CN106874406A - A kind of interactive output intent for robot - Google Patents
A kind of interactive output intent for robot Download PDFInfo
- Publication number
- CN106874406A CN106874406A CN201710035026.3A CN201710035026A CN106874406A CN 106874406 A CN106874406 A CN 106874406A CN 201710035026 A CN201710035026 A CN 201710035026A CN 106874406 A CN106874406 A CN 106874406A
- Authority
- CN
- China
- Prior art keywords
- request
- similar problems
- answer
- matching
- robot
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- 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
- G06F16/3331—Query processing
Abstract
The invention discloses a kind of interactive output intent and robot for robot.Methods described includes：Acquisition is multi-modal to interactively enter data；Parse it is described it is multi-modal interactively enter data, extract request problem；The matching problem of request problem described in knowledge base is obtained, based on the Similar Problems for obtaining request problem described in the knowledge base when in the absence of the matching problem；According to the answer of the request problem and the Similar Problems generation correspondence request problem, the correspondence multi-modal multi-modal interaction output for interactively entering data is generated and exported according to the answer.Method according to the invention it is possible to when robot knowledge based library searching is less than suitable answer, the answer for voluntarily generating matching returns to user.Compared to prior art, the method for the present invention expanded robot can response scope, improve the response degree of accuracy of intelligent robot, improve the Consumer's Experience of robot.
Description
Technical field
The present invention relates to robot field, and in particular to a kind of interactive output intent for robot.
Background technology
With continuing to develop for artificial intelligence technology, automatic answering system is more and more applied to the daily production of the mankind
In life.
In the prior art, automatic answering system leads to the response search strategy using knowledge based storehouse.Needing to carry out
During automatic-answering back device, obtain current user and put question to, search for what matching user putd question to from knowledge base using knowledge base matching algorithm
Answer.
In above-mentioned acknowledgment strategy, obtain the premise that the current user of correspondence puts question to be preserved in knowledge base with
Put question to the answer of matching in family.But, actual interaction scenarios are complicated and changeable, can not possibly be preserved in knowledge base and are possible to
The corresponding answer of the interaction request of generation.When user puts question to the scope beyond knowledge base, answering system just cannot be to user
Enquirement is responded.
The content of the invention
The invention provides a kind of interactive output intent for robot, methods described includes：
Acquisition is multi-modal to interactively enter data；
Parse it is described it is multi-modal interactively enter data, extract request problem；
The matching problem of request problem described in knowledge base is obtained, it is described based on obtaining when in the absence of the matching problem
The Similar Problems of problem are asked described in knowledge base；
According to the answer of the request problem and the Similar Problems generation correspondence request problem, answered according to described
Case is generated and exports the correspondence multi-modal multi-modal interaction output for interactively entering data.
In one embodiment, answering for the correspondence request problem is generated according to the request problem and the Similar Problems
Case, wherein：
The request problem and the Similar Problems are carried out into participle coding respectively；
Coding result to the request problem and the Similar Problems is decoded to generate the answer.
In one embodiment, the coding result to the request problem and the Similar Problems is decoded, wherein, adopt
The coding result of the request problem and the Similar Problems is decoded with notice mechanism.
In one embodiment, the Similar Problems of the request problem are obtained, wherein, the number of the Similar Problems is one
It is individual.
In one embodiment, the request problem is obtained based on the knowledge base searching when in the absence of the matching problem
Similar Problems, wherein, determine that described matching is asked with the similarity of the request problem based on each problem in the knowledge base
Topic or the Similar Problems, including：
Judge to be matched in the absence of described when the problem of first threshold is reached in the absence of the similarity with the request problem
Problem；
Judge in the absence of described similar when the problem of Second Threshold is reached in the absence of the similarity to the request problem
Problem.
