CN112836098B - Multi-role-based conversation assistance method and device - Google Patents
Multi-role-based conversation assistance method and device Download PDFInfo
- Publication number
- CN112836098B CN112836098B CN202110133476.2A CN202110133476A CN112836098B CN 112836098 B CN112836098 B CN 112836098B CN 202110133476 A CN202110133476 A CN 202110133476A CN 112836098 B CN112836098 B CN 112836098B
- Authority
- CN
- China
- Prior art keywords
- user
- assistant
- role
- dialogue
- information
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/903—Querying
- G06F16/9032—Query formulation
- G06F16/90332—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/903—Querying
- G06F16/9035—Filtering based on additional data, e.g. user or group profiles
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Machine Translation (AREA)
- User Interface Of Digital Computer (AREA)
Abstract
The application discloses a multi-role-based conversation assistance method and device, wherein the method comprises the following steps: in a normal interactive working mode, based on real-time dialogue voices of the user and the assistant roles, dialogue content, user emotion information and environment information are obtained; judging whether a conversation is in a dead state or a preset emergency environmental event currently occurs according to the conversation content, the user emotion information and the environmental information; and if so, selecting a switched target assistant role according to the dialogue content, and switching the assistant role of the user into the target assistant role. By adopting the method and the device, the intelligence of role switching can be improved.
Description
Technical Field
The present invention relates to computer application technology, and in particular, to a multi-role-based dialog assistance method and apparatus.
Background
With the development and popularization of intelligent devices, people have more and more demands for intelligent voice interaction, and intelligent voice interaction systems gradually become research hotspots.
A solution is currently proposed for training and configuring a team of intelligent assistants and switching roles of intelligent assistants. In this scenario, a team of intelligent assistants is provided with multiple helper roles, each with different features (e.g., different functions, basic language models, training levels, avatars, personalities, etc.). The characteristics of each assistant may be configured and trained by a trainer, end user, or assistant server. The assistant may perform different functions in the form of dialogues with the end user through the user interaction interface. Different intelligent assistants may be adapted for different dialog contexts. Meanwhile, different assistants can interact with each other to jointly complete tasks related to users, and the interactions are displayed on a user interaction interface.
The inventors found in the course of implementing the present invention that: the scheme has the problem that the intelligent switching mechanism of the assistant role is poor, and the specific analysis is as follows:
the assistant role switching mechanism adopted in the scheme is as follows: when the assistant role interacts with the user, whether the assistant role needs to be switched is judged according to the keywords input by the current user and the change of the objective background. In practical applications, when a user and a real person talk, the user needs to switch the roles of the assistants due to the influence of emotion and talking atmosphere. For example, when a conversation is trapped in a dead office, the user may not want to continue talking again, but may wish to swap the helper role. At this time, the above-mentioned switching mechanism does not distinguish the emotion and talking state of the user in real time, so that the assistant role is not automatically switched for the user, so that a better talking experience is obtained.
Disclosure of Invention
In view of the above, a main object of the present invention is to provide a multi-role-based dialog assistance method and apparatus, which can improve the intelligence of role switching.
In order to achieve the above purpose, the technical scheme provided by the invention is as follows:
a multi-role based dialog assistance method, comprising:
in a normal interactive working mode, based on real-time dialogue voices of the user and the assistant roles, dialogue content, user emotion information and environment information are obtained;
judging whether a conversation is in a dead state or a preset emergency environmental event currently occurs according to the conversation content, the user emotion information and the environmental information; and if so, selecting a switched target assistant role according to the dialogue content, and switching the assistant role of the user into the target assistant role.
In one embodiment, the determining whether the session is currently in the session tie state includes:
judging whether the dialogue content exists or not: the number of times of the same keywords in the dialogue information of the user reaches a preset repetition number threshold, and if so, the current dialogue dead office state is judged;
judging whether the dialogue content exists or not: the dialogue information of the assistant role comprises all preset candidate feedback information of the user consultation problem, and if yes, the current dialogue stiff state is judged;
judging whether the user has negative emotion currently according to the emotion information of the user, and judging that the user is in a conversation dead state currently if the corresponding emotion fluctuation degree reaches a preset emotion threshold value.
In one embodiment, the selecting the target helper role for the handoff includes:
determining the relativity of the background story of the assistant role with the dialogue content or the emergency event based on the archives of the assistant role in the assistant role database; taking the assistant roles of the background stories with the relevance degree larger than a preset relevance threshold as candidate assistant roles;
determining an affinity of the candidate helper role with the user based on social relationship configuration information in the profile of the candidate helper role;
determining a degree of matching of character features of the candidate helper character with the dialog content or the emergency event based on character features in the archive of the candidate helper character;
based on the relevance, the intimacy, the matching degree and corresponding weight parameters, calculating the comprehensive matching score of the candidate assistant roles according to a weight calculation method;
and selecting the candidate assistant role corresponding to the largest comprehensive matching score as the target assistant role.
