The text hidden vector sum described problem is hidden Vector Fusion and got up by S2, information fusion layer, and after being merged Interconnection vector group;S3, decoding layer decode according to the interconnection vector group to the text, obtain the answer of described problem, and export institute State answer.
- It is 2. according to claim 1 a kind of towards the method without structure text intelligent answer, it is characterised in that the S1 tools Body method is:S11, obtain the text and problem of input;S12:The text and described problem are segmented, obtain textual phrase and problem phrase;S13:The textual phrase and described problem phrase are respectively mapped to corresponding term vector, obtain textual phrase vector sum Problem phrase vector;S14:The textual phrase vector sum described problem phrase vector is encoded using bidirectional circulating neutral net, obtained Text hidden vector sum problem hides vector.
- It is 3. according to claim 2 a kind of towards the method without structure text intelligent answer, it is characterised in that the S12 Specific method is:Text C and problem Q are segmented respectively, obtain textual phrase C1:With problem phrase Q1:WhereinFor i-th of word in textual phrase,For j-th of word in problem phrase, n is the total number of word in textual phrase, and m is problem The total number of word in phrase.
- It is 4. according to claim 3 a kind of towards the method without structure text intelligent answer, it is characterised in that the S13 Specific method is:By textual phrase C1:With problem phrase Q1:Corresponding term vector is respectively mapped to, obtains text word The vectorial C of group2:With problem phrase vector Q2:WhereinForCorresponding term vector,ForIt is corresponding Term vector.
- It is 5. according to claim 4 a kind of towards the method without structure text intelligent answer, it is characterised in that the S14 Specially:Using bidirectional circulating neutral net to textual phrase vector C2:With problem phrase vector Q2:Enter respectively Row coding, obtains text hidden vector C3:Vectorial Q is hidden with problem3:WhereinForIt is corresponding to hide Vector,ForCorresponding hiding vector, wherein
- It is 6. according to claim 5 a kind of towards the method without structure text intelligent answer, it is characterised in that the S2 tools Body is:Problem is hidden to last hiding vector of vectorWith text hidden vectorIn each hiding vector enter Row weighting operations, calculate similar value<mrow> <msub> <mi>a</mi> <mi>i</mi> </msub> <mo>=</mo> <mi>s</mi> <mi>o</mi> <mi>f</mi> <mi>t</mi> <mi>m</mi> <mi>a</mi> <mi>x</mi> <mrow> <mo>(</mo> <msubsup> <mi>h</mi> <mi>m</mi> <msup> <mi>Q</mi> <mi>T</mi> </msup> </msubsup> <msub> <mi>w</mi> <mi>s</mi> </msub> <msubsup> <mi>h</mi> <mi>i</mi> <mi>C</mi> </msubsup> <mo>)</mo> </mrow> </mrow>Wherein,Represent to hide vectorTransposition, wsFor parameter matrix;By similar value aiWith text hidden vectorIn each hiding multiplication of vectors, calculate the interconnection vector group after fusion<mrow> <msub> <mi>H</mi> <mi>i</mi> </msub> <mo>=</mo> <msub> <mi>a</mi> <mi>i</mi> </msub> <msubsup> <mi>h</mi> <mi>i</mi> <mi>C</mi> </msubsup> <mo>.</mo> </mrow>
- It is 7. according to claim 6 a kind of towards the method without structure text intelligent answer, it is characterised in that the S3 tools Body is:Interconnection vector group after fusion is merged, g=concat (Hi), the vectorial g after merging is different as two The input of fully-connected network, two different fully-connected networks include the first fully-connected network and the second fully-connected network, and first The output valve of fully-connected network is the probability distribution p of prediction answer starting position1, the output valve of second fully-connected network is pre- Survey the probability distribution p of answer end position2,p1=softmax (w1g)p2=softmax (w2g)Wherein w1And w2For parameter, the starting position p of answer is calculatedsWith end position pe,ps=argmax (p1)pe=argmax (p2)In the text, starting position p is extractedsWith end position peBetween answer and output of the content of text as problem Answer.