The invention allows for a kind of intelligent robot, the robot includes：
Input acquisition module, it is configured to, and acquisition is multi-modal to interactively enter data；
Problem extraction module, its be configured to parsing it is described it is multi-modal interactively enter data, extract request problem；
Problem matching module, it is configured to obtain the matching problem of request problem described in knowledge base, also, ought not exist
Based on the Similar Problems for obtaining request problem described in the knowledge base during matching problem；
Answer generation module, it is configured to according to the request problem and the Similar Problems generation correspondence request
The answer of problem；
Output module, it is configured to be generated according to the answer and exported, and correspondence is described multi-modal interactively enters many of data
Mode interaction output.
In one embodiment, the answer generation module is included：
Coding unit, it is configured to for the request problem and the Similar Problems to carry out participle coding respectively；
Decoding unit, it is configured to decode with life the coding result of the request problem and the Similar Problems
Into the answer.
In one embodiment, the decoding unit is configured to using notice mechanism to the request problem and the phase
Decoded like the coding result of problem.
In one embodiment, described problem matching module is configured to obtain the Similar Problems of the request problem, wherein, institute
The number for stating Similar Problems is one.
In one embodiment, described problem matching module is configured to each problem and the request in the knowledge base
The similarity of problem determines the matching problem or the Similar Problems, including：
Judge to be matched in the absence of described when the problem of first threshold is reached in the absence of the similarity with the request problem
Problem；
Judge in the absence of described similar when the problem of Second Threshold is reached in the absence of the similarity to the request problem
Problem.
Method according to the invention it is possible to when robot knowledge based library searching is less than suitable answer, voluntarily generate
The answer of matching returns to user.Compared to prior art, the method for the present invention expanded robot can response scope, improve
The response degree of accuracy of intelligent robot, improves the Consumer's Experience of robot.
Further feature of the invention or advantage will be illustrated in the following description.Also, Partial Feature of the invention or
Advantage will be become apparent by specification, or be appreciated that by implementing the present invention.The purpose of the present invention and part
Advantage can be realized or obtained by specifically noted step in specification, claims and accompanying drawing.
Brief description of the drawings
Accompanying drawing is used for providing a further understanding of the present invention, and constitutes a part for specification, with reality of the invention
Apply example to be provided commonly for explaining the present invention, be not construed as limiting the invention.In the accompanying drawings：
Fig. 1 is method flow diagram according to an embodiment of the invention；
Fig. 2, Fig. 3 and Fig. 4 are the partial process views of method according to embodiments of the present invention；
Fig. 5 is robot system architecture's sketch according to an embodiment of the invention；
Fig. 6 and Fig. 7 are robot system part-structure sketches according to embodiments of the present invention.
Specific embodiment
Describe embodiments of the present invention in detail below with reference to drawings and Examples, implementation personnel of the invention whereby
Can fully understand how application technology means solve technical problem for the present invention, and reach technique effect implementation process and according to
The present invention is embodied according to above-mentioned implementation process.If it should be noted that do not constitute conflict, each embodiment in the present invention
And each feature in each embodiment can be combined with each other, the technical scheme for being formed protection scope of the present invention it
It is interior.
With continuing to develop for artificial intelligence technology, automatic answering system is more and more applied to the daily production of the mankind
In life.
In the prior art, automatic answering system is generally using the response search strategy in knowledge based storehouse.I.e. need into
During row automatic-answering back device, obtain current user and put question to, search for matching user from knowledge base using knowledge base matching algorithm and put question to
Answer.
In above-mentioned acknowledgment strategy, obtain the premise that the current user of correspondence puts question to be preserved in knowledge base with
Put question to the answer of matching in family.But, actual interaction scenarios are complicated and changeable, can not possibly be preserved in knowledge base and are possible to
The corresponding answer of the interaction request of generation.When user puts question to the scope beyond knowledge base, answering system just cannot be to user
Enquirement is responded.
For problems of the prior art, the present invention proposes a kind of interactive output intent for robot.