In one embodiment, the method further comprises:
when the real-time dialogue voice of the user and the assistant role is detected for the first time, judging whether the user is a new user or not by utilizing a voice print identity recognition method, and if so, creating a file for the user.
In one embodiment, the method further comprises:
and after the assistant role is switched, the dialogue content triggering the switching is used as a background story and is stored in a user file of the user, and the background story is stored in a corresponding file of the assistant role with social relation with the user according to a preset event sharing rule.
In one embodiment, the method further comprises:
searching feedback information related to the current real-time dialogue voice of the user in the Internet and background stories in respective files of the user and the current assistant role according to the current dialogue content of the user when judging that the current dialogue is not in a dialogue dead state and the emergency environment event does not occur or when judging that the user is satisfied with the target assistant role after the switching is completed; generating a feedback information candidate set according to the searching result;
for each piece of feedback information in the feedback information candidate set, calculating character feature matching degree and dialogue logic feature matching degree of the feedback information and the current assistant character, and calculating information comprehensive matching degree of the feedback information based on the character feature matching degree and the dialogue logic feature matching degree;
based on dialogue logic characteristics of the current assistant role, carrying out word modification on feedback information corresponding to the maximum value of the information comprehensive matching degree to generate text feedback information;
and processing the text feedback information according to language characteristics in the file of the current assistant role, generating corresponding voice feedback information and playing the voice feedback information to the user.
In one embodiment, the method further comprises:
in a role generation working mode, when non-real-time voice is detected, voice pattern identification is carried out according to the non-real-time voice; if the identification fails, creating a file for the corresponding user or assistant role;
analyzing the external voice to obtain language characteristics of corresponding user or assistant roles and adding the language characteristics into corresponding files;
extracting dialogue content and emotion information from the external voice; analyzing the extracted dialogue content and emotion information to obtain dialogue logic characteristics and character characteristics of corresponding user or assistant roles, and adding the dialogue logic characteristics and character characteristics into corresponding files;
judging whether the emotion fluctuation degree of the corresponding user or assistant role reaches a preset emotion threshold according to the emotion information, if so, adding the corresponding dialogue content as the background story into the archive of the corresponding user or assistant role;
triggering a user to configure social relations for the user or the assistant roles corresponding to the external voice, and adding the social relations into files of the corresponding user or assistant roles.
In one embodiment, the method further comprises:
when an interaction scene mode setting instruction of a user is received, setting a current interaction scene mode as a target interaction scene mode indicated by the instruction according to the interaction scene mode setting instruction; and switching the helper role to a helper role matched with the target interaction scene mode.
In one embodiment, the method further comprises:
based on the environment information, judging whether the interaction scene mode needs to be switched, and if so, switching the current interaction scene mode into the interaction scene mode matched with the environment information.
The embodiment of the application also discloses a dialogue auxiliary device based on multiple roles, which comprises:
the information extraction unit is used for acquiring dialogue content, user emotion information and environment information based on real-time dialogue voices of the user and the assistant roles in a normal interactive working mode;
the role control unit is used for judging whether the current dialogue is in a dialogue dead state or a preset emergency environment event occurs according to the dialogue content, the user emotion information and the environment information; and if so, selecting a switched target assistant role according to the dialogue content, and switching the assistant role of the user into the target assistant role.
Also disclosed are non-transitory computer readable storage media storing instructions that, when executed by a processor, cause the processor to perform the steps of the multi-role based dialog assistance method as described previously.
Embodiments of the present application also disclose an electronic device comprising a non-volatile computer readable storage medium as described above, and the processor having access to the non-volatile computer readable storage medium.
According to the technical scheme, in the multi-role-based conversation assistance method and device provided by the invention, under a normal interactive working mode, conversation content, user emotion information and environment information are obtained from conversation voices of users and assistant roles in real time, whether the current conversation is in a conversation dead office state or a preset emergency environment event occurs is monitored based on the obtained content and information, when the conversation dead office state or the preset emergency environment event occurs is monitored, a switched target assistant role is selected according to the obtained conversation content, and the assistant role of the user is switched to the selected target assistant role. Therefore, when the user interacts with the assistant role, the emotion and the talking state of the user are distinguished in real time, and when the user is in the dead office state or an emergency environment event occurs, the assistant role is automatically triggered to be switched for the user, so that the assistant role matched with the current dialogue scene is switched for the user, the intelligence of role switching can be improved, and the user can obtain better talking experience.
Drawings
FIG. 1 is a schematic flow chart of a method according to an embodiment of the invention;
fig. 2 is a schematic diagram of a device structure according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and the embodiments, in order to make the objects, technical solutions and advantages of the present invention more apparent.
Fig. 1 is a flow chart of a method according to an embodiment of the present invention, as shown in fig. 1, a multi-role-based conversation assistance method implemented in this embodiment mainly includes:
The step is used for acquiring dialogue content, user emotion information and environment information from two dialogues in real time when the user and the assistant roles normally interact, so as to monitor whether the dialogue is in a dialogue impatience state or a preset emergency environment event occurs or not based on the content and the information in step 102, and automatically trigger the assistant roles to switch for the user when the dialogue is in the dialogue impatience state or the preset emergency environment event occurs, thereby improving the intelligence of role switching.
Specifically, in this step, the current emotion information of the user and the surrounding environment information can be obtained by monitoring the language change (including but not limited to, language change, mood change, accent change, and use of specific spoken Buddhist etc.) of the user in the dialogue.
In practical applications, the specific method for obtaining dialogue content, user emotion information and environment information from real-time dialogue speech is known to those skilled in the art, and will not be described herein.
In this step, according to the dialogue content, the user emotion information and the environment information acquired from the real-time dialogue content in step 101, it is determined whether the dialogue is currently in a dialogue impatience state or a preset emergency environment event occurs, so as to trigger the switching of the assistant roles in time when the dialogue impatience or the emergency environment event occurs.
Here, when the assistant role is required to be switched, the target switching assistant role is selected according to the current dialogue content, so that the switched assistant role can be matched with the current dialogue scene, the assistant role which can better provide interactive service for the user can be used as a new assistant role, the interactive experience of the user and the assistant role can be improved, and the intelligence of the assistant role switching is enhanced.
In one embodiment, in order to accurately identify the session bot state, the following method may be specifically adopted to determine whether the session bot state is currently in:
judging whether the dialogue content exists or not: the number of times of the same keywords in the dialogue information of the user reaches a preset repetition number threshold, and if so, the current dialogue dead office state is judged;
judging whether the dialogue content exists or not: the dialogue information of the assistant role comprises all preset candidate feedback information of the user consultation problem, and if yes, the current dialogue stiff state is judged;
judging whether the user has negative emotion currently according to the emotion information of the user, and judging that the user is in a conversation dead state currently if the corresponding emotion fluctuation degree reaches a preset emotion threshold value.
In the above method, when the number of times of the same keyword appearing in the dialogue information of the user reaches the preset repetition number threshold, it means that the user may ask questions for the same question multiple times in the dialogue, and at this time, no effective information is provided with a high probability through multiple interactive assistant roles, so as to further explain that the dialogue enters into the dead office.
When the dialogue information of the assistant role contains all preset candidate feedback information of the user consultation problem, the assistant role is used for providing all preset candidate feedback information of the user consultation problem to the user, so that effective feedback information can not be provided for the user for the current dialogue any more, and the situation that the dialogue enters the dead office is explained.
When negative emotions (such as anger, fear, sadness and the like) of the user are judged to occur currently, and the corresponding emotion fluctuation degree reaches a preset emotion threshold value, the user is not satisfied with the current interaction with the assistant role, the emotion fluctuation is severe, the conversation enters a dead office at the moment, and the assistant role switching needs to be triggered to switch the assistant role capable of better providing services for the user.
In the above method, the keywords may be preset by those skilled in the art according to actual service requirements. The repetition number threshold and the emotion threshold can be set by a person skilled in the art according to actual application scene requirements.
In practical application, the emergency environmental event may be preset according to a practical application scenario, for example, but not limited to, an event such as injury of a user or occurrence of a huge change in surrounding environment.
In one embodiment, to select a helper role that more closely matches the current dialog scenario, the following method may be used to select the target helper role for the handoff:
step x1, determining the relativity of the background story of the assistant role and the dialogue content or the emergency event based on the archives of the assistant role in the assistant role database; and taking the assistant roles of the background story with the relevance degree larger than a preset relevance threshold as candidate assistant roles.
And step x2, determining the affinity between the candidate assistant role and the user based on social relation configuration information in the archive of the candidate assistant role.
And step x3, determining the matching degree of the character features of the candidate assistant roles and the dialogue content or the emergency event based on the character features in the archives of the candidate assistant roles.
