CN109510753A - Construction method, interaction response method and apparatus, the storage medium, server of group, IP robot - Google Patents
Construction method, interaction response method and apparatus, the storage medium, server of group, IP robot Download PDFInfo
- Publication number
- CN109510753A CN109510753A CN201710833842.9A CN201710833842A CN109510753A CN 109510753 A CN109510753 A CN 109510753A CN 201710833842 A CN201710833842 A CN 201710833842A CN 109510753 A CN109510753 A CN 109510753A
- Authority
- CN
- China
- Prior art keywords
- robot
- corpus
- role
- group
- works
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L51/00—User-to-user messaging in packet-switching networks, transmitted according to store-and-forward or real-time protocols, e.g. e-mail
- H04L51/02—User-to-user messaging in packet-switching networks, transmitted according to store-and-forward or real-time protocols, e.g. e-mail using automatic reactions or user delegation, e.g. automatic replies or chatbot-generated messages
Abstract
Construction method, interaction response method and apparatus, the storage medium, server of a kind of group, IP robot, the building comprise determining that works corpus and multiple IP robots with interaction response method;Using the multiple IP robot of role's corpus training of each role in the works corpus, to obtain the IP robot after multiple training;Group, IP robot is added in IP robot after the multiple training;The input information of interaction person is sent to each IP robot in group, IP robot, matching degree sequence is carried out in response to the reply message of the input information to each IP robot;The highest reply message of matching degree is exported to the interactive person.Technical solution of the present invention solves the problems, such as that IP robot existing in the prior art corpus is insufficient, improves the user experience in human-computer interaction.
Description
Technical field
The present invention relates to technical field of information processing, and in particular to a kind of construction method of group, IP robot, interaction are rung
Induction method and device, storage medium, server.
Background technique
In recent years, under the promotion of artificial intelligence study's upsurge, intelligent chat robots are oriented the following various products
With the entrance of service, it has also become the primary study project of each related fields company.Main side of the intelligent answer as human-computer interaction
Formula has a wide range of applications scene.Intellectual property (Intellectual Property, IP) robot is in current powerful network
It comes into being under scientific and technical background.So-called IP robot, refers to using the role in the original artworks such as novel, drama, caricature
Role's corpus of (i.e. figure image) is trained, and can imitate the chat machine that original artwork role and user interact
People.
In the prior art, the step of training IP robot is as follows: (1) trainer passes through robot teaching interface to IP machine
People inputs problem and provides answer, answers using problem as the context of corpus in this, as the corpus of IP robot and is used as language
The answer of material;(2) step 1 is repeated, so that IP robot includes a large amount of corpus.When carrying out human-computer interaction, user is to IP machine
People inputs problem, and IP robot will compare the problem of input with the context of corpus, looks for the highest corpus of matching degree
Then context exports the answer of the corpus to user as the reply message of IP robot.
The method defect of existing trained IP robot is as follows: firstly, since the single role in original artwork is in works language
The number of the workable role's corpus occurred in material is less, it is difficult to reach corpus number required for one IP robot of training
Mesh, therefore, if the user IP robot less with corpus engages in the dialogue, the latter is difficult to provide and make us in many cases,
Satisfied answer, this will lead to chat content cavity, and session is difficult to continue;Secondly, if using each in any works corpus
Role's corpus of role enriches the corpus of IP robot, and the personality consistency that will lead to IP robot is poor.
Summary of the invention
Present invention solves the technical problem that being that corpus is insufficient when how to solve the problems, such as to train IP robot, improve man-machine friendship
User experience in mutually.
In order to solve the above technical problems, the embodiment of the present invention provides a kind of construction method of group, IP robot, IP machine
The construction method of crowd's group comprises determining that works corpus and multiple IP robots;Use each role in the works corpus
The multiple IP robot of role's corpus training, to obtain the IP robot after multiple training;By the IP after the multiple training
Group, IP robot is added in robot.
Optionally, role's corpus of the role is the set of dialogue of the role in the works corpus.
Optionally, role's corpus training the multiple IP robot packet using each role in the works corpus
It includes: using the single IP robot of role's corpus training of at least one role.
Optionally, the multiple IP robot of role's corpus training using each role in the works corpus it
Before further include: the works corpus is cleaned and marked, to obtain role's corpus of each role.