In an embodiment of the invention, when user put question to beyond knowledge base scope when (do not preserved in knowledge base with
Put question to the problem of matching in family), enable dialogue generation model and put question to generation suitable answer according to user.So, putd question in user
Beyond knowledge base scope in the case of (when robot knowledge based library searching is less than suitable answer), it is also possible to voluntarily
The answer for generating matching returns to user.Compared to prior art, the method for the present invention has expanded robot can response scope.
Further, in one embodiment, during using the answer for talking with generation model generation correspondence user, also
Cause that the understanding expression that dialogue generation model is putd question to user is more rich by way of adding priori, so as to generate more
The answer of Semantic so that the answer of dialogue generation model generation more suits user's enquirement.Specifically, being obtained from knowledge base
The Similar Problems similar to user's enquirement, are then putd question to and from knowledge base using talking with the script-based user of generation model
The Similar Problems of acquisition correspond to the answer that the user of script puts question to generate.Answering for intelligent robot can thus be greatly improved
Answer the degree of accuracy.
By taking a specific application scenarios as an example, as one problem of user's request (primal problem, Original Question)
When, robot is found and Original Question identical Question first in knowledge base, if finding identical
Question (here identical refer to implication essentially identical), by the decision model of its corresponding A incoming robot systems of nswer
Type, response of the Answer realizations to user is based on by decision model.
If not finding identical Question, will be with the immediate Similar Problems of Original Question
Incoming dialogue generation model after (Similar Question) and Original Question participles, dialogue generation model
Used as the incoming decision model of answer, decision model judges whether to return to user using Answer as answer the Answer of generation.
For example, one Question of user's request " annual New Year's Day all carrys out high fever ", talks with one Answer of generation model return and " wishes early
Rehabilitation, it is healthy！" as incoming decision-making level is replied, judged to export by decision-making level, the reply is returned into user.
Next the detailed process of method according to embodiments of the present invention is described in detail based on accompanying drawing, in the flow chart of accompanying drawing
The step of showing can perform in comprising the such as one group computer system of computer executable instructions.Although in flow charts
The logical order of each step is shown, but in some cases, can be performing shown different from order herein or retouch
The step of stating.
As shown in figure 1, in one embodiment, robot obtains and multi-modal interactively enter data (such as user's asks first
Topic input) (step S100)；Then be analyzed and acquired by it is multi-modal interactively enter data (step S110), therefrom extract request ask
Topic (user wants what is asked) (primal problem, Original Question) (step S120).
Next the matching problem of request problem is obtained from knowledge base, specifically, i.e. searching request is asked from knowledge base
Can the matching problem (step S130) of topic, judge whether matching problem (search the matching problem) (step of request problem
Rapid S140).
If there is matching problem, then corresponding answer (step S150) is obtained according to the matching problem；And following base
The multi-modal output for responding multi-modal input data is generated and exported in the answer (answer corresponding with matching problem) for getting
(enquirement of response user) (step S160).
If there is no matching problem, then Similar Problems (the step similar to request problem is obtained from knowledge base
S170)；And it is next corresponding according to original request problem and the Similar Problems for getting acquisition using generation model is talked with
The answer (step S180) of original request problem；It is finally based on answer (comprehensive request problem and the Similar Problems for getting
The answer for getting) generate and export multi-modal output (enquirement of response the user) (step for responding multi-modal input data
S160)。
Based on the method for the present invention, not only the answer that matching user puts question to, Er Qieke can be obtained with knowledge based library searching
Corresponding answer is automatically generated with when knowledge base searching is less than corresponding answer (scope of the enquirement of user beyond knowledge base),
Expanded significantly robot can response scope, improve the Consumer's Experience of robot.
Flow according to Fig. 1, when user is putd question in knowledge base scope, the answer (step S150) for getting is
It is pre-stored in answer corresponding with matching problem in knowledge base.Therefore the correctness of answer is that it is pin to set up knowledge base
Whether the answer pre-saved to matching problem is correct.And work as user and put question to the answer (step for inside and outside knowledge base scope, getting
Rapid S150) it is the answer (step S180) talked with generation model synthesis Similar Problems and request problem and generate.Answer is just
Really with the generating process (step S180) for otherwise depending primarily on answer.
In order to ensure the accuracy of answer, in an embodiment of the present invention, dialogue generation model employs encoding and decoding framework
(Encoder-Decoder frameworks) generates answer.Request problem and Similar Problems are carried out into participle coding respectively first；So
The coding result to request problem and Similar Problems is decoded afterwards；Answer is generated finally according to decoded result.
Specifically, Encoder-Decoder frameworks include two parts, Encoder layers and Decoder layers.Encoder layers of master
Refer to the sentence of input variable length, carry out semantic understanding, the vector table for exporting fixed length reaches, and the layer is usually RNN networks.
Decoder layers is primarily referred to as being input into the vector table of fixed length up to (i.e. the output of Encoder), and the difference output according to task is elongated
Sentence.By Similar Question and the Original Question after participle by two Encoder layers, one is generated
Selective vector table to Question middle parts participle (words) reaches, by one Answer of its incoming Decoder layers of generation.
As shown in Fig. 2 in one embodiment, carrying out word segmentation processing (step to request problem and Similar Problems first
S201 and step S202), then the word segmentation result of request problem and Similar Problems is encoded and is obtained request respectively and is asked
The coding result (step S211 and step S212) of topic and Similar Problems.Next it is comprehensive to ask problem and Similar Problems
Coding result decoded (step S220) and ultimately generated answer (step S230)
Further, in an embodiment of the present invention, robot uses notice mechanism (Attention mechanism)
Coding result to request problem and Similar Problems is decoded.And Decoder layers of use of dialogue generation model
Attention mechanism are decoded.Herein, Attention mechanism are a kind of by mode input sentence
The mechanism that j-th word of i-th word (word) and model output sentence is associated.Attention mechanism can
So that the sentence of input expressing by selectivity, rather than whole expression.
Further, in an embodiment of the present invention, the implementation process (step S212) of the coding of Similar Problems is input
It is the Similar Question after participle, semantic meaning representation is carried out by Encoder layers of coding, by coding result and Original
Question coding results carry out selective expression to Similar Question and will tie by Attention mechanism
Really incoming Decoder layers carrying out generation Answer.It is in order to more preferable wherein to add Original Question coding results
To the semantic selection of Similar Question.
Further, in an embodiment of the present invention, coding implementation process (step S211) of request problem is that input is
Original Question after participle, carry out semantic meaning representation, by coding result and Similar by Encoder layers of coding
Question coding results carry out selective expression to Original Question and incite somebody to action by Attention mechanism
Incoming Decoder layers of result is carrying out generation Answer.It is in order to more preferable wherein to add Similar Question coding results
The semantic understanding to Original Question.
As shown in figure 3, carrying out word segmentation processing (step S301 and step to request problem and Similar Problems first
S302)；Then to the Original Question (request problem) after participle, semantic meaning representation is carried out by Encoder layers of coding
(step S311)；Meanwhile, to the Similar Question (Similar Problems) after participle, language is carried out by Encoder layers of coding
Justice expression (step S312).
By the coding result (coding result of request problem) and the coding result (Similar Problems of step S312 of step S311
Coding result) selective expression is carried out to Original Question (request problem) by Attention mechanism
(step S313).Meanwhile, by the coding result (coding result of request problem) and the coding result of step S312 of step S311
(coding results of Similar Problems) are selected Similar Question (Similar Problems) by Attention mechanism
Selecting property expresses (step S314).
(step S320) is decoded by the Decoder layers of result based on step S313 and step S314, with final
Generation answer (Answer) (step S330).
Further, in one embodiment, the degree of accuracy of the answer that generation model is generated is talked with to improve, in Fig. 1 institutes
In the step of showing S170, a Similar Problems of request problem are only obtained.