- It is 8. a kind of towards the system without structure text intelligent answer, it is characterised in that including:Coding layer module, for the text got and problem to be encoded respectively, it is hidden to obtain text hidden vector sum problem Hide vector;Information fusion layer module, get up for the text hidden vector sum described problem to be hidden into Vector Fusion, and melted Interconnection vector group after conjunction;Decoding layer module, for being decoded according to the interconnection vector group to the text, the answer of described problem is obtained, and Export the answer.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710985745.1A CN107679224B (en) | 2017-10-20 | 2017-10-20 | Intelligent question and answer method and system for unstructured text |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710985745.1A CN107679224B (en) | 2017-10-20 | 2017-10-20 | Intelligent question and answer method and system for unstructured text |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107679224A true CN107679224A (en) | 2018-02-09 |
CN107679224B CN107679224B (en) | 2020-09-08 |
Family
ID=61141785
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710985745.1A Active CN107679224B (en) | 2017-10-20 | 2017-10-20 | Intelligent question and answer method and system for unstructured text |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107679224B (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108846138A (en) * | 2018-07-10 | 2018-11-20 | 苏州大学 | A kind of the problem of fusion answer information disaggregated model construction method, device and medium |
CN108959388A (en) * | 2018-05-31 | 2018-12-07 | 科大讯飞股份有限公司 | information generating method and device |
CN108959387A (en) * | 2018-05-31 | 2018-12-07 | 科大讯飞股份有限公司 | Information acquisition method and device |
CN108959467A (en) * | 2018-06-20 | 2018-12-07 | 华东师范大学 | A kind of calculation method of question sentence and the Answer Sentence degree of correlation based on intensified learning |
CN108959396A (en) * | 2018-06-04 | 2018-12-07 | 众安信息技术服务有限公司 | Machine reading model training method and device, answering method and device |
CN109408624A (en) * | 2018-11-06 | 2019-03-01 | 江西师范大学 | A kind of vision chat robots session generation method |
CN109492227A (en) * | 2018-11-16 | 2019-03-19 | 大连理工大学 | It is a kind of that understanding method is read based on the machine of bull attention mechanism and Dynamic iterations |
CN109543200A (en) * | 2018-11-30 | 2019-03-29 | 腾讯科技(深圳)有限公司 | A kind of text interpretation method and device |
CN110348462A (en) * | 2019-07-09 | 2019-10-18 | 北京金山数字娱乐科技有限公司 | A kind of characteristics of image determination, vision answering method, device, equipment and medium |
CN110765254A (en) * | 2019-10-21 | 2020-02-07 | 北京理工大学 | Multi-document question-answering system model integrating multi-view answer reordering |
CN110990547A (en) * | 2019-11-29 | 2020-04-10 | 支付宝(杭州)信息技术有限公司 | Phone operation generation method and system |
CN111309875A (en) * | 2018-12-10 | 2020-06-19 | 百度在线网络技术(北京)有限公司 | Question answering method, device, equipment and storage medium |
WO2021072892A1 (en) * | 2019-10-18 | 2021-04-22 | 平安科技(深圳)有限公司 | Legal provision search method based on neural network hybrid model, and related device |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6083173A (en) * | 1998-03-06 | 2000-07-04 | Research Foundation Of State University Of New York | Artificial neural network for predicting respiratory disturbances and method for developing the same |
CN1924995A (en) * | 2005-08-31 | 2007-03-07 | 中国科学院声学研究所 | Content analysis based short message ask/answer system and implementing method thereof |
CN101770512A (en) * | 2010-01-20 | 2010-07-07 | 何吴迪 | Cloud stored hypertext file storage and framework method of WEB window expression thereof |
CN104715168A (en) * | 2015-02-13 | 2015-06-17 | 陈佳阳 | File security control and trace method and system based on digital fingerprints |
CN106569998A (en) * | 2016-10-27 | 2017-04-19 | 浙江大学 | Text named entity recognition method based on Bi-LSTM, CNN and CRF |
-
2017
- 2017-10-20 CN CN201710985745.1A patent/CN107679224B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6083173A (en) * | 1998-03-06 | 2000-07-04 | Research Foundation Of State University Of New York | Artificial neural network for predicting respiratory disturbances and method for developing the same |
CN1924995A (en) * | 2005-08-31 | 2007-03-07 | 中国科学院声学研究所 | Content analysis based short message ask/answer system and implementing method thereof |
CN101770512A (en) * | 2010-01-20 | 2010-07-07 | 何吴迪 | Cloud stored hypertext file storage and framework method of WEB window expression thereof |
CN104715168A (en) * | 2015-02-13 | 2015-06-17 | 陈佳阳 | File security control and trace method and system based on digital fingerprints |
CN106569998A (en) * | 2016-10-27 | 2017-04-19 | 浙江大学 | Text named entity recognition method based on Bi-LSTM, CNN and CRF |
Cited By (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108959388B (en) * | 2018-05-31 | 2020-09-11 | 科大讯飞股份有限公司 | Information generation method and device |
CN108959388A (en) * | 2018-05-31 | 2018-12-07 | 科大讯飞股份有限公司 | information generating method and device |