Optionally, the works corpus is determined according to following one or more original artworks: novel, drama, caricature.
In order to solve the above technical problems, the embodiment of the present invention also provides a kind of interaction response method, interaction response method packet
It includes: receiving the input information of interaction person;The input information is sent to each IP robot in group, IP robot, it is described
Group, IP robot is constructed using the construction method;To each IP robot in response to the reply message of the input information
Carry out matching degree sequence;The highest reply message of matching degree is exported to the interactive person.
Optionally, each IP robot determines in the following way in response to the reply message of the input information: if
The similarity of the corpus context of the input information and the IP robot is more than preset threshold, then using above and below the corpus
Text is corresponding to reply the reply message for generating the IP robot.
Optionally, described to utilize the corresponding reply message packet for replying the generation IP robot of the corpus context
It includes: role's title in the corresponding answer of the corpus context being replaced with and is referred in the title or dialogue of the interactive person
Role's title, to obtain the reply message of the IP robot.
In order to solve the above technical problems, the embodiment of the present invention also provides a kind of construction device of group, IP robot, IP machine
The construction device of device crowd group comprises determining that module, is adapted to determine that works corpus and multiple IP robots;Training module is suitable for
Using the multiple IP robot of role's corpus training of each role in the works corpus, to obtain the IP after multiple training
Robot;Package module, suitable for group, IP robot is added in the IP robot after the multiple training.
Optionally, role's corpus of the role is the set of dialogue of the role in the works corpus.
Optionally, the training module is suitable for the single IP robot of role's corpus training using at least one role.
Optionally, the construction device further include: corpus cleaning module, suitable for the works corpus carry out cleaning and
Mark, to obtain role's corpus of each role.
Optionally, the works corpus is determined according to following one or more original artworks: novel, drama, caricature.
In order to solve the above technical problems, the embodiment of the present invention also provides a kind of interaction response device, interaction response device packet
It includes: information receiving module, suitable for receiving the input information of interaction person;Sending module, suitable for the input information is sent to IP
Each IP robot in group, robot, group, the IP robot are constructed using construction device described in any of the above embodiments;
Answer sorting module, suitable for carrying out matching degree sequence in response to the reply message of the input information to each IP robot;Letter
Output module is ceased, suitable for exporting the highest reply message of matching degree to the interactive person.
Optionally, each IP robot determines in the following way in response to the reply message of the input information: if
The similarity of the corpus context of the input information and the IP robot is more than preset threshold, then using above and below the corpus
Text is corresponding to reply the reply message for generating the IP robot.
Optionally, role's title in the corresponding answer of the corpus context is replaced with institute by the message output module
The role's title referred in the title or dialogue of interactive person is stated, to obtain the reply message of the IP robot.
The embodiment of the invention also discloses a kind of storage mediums, are stored thereon with computer instruction, the computer instruction
The step of construction method based on group, IP robot or the interaction response method are executed when operation.
The embodiment of the invention also discloses a kind of server, including memory and processor, it is stored on the memory
The computer instruction that can be run on the processor, the processor execute the IP machine when running the computer instruction
The step of construction method of crowd's group or the interaction response method.
Compared with prior art, the technical solution of the embodiment of the present invention has the advantages that
Technical solution of the present invention constructs group, IP robot in the following way: determining works corpus and multiple IP machines
People;Using the multiple IP robot of role's corpus training of each role in the works corpus, after obtaining multiple training
IP robot;Group, IP robot is added in IP robot after the multiple training.Using single in compared with the existing technology
For the scheme of role's corpus training IP robot of a role, technical solution of the present invention uses multiple angles of works corpus offer
The multiple IP robots of role's corpus training of color.For group, entire IP robot, workable corpus quantity is increased,
Therefore, when group, entire IP robot being used for single man-machine conversation, the quality of session can be improved.Specifically, for
The context of user's input can choose a suitable IP robot from group, IP robot and answer, so that answer
Content and context keep higher continuity.
Further, before using the multiple IP robots of role's corpus training of each role in works corpus further include:
The works corpus is cleaned and marked, to obtain role's corpus of each role.Technical solution of the present invention passes through to work
Product corpus is cleaned and is marked, and is deleted repetition and redundancy, is identified, marks, extracting role's corpus of each role, avoiding
There is the problem of personality consistency difference in IP robot.