The character features may specifically include: the characteristics of impatience, inward direction, gentle and calm, etc., but are not limited thereto.
And step x4, calculating the comprehensive matching score of the candidate assistant role according to a weight calculation method based on the relevance, the intimacy, the matching degree and corresponding weight parameters.
And step x5, selecting the candidate assistant role corresponding to the largest comprehensive matching score as the target assistant role.
By adopting the method, the background stories closely related to the current dialogue content or the emergency event are selected from the assistant role database, namely, the background stories with the relativity larger than the preset threshold value, the assistant roles of the background stories are taken as candidate assistant roles, the affinity of the assistant roles and the user and the matching degree of the assistant roles and the character characteristics of the user are combined, the comprehensive matching score of each candidate assistant role is calculated, the candidate assistant role with the largest comprehensive matching score is selected as the target assistant role for switching, and therefore, the assistant role which is more matched with the current dialogue content or the emergency event can be selected, and the dialogue experience of the user after the roles are switched can be improved.
In one embodiment, when the real-time conversational speech of the user and the helper character is detected for the first time, it may be determined whether the user is a new user based on a voiceprint recognition method, and if so, a profile is created for the new user for user management.
In one embodiment, after the helper role switch is completed, the user profile may be further perfected to improve the accuracy of the subsequent helper role switch, and specifically, the method may further include the following steps:
and after the assistant role is switched, the dialogue content triggering the switching is used as a background story and is stored in a user file of the user, and the background story is stored in a corresponding file of the assistant role with social relation with the user according to a preset event sharing rule.
Specifically, the event sharing rules may be set according to the actual application scenario requirements, so as to define which helper roles of the user may (or may not) be shared with the background story, for example, events generated between the user and certain specific roles (such as psychological consultants, lawyers, etc.) will not be shared as appropriate; for experiences that the user explicitly indicates is not willing to share, it will not share.
In one embodiment, to improve the sense of realism of the interaction between the user and the assistant character, feedback information to the user in the interaction process may also be generated in real time according to the dialogue logic feature and the character feature of the current assistant character, and specifically, the method further includes:
step y1, searching feedback information related to the current real-time dialogue voice of the user in the internet and in the background stories in the files of the user and the current assistant role respectively according to the current dialogue content of the user when judging that the current dialogue is not in a dialogue dead state and the emergency environment event does not occur or when judging that the user is satisfied with the target assistant role after the switching is completed; and generating a feedback information candidate set according to the searching result.
This step is used to search for feedback information related to the user's current real-time conversational speech to construct a candidate set of feedback information.
And step y2, calculating character feature matching degree and dialogue logic feature matching degree of the feedback information and the current assistant character for each piece of feedback information in the feedback information candidate set, and calculating information comprehensive matching degree of the feedback information based on the character feature matching degree and the dialogue logic feature matching degree.
The step is used for calculating the information comprehensive matching degree of each piece of feedback information in the feedback information candidate set so as to select the information most suitable for the role feedback of the assistant in the subsequent step.
The dialogue logic feature may specifically include: whether or not certain features such as whether or not to identify or even more cater to the user, whether or not to lie well in a particular situation, whether or not the direction of the conversation is changed due to the emotion of the user, etc., are not limited thereto.
In this step, specifically, a weight calculation manner may be adopted to obtain an information comprehensive matching degree of each piece of feedback information.
And step y3, carrying out wording modification on the feedback information corresponding to the maximum value of the information comprehensive matching degree based on the dialogue logic characteristics of the current assistant role, and generating text feedback information.
In the step, the feedback information with the highest comprehensive matching degree of the information is selected, and the feedback information is subjected to wording modification by combining with the dialogue logic characteristics of the assistant role, so that the corresponding text feedback information is obtained.
The specific method for generating text feedback information according to the dialog logic features is known to those skilled in the art and will not be described in detail herein.
And y4, processing the text feedback information according to language characteristics in the file of the current assistant role, generating corresponding voice feedback information and playing the voice feedback information to the user.
The language features may specifically include: sentence making style, oral Buddhist, speed and accent.
In this step, the specific method for processing the text feedback information according to the language features to obtain the voice feedback information is known to those skilled in the art, and will not be described herein.
In one embodiment, the following method may be further employed to create a profile for a user or helper character based on non-real-time speech in a character generation mode of operation:
step z1, in a role generation working mode, when non-real-time voice is detected, voice pattern identification is carried out according to the non-real-time voice; if the identification fails, a profile is created for the corresponding user or helper role.
Here, the non-real-time voice may be a voice call provided for a user, a recorded broadcast, or the like.
In the step, when the voiceprint identity is identified, voiceprint features are acquired from the played voice, if no user matched with the voiceprint features exists at present, the identification fails, and the fact that files are not established for the corresponding voice belongs to is indicated, so that the files are required to be triggered to be established, language features, dialogue logic features, character features and social relations are acquired from the voice by utilizing the follow-up steps, and the language features, dialogue logic features, character features and social relations are added into the corresponding files.
And step z2, analyzing the external voice to obtain language features of corresponding user or assistant roles and adding the language features into corresponding files.
Step z3, extracting dialogue content and emotion information from the external voice; and analyzing the extracted dialogue content and emotion information to obtain dialogue logic characteristics and character characteristics of the corresponding user or assistant roles, and adding the dialogue logic characteristics and character characteristics into the corresponding archive.
And step z4, judging whether the emotion fluctuation degree of the corresponding user or assistant role reaches a preset emotion threshold value according to the emotion information, and if so, adding the corresponding dialogue content as the background story into the archive of the corresponding user or assistant role.
In this step, dialogue content when emotion fluctuation exists in non-real-time voice is added into the archive as a background story, so that the archive can contain the background story reflecting character characteristics and emotion information of the user or the character, thereby facilitating screening of assistant characters.
And step z5, triggering a user to configure social relations for the user or the assistant roles corresponding to the external voice, and adding the social relations to files of the corresponding user or assistant roles.
Here, the social relationship may specifically include: relationships such as relatedness, friends, teachers, students, colleagues, and the like, degrees of relatedness and sparseness of the relationships, whether experience stories can be shared, and the like, but are not limited thereto.
In one embodiment, the interaction scene mode may be further set according to a user instruction, and the switching of the roles of the assistant may be triggered according to the current interaction scene mode, which may be implemented by the following method:
when an interaction scene mode setting instruction of a user is received, setting a current interaction scene mode as a target interaction scene mode indicated by the instruction according to the interaction scene mode setting instruction; and switching the helper role to a helper role matched with the target interaction scene mode.
In one embodiment, the switching of the interactive scene mode may be automatically triggered when the current dialogue background changes (e.g. from an entertainment scene to a meeting or a working scene) further according to the environment information extracted from the current real-time dialogue content, and the following method may be specifically adopted:
based on the environment information, judging whether the interaction scene mode needs to be switched, and if so, switching the current interaction scene mode into the interaction scene mode matched with the environment information.
Specifically, if the current dialogue background is known to change based on the environment information, the current need of switching the interaction scene mode is judged, otherwise, the interaction scene mode does not need to be switched.
Based on the above method embodiments, it can be seen that the intelligence of role switching can be effectively improved by using the above method embodiments, and the specific implementation of the above embodiments is further described below with reference to several specific application scenarios.
Scene one:
when a user is in interaction with a voice assistant, different voice assistant roles will have different voice feedback in different interaction modes for the user's input. If the user inquires about what the user intends to do on the weekend in the entertainment mode, an assistant character with an outward character feature and hobbies to play outdoor sports may reply to "go to kick bar"; an assistant character with an inward character feature that likes silence may reply to "go to library bar"; family member roles such as mother role may reply to "go to park bar"; and a female friend character may reply "to a female user to shop for a shopping bar". In another example, a user speaking "play music" in a learning mode, an assistant character with an outward character may play relatively happy music, an assistant character with an inward character may play relatively soft pure music, a character assistant with a more severe character may reply to "learn attentively, not to listen to music", a friend character with a common music taste may play music consistent with the user taste, if a currently served assistant character has a background story about the user's learning, the assistant character may reply "according to a conversation context, etc. to listen to music again when a rest is needed, otherwise just listen to music like xx, and cannot complete a lesson.
When the user is in entertainment mode and is communicating with the voice assistant, if the current user suddenly enters a telephone conference, the system is switched to a conference mode through background analysis and is switched to the voice assistant corresponding to the conference mode.
And in the second scene, when the user interacts with the voice system in the entertainment mode and intends to hold a birthday party for a friend, the user can switch the current role of the voice assistant to the role of the friend. The party-related preparation (e.g., location, placement, food, background music, game piece, etc.) communicates with the voice assistant, who will reply according to the character's personality profile settings in the system. Such as when the user asks "what background music should be selected", the voice assistant replies to the type of music that the buddy may prefer (music to roll) based on personality settings (outward, impatience). As another example, when the user asks about a party game link, the voice assistant may reply to the type of game that the friend may enjoy (e.g., a comparison of a happy game that was experienced together) based on the previous background story settings.
And in a third scenario, when the user interacts with the voice system in the learning mode to prepare for a paper, the user can communicate with a voice assistant (learning assistant) with default learning mode preset in the system. When the user is annoyed by the thinking of the current learning-class voice assistant, repeatedly inquires about the same questions and is not satisfied with the answer representation, the user can automatically switch to a friend role. The friend character may play a background story (e.g., scenic spots once taken together, shows once viewed together, etc.) related to the paper topic based on the background story settings.
Scene four, when the child user interacts with the voice system, if accidental injury occurs, the user can be switched into a role assistant (such as a family member mother) with pacifying property by strategy control, guiding voice is provided according to the prior setting (such as providing the position of a first-aid kit in home according to the background story of the mother role), and the background story of the role assistant is triggered to play (such as the stories about courage and pacifying in the mother role) when the child user is in a more vigorous need of pacifying. The user injury experience will supplement the personality and language characteristics of the child user (response to accidental injury, performance of speech characteristics upon agitation, etc.) as well as the child user's background story (time and place of occurrence of the experience and subsequent processing). While the mom character will update the background story (unexpected experience and subsequent processing of the child user in time and place, reaction to a pacifying story, etc.), and at the same time decide whether to share the experience between characters (whether to share the story experience in family members, whether to inform child user's friend characters, etc.) based on the relationship between the service characters corresponding to the child user. In addition, it may be decided whether to trigger a handoff helper role based on the child user's reaction to the mom role (whether the child user is satisfied with the voice interaction of the mom role, should be changed to dad or other buddy roles, etc.).
And when the user is opposite to the emotion and reaches a certain threshold value, and the user resists the willingness to cause communication to be unsmooth or stagnate, the assistant role can be switched to other roles with more convincing (a doctor role) or soothing (a family member) so as to achieve the aim of continuing communication if the user is annoyed to the voice assistant feedback of the current nursing role (such as unwilling to move on time or take medicine according to the advice of the nursing role) when the old user interacts with the voice system in the monitoring mode. Meanwhile, the context causing the switching is used as a background story and is supplemented into files of the user and the assistant roles which can be shared.
Corresponding to the above method embodiment, the embodiment of the present application also discloses a multi-role-based dialog assistance device, as shown in fig. 2, including:
an information extraction unit 201 for acquiring dialogue content, user emotion information, and environmental information based on real-time dialogue voices of the user and assistant roles in a normal interworking mode;
a role control unit 202, configured to determine whether a session is currently in a dead-office state or a preset emergency environment event occurs according to the session content, the user emotion information and the environment information; and if so, selecting a switched target assistant role according to the dialogue content, and switching the assistant role of the user into the target assistant role.
Embodiments also provide a non-transitory computer readable storage medium storing instructions that, when executed by a processor, cause the processor to perform the steps of a multi-role based dialog assistance method as described above.
Embodiments of the present application also provide an electronic device comprising a non-volatile computer readable storage medium as described above, and the processor having access to the non-volatile computer readable storage medium.
In summary, the above embodiments are only preferred embodiments of the present invention, and are not intended to limit the scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (11)
1. A multi-role based dialog assistance method, comprising:
in a normal interactive working mode, based on real-time dialogue voices of the user and the assistant roles, dialogue content, user emotion information and environment information are obtained;
judging whether a conversation is in a dead state or a preset emergency environmental event currently occurs according to the conversation content, the user emotion information and the environmental information; if so, selecting an assistant role matched with the current dialogue content or the emergency event from an assistant role database based on the affinity of the assistant role and the user and the matching degree of the assistant role and the character characteristics of the user as a target assistant role for switching, and switching the assistant role of the user into the target assistant role;
searching feedback information related to the current real-time dialogue voice of the user in the Internet and background stories in respective files of the user and the current assistant role according to the current dialogue content of the user when judging that the current dialogue is not in a dialogue dead state and the emergency environment event does not occur or when judging that the user is satisfied with the target assistant role after the switching is completed; generating a feedback information candidate set according to the searching result;
for each piece of feedback information in the feedback information candidate set, calculating character feature matching degree and dialogue logic feature matching degree of the feedback information and the current assistant character, and calculating information comprehensive matching degree of the feedback information based on the character feature matching degree and the dialogue logic feature matching degree;
based on dialogue logic characteristics of the current assistant role, carrying out word modification on feedback information corresponding to the maximum value of the information comprehensive matching degree to generate text feedback information;
and processing the text feedback information according to language characteristics in the file of the current assistant role, generating corresponding voice feedback information and playing the voice feedback information to the user.
2. The method of claim 1, wherein said determining whether the current session is in a session tie state comprises:
judging whether the dialogue content exists or not: the number of times of the same keywords in the dialogue information of the user reaches a preset repetition number threshold, and if so, the current dialogue dead office state is judged;
judging whether the dialogue content exists or not: the dialogue information of the assistant role comprises all preset candidate feedback information of the user consultation problem, and if yes, the current dialogue stiff state is judged;
judging whether the user has negative emotion currently according to the emotion information of the user, and judging that the user is in a conversation dead state currently if the corresponding emotion fluctuation degree reaches a preset emotion threshold value.
3. The method of claim 1, wherein the selecting of the target helper role for the handoff comprises:
determining the relativity of the background story of the assistant role with the dialogue content or the emergency event based on the archives of the assistant role in the assistant role database; taking the assistant roles of the background stories with the relevance degree larger than a preset relevance threshold as candidate assistant roles;
determining an affinity of the candidate helper role with the user based on social relationship configuration information in the profile of the candidate helper role;
determining a degree of matching of character features of the candidate helper character with the dialog content or the emergency event based on character features in the archive of the candidate helper character;
based on the relevance, the intimacy, the matching degree and corresponding weight parameters, calculating the comprehensive matching score of the candidate assistant roles according to a weight calculation method;
and selecting the candidate assistant role corresponding to the largest comprehensive matching score as the target assistant role.
4. The method according to claim 1, wherein the method further comprises:
when the real-time dialogue voice of the user and the assistant role is detected for the first time, judging whether the user is a new user or not by utilizing a voice print identity recognition method, and if so, creating a file for the user.
5. A method according to claim 3, wherein the method further comprises:
and after the assistant role is switched, the dialogue content triggering the switching is used as a background story and is stored in a user file of the user, and the background story is stored in a corresponding file of the assistant role with social relation with the user according to a preset event sharing rule.
6. The method according to claim 1, wherein the method further comprises:
in a role generation working mode, when non-real-time voice is detected, voice pattern identification is carried out according to the non-real-time voice; if the identification fails, creating a file for the corresponding user or assistant role;
analyzing the external voice to obtain language characteristics of corresponding user or assistant roles and adding the language characteristics into corresponding files;
extracting dialogue content and emotion information from the external voice; analyzing the extracted dialogue content and emotion information to obtain dialogue logic characteristics and character characteristics of corresponding user or assistant roles, and adding the dialogue logic characteristics and character characteristics into corresponding files;
judging whether the emotion fluctuation degree of the corresponding user or assistant role reaches a preset emotion threshold value according to the emotion information, if so, adding the corresponding dialogue content as a background story into the archive of the corresponding user or assistant role;
triggering a user to configure social relations for the user or the assistant roles corresponding to the external voice, and adding the social relations into files of the corresponding user or assistant roles.
7. The method according to claim 1, wherein the method further comprises:
when an interaction scene mode setting instruction of a user is received, setting a current interaction scene mode as a target interaction scene mode indicated by the instruction according to the interaction scene mode setting instruction; and switching the helper role to a helper role matched with the target interaction scene mode.
8. The method according to claim 1, wherein the method further comprises:
based on the environment information, judging whether the interaction scene mode needs to be switched, and if so, switching the current interaction scene mode into the interaction scene mode matched with the environment information.
9. A multi-role based dialog assistance device, comprising:
the information extraction unit is used for acquiring dialogue content, user emotion information and environment information based on real-time dialogue voices of the user and the assistant roles in a normal interactive working mode;
the role control unit is used for judging whether the current dialogue is in a dialogue dead state or a preset emergency environment event occurs according to the dialogue content, the user emotion information and the environment information; if so, selecting an assistant role matched with the current dialogue content or the emergency event from an assistant role database based on the affinity of the assistant role and the user and the matching degree of the assistant role and the character characteristics of the user as a target assistant role for switching, and switching the assistant role of the user into the target assistant role; searching feedback information related to the current real-time dialogue voice of the user in the Internet and background stories in respective files of the user and the current assistant role according to the current dialogue content of the user when judging that the current dialogue is not in a dialogue dead state and the emergency environment event does not occur or when judging that the user is satisfied with the target assistant role after the switching is completed; generating a feedback information candidate set according to the searching result; for each piece of feedback information in the feedback information candidate set, calculating character feature matching degree and dialogue logic feature matching degree of the feedback information and the current assistant character, and calculating information comprehensive matching degree of the feedback information based on the character feature matching degree and the dialogue logic feature matching degree; based on dialogue logic characteristics of the current assistant role, carrying out word modification on feedback information corresponding to the maximum value of the information comprehensive matching degree to generate text feedback information; and processing the text feedback information according to language characteristics in the file of the current assistant role, generating corresponding voice feedback information and playing the voice feedback information to the user.
10. A non-transitory computer-readable storage medium storing instructions that, when executed by a processor, cause the processor to perform the steps of the multi-role based dialog assistance method of any one of claims 1 to 8.
11. An electronic device comprising the non-volatile computer-readable storage medium of claim 10, and the processor having access to the non-volatile computer-readable storage medium.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110133476.2A CN112836098B (en) | 2021-02-01 | 2021-02-01 | Multi-role-based conversation assistance method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110133476.2A CN112836098B (en) | 2021-02-01 | 2021-02-01 | Multi-role-based conversation assistance method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112836098A CN112836098A (en) | 2021-05-25 |
CN112836098B true CN112836098B (en) | 2023-07-07 |
Family
ID=75932715
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110133476.2A Active CN112836098B (en) | 2021-02-01 | 2021-02-01 | Multi-role-based conversation assistance method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112836098B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2023159536A1 (en) * | 2022-02-28 | 2023-08-31 | 华为技术有限公司 | Human-computer interaction method and apparatus, and terminal device |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110265021A (en) * | 2019-07-22 | 2019-09-20 | 深圳前海微众银行股份有限公司 | Personalized speech exchange method, robot terminal, device and readable storage medium storing program for executing |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106776926A (en) * | 2016-12-01 | 2017-05-31 | 竹间智能科技(上海)有限公司 | Improve the method and system of responsibility when robot talks with |
CN107340991B (en) * | 2017-07-18 | 2020-08-25 | 百度在线网络技术(北京)有限公司 | Voice role switching method, device, equipment and storage medium |
CN111026932B (en) * | 2019-12-20 | 2023-05-26 | 北京百度网讯科技有限公司 | Man-machine dialogue interaction method and device, electronic equipment and storage medium |
-
2021
- 2021-02-01 CN CN202110133476.2A patent/CN112836098B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110265021A (en) * | 2019-07-22 | 2019-09-20 | 深圳前海微众银行股份有限公司 | Personalized speech exchange method, robot terminal, device and readable storage medium storing program for executing |
Also Published As
Publication number | Publication date |
---|---|
CN112836098A (en) | 2021-05-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10388281B2 (en) | Dialogue control apparatus, dialogue control method, and non-transitory recording medium | |
US9953078B2 (en) | Dialogue control device, dialogue control method and non-transitory computer-readable information recording medium | |
JP6719747B2 (en) | Interactive method, interactive system, interactive device, and program | |
CN105991847B (en) | Call method and electronic equipment | |
JP6699010B2 (en) | Dialogue method, dialogue system, dialogue device, and program | |
Traum et al. | Incremental dialogue understanding and feedback for multiparty, multimodal conversation | |
US11232789B2 (en) | Dialogue establishing utterances without content words | |
Fang et al. | Sounding board–university of washington’s alexa prize submission | |
WO2017200080A1 (en) | Intercommunication method, intercommunication device, and program | |
JP6970413B2 (en) | Dialogue methods, dialogue systems, dialogue devices, and programs | |
JP6551793B2 (en) | Dialogue method, dialogue system, dialogue apparatus, and program | |
CN112632242A (en) | Intelligent conversation method and device and electronic equipment | |
CN111557001A (en) | Method, computer device and computer readable storage medium for providing natural language dialog by providing instant responsive language response | |
CN112836098B (en) | Multi-role-based conversation assistance method and device | |
CN111063346A (en) | Cross-media star emotion accompany interaction system based on machine learning | |
Yang et al. | User behavior fusion in dialog management with multi-modal history cues | |
Hasani et al. | Immersive experience with non-player characters dynamic dialogue | |
JP7160076B2 (en) | Dialogue control device, dialogue control method and program | |
US12002470B1 (en) | Multi-source based knowledge data for artificial intelligence characters | |
Kurniasari | A pragmatic analysis of conversational implicature of humor practices in How to Train Your Dragon | |
US11960983B1 (en) | Pre-fetching results from large language models | |
US20230351142A1 (en) | Relationship graphs for artificial intelligence character models | |
JP2022133188A (en) | Interaction program, device and method which forget noun spoken by user | |
Prendinger et al. | Agents that talk back (sometimes): Filter programs for affective communication | |
WO2024145675A1 (en) | Language model orchestration |
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 |