Further, in one embodiment, each problem determines to match with the similarity of request problem in knowledge based storehouse
Problem or Similar Problems.Specifically, the first similarity threshold is set, when the similarity that there are problems that in knowledge base with ask reaches
During the problem of the first similarity threshold (being more than or equal to), there is matching problem in judgement, similar more than the first similarity threshold
Spend maximum problem and be matching problem.It is more than the first similarity threshold with the similarity of request problem when not existing in knowledge base
Problem when, judge do not exist matching problem, the maximum problem of similarity is Similar Problems.
Further, in practical application, there are problems that in knowledge base it is all of with the request problem degree of association very
Small situation.If now will the maximum problem of similarity (although its similarity is compared, other problemses are maximum, and its is similar
The value (associating system with request problem) of degree very little in fact) exported as Similar Problems and give dialogue generation model, it is possible to can give birth to
Into the answer unrelated with request problem.For such case, in one embodiment, the Similar Problems to asking problem be provided with into
The similarity of one step is limited.Specifically, set the second similarity threshold (it is less than the first similarity threshold), when in the absence of with please
Judge in the absence of Similar Problems when asking the similarity of problem to reach (be more than or equal to) problem of Second Threshold.
Specifically, as shown in figure 4, in one embodiment, robot extracts request problem (step S420) and searches for afterwards to be known
Know storehouse (step S430), judge whether that the matching problem of request problem (whether there is and reach the with request problem similarity
The problem of one similarity threshold) (step S440).
If there is matching problem (there is a problem of that similarity reaches the first similarity threshold), then wherein similarity is obtained
The maximum corresponding answer (step S450) of problem (matching problem)；And next based on the answer for getting (with matching problem
Corresponding answer) generate and export multi-modal output (enquirement of response the user) (step for responding multi-modal input data
S460)。
If there is no matching problem (reaching the problem of the first similarity threshold in the absence of similarity), then judge whether to deposit
In Similar Problems (with the presence or absence of the problem that the second similarity threshold is reached with request problem similarity) (step of request problem
S470)；If there is Similar Problems (there is a problem of that similarity reaches the second similarity threshold), then obtain wherein that similarity is most
Big problem (Similar Problems) (step S471)；And next using talking with generation model according to original request problem and obtain
The Similar Problems got obtain the answer (step S480) of the original request problem of correspondence；It is finally based on the answer for getting (comprehensive
Close the answer that request problem and Similar Problems get) generate and export the multi-modal output for responding multi-modal input data
(enquirement of response user) (step S460).
If there is no Similar Problems, then cannot understand that active user puts question to the instruction of (request problem) to user's output,
Or be based only on the corresponding answer of request problem generation and export (step S460) using generation model is talked with.
The method of the present invention is thoroughly done away with, the invention allows for a kind of robot.As shown in figure 5, in one embodiment, machine
People includes：
Input acquisition module 500, it is configured to, and acquisition is multi-modal to interactively enter data；
Problem extraction module 510, it is configured to, and parsing is multi-modal to interactively enter data, extracts request problem；
Problem matching module 520, it is configured to obtain the matching problem of request problem in knowledge base, also, ought not exist
Similar Problems based on the request problem in knowledge base that obtains during matching problem；
Answer generation module 530, it is configured to generate answering for corresponding requests problem according to request problem and Similar Problems
Case, or corresponding answer is obtained according to matching problem；
Output module 540, it is configured to be generated according to answer and exported, and correspondence is multi-modal interactively enters the multi-modal of data
Interaction output.
Specifically, as shown in fig. 6, problem matching module 620 include problem search unit 621, its obtain problem extract mould
The request problem of the output of block 610 is simultaneously searched in the knowledge base that knowledge base memory 600 (it can also be the webserver) is preserved
Rope.When problem search unit 621 searches matching problem, the answer acquiring unit 631 of answer generation module 630 is from knowledge base
The corresponding answer of matching problem is obtained in the knowledge base that memory 600 is preserved, and answer is exported to output module 540.When asking
Topic search unit 621 is searched for during less than matching problem, and Similar Problems are obtained in its knowledge base preserved from knowledge base memory 600
And export Similar Problems and request problem to the dialogue generation unit 632 of answer generation module 630, talk with generation unit
632 export to output module 640 according to Similar Problems and the generation answer of request problem and by answer.
Further, in one embodiment, problem matching module 620 is configured to obtain the Similar Problems of request problem, its
In, the number of Similar Problems is one.
Further, in one embodiment, problem matching module 620 is configured to each problem and request in knowledge base
The similarity of problem determines matching problem or Similar Problems, including：
Judge in the absence of matching problem when the problem of first threshold is reached in the absence of the similarity with request problem；
Judge in the absence of Similar Problems when the problem of Second Threshold is reached in the absence of the similarity with request problem.
Further, as shown in fig. 7, in one embodiment, answer generation module 700 is included：
Coding unit 701, it is configured to for request problem and Similar Problems to carry out participle coding respectively；
Decoding unit 702, it is configured to decode the coding result of request problem and Similar Problems and is answered with generation
Case.
Further, in one embodiment, decoding unit 702 is configured to using notice mechanism to request problem and phase
Decoded like the coding result of problem.
While it is disclosed that implementation method as above, but described content is only to facilitate understanding the present invention and adopting
Implementation method, is not limited to the present invention.Method of the present invention can also have other various embodiments.Without departing substantially from
In the case of essence of the present invention, those of ordinary skill in the art work as can make various corresponding changes or change according to the present invention
Shape, but these corresponding changes or deformation should all belong to scope of the claims of the invention.
Claims (10)
1. a kind of interactive output intent for robot, it is characterised in that methods described includes：
Acquisition is multi-modal to interactively enter data；
Parse it is described it is multi-modal interactively enter data, extract request problem；
The matching problem of request problem described in knowledge base is obtained, based on the acquisition knowledge when in the absence of the matching problem
The Similar Problems of problem are asked described in storehouse；
According to the answer of the request problem and the Similar Problems generation correspondence request problem, given birth to according to the answer
Into and export correspondence it is described it is multi-modal interactively enter data it is multi-modal interaction output.
2. method according to claim 1, it is characterised in that generated according to the request problem and the Similar Problems
The answer of the correspondence request problem, wherein：
The request problem and the Similar Problems are carried out into participle coding respectively；
Coding result to the request problem and the Similar Problems is decoded to generate the answer.
3. method according to claim 2, it is characterised in that to the request problem and the coding of the Similar Problems
Result is decoded, wherein, the coding result of the request problem and the Similar Problems is carried out using notice mechanism
Decoding.
4. method according to claim 1, it is characterised in that obtain the Similar Problems of the request problem, wherein, it is described
The number of Similar Problems is one.
5. method according to claim 1, it is characterised in that the knowledge base is based on when in the absence of the matching problem
Search obtains the Similar Problems of the request problem, wherein, based on each problem in the knowledge base and the request problem
Similarity determines the matching problem or the Similar Problems, including：
Judge do not exist the matching problem when the problem of first threshold is reached in the absence of the similarity with the request problem；
Judge do not exist the Similar Problems when the problem of Second Threshold is reached in the absence of the similarity with the request problem.
6. a kind of intelligent robot, it is characterised in that the robot includes：
Input acquisition module, it is configured to, and acquisition is multi-modal to interactively enter data；
Problem extraction module, its be configured to parsing it is described it is multi-modal interactively enter data, extract request problem；
Problem matching module, its matching problem for being configured to obtain request problem described in knowledge base, also, when in the absence of described
Based on the Similar Problems for obtaining request problem described in the knowledge base during matching problem；
Answer generation module, it is configured to according to the request problem and the Similar Problems generation correspondence request problem
Answer；
Output module, it is configured to be generated according to the answer and exported, and correspondence is described multi-modal interactively enters the multi-modal of data
Interaction output.
7. robot according to claim 6, it is characterised in that the answer generation module is included：
Coding unit, it is configured to for the request problem and the Similar Problems to carry out participle coding respectively；
Decoding unit, it is configured to decode to generate the coding result of the request problem and the Similar Problems
State answer.
8. robot according to claim 7, it is characterised in that the decoding unit is configured to using notice mechanism pair
The coding result of the request problem and the Similar Problems is decoded.
9. robot according to claim 6, it is characterised in that described problem matching module is configured to obtain the request
The Similar Problems of problem, wherein, the number of the Similar Problems is one.
10. robot according to claim 6, it is characterised in that described problem matching module is configured to described know
Each problem determines the matching problem or the Similar Problems with the similarity of the request problem in knowing storehouse, including：
Judge do not exist the matching problem when the problem of first threshold is reached in the absence of the similarity with the request problem；
Judge do not exist the Similar Problems when the problem of Second Threshold is reached in the absence of the similarity with the request problem.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710035026.3A CN106874406A (en) | 2017-01-18 | 2017-01-18 | A kind of interactive output intent for robot |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710035026.3A CN106874406A (en) | 2017-01-18 | 2017-01-18 | A kind of interactive output intent for robot |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106874406A true CN106874406A (en) | 2017-06-20 |
Family
ID=59157863
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710035026.3A Pending CN106874406A (en) | 2017-01-18 | 2017-01-18 | A kind of interactive output intent for robot |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106874406A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108133051A (en) * | 2018-01-17 | 2018-06-08 | 成都小多科技有限公司 | service consultation method and device |
| CN110209789A (en) * | 2019-05-29 | 2019-09-06 | 山东大学 | A kind of multi-modal dialog system and method for user's attention guidance |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090157482A1 (en) * | 2007-12-13 | 2009-06-18 | Searete Llc, A Limited Liability Corporation Of The State Of Delaware | Methods and systems for indicating behavior in a population cohort |
| CN104657346A (en) * | 2015-01-15 | 2015-05-27 | 深圳市前海安测信息技术有限公司 | Question matching system and question matching system in intelligent interaction system |
| CN104679910A (en) * | 2015-03-25 | 2015-06-03 | 北京智齿博创科技有限公司 | Intelligent answering method and system |
| CN105787560A (en) * | 2016-03-18 | 2016-07-20 | 北京光年无限科技有限公司 | Dialogue data interaction processing method and device based on recurrent neural network |
| CN105824935A (en) * | 2016-03-18 | 2016-08-03 | 北京光年无限科技有限公司 | Method and system for information processing for question and answer robot |
| CN106021273A (en) * | 2016-04-25 | 2016-10-12 | 北京光年无限科技有限公司 | Method and system for processing information facing question answering robot |
| CN106205612A (en) * | 2016-07-08 | 2016-12-07 | 北京光年无限科技有限公司 | Information processing method and system towards intelligent robot |
| CN106295792A (en) * | 2016-08-05 | 2017-01-04 | 北京光年无限科技有限公司 | Dialogue data interaction processing method based on multi-model output and device |
-
2017
- 2017-01-18 CN CN201710035026.3A patent/CN106874406A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090157482A1 (en) * | 2007-12-13 | 2009-06-18 | Searete Llc, A Limited Liability Corporation Of The State Of Delaware | Methods and systems for indicating behavior in a population cohort |
| CN104657346A (en) * | 2015-01-15 | 2015-05-27 | 深圳市前海安测信息技术有限公司 | Question matching system and question matching system in intelligent interaction system |
| CN104679910A (en) * | 2015-03-25 | 2015-06-03 | 北京智齿博创科技有限公司 | Intelligent answering method and system |
| CN105787560A (en) * | 2016-03-18 | 2016-07-20 | 北京光年无限科技有限公司 | Dialogue data interaction processing method and device based on recurrent neural network |
| CN105824935A (en) * | 2016-03-18 | 2016-08-03 | 北京光年无限科技有限公司 | Method and system for information processing for question and answer robot |
| CN106021273A (en) * | 2016-04-25 | 2016-10-12 | 北京光年无限科技有限公司 | Method and system for processing information facing question answering robot |
| CN106205612A (en) * | 2016-07-08 | 2016-12-07 | 北京光年无限科技有限公司 | Information processing method and system towards intelligent robot |
| CN106295792A (en) * | 2016-08-05 | 2017-01-04 | 北京光年无限科技有限公司 | Dialogue data interaction processing method based on multi-model output and device |
Non-Patent Citations (5)
| Title |
|---|
| BING LIU等: "Attention-Based Recurrent Neural Network Models for Joint Intent Detection and slot filling", 《RESEARCHGATE》 * |
| LI DONG等: "Language to Logical Form with Neural Attention", 《RESEARCHGATE》 * |
| SUBHASHINI VENUGOPALAN等: "Sequence to Sequence-Video to Text", 《2015 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION》 * |
| ZHILIN YANG等: "Encode, Review, and Decode: Reviewer Module for Caption Generation", 《RESEARCHGATE》 * |
| 黄积杨: "基于双向LSTMN神经网络的中文分词研究分析", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108133051A (en) * | 2018-01-17 | 2018-06-08 | 成都小多科技有限公司 | service consultation method and device |
| CN110209789A (en) * | 2019-05-29 | 2019-09-06 | 山东大学 | A kind of multi-modal dialog system and method for user's attention guidance |
| CN110209789B (en) * | 2019-05-29 | 2021-08-24 | 山东大学 | Multi-modal dialog system and method for guiding user attention |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111309889B (en) | Method and device for text processing | |
| US11281945B1 (en) | Multimodal dimensional emotion recognition method | |
| CN106297789B (en) | Personalized interaction method and system for intelligent robot | |
| TW201833903A (en) | Conversation generating method, device, and storage medium | |
| CN108829757A (en) | A kind of intelligent Service method, server and the storage medium of chat robots | |
| CN108804698A (en) | Man-machine interaction method, system, medium based on personage IP and equipment | |
| CN111538848A (en) | Knowledge representation learning method fusing multi-source information | |
| CN109992669B (en) | Keyword question-answering method based on language model and reinforcement learning | |
| CN110851575B (en) | Dialogue generating system and dialogue realizing method | |
| CN113761893B (en) | Relation extraction method based on mode pre-training | |
| CN112463924B (en) | Text intention matching method for intelligent question answering based on internal correlation coding | |
| CN111897929B (en) | Method and device for processing multiple rounds of questions, storage medium and electronic equipment | |
| CN110942774A (en) | Man-machine interaction system, and dialogue method, medium and equipment thereof | |
| CN115293348A (en) | Pre-training method and device for multi-mode feature extraction network | |
| CN113553418A (en) | Visual dialog generation method and device based on multi-modal learning | |
| CN106874406A (en) | A kind of interactive output intent for robot | |
| CN109933773A (en) | A kind of multiple semantic sentence analysis system and method | |
| CN108509539A (en) | Information processing method electronic equipment | |
| CN114372454A (en) | Text information extraction method, model training method, device and storage medium | |
| Xu et al. | Enabling language representation with knowledge graph and structured semantic information | |
| CN109522920A (en) | Training method and equipment based on the synonymous discrimination model for combining semantic feature | |
| CN113569585A (en) | Translation method and device, storage medium and electronic equipment | |
| CN109739970A (en) | Information processing method and device and electronic equipment | |
| CN114840697B (en) | Visual question-answering method and system for cloud service robot | |
| Yao et al. | Context-guided triple matching for multiple choice question answering |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170620 |