In addition, technical solution of the present invention also provides a kind of interaction response method, this method comprises: receiving the input of interaction person
Information;The input information is sent to each IP robot in group, IP robot, group, the IP robot is using upper
State construction method building;Matching degree sequence is carried out in response to the reply message of the input information to each IP robot;General
It exports with highest reply message is spent to the interactive person.It is sorted by matching degree, technical solution of the present invention can be from IP machine
Crowd, which organizes, obtains one and the highest reply message of matching degree of the input information of the person of interaction, further increases group, IP robot
Answer quality, so that reply message is more met session requirements.
Detailed description of the invention
Fig. 1 is a kind of flow chart of the construction method of group, IP robot of the embodiment of the present invention;
Fig. 2 is a kind of flow chart of the interaction response method of group, IP robot of the embodiment of the present invention;
Fig. 3 is a kind of structural schematic diagram of the construction device of group, IP robot of the embodiment of the present invention;
Fig. 4 is a kind of structural schematic diagram of the interaction response device of group, IP robot of the embodiment of the present invention.
Specific embodiment
As indicated in the background, there is corpus deficiency in the technology of existing trained IP robot, and IP robot is caused to be answered
Change ability is poor, can not provide satisfactory answer.
Technical solution of the present invention constructs group, IP robot in the following way: determining works corpus and multiple IP machines
People;Using the multiple IP robot of role's corpus training of each role in the works corpus, after obtaining multiple training
IP robot;Group, IP robot is added in IP robot after the multiple training.Using single in compared with the existing technology
For the scheme of role's corpus training IP robot of a role, technical solution of the present invention uses multiple angles of works corpus offer
The multiple IP robots of role's corpus training of color.For group, entire IP robot, workable corpus quantity is increased,
Therefore, when group, entire IP robot being used for single man-machine conversation, the quality of dialogue can be improved.Specifically, for
The context of user's input can choose a suitable IP robot from group, IP robot and answer, so that answer
Content and context keep higher continuity.
It is understandable to enable above-mentioned purpose of the invention, feature and beneficial effect to become apparent, with reference to the accompanying drawing to this
The specific embodiment of invention is described in detail.
Fig. 1 is a kind of flow chart of the construction method of group, IP robot of the embodiment of the present invention.
The construction method of group, IP robot shown in FIG. 1 may comprise steps of:
Step S101: works corpus and multiple IP robots are determined;
Step S102: training the multiple IP robot using role's corpus of each role in the works corpus, with
IP robot after obtaining multiple training;
Step S103: group, IP robot is added in the IP robot after the multiple training.
In the specific implementation of step S101, works corpus can be determined according to following one or more original artworks: small
It says, drama, caricature, but not limited to this, works corpus can also be carried out using the type in other any original artworks appropriate
It determines, no longer illustrates one by one herein.Wherein, works corpus can be the collection of the dialogue for each role being present in original artwork
It closes, in addition also may include context relevant to dialogue.It include role's corpus of multiple roles in works corpus.Wherein, often
Role's corpus of a role is the set of dialogue of the role in the works corpus, in addition also may include and dialogue phase
The context of pass.
It in step s 102, specifically can be with using role's corpus training IP robot of each role in works corpus
Using well known to a person skilled in the art the IP robots that various training method training appropriate correspond to different role.For example,
All role's corpus of the same role in works corpus can be passed through to robot teaching interface with the side of " question-response "
Formula trains IP robot, and then the IP robot after being trained.It is right in original artwork that IP robot after training can imitate
The role answered and user engage in the dialogue.
In specific implementation, single IP robot can be used a role in works corpus role's corpus or multiple angles
Role's corpus of color is trained.It is understood that in works corpus unless each except role's corpus of role, it is also possible to exist
Various unrelated contents.Therefore, before role's corpus training IP robot using each role in the works corpus, also
The works corpus can be cleaned and be marked, delete repetition and redundancy, with identification, mark, extract each role
Role's corpus.For example, the works corpus of TV play drama can be cleaned, role's language of mark and typing different role
Material and scene information relevant to role's corpus.Furthermore, it can will be obtained after cleaning and mark according to work title
Each role role's corpus and scene information saved respectively to role's corpus and scene library, wherein role's corpus can
To be stored for training IP robot, scene library can be stored for distinguishing different works corpus and/or role's language
The application range of material.After obtaining role's corpus, IP robot can be trained using role's corpus.
Further, according to same portion's original artwork determine works corpus in may containing multiple dominant roles and time
Role's corpus of role is wanted, therefore, can be filtered out role's corpus of secondary part, on the one hand before training IP robot
Trained corpus input quantity can be reduced to improve efficiency, it on the other hand can also be to avoid role's corpus of secondary part and main
After role's corpus of role trains IP robot together, lead to the problem of IP robot personality consistency difference.
In step s 103, group, IP robot is added in the IP robot by multiple after training.Specifically, single IP
Group, robot can correspond to individual original artwork, and also as single original artwork creates group, individual IP robot.Example
Such as, role's corpus training IP robot of whole roles of multiple collection of dramas of TV play Pilgrimage to the West can be used.Specifically, first
Works corpus is first determined from Pilgrimage to the West original novel or drama, it is all then to obtain Pilgrimage to the West by cleaning and mark
Sun Wukong, Tang's monk, one of the chief characters in "Pilgrimage To The West" who was supposedly incarnated through the spirit of pig, a symbol of man's cupidity and Sha are realized the role's corpus and scene information of each role such as net by role's corpus of role
It is saved respectively into the role's corpus and scene library of Pilgrimage to the West;Then using each angle in role's corpus of Pilgrimage to the West
The multiple IP robots of role's corpus training of color, to obtain the IP machine of each role in the correspondence Pilgrimage to the West after multiple training
People;Finally, the IP robot after multiple training is configured to group, single Pilgrimage to the West robot.Group, single IP robot
A series of original artwork can be corresponded to, that is, creating group, single IP robot for multiple original artworks.For example, can make
With role's corpus training IP robot of the dominant role of the multi-section novel in Jin Yong's portfolio, the IP robot after training is added
Enter group, same IP robot, so that multi-sections such as " Swordmans ", " the semi-gods and the semi-devils " and " Hero Shooting Vulture " are covered in building
The group, IP robot of original artwork.
Fig. 2 is a kind of flow chart of the interaction response method of group, IP robot of the embodiment of the present invention.
The interaction response method of group, IP robot shown in Fig. 2 may comprise steps of:
Step S201: the input information of interaction person is received;
Step S202: the input information is sent to each IP robot in group, IP robot, the IP machine
Crowd's group is using the building of construction method described in embodiment illustrated in fig. 1;
Step S203: matching degree sequence is carried out in response to the reply message of the input information to each IP robot;
Step S204: the highest reply message of matching degree is exported to the interactive person.
In step s 201, the interactive person can be user, be also possible to chat robots, and the chat robots are for example
It can be the IP robot in group, same IP robot, be also possible to any other robot for chat.
In step S202, the input information is sent to each IP robot in group, IP robot.The IP
Group, robot can be using the building of construction method described in embodiment illustrated in fig. 1.In specific implementation, each IP robot difference
Generate the reply message in response to the input information.If an IP robot is the IP obtained by the training of multiple role's corpus
Robot, then the IP robot will generate multiple reply messages.The reply message can determine in the following way: if institute
The similarity for stating the corpus context of input information and the IP robot is more than preset threshold, then utilizes the corpus context
It is corresponding to reply the reply message for generating the IP robot.For example, robot little Ai is instructed using role's corpus of Sun Wukong
The single role IP robot obtained after white silk, robot little Bao are double using obtaining after the training of role's corpus of Tang's monk and one of the chief characters in "Pilgrimage To The West" who was supposedly incarnated through the spirit of pig, a symbol of man's cupidity
The common structure of IP robot that role IP robot, the two and role's corpus using other roles in Pilgrimage to the West obtain after training
Group, Jian Liaoyige Pilgrimage to the West robot.When the small long-drawn-out input " White Bone Demon is monster " of user, the online robot of network is small
Chinese mugwort and little Bao can provide reply message respectively.Assuming that the answer of little Ai is " where monster escapes? " the answer of little Bao is two total,
It is " to realize sky, stop impolite respectively!" and " monkey brother!Master is caught away by monster." so far, which obtains in total
Three the candidate of IP robot are answered.
In step S203, group, IP robot carries out matching degree sequence to all reply messages of IP robot.Specifically
In implementation, matching degree sequence can be according to the continuity degree of reply message and current context on grammer or dialogue scene
According to being ranked up from high to low, the highest reply message of matching degree is as reply message.
In step S204, the highest reply message of the matching degree is exported to interaction person.Those skilled in the art answer
When understanding, the input information of interactive person and the matching of role's corpus context can be calculated using any enforceable mode
Degree, the embodiment of the present invention to this with no restriction.
Preferably, a degree of modification can also be carried out to the reply message that group, IP robot exports, specifically,
Role's title in the corresponding answer of role's corpus context can be replaced with to the title of the interactive person, to obtain
State the reply message of IP robot.For example, by taking group, Pilgrimage to the West robot as an example, when the small long-drawn-out input " what " of user
When, the answer of little Ai is that " master, I is to be one day Buddhist monk, hits one day clock!" answer of little Bao is " to realize sky, you walk too fast
?." and " senior apprentice!What's the matter for your eyes? " so far, group, Pilgrimage to the West robot obtains the time of three IP robots in total
Choosing is answered.Assuming that being sorted according to matching degree, " master, I is to be one day Buddhist monk, hits one day clock for selection!" as the answer exported
Information.In order to make dialogue more remarkable fluency, output to role's title in the reply message of interaction person can use interaction person
Role's title be replaced, i.e., " small long-drawn-out, I is to be one day Buddhist monk, hits one day clock!".
Furthermore, role's title in the corresponding answer of role's corpus context can also be replaced with dialogue
In role's title for referring to, to obtain the reply message of the IP robot.For example, it is assumed that " where monster escapes? " as current
The highest reply message of matching degree can be exported to interaction person.However, in order to close to human conversation mode, in the reply message
Role's title can be replaced using role's title for referring in dialogue, i.e., by " where monster escapes? " be changed to " White Bone Demon which
In escape? ".
It is understood that if the corpus context of one of IP robot in group, IP robot with it is described defeated
When entering the similarity of information less than preset threshold, then the IP robot does not generate reply message.If in group, IP robot
All IP robots role's corpus context and the input information similarity when being respectively less than preset threshold, then the IP
Group, robot does not generate reply message, alternatively, the group, IP robot can export the reply message of default, such as " I not
Understand what you are saying ".
It, can also be by the popular characters in works corpus in order to facilitate user's input in a unrestricted example
The context of role's corpus be set to the signal language clicked for user, or the scene information of works corpus can also be set
It is set to the signal language to click for user.When the user clicks when the signal language, corresponding context or scene information are just
As the quick input of user, group, IP robot can the corresponding answer of context directly by role's corpus or field
Corresponding reply of scape information exports as reply message to user.For example, using the directory name of Pilgrimage to the West collection of drama as scene
Information is set as the signal language clicked for user, if user clicks the signal language, IP robot can directly display " west
Travel notes " collection of drama all directory names.
Fig. 3 is a kind of structural schematic diagram of the construction device of group, IP robot of the embodiment of the present invention.
The construction device 30 of group, IP robot shown in Fig. 3 can include determining that module 301, corpus cleaning module 302,
Training module 303 and package module 304.
In specific implementation, determining module 301 is to determine works corpus and multiple IP robots.Works corpus is according to following
One or more original artworks determine: novel, drama and caricature.Works corpus can be each angle being present in original artwork
The dialogue set of color and the context closely related with dialogue.Works corpus includes role's corpus of multiple roles.Wherein, often
Role's corpus of a role is the set of dialogue of the role in the works corpus.
In specific implementation, training module 303 is described more using role's corpus training of each role in the works corpus
A IP robot, to obtain the IP robot after multiple training.For example, can use novel, drama, in the works corpus such as caricature
Role's corpus training IP robot specifically can be using well known to a person skilled in the art various training methods appropriate
Training IP robot.For example, all role's corpus of the same role in works corpus can be passed through robot teaching circle
Face training IP robot in a manner of " question-response ", the IP robot after being trained can imitate original artwork role and hand over
Mutual person engages in the dialogue.Furthermore, single IP robot can be used a role in works corpus role's corpus or
Use the single IP robot of role's corpus training of multiple roles.
In a unrestricted example, the construction device 30 of group, the IP robot can also include that corpus cleans
Module 302, the module can be right before role's corpus training IP robot using each role in the works corpus
The works corpus is cleaned and is marked, and repetition and redundancy are deleted, and is marked from works corpus, identifies, is extracted role
Corpus and associated scenario information, in case subsequent trained IP robot.
Package module 304 is to be packaged into a group, IP robot for the multiple IP robot after training.
Specifically, group, IP robot can be the group, IP robot of corresponding independent original artwork role, it is also possible to corresponding series
The group, IP robot of original artwork role.It is specifically referred to the construction method of embodiment illustrated in fig. 1, details are not described herein again.
Fig. 4 is a kind of structural schematic diagram of the interaction response device of group, IP robot of the embodiment of the present invention.
The interaction response device 40 of group, IP robot shown in Fig. 4 may include information receiving module 401, sending module
402, answer sorting module 403 and message output module 404.
Information receiving module 401 is used to receive the input information of interactive person.Interaction person can be user, be also possible to chat
Robot, the chat robots for example can be the IP robot in group, same IP robot, are also possible to any other
Robot for chat.
Sending module 402 is used to the input information being sent to each IP robot in group, IP robot, described
Group, IP robot can be constructed using the construction device of group, IP robot shown in Fig. 3.
Answer sorting module 403 is used to match each IP robot in response to the reply message of the input information
Degree sequence.In specific implementation, each IP robot can in the following way really in response to the reply message of the input information
It is fixed: if the similarity of the corpus context of the input information and the IP robot is more than preset threshold, to utilize described
Corpus context is corresponding to reply the reply message for generating the IP robot.It is understood that if group, IP robot
When the similarity of the corpus context of interior one of IP robot and the input information is less than preset threshold, then the IP
Robot does not generate reply message.Predictably, if above and below the corpus of all IP robots in group, IP robot
When the literary similarity with the input information is respectively less than preset threshold, then the answer sorting module 403 does not generate reply message.
Message output module 404 is suitable for exporting the highest reply message of matching degree to the interactive person.Preferably, may be used also
A degree of modification is carried out with the reply message that exports to group, IP robot, such as is answered the corpus context is corresponding
Role's title in multiple replaces with the role's title referred in the title or dialogue of the interactive person, so that the IP robot
The answer of group more meets the speech habits that the mankind are exchanged, and then improves the quality of man-machine conversation.
It, can also be by the popular characters in works corpus in order to facilitate user's input in a unrestricted example
Role's corpus context or scene information be treated as user click signal language.The signal language is made when the user clicks
For user quick input when, group, IP robot can be directly by the corresponding answer of role's corpus context or correspondence
Scene information export as reply message to user.
It is understood that if answer sorting module 403 does not generate reply message.At this moment, message output module 404 does not have
There is the message for output, or the reply message of default can be exported, such as " I does not understand what you are saying ".
Further, the embodiment of the invention also discloses a kind of storage mediums, are stored thereon with computer instruction, the meter
The construction method of group, IP robot or interaction response side in above-mentioned Fig. 1 or embodiment illustrated in fig. 2 are executed when the instruction operation of calculation machine
The step of method.Preferably, the storage medium may include computer readable storage medium.In specific implementation, the storage
Medium may include ROM, RAM, disk or CD etc..
Further, the embodiment of the invention also discloses a kind of server, including memory and processor, the memories
On be stored with the computer instruction that can be run on the processor, the processor executes when running the computer instruction
In above-mentioned Fig. 1 or embodiment illustrated in fig. 2 the step of the construction method or interaction response method of group, IP robot.
Although present disclosure is as above, present invention is not limited to this.Anyone skilled in the art are not departing from this
It in the spirit and scope of invention, can make various changes or modifications, therefore protection scope of the present invention should be with claim institute
Subject to the range of restriction.
Claims (18)
1. a kind of construction method of group, IP robot characterized by comprising
Determine works corpus and multiple IP robots;
Using the multiple IP robot of role's corpus training of each role in the works corpus, after obtaining multiple training
IP robot;
Group, IP robot is added in IP robot after the multiple training.
2. construction method according to claim 1, which is characterized in that role's corpus of the role is the role in institute
State the set of the dialogue in works corpus.
3. construction method according to claim 2, which is characterized in that described to use each role in the works corpus
The multiple IP robot of role's corpus training includes: the single IP robot of role's corpus training using at least one role.
4. construction method according to claim 1,2 or 3, which is characterized in that described using each in the works corpus
Before the multiple IP robot of role's corpus training of role further include:
The works corpus is cleaned and marked, to obtain role's corpus of each role.
5. construction method according to claim 1,2 or 3, which is characterized in that the works corpus according to the next item down or
Multinomial original artwork determines: novel, drama, caricature.
6. a kind of interaction response method characterized by comprising
Receive the input information of interaction person;
The input information is sent to each IP robot in group, IP robot, group, the IP robot uses right
It is required that construction method described in any one of 1 to 5 constructs;
Matching degree sequence is carried out in response to the reply message of the input information to each IP robot;
The highest reply message of matching degree is exported to the interactive person.
7. interaction response method according to claim 6, which is characterized in that each IP robot believes in response to the input
The reply message of breath determines in the following way: if the input information is similar to the corpus context of the IP robot
Degree is more than preset threshold, then utilizes the corresponding reply message for replying the generation IP robot of the corpus context.
8. interaction response method according to claim 7, which is characterized in that described corresponding using the corpus context
Answer to generate the reply message of the IP robot include: to replace role's title in the corresponding answer of the corpus context
It is changed to the role's title referred in the title or dialogue of the interactive person, to obtain the reply message of the IP robot.
9. a kind of construction device of group, IP robot characterized by comprising
Determining module is adapted to determine that works corpus and multiple IP robots;
Training module, suitable for using the multiple IP robot of role's corpus training of each role in the works corpus, with
IP robot after obtaining multiple training;
Package module, suitable for group, IP robot is added in the IP robot after the multiple training.
10. construction device according to claim 9, which is characterized in that role's corpus of the role is that the role exists
The set of dialogue in the works corpus.
11. construction device according to claim 10, which is characterized in that the training module is suitable for using at least one angle
The single IP robot of role's corpus training of color.
12. according to construction device described in claim 9,10 or 11, which is characterized in that further include:
Corpus cleaning module, suitable for the works corpus is cleaned and marked, to obtain role's corpus of each role.
13. according to construction device described in claim 9,10 or 11, which is characterized in that the works corpus is according to the next item down
Or multinomial original artwork determines: novel, drama, caricature.
14. a kind of interaction response device characterized by comprising
Information receiving module, suitable for receiving the input information of interaction person;
Sending module, suitable for the input information is sent to each IP robot in group, IP robot, the IP machine
Crowd's group is using construction device building described in any one of claim 9 to 13;
Answer sorting module, suitable for carrying out matching degree row in response to the reply message of the input information to each IP robot
Sequence;
Message output module, suitable for exporting the highest reply message of matching degree to the interactive person.
15. interaction response device according to claim 14, which is characterized in that each IP robot is in response to the input
The reply message of information determines in the following way: if the phase of the input information and the corpus context of the IP robot
It is more than preset threshold like degree, then utilizes the corresponding reply message for replying the generation IP robot of the corpus context.
16. interaction response device according to claim 14, which is characterized in that the message output module is by institute's predicate
Role's title in the corresponding answer of material context replaces with the role's title referred in the title or dialogue of the interactive person,
To obtain the reply message of the IP robot.
17. a kind of storage medium, is stored thereon with computer instruction, which is characterized in that the computer instruction executes when running
Friendship described in any one of the construction method of group, IP robot or claim 6 to 8 described in any one of claims 1 to 5
The step of mutual response method.
18. a kind of server, including memory and processor, be stored on the memory to run on the processor
Computer instruction, which is characterized in that perform claim requires any one of 1 to 5 when the processor runs the computer instruction
The step of interaction response method described in any one of the construction method of group, the IP robot or claim 6 to 8.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710833842.9A CN109510753A (en) | 2017-09-15 | 2017-09-15 | Construction method, interaction response method and apparatus, the storage medium, server of group, IP robot |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710833842.9A CN109510753A (en) | 2017-09-15 | 2017-09-15 | Construction method, interaction response method and apparatus, the storage medium, server of group, IP robot |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109510753A true CN109510753A (en) | 2019-03-22 |
Family
ID=65745017
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710833842.9A Pending CN109510753A (en) | 2017-09-15 | 2017-09-15 | Construction method, interaction response method and apparatus, the storage medium, server of group, IP robot |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109510753A (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050049752A1 (en) * | 2003-08-28 | 2005-03-03 | Sony Corporation | Robot apparatus, control method for robot apparatus, and toy for robot apparatus |
CN101282314A (en) * | 2008-05-23 | 2008-10-08 | 中国移动通信集团北京有限公司 | Group communication method, equipment and system for instant communication |
CN103078867A (en) * | 2013-01-15 | 2013-05-01 | 深圳市紫光杰思谷科技有限公司 | Automatic chatting method and chatting system among robots |
CN105141503A (en) * | 2015-08-13 | 2015-12-09 | 北京北信源软件股份有限公司 | Novel instant messaging intelligent robot |
CN106933344A (en) * | 2017-01-18 | 2017-07-07 | 北京光年无限科技有限公司 | Realize the method and device of multi-modal interaction between intelligent robot |
CN106934068A (en) * | 2017-04-10 | 2017-07-07 | 江苏东方金钰智能机器人有限公司 | The method that robot is based on the semantic understanding of environmental context |
-
2017
- 2017-09-15 CN CN201710833842.9A patent/CN109510753A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050049752A1 (en) * | 2003-08-28 | 2005-03-03 | Sony Corporation | Robot apparatus, control method for robot apparatus, and toy for robot apparatus |
CN101282314A (en) * | 2008-05-23 | 2008-10-08 | 中国移动通信集团北京有限公司 | Group communication method, equipment and system for instant communication |
CN103078867A (en) * | 2013-01-15 | 2013-05-01 | 深圳市紫光杰思谷科技有限公司 | Automatic chatting method and chatting system among robots |
CN105141503A (en) * | 2015-08-13 | 2015-12-09 | 北京北信源软件股份有限公司 | Novel instant messaging intelligent robot |
CN106933344A (en) * | 2017-01-18 | 2017-07-07 | 北京光年无限科技有限公司 | Realize the method and device of multi-modal interaction between intelligent robot |
CN106934068A (en) * | 2017-04-10 | 2017-07-07 | 江苏东方金钰智能机器人有限公司 | The method that robot is based on the semantic understanding of environmental context |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110647636B (en) | Interaction method, interaction device, terminal equipment and storage medium | |
Jin et al. | A good prompt is worth millions of parameters: Low-resource prompt-based learning for vision-language models | |
CN107870994A (en) | Man-machine interaction method and system for intelligent robot | |
CN107291867B (en) | Dialog processing method, device and equipment based on artificial intelligence and computer readable storage medium | |
CN107766506A (en) | A kind of more wheel dialog model construction methods based on stratification notice mechanism | |
CN110188182A (en) | Model training method, dialogue generation method, device, equipment and medium | |
CN108920622A (en) | A kind of training method of intention assessment, training device and identification device | |
CN107728780A (en) | A kind of man-machine interaction method and device based on virtual robot | |
CN110489755A (en) | Document creation method and device | |
CN106650756A (en) | Image text description method based on knowledge transfer multi-modal recurrent neural network | |
CN107679225B (en) | Reply generation method based on keywords | |
CN110209789A (en) | A kind of multi-modal dialog system and method for user's attention guidance | |
CN106294854A (en) | A kind of man-machine interaction method for intelligent robot and device | |
CN106355446B (en) | A kind of advertisement recommender system of network and mobile phone games | |
CN110234018A (en) | Multimedia content description generation method, training method, device, equipment and medium | |
CN110459211A (en) | Interactive method, client, electronic equipment and storage medium | |
CN112131368A (en) | Dialog generation method and device, electronic equipment and storage medium | |
CN109977238A (en) | Generate the system for drawing this, method and apparatus | |
CN113761220A (en) | Information acquisition method, device, equipment and storage medium | |
CN106557165A (en) | The action simulation exchange method of smart machine and device and smart machine | |
Wu et al. | Hierarchical memory decoder for visual narrating | |
CN108810132A (en) | Cartoon display method, device, terminal, server and storage medium | |
CN117271745A (en) | Information processing method and device, computing equipment and storage medium | |
CN110491372A (en) | A kind of feedback information generating method, device, storage medium and smart machine | |
CN110020007A (en) | Robot interactive control method and device, storage medium, server |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190322 |