CN108959387A (en) * | 2018-05-31 | 2018-12-07 | 科大讯飞股份有限公司 | Information acquisition method and device |
CN108959396A (en) * | 2018-06-04 | 2018-12-07 | 众安信息技术服务有限公司 | Machine reading model training method and device, answering method and device |
CN108959467A (en) * | 2018-06-20 | 2018-12-07 | 华东师范大学 | A kind of calculation method of question sentence and the Answer Sentence degree of correlation based on intensified learning |
CN108959467B (en) * | 2018-06-20 | 2021-10-15 | 华东师范大学 | Method for calculating correlation degree of question sentences and answer sentences based on reinforcement learning |
CN108846138A (en) * | 2018-07-10 | 2018-11-20 | 苏州大学 | A kind of the problem of fusion answer information disaggregated model construction method, device and medium |
CN108846138B (en) * | 2018-07-10 | 2022-06-07 | 苏州大学 | Question classification model construction method, device and medium fusing answer information |
CN109408624A (en) * | 2018-11-06 | 2019-03-01 | 江西师范大学 | A kind of vision chat robots session generation method |
CN109408624B (en) * | 2018-11-06 | 2020-11-27 | 江西师范大学 | Visual chat robot session generation method |
CN109492227A (en) * | 2018-11-16 | 2019-03-19 | 大连理工大学 | It is a kind of that understanding method is read based on the machine of bull attention mechanism and Dynamic iterations |
CN109543200A (en) * | 2018-11-30 | 2019-03-29 | 腾讯科技(深圳)有限公司 | A kind of text interpretation method and device |
CN109543200B (en) * | 2018-11-30 | 2022-04-15 | 腾讯科技(深圳)有限公司 | Text translation method and device |
CN111309875A (en) * | 2018-12-10 | 2020-06-19 | 百度在线网络技术(北京)有限公司 | Question answering method, device, equipment and storage medium |
CN111309875B (en) * | 2018-12-10 | 2023-08-04 | 百度在线网络技术(北京)有限公司 | Method, device, equipment and storage medium for answering questions |
CN110348462A (en) * | 2019-07-09 | 2019-10-18 | 北京金山数字娱乐科技有限公司 | A kind of characteristics of image determination, vision answering method, device, equipment and medium |
CN110348462B (en) * | 2019-07-09 | 2022-03-04 | 北京金山数字娱乐科技有限公司 | Image feature determination and visual question and answer method, device, equipment and medium |
WO2021072892A1 (en) * | 2019-10-18 | 2021-04-22 | 平安科技(深圳)有限公司 | Legal provision search method based on neural network hybrid model, and related device |
CN110765254A (en) * | 2019-10-21 | 2020-02-07 | 北京理工大学 | Multi-document question-answering system model integrating multi-view answer reordering |
CN110990547A (en) * | 2019-11-29 | 2020-04-10 | 支付宝(杭州)信息技术有限公司 | Phone operation generation method and system |
CN110990547B (en) * | 2019-11-29 | 2023-03-14 | 支付宝(杭州)信息技术有限公司 | Phone operation generation method and system |
Also Published As
Publication number | Publication date |
---|---|
CN107679224B (en) | 2020-09-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107679224A (en) | It is a kind of towards the method and system without structure text intelligent answer | |
CN104239897B (en) | Visual feature representing method based on autoencoder word bag | |
Sneegas | Making the case for critical Q methodology | |
CN110134771A (en) | A kind of implementation method based on more attention mechanism converged network question answering systems | |
Heikkila | Are municipalities Tieboutian clubs? | |
Schiller et al. | Nicaragua’s agroecological transition: Transformation or reconfiguration of the agri-food regime? | |
CN106095872A (en) | Answer sort method and device for Intelligent Answer System | |
CN108664632A (en) | A kind of text emotion sorting algorithm based on convolutional neural networks and attention mechanism | |
CN108345690A (en) | Intelligent answer method and system | |
CN107818164A (en) | A kind of intelligent answer method and its system | |
CN109753567A (en) | A kind of file classification method of combination title and text attention mechanism | |
CN110309503A (en) | A kind of subjective item Rating Model and methods of marking based on deep learning BERT--CNN | |
Serva et al. | Malagasy dialects and the peopling of Madagascar | |
Otto et al. | The interrelation between collective identities and place concepts in local wind energy conflicts | |
Hultman et al. | Ecological masculinities: a response to the manthropocene question? | |
CN106778882A (en) | A kind of intelligent contract automatic classification method based on feedforward neural network | |
CN110321862B (en) | Pedestrian re-identification method based on compact ternary loss | |
CN110309300B (en) | Method for identifying knowledge points of physical examination questions | |
CN106022287A (en) | Over-age face verification method based on deep learning and dictionary representation | |
Arifi et al. | Wind energy counter-conducts in Germany: understanding a new wave of socio-environmental grassroots protest | |
Roman et al. | Civic science in urban forestry: An introduction | |
Arceo et al. | Evaluating sustainability criteria for a marine protected area in Veracruz, Mexico | |
Sarjant et al. | " All You Can Eat" Ontology-Building: Feeding Wikipedia to Cyc | |
Qayumjonovich | In Particular of the Formation of the Oronims | |
CN103440332B (en) | A kind of image search method strengthening expression based on relational matrix regularization |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |