CN113378583A - Dialogue reply method and device, dialogue model training method and device, and storage medium - Google Patents
Dialogue reply method and device, dialogue model training method and device, and storage medium Download PDFInfo
- Publication number
- CN113378583A CN113378583A CN202110802896.5A CN202110802896A CN113378583A CN 113378583 A CN113378583 A CN 113378583A CN 202110802896 A CN202110802896 A CN 202110802896A CN 113378583 A CN113378583 A CN 113378583A
- Authority
- CN
- China
- Prior art keywords
- information
- dialogue
- virtual character
- model
- conversation
- 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
- 238000012549 training Methods 0.000 title claims abstract description 89
- 238000000034 method Methods 0.000 title claims abstract description 57
- 230000004044 response Effects 0.000 claims abstract description 97
- 238000003780 insertion Methods 0.000 claims description 25
- 230000037431 insertion Effects 0.000 claims description 25
- 238000012545 processing Methods 0.000 claims description 16
- 238000005065 mining Methods 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 6
- 238000001914 filtration Methods 0.000 claims description 6
- 238000002372 labelling Methods 0.000 claims description 5
- 230000003993 interaction Effects 0.000 description 25
- 230000000694 effects Effects 0.000 description 21
- 238000004891 communication Methods 0.000 description 11
- 238000005516 engineering process Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 7
- 241000282414 Homo sapiens Species 0.000 description 6
- 101100264174 Mus musculus Xiap gene Proteins 0.000 description 5
- 235000010627 Phaseolus vulgaris Nutrition 0.000 description 4
- 244000046052 Phaseolus vulgaris Species 0.000 description 4
- 230000003287 optical effect Effects 0.000 description 4
- 230000005236 sound signal Effects 0.000 description 4
- 208000024891 symptom Diseases 0.000 description 4
- 238000007726 management method Methods 0.000 description 3
- 206010044565 Tremor Diseases 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000007654 immersion Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 239000003550 marker Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010422 painting Methods 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 239000000047 product Substances 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
- G06F40/35—Discourse or dialogue representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Machine Translation (AREA)
Abstract
The disclosure relates to a dialogue reply method and device, a dialogue model training method and device and a storage medium. The method comprises the following steps: acquiring first dialogue information input by a user; inputting the first dialogue information and the target role information into a pre-trained dialogue model to obtain target response information which is output by the dialogue model and used for replying the first dialogue information; the target response information is information representing the conversation style of the target role. By adopting the method disclosed by the invention, more natural and accurate target response information can be replied in the conversation style of the target role aiming at any first conversation information input by the user.
Description
Technical Field
The present disclosure relates to the field of human-computer conversation technologies, and in particular, to a conversation reply method and apparatus, a training method and apparatus for a conversation model, and a storage medium.
Background
With the development of internet technology, information communication technology and artificial intelligence technology, the natural convenience of the human-computer conversation Systems (conversation Systems) makes the human-computer conversation Systems become a new generation of interactive paradigm. Human-machine conversation techniques have been applied by the industry to various types of product services. In the application of the man-machine conversation system, as the AI intelligence is gradually improved, people are also gradually raising higher-level requirements on the man-machine conversation system, for example, the conversation robot is required to be capable of naturally interacting with the user more like a real person, or the conversation robot is required to have characters/personalities like a real person.
Disclosure of Invention
In order to overcome the problems in the related art, the present disclosure provides a dialog reply method and apparatus, a training method and apparatus of a dialog model, and a storage medium, so as to solve the problems in the related art.
According to a first aspect of the embodiments of the present disclosure, there is provided a dialog reply method, the method including:
acquiring first dialogue information input by a user;
inputting the first dialogue information and the target role information into a pre-trained dialogue model to obtain target response information which is output by the dialogue model and used for replying the first dialogue information;
the target response information is information representing the conversation style of the target role.
In some embodiments, the conversation model comprises a conversation markup network;
the inputting the first dialogue information and the target role information into a pre-trained dialogue model to obtain target response information which is output by the dialogue model and used for replying the first dialogue information comprises:
inputting the first dialogue information and the target role information into a pre-trained dialogue marking network to obtain first response information which is output by the dialogue marking network and contains label information;
the label information is used for representing labeling information of each unit character in the first response information, and the label information comprises at least one of a deleting label, a reserving label and a text inserting position label.
In some embodiments, the conversation model includes a conversation insertion network;
the inputting the first dialogue information and the target role information into a pre-trained dialogue model to obtain target response information which is output by the dialogue model and used for replying the first dialogue information, further comprising:
and inputting the first response information into the dialog insertion network to obtain the target response information which is output by the dialog insertion network and is adjusted according to the label information.
In some embodiments, the dialogue model is trained by:
constructing a sample training set according to a virtual character sample and a dialogue corpus pair corresponding to the virtual character sample, wherein the dialogue corpus pair comprises a first corpus pair which is screened from a target platform and is associated with a virtual character corresponding to the virtual character sample and a second corpus pair of the virtual character in a source work;
and training the conversation model according to the training sample set until the conversation model has the capacity of generating information representing the conversation style of the virtual character sample.
In some embodiments, the filtering, from the target platform, the first corpus pair associated with the virtual role comprises:
mining a candidate corpus pair associated with the virtual role from the target platform, wherein the candidate corpus pair comprises candidate dialogue information and candidate reply information;
and determining the candidate corpus pair in which the candidate reply information is located as the first corpus pair under the condition that the similarity between the candidate reply information and second reply information included in the second corpus pair is greater than a first preset threshold.
In some embodiments, the filtering, from the target platform, the first corpus pair associated with the virtual role comprises:
mining a candidate corpus pair associated with the virtual role from the target platform, wherein the candidate corpus pair comprises candidate dialogue information and candidate reply information;
inputting the candidate reply information into a conversation identification model to obtain an identification value representing whether the candidate reply information is the source reply information included in the source work;
and determining the candidate corpus pair in which the candidate reply message with the discrimination value larger than a second preset threshold is positioned as the first corpus pair.
According to a second aspect of the embodiments of the present disclosure, there is provided a training method of a dialogue model, the method including:
constructing a sample training set according to a virtual character sample and a dialogue corpus pair corresponding to the virtual character sample, wherein the dialogue corpus pair comprises a first corpus pair which is screened from a target platform and is associated with a virtual character corresponding to the virtual character sample and a second corpus pair of the virtual character in a source work;
and training the conversation model according to the training sample set until the conversation model has the capacity of generating information representing the conversation style of the virtual character sample.
According to a third aspect of the embodiments of the present disclosure, there is provided a dialog replying device, including:
the system comprises an acquisition module, a processing module and a processing module, wherein the acquisition module is configured to acquire first dialogue information input by a user;
the input module is configured to input the first dialogue information and the target role information into a pre-trained dialogue model to obtain target response information which is output by the dialogue model and used for replying the first dialogue information;
the target response information is information representing the conversation style of the target role.
In some embodiments, the conversation model comprises a conversation markup network; the input module includes:
the first input submodule is configured to input the first dialogue information and the target role information into a pre-trained dialogue marking network to obtain first response information which is output by the dialogue marking network and contains label information;
the label information is used for representing labeling information of each unit character in the first response information, and the label information comprises at least one of a deleting label, a reserving label and a text inserting position label.
In some embodiments, the conversation model includes a conversation insertion network;
the input module further comprises:
and the second input submodule is configured to input the first response information into the dialog insertion network, so as to obtain the target response information which is output by the dialog insertion network and is obtained after the first response information is adjusted according to the tag information.
In some embodiments, the apparatus further comprises a training module configured to train the dialogue model by:
constructing a sample training set according to a virtual character sample and a dialogue corpus pair corresponding to the virtual character sample, wherein the dialogue corpus pair comprises a first corpus pair which is screened from a target platform and is associated with a virtual character corresponding to the virtual character sample and a second corpus pair of the virtual character in a source work; and training the conversation model according to the training sample set until the conversation model has the capacity of generating information representing the conversation style of the virtual character sample. .
In some embodiments, the training module comprises:
a first mining submodule configured to mine a candidate corpus pair associated with the virtual character from the target platform, the candidate corpus pair including candidate dialogue information and candidate reply information;
the first determining submodule is configured to determine the corpus candidate pair in which the reply candidate information is located as the first corpus pair when the similarity between the reply candidate information and second reply information included in the second corpus pair is greater than a first preset threshold.
In some embodiments, the training module comprises:
a second mining submodule configured to mine a candidate corpus pair associated with the virtual character from the target platform, the candidate corpus pair including candidate dialogue information and candidate reply information;
a third input submodule configured to input the candidate reply message into a dialogue identification model, to obtain an identification value representing whether the candidate reply message is a source reply message included in the source work;
a second determining submodule configured to determine the corpus candidate pair in which the candidate reply message with the discrimination value larger than a second preset threshold is located as the first corpus pair.
According to a fourth aspect of the embodiments of the present disclosure, there is provided an apparatus for training a dialogue model, the apparatus including:
the system comprises a first execution module, a second execution module and a third execution module, wherein the first execution module is configured to construct a sample training set according to a virtual character sample and a dialogue corpus pair corresponding to the virtual character sample, and the dialogue corpus pair comprises a first corpus pair which is screened from a target platform and is associated with a virtual character corresponding to the virtual character sample and a second corpus pair of the virtual character in a source work;
a second execution module configured to train the conversation model according to the training sample set until the conversation model has the capability of generating information characterizing a conversation style of the virtual character sample.
According to a fifth aspect of the embodiments of the present disclosure, there is provided a dialog replying device, including:
a processor;
a memory for storing processor-executable instructions;
wherein the processor is configured to:
acquiring first dialogue information input by a user;
inputting the first dialogue information and the target role information into a pre-trained dialogue model to obtain target response information which is output by the dialogue model and used for replying the first dialogue information;
the target response information is information representing the conversation style of the target role.
According to a sixth aspect of the embodiments of the present disclosure, there is provided a training apparatus for a dialogue model, including:
a processor;
a memory for storing processor-executable instructions;
wherein the processor is configured to:
constructing a sample training set according to a virtual character sample and a dialogue corpus pair corresponding to the virtual character sample, wherein the dialogue corpus pair comprises a first corpus pair which is screened from a target platform and is associated with a virtual character corresponding to the virtual character sample and a second corpus pair of the virtual character in a source work;
and training the conversation model according to the training sample set until the conversation model has the capacity of generating information representing the conversation style of the virtual character sample.
According to a seventh aspect of embodiments of the present disclosure, there is provided a computer readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement the steps of the dialogue reply method provided by the first aspect of the present disclosure or implement the steps of the training method of the dialogue model provided by the second aspect of the present disclosure.
The technical scheme provided by the embodiment of the disclosure can have the following beneficial effects:
and obtaining target response information which is output by the dialogue model and used for replying the first dialogue information by acquiring the first dialogue information input by the user and inputting the first dialogue information and the target role information into a pre-trained dialogue model. The target response information is information representing a dialog style of the target character. By adopting the method, more natural and accurate target response information can be replied by the conversation style (conversation characteristics) of the target role aiming at any first conversation information input by the user.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and together with the description, serve to explain the principles of the disclosure.
Fig. 1 is a flow chart illustrating a dialog reply method according to an exemplary embodiment of the present disclosure.
FIG. 2 is a flow diagram illustrating a dialogue model training process according to an exemplary embodiment of the present disclosure.
FIG. 3 is a framework of a dialogue model shown in accordance with an exemplary embodiment of the present disclosure.
FIG. 4 is a flowchart illustrating a method for training a dialogue model according to an exemplary embodiment of the present disclosure.
Fig. 5 is a block diagram illustrating a dialog reply device according to an exemplary embodiment of the present disclosure.
FIG. 6 is a block diagram illustrating a training apparatus for a dialogue model according to an exemplary embodiment of the present disclosure.
Fig. 7 is a block diagram illustrating an apparatus (a general structure of a mobile terminal) according to an exemplary embodiment of the present disclosure.
Fig. 8 is a block diagram illustrating another apparatus according to an exemplary embodiment of the present disclosure. (general structure of server).
Detailed Description
Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the exemplary embodiments below are not intended to represent all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure, as detailed in the appended claims.
In the related art, the technical solution of the intelligent conversation of the virtual character is like the Avatar frame of microsoft ice. Based on reading and learning the original works of the novels, Microsoft Avatar Framework reconstructs the architecture and knowledge system of the virtual world in 100 novels, and expands and supplements the architecture and knowledge system, thereby establishing a knowledge graph containing a large number of people, entities and various knowledge associations. Based on the established knowledge graph, Microsoft ice Avatar Framework brings 100 fiction owner public IPs into the real world, and brings dialogues, sounds and skills based on their unique characters for virtual characters, and the creation capability of literary works, music works, painting works and the like and a corresponding knowledge system. However, any master's public IP obtained in this way still has limitations in free dialogue with the user.
Before describing the technical solution of the present disclosure in detail, a brief introduction will be made to microsoft small ice Avatar Framework in the related art.
The Avatar Framework consists essentially of four components, one is the Profile tool, which defines the personality for AI notes in terms of both IQ and EQ. And the second is a calculation tool which is used for adjusting the capabilities of a core dialogue engine, computer voice, computer vision and the like for the AI beans. And thirdly, a knowledge tool (namely a knowledge graph) is used for enabling the AI beans to have three views, knowledge, skills and the like, so that the AI beans are closer to the characteristics of human beings. And fourthly, 3D image driving is used for driving a 3D model, so that virtual AI beans appear in front of people.
In the course of training of the Avatar Framework, 100 parts of novel text are used. However, the dialog corpus related to the role in the novel is very limited, and the dialog corpus related to the role in the novel cannot be directly used as the training corpus of the Avatar Framework, so that the virtual character model obtained by training in the way always has the dialog limitation when the open type man-machine dialog is carried out, namely the virtual character model cannot cope with other application scenes except the architecture and the knowledge system of the virtual world in the novel.
The limitation of the character dialogue corpora means that the language corpora need to be written manually or the language corpora need to be collected manually, but the former means that the intelligence and the labor cost are not good, and the latter means that the consistency and immersion feeling of the virtual character model with the character background are lost, because the general dialogue corpora can make the reply generated by the model become general.
In view of the above, embodiments of the present disclosure provide a method and an apparatus for dialog reply, a method and an apparatus for training a dialog model, and a storage medium to solve the above problems.
The following provides a detailed description of embodiments of the present disclosure.
Fig. 1 is a flowchart illustrating a dialog reply method according to an exemplary embodiment, which may be applied to an electronic device, as shown in fig. 1, for example: in the mobile terminal or the server, the mobile terminal comprises a mobile phone, a notebook, a tablet computer, a desktop computer, a POS machine and the like. The server includes: local server and cloud ware. The dialog reply method may specifically include the following steps.
In step S11, first dialogue information input by the user is acquired.
The first dialogue information input by the user can be voice information or text information. It is readily understood that conversion between speech information and text information may be performed.
The first dialogue information can be dialogue information in a question form or non-question form.
In step S12, the first dialogue information and the target character information are input into a pre-trained dialogue model, and target response information, which is output by the dialogue model and replies to the first dialogue information, is obtained.
The target response information is information representing the conversation style of the target role.
It should be noted that inputting the first dialog information and the target character information into a pre-trained dialog model to obtain target response information, which is output by the dialog model and replies to the first dialog information, includes: under the condition that a conversation model is obtained by training a sample training set corresponding to a virtual character, directly inputting the first conversation information into the conversation model to obtain target response information of the target character for replying the first conversation information; or, when the dialog model is obtained by training a plurality of sample training sets corresponding to a plurality of virtual characters, inputting the first dialog information and the target character information into the dialog model to obtain the target response information for replying to the first dialog information. That is, the target character information may be information for selecting a target character to perform a conversation inputted by the user, or may be target character information preset in the conversation model. In this regard, the present disclosure is not particularly limited.
Wherein, the dialogue model in the dialogue reply method is obtained by training in the following way:
constructing a sample training set according to a virtual character sample and a dialogue corpus pair corresponding to the virtual character sample, wherein the dialogue corpus pair comprises a first corpus pair which is screened from a target platform and is associated with a virtual character corresponding to the virtual character sample and a second corpus pair of the virtual character in a source work; and training the conversation model according to the training sample set until the conversation model has the capacity of generating information representing the conversation style of the virtual character sample.
In some possible embodiments, a user group corresponding to a virtual role sample may be determined first, and dialog interaction information when the user group plays a virtual role corresponding to the virtual role sample is obtained, where a platform where the user group is located is a target platform, the target platform may be one or more platforms, and the dialog interaction information may represent a first corpus pair associated with the virtual role. Generating a sample training set corresponding to the virtual character sample according to the first corpus pair and source conversation interaction information of the virtual character in the source work, wherein the source conversation interaction information is a second corpus pair of the virtual character in the source work; and training according to the virtual character sample and the corresponding sample training set to obtain the conversation model.
In the present disclosure, the virtual character may be a character in a novel work, a character in a game work, a virtual human such as a virtual anchor (vtuber), and in a possible case, the virtual character may also be regarded as a virtual human for imitating a real character, and the present disclosure is not particularly limited thereto.
An electronic device determines a user group corresponding to a virtual character sample, searches keywords of the virtual character sample based on a database such as a virtual character knowledge graph and/or a Wikipedia database, and determines a source work type and a source work name corresponding to the virtual character sample according to a search result, wherein the source work type comprises a novel, a game, a virtual anchor and the like. And determining a user group corresponding to the virtual character sample according to the source work type and the source work name of the virtual character sample, wherein the user group can be a novel reader, a game player, an anchor concern/subscriber and the like.
The electronic equipment determines the platform where the user group is located based on the preset corresponding relation between the user group type and the platform, and determines the target platform where the source work is issued from the platforms according to the source work name corresponding to the virtual character sample. Example of preset correspondence of user group type to platform: under the condition that the user group type is a novel reader, the platform is a novel reading platform, a novel bar, a novel discussion community and the like; for example, when the user group type is a game player, the platform is a community corresponding to the game, an official website or an official message area of the game, a tribe, or the like. For another example, in the case where the user group type is a follower of the virtual anchor, the platform may be a live platform on which the virtual anchor is located.
After the target platform is determined, the user group interacting with the source work of the virtual character sample (such as user clicking, watching, downloading, commenting, forwarding and subscribing the source work) can be screened from a user database of the target platform under the condition that the target platform is legally authorized.
Further, the electronic device obtains historical dialogue data of the user groups from a background database of the target platform. And screening candidate reply information similar to the source reply information from historical dialogue data based on the existing source reply information of a small part of virtual character samples, and taking the candidate dialogue information associated with the candidate reply information and the candidate reply information as candidate corpus pairs which are mined from a target platform and associated with virtual characters. The first corpus pair satisfying the requirement can be further screened from the corpus candidate pair.
For example, in the case where the sample of virtual characters is a persona in a fictional work, readers of the fictional work may be determined as a user group to which the sample of virtual characters corresponds. In the case where the virtual character sample is a character in a game work, a player of the game work may be determined as a user group to which the virtual character sample corresponds. In the case that the virtual character sample is a vtube, the attendees/viewers/consumers of the vtube can be determined as the user group to which the virtual character sample corresponds. Taking the game manufacturer miha game as an example, the self-developed miha app aggregates a large number of game player groups of the miha game, and therefore, the user of the miha app can be determined as the user group of the miha game. Accordingly, the Mitsu agency app where the user group is located is a target platform. Furthermore, for social or video platforms such as Youtube, Twitter, microblog, BiliBili, tremble, etc., mihao also has millions of fan public numbers or official channels, and the followers or subscribers of these public numbers or official channels can also be the user groups of mihao. Accordingly, social or video platforms such as Youtube, Twitter, microblog, BiliBili, tremble, etc. are target platforms.
After determining the user group corresponding to the virtual role sample, the dialog interaction information under the condition that the user group plays the virtual role corresponding to the virtual role sample can be obtained. And generating a corresponding sample training set according to the acquired conversation interaction information and the source conversation interaction information of the virtual character in the source work, and training to obtain a conversation model according to the virtual character sample and the sample training set.
By adopting the method disclosed by the invention, the target response information which is output by the dialogue model and used for replying the first dialogue information is obtained by acquiring the first dialogue information input by the user and inputting the first dialogue information and the target role information into the pre-trained dialogue model. The target response information is information representing a dialog style of the target character. By adopting the mode of the disclosure, more natural and accurate target response information can be replied by the conversation style (human setting characteristics) of the target role aiming at any first conversation information input by the user.
Moreover, the sample training set of the conversation model can be generated according to the obtained conversation interaction information when the user group plays the virtual role corresponding to the virtual role sample and the source conversation interaction information of the virtual role sample in the source work, so that the conversation model obtained according to the sample training set has a better effect compared with the model obtained by training only according to the source conversation interaction information in the source work. By adopting the mode of the disclosure, the limitation of man-machine conversation can be avoided. Aiming at any first dialogue information input by the user, more natural and accurate target response information can be replied by the human-set characteristics of the target role.
In some embodiments, the obtaining of the dialog interaction information when the user group plays the virtual role corresponding to the virtual role sample includes:
the electronic equipment pushes the collection activity information to accounts or terminal equipment (or a mailbox) associated with the user group; and acquiring the dialogue interaction information fed back by the user group in response to the gathering activity information. In some possible embodiments, the subset activity information may be role-playing activity information, and the subset activity information is used for obtaining the corpus candidate pairs from the user group.
Since the user of the sound quality player, reader, player, fan, etc. of the virtual character likes the virtual character, the user may actively imitate the features of the virtual character (or quote/rephrase a sentence of the virtual character) in his/her homepage to make relevant and appropriate dynamic publishing or commenting. Therefore, in some embodiments, the obtaining of the dialog interaction information when the user group plays the virtual role corresponding to the virtual role sample further includes:
and identifying and acquiring conversation interaction information related to the virtual character from account number issuing content associated with the user group. In some possible embodiments, the content of the account posting associated with the user group may be content of a microblog, a blog, a WeChat dynamic state, and the like posted by a social media account associated with the user group.
The pushing of the collection activity information to the terminal devices of the user group includes: and pushing preset dialogue information related to the gathering activity information to terminal equipment of the user group, wherein the preset dialogue information is used for guiding the user group to reply the preset dialogue information by imitating dialogue characteristics of a virtual role. Correspondingly, the obtaining of the dialogue interaction information of the user group responding to the symptom set activity information comprises: and acquiring response information of the user group responding to the collection activity information and replying the preset conversation information, wherein the conversation interaction information comprises the preset conversation information and the response information, and the response information is the response information of the user imitating the virtual character.
Specifically, the electronic device can push preset dialogue information of the gathering activity information to the user group so as to guide the user group to reply to the preset dialogue information on the basis of the human-set style imitating the virtual character sample. By acquiring the response information used by the user group for replying the preset dialog information, the dialog interaction information comprising the preset dialog information and the response information can be obtained.
In some embodiments, the pushing of the solicitation activity information to the terminals of the group of users comprises: pushing the collection activity information to the user group and indicating the user group to participate in the collection activity in a mode of carrying preset identification release information; accordingly, the obtaining of the dialogue interaction information of the user group responding to the symptom set activity comprises: and acquiring the dialogue interaction information which is issued by the user group and carries the mark of the symptom set activity in a background database of a target platform, wherein the dialogue interaction information comprises preset dialogue information and response information issued by the user, and the response information is response information of the user imitating the virtual role.
For example, in a case where preset session information of the solicitation activity is not actively pushed to the terminals of the user group, only the solicitation activity information may be pushed to the user group, and the user group may be instructed to participate in the role-playing activity in a manner of carrying a preset identifier. In this way, the dialogue interaction information which is issued by the user group and carries the identification of the symptom set activity can be obtained by identifying the topic identification of the role playing activity. The style of the preset dialog information issued by the user in the dialog interaction information can be the style of any character (including the virtual character sample), but the response information issued by the user in the dialog interaction information should imitate the response information of the virtual character corresponding to the virtual character sample for the user. In some possible embodiments, the preset identifier may be a predetermined symbol, such as # number, and the like, and the preset identifier may be used to characterize the topic related to the preset virtual character.
In some embodiments, the filtering, from the target platform, the first corpus pair associated with the virtual role comprises:
mining a candidate corpus pair associated with the virtual role from the target platform, wherein the candidate corpus pair comprises candidate dialogue information and candidate reply information; and determining the candidate corpus pair in which the candidate reply information is located as the first corpus pair under the condition that the similarity between the candidate reply information and second reply information included in the second corpus pair is greater than a first preset threshold.
In one possible implementation, the second corpus pair includes second dialogue information and second reply information. The corpus candidate pair comprises candidate dialogue information and candidate reply information. After the candidate corpus pair associated with the virtual role is mined from the target platform, the candidate corpus pair in which the candidate reply information is located can be determined as the first corpus pair under the condition that the similarity between the candidate reply information and the second reply information is greater than a first preset threshold value.
Since the candidate answer information in the corpus candidate pair is collected from the user group side, the candidate answer information may include information that does not meet the human requirements (or dialog style) of the virtual character sample. Therefore, the first response information meeting the requirements of the virtual character sample is required to be screened from the candidate response information, so as to improve the accuracy of the training of the dialogue model.
Illustratively, as shown in fig. 2, the step of training the dialogue model may include:
in step S21, mining a corpus candidate pair associated with the virtual character from a target platform, where the corpus candidate pair includes candidate dialog information and candidate reply information;
in step S22, when the similarity between the candidate reply information and the second reply information included in the second corpus pair is greater than a first preset threshold, determining the candidate corpus pair in which the candidate reply information is located as the first corpus pair;
in step S23, obtaining a second corpus pair of the virtual character in the source work;
in step S24, a sample training set is constructed according to a virtual character sample and a dialog corpus pair corresponding to the virtual character sample, where the dialog corpus pair includes a first corpus pair associated with a virtual character corresponding to the virtual character sample and screened from a target platform, and a second corpus pair of the virtual character in a source work;
in step S25, the dialogue model is trained according to the training sample set until the dialogue model has the capability of generating information characterizing the dialogue style of the virtual character sample.
The present disclosure does not limit the sequence between the above steps S21 and S23.
In yet another possible implementation, the filtering, from the target platform, the first corpus pair associated with the virtual character includes:
mining a candidate corpus pair associated with the virtual role from the target platform, wherein the candidate corpus pair comprises candidate dialogue information and candidate reply information; inputting the candidate reply information into a conversation identification model to obtain an identification value representing whether the candidate reply information is the source reply information included in the source work; and determining the candidate corpus pair in which the candidate reply message with the discrimination value larger than a second preset threshold is positioned as the first corpus pair.
In a possible implementation manner, a first candidate reply message whose similarity to the source reply message of the virtual character sample is greater than a first preset threshold may be determined from the candidate reply messages. And then inputting the first candidate reply message into a dialogue identification model (a correlation dialogue model) to obtain an identification value representing whether the first candidate reply message is the source reply message included in the source work, and determining a candidate corpus pair in which the first candidate reply message with the identification value larger than a second preset threshold is positioned as a first corpus pair.
It is worth explaining that the dialogue identification model is obtained by training positive sample corpora and negative sample corpora of the virtual character samples.
In some embodiments, the conversation model comprises a conversation markup network; the inputting the first dialogue information and the target role information into a pre-trained dialogue model to obtain target response information which is output by the dialogue model and used for replying the first dialogue information comprises:
inputting the first dialogue information and the target role information into a pre-trained dialogue marking network to obtain first response information which is output by the dialogue marking network and contains label information;
the label information is used for representing labeling information of each unit character in the first response information, and the label information comprises at least one of a deleting label, a reserving label and a text inserting position label.
In some embodiments, the conversation model includes a conversation insertion network;
the inputting the first dialogue information and the target role information into a pre-trained dialogue model to obtain target response information which is output by the dialogue model and used for replying the first dialogue information, further comprising:
and inputting the first response information into the dialog insertion network to obtain the target response information which is output by the dialog insertion network and is adjusted according to the label information.
In one possible embodiment, referring to fig. 3, in the present disclosure, the dialogue model includes an encoder, a role template matcher, and a decoder; the inputting the first dialogue information and the target role information into a dialogue model includes:
inputting the first dialogue information into the encoder to obtain a first dialogue characteristic vector; inputting the first dialogue characteristic vector and the number of the target role into the role template matcher to obtain a second response information template of the target role; and inputting the second response information template into the decoder to obtain the second response information.
Wherein the decoder comprises a dialog marker network and a dialog insertion network, the decoder being configured to:
inputting the second response information template into the dialog insertion network to label each unit text in the second response information template to obtain a labeled second response information template, wherein the label comprises at least one of a delete label, a reserve label and a text insertion position label; and inputting the labeled second response information template into the dialogue insertion network, and performing corresponding operation on the labeled second response information template according to each labeled tag to obtain the target response information.
For example, the first dialog information X ═ { X1., xN } is input to the encoder, resulting in a first dialog feature vector H ═ H1., hN }, which may be characterized as H ═ encoder (X). Inputting a first dialog feature vector H ═ { H1., hN } and a number of a target character, for example, a number, into a character template matcher, which matches (essentially scores H and each dialog model for matching by a text matching model) a second response information template from M previously stored original dialog templates (lines) associated with character a based on constraints of the first dialog feature vector H ═ { H1., hN }, which can be characterized as T ═ argmaxt∈[1,M](RoleMatch (H, t)) H. Inputting the second response message template into a dialogue marking network in the decoder, wherein the dialogue marking network labels each unit text (such as unit character, unit word, sentence, etc.) in the second response message template, and the label can be a deletion labelLabel, reserve label, text insertion position label. And inputting the labeled second response information template into a dialog insertion network in a Decoder, and performing corresponding operation on the labeled second response information template by the dialog insertion network according to each labeled tag to obtain second response information, wherein the process can be characterized as Y ═ Decoder (H, T).
It should be noted that, in the case that the labeled second response information template includes a text insertion position tag, the new content inserted at the text insertion position is predicted by the dialog insertion network.
In summary, an advantageous dialog model is provided in that the output of the dialog model enables the user to immediately imagine the target character of the user-selected dialog. That is, the output of the dialogue model is highly matched with the speech or the human-set style of the target character. In the related art, response information is generated word by word from zero by using a conventional seq2seq dialogue model in an autoregressive generation manner until a complete response sentence is obtained. This is very time consuming. By adopting the dialogue model disclosed by the disclosure, as the dialogue model disclosed by the disclosure is a non-autoregressive model, namely the dialogue model disclosed by the disclosure is an editing mode of selecting one template from a plurality of response templates to delete, reserve and insert new words to obtain target response information, the dialogue model disclosed by the disclosure can improve the efficiency of dialogue reply. The response speed of such a dialogue model of the present disclosure is much faster than the seq2 seq-based dialogue model in the related art.
Based on the same inventive concept, an embodiment of the present disclosure further provides a training method of a dialogue model, as shown in fig. 4, the method includes the following steps:
in step S31, a sample training set is constructed according to a virtual character sample and a dialog corpus pair corresponding to the virtual character sample, where the dialog corpus pair includes a first corpus pair associated with a virtual character corresponding to the virtual character sample and screened from a target platform, and a second corpus pair of the virtual character in a source work;
in step S32, the dialogue model is trained according to the training sample set until the dialogue model has the capability of generating information characterizing the dialogue style of the virtual character sample.
By adopting the training method, a dialogue model which can reply more natural and accurate target response information by the dialogue style (human setting characteristics) of the target role according to any first dialogue information input by a user can be trained.
Based on the same inventive concept, the disclosed embodiment further provides a dialog reply device, as shown in fig. 5, the dialog reply device 300 includes an obtaining module 310 and an inputting module 320.
An obtaining module 310 configured to obtain first dialogue information input by a user;
an input module 320 configured to input the first dialog information and the target role information into a pre-trained dialog model, so as to obtain target response information, which is output by the dialog model and replies to the first dialog information; the target response information is information representing the conversation style of the target role.
By using the dialog reply device 300 of the present disclosure, the target response information for replying to the first dialog information output by the dialog model is obtained by acquiring the first dialog information input by the user and inputting the first dialog information and the target character information into the pre-trained dialog model. The target response information is information representing a dialog style of the target character. By adopting the mode of the disclosure, more natural and accurate target response information can be replied by the conversation style (human setting characteristics) of the target role aiming at any first conversation information input by the user.
In some embodiments, the conversation model comprises a conversation markup network; the input module 320 includes:
the first input submodule is configured to input the first dialogue information and the target role information into a pre-trained dialogue marking network to obtain first response information which is output by the dialogue marking network and contains label information;
the label information is used for representing labeling information of each unit character in the first response information, and the label information comprises at least one of a deleting label, a reserving label and a text inserting position label.
In some embodiments, the conversation model includes a conversation insertion network;
the input module 320 further includes:
and the second input submodule is configured to input the first response information into the dialog insertion network, so as to obtain the target response information which is output by the dialog insertion network and is obtained after the first response information is adjusted according to the tag information.
In some embodiments, the apparatus further comprises a training module 330 configured to train the dialogue model by:
constructing a sample training set according to a virtual character sample and a dialogue corpus pair corresponding to the virtual character sample, wherein the dialogue corpus pair comprises a first corpus pair which is screened from a target platform and is associated with a virtual character corresponding to the virtual character sample and a second corpus pair of the virtual character in a source work; and training the conversation model according to the training sample set until the conversation model has the capacity of generating information representing the conversation style of the virtual character sample. .
In some embodiments, the training module 330 includes:
a first mining submodule configured to mine a candidate corpus pair associated with the virtual character from the target platform, the candidate corpus pair including candidate dialogue information and candidate reply information;
the first determining submodule is configured to determine the corpus candidate pair in which the reply candidate information is located as the first corpus pair when the similarity between the reply candidate information and second reply information included in the second corpus pair is greater than a first preset threshold.
In some embodiments, the training module 330 includes:
a second mining submodule configured to mine a candidate corpus pair associated with the virtual character from the target platform, the candidate corpus pair including candidate dialogue information and candidate reply information;
a third input submodule configured to input the candidate reply message into a dialogue identification model, to obtain an identification value representing whether the candidate reply message is a source reply message included in the source work;
a second determining submodule configured to determine the corpus candidate pair in which the candidate reply message with the discrimination value larger than a second preset threshold is located as the first corpus pair.
With regard to the apparatus in the above-described embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment related to the method, and will not be elaborated here.
Based on the same inventive concept, the embodiment of the present disclosure further provides a training apparatus for a dialogue model, as shown in fig. 6, the apparatus 400 includes:
a first execution module 410, configured to construct a sample training set according to a virtual character sample and a dialog corpus pair corresponding to the virtual character sample, where the dialog corpus pair includes a first corpus pair associated with a virtual character corresponding to the virtual character sample screened from a target platform and a second corpus pair of the virtual character in a source work;
a second execution module 420 configured to train the conversation model according to the training sample set until the conversation model has the capability of generating information characterizing a conversation style of the virtual character sample.
By using the training device, a dialogue model which can reply more natural and accurate target response information according to the dialogue style (human setting characteristics) of the target role aiming at any first dialogue information input by a user can be trained.
The present disclosure also provides a computer-readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement the steps of the dialog reply method or the training method of the dialog model provided by the present disclosure.
FIG. 7 is a block diagram illustrating an apparatus 800 for dialog reply or for training of a dialog model, according to an example embodiment. For example, the apparatus 800 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, an exercise device, a personal digital assistant, and the like.
Referring to fig. 7, the apparatus 800 may include one or more of the following components: a processing component 802, a memory 804, a power component 806, a multimedia component 808, an audio component 810, an input/output (I/O) interface 812, a sensor component 814, and a communication component 816.
The processing component 802 generally controls overall operation of the device 800, such as operations associated with display, telephone calls, data communications, camera operations, and recording operations. The processing components 802 may include one or more processors 820 to execute instructions to perform all or a portion of the steps of the dialog reply method or the training method of the dialog model described above. Further, the processing component 802 can include one or more modules that facilitate interaction between the processing component 802 and other components. For example, the processing component 802 can include a multimedia module to facilitate interaction between the multimedia component 808 and the processing component 802.
The memory 804 is configured to store various types of data to support operations at the apparatus 800. Examples of such data include instructions for any application or method operating on device 800, contact data, phonebook data, messages, pictures, videos, and so forth. The memory 804 may be implemented by any type or combination of volatile or non-volatile memory devices such as Static Random Access Memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic or optical disks.
The multimedia component 808 includes a screen that provides an output interface between the device 800 and a user. In some embodiments, the screen may include a Liquid Crystal Display (LCD) and a Touch Panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive an input signal from a user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundary of a touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 808 includes a front facing camera and/or a rear facing camera. The front camera and/or the rear camera may receive external multimedia data when the device 800 is in an operating mode, such as a shooting mode or a video mode. Each front camera and rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.
The audio component 810 is configured to output and/or input audio signals. For example, the audio component 810 includes a Microphone (MIC) configured to receive external audio signals when the apparatus 800 is in an operational mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signals may further be stored in the memory 804 or transmitted via the communication component 816. In some embodiments, audio component 810 also includes a speaker for outputting audio signals.
The I/O interface 812 provides an interface between the processing component 802 and peripheral interface modules, which may be keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to: a home button, a volume button, a start button, and a lock button.
The sensor assembly 814 includes one or more sensors for providing various aspects of state assessment for the device 800. For example, the sensor assembly 814 may detect the open/closed status of the device 800, the relative positioning of components, such as a display and keypad of the device 800, the sensor assembly 814 may also detect a change in the position of the device 800 or a component of the device 800, the presence or absence of user contact with the device 800, the orientation or acceleration/deceleration of the device 800, and a change in the temperature of the device 800. Sensor assembly 814 may include a proximity sensor configured to detect the presence of a nearby object without any physical contact. The sensor assembly 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.
The communication component 816 is configured to facilitate communications between the apparatus 800 and other devices in a wired or wireless manner. The device 800 may access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 further includes a Near Field Communication (NFC) module to facilitate short-range communications. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID) technology, infrared data association (IrDA) technology, Ultra Wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.
In an exemplary embodiment, the apparatus 800 may be implemented by one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), controllers, micro-controllers, microprocessors or other electronic components for performing the above-described dialog reply method or the training method of the dialog model.
In an exemplary embodiment, a non-transitory computer-readable storage medium comprising instructions, such as the memory 804 comprising instructions, executable by the processor 820 of the device 800 to perform the above-described dialog reply method or training method of a dialog model is also provided. For example, the non-transitory computer readable storage medium may be a ROM, a Random Access Memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.
In another exemplary embodiment, a computer program product is also provided, which comprises a computer program executable by a programmable apparatus, the computer program having code portions for performing the above-described dialog reply method or the training method of the dialog model when executed by the programmable apparatus.
Fig. 8 is a block diagram illustrating an apparatus 1900 for another dialog reply in accordance with an example embodiment. For example, the apparatus 1900 may be provided as a server. Referring to FIG. 8, the device 1900 includes a processing component 1922 further including one or more processors and memory resources, represented by memory 1932, for storing instructions, e.g., applications, executable by the processing component 1922. The application programs stored in memory 1932 may include one or more modules that each correspond to a set of instructions. Further, the processing component 1922 is configured to execute instructions to perform the dialog reply method described above.
The device 1900 may also include a power component 1926 configured to perform power management of the device 1900, a wired or wireless network interface 1950 configured to connect the device 1900 to a network, and an input/output (I/O) interface 1958. The device 1900 may operate based on an operating system, such as Windows Server, stored in memory 1932TM，Mac OS XTM，UnixTM，LinuxTM，FreeBSDTMOr the like.
Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure. This application is intended to cover any variations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.
It will be understood that the present disclosure is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.
Claims (12)
1. A dialog reply method, characterized in that the method comprises:
acquiring first dialogue information input by a user;
inputting the first dialogue information and the target role information into a pre-trained dialogue model to obtain target response information which is output by the dialogue model and used for replying the first dialogue information;
the target response information is information representing the conversation style of the target role.
2. The method of claim 1, wherein the conversation model comprises a conversation markup network;
the inputting the first dialogue information and the target role information into a pre-trained dialogue model to obtain target response information which is output by the dialogue model and used for replying the first dialogue information comprises:
inputting the first dialogue information and the target role information into a pre-trained dialogue marking network to obtain first response information which is output by the dialogue marking network and contains label information;
the label information is used for representing labeling information of each unit character in the first response information, and the label information comprises at least one of a deleting label, a reserving label and a text inserting position label.
3. The method of claim 2, wherein the conversation model further comprises a conversation insertion network;
the inputting the first dialogue information and the target role information into a pre-trained dialogue model to obtain target response information which is output by the dialogue model and used for replying the first dialogue information, further comprising:
and inputting the first response information into the dialog insertion network to obtain the target response information which is output by the dialog insertion network and is adjusted according to the label information.
4. The method of claim 1, wherein the dialogue model is trained by:
constructing a sample training set according to a virtual character sample and a dialogue corpus pair corresponding to the virtual character sample, wherein the dialogue corpus pair comprises a first corpus pair which is screened from a target platform and is associated with a virtual character corresponding to the virtual character sample and a second corpus pair of the virtual character in a source work;
and training the conversation model according to the training sample set until the conversation model has the capacity of generating information representing the conversation style of the virtual character sample.
5. The method of claim 4, wherein the filtering the first corpus pair associated with the virtual character from the target platform comprises:
mining a candidate corpus pair associated with the virtual role from the target platform, wherein the candidate corpus pair comprises candidate dialogue information and candidate reply information;
and determining the candidate corpus pair in which the candidate reply information is located as the first corpus pair under the condition that the similarity between the candidate reply information and second reply information included in the second corpus pair is greater than a first preset threshold.
6. The method of claim 4, wherein the filtering the first corpus pair associated with the virtual character from the target platform comprises:
mining a candidate corpus pair associated with the virtual role from the target platform, wherein the candidate corpus pair comprises candidate dialogue information and candidate reply information;
inputting the candidate reply information into a conversation identification model to obtain an identification value representing whether the candidate reply information is the source reply information included in the source work;
and determining the candidate corpus pair in which the candidate reply message with the discrimination value larger than a second preset threshold is positioned as the first corpus pair.
7. A method for training a dialogue model, the method comprising:
constructing a sample training set according to a virtual character sample and a dialogue corpus pair corresponding to the virtual character sample, wherein the dialogue corpus pair comprises a first corpus pair which is screened from a target platform and is associated with a virtual character corresponding to the virtual character sample and a second corpus pair of the virtual character in a source work;
and training the conversation model according to the training sample set until the conversation model has the capacity of generating information representing the conversation style of the virtual character sample.
8. A conversation reply device, comprising:
the system comprises an acquisition module, a processing module and a processing module, wherein the acquisition module is configured to acquire first dialogue information input by a user;
the input module is configured to input the first dialogue information and the target role information into a pre-trained dialogue model to obtain target response information which is output by the dialogue model and used for replying the first dialogue information;
the target response information is information representing the conversation style of the target role.
9. An apparatus for training a dialogue model, the apparatus comprising:
the system comprises a first execution module, a second execution module and a third execution module, wherein the first execution module is configured to construct a sample training set according to a virtual character sample and a dialogue corpus pair corresponding to the virtual character sample, and the dialogue corpus pair comprises a first corpus pair which is screened from a target platform and is associated with a virtual character corresponding to the virtual character sample and a second corpus pair of the virtual character in a source work;
a second execution module configured to train the conversation model according to the training sample set until the conversation model has a capability of generating information characterizing a conversation style of the virtual character sample.
10. A conversation reply device, comprising:
a processor;
a memory for storing processor-executable instructions;
wherein the processor is configured to:
acquiring first dialogue information input by a user;
inputting the first dialogue information and the target role information into a pre-trained dialogue model to obtain target response information which is output by the dialogue model and used for replying the first dialogue information;
the target response information is information representing the conversation style of the target role.
11. An apparatus for training a dialogue model, comprising:
a processor;
a memory for storing processor-executable instructions;
wherein the processor is configured to:
constructing a sample training set according to a virtual character sample and a dialogue corpus pair corresponding to the virtual character sample, wherein the dialogue corpus pair comprises a first corpus pair which is screened from a target platform and is associated with a virtual character corresponding to the virtual character sample and a second corpus pair of the virtual character in a source work;
and training the conversation model according to the training sample set until the conversation model has the capacity of generating information representing the conversation style of the virtual character sample.
12. A computer-readable storage medium, on which computer program instructions are stored, which program instructions, when executed by a processor, carry out the steps of the method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110802896.5A CN113378583A (en) | 2021-07-15 | 2021-07-15 | Dialogue reply method and device, dialogue model training method and device, and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110802896.5A CN113378583A (en) | 2021-07-15 | 2021-07-15 | Dialogue reply method and device, dialogue model training method and device, and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113378583A true CN113378583A (en) | 2021-09-10 |
Family
ID=77582231
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110802896.5A Pending CN113378583A (en) | 2021-07-15 | 2021-07-15 | Dialogue reply method and device, dialogue model training method and device, and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113378583A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114247143A (en) * | 2021-12-21 | 2022-03-29 | 北京蔚领时代科技有限公司 | Digital human interaction method, device, equipment and storage medium based on cloud server |
| CN114490967A (en) * | 2021-12-28 | 2022-05-13 | 北京百度网讯科技有限公司 | Training method of dialogue model, dialogue method and device of dialogue robot and electronic equipment |
| WO2024066920A1 (en) * | 2022-09-30 | 2024-04-04 | 腾讯科技（深圳）有限公司 | Processing method and apparatus for dialogue in virtual scene, and electronic device, computer program product and computer storage medium |
Citations (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120239390A1 (en) * | 2011-03-18 | 2012-09-20 | Kabushiki Kaisha Toshiba | Apparatus and method for supporting reading of document, and computer readable medium |
| US20180061400A1 (en) * | 2016-08-30 | 2018-03-01 | Google Inc. | Using textual input and user state information to generate reply content to present in response to the textual input |
| US20180293483A1 (en) * | 2017-04-11 | 2018-10-11 | Microsoft Technology Licensing, Llc | Creating a Conversational Chat Bot of a Specific Person |
| CN108984655A (en) * | 2018-06-28 | 2018-12-11 | 厦门快商通信息技术有限公司 | A kind of customer service robot intelligent customer service bootstrap technique |
| CN110197279A (en) * | 2019-06-10 | 2019-09-03 | 北京百度网讯科技有限公司 | Transformation model training method, device, equipment and storage medium |
| US20190286996A1 (en) * | 2016-09-05 | 2019-09-19 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Human-machine interactive method and device based on artificial intelligence |
| CN110399474A (en) * | 2019-07-18 | 2019-11-01 | 腾讯科技（深圳）有限公司 | A kind of Intelligent dialogue method, apparatus, equipment and storage medium |
| CN111143535A (en) * | 2019-12-27 | 2020-05-12 | 北京百度网讯科技有限公司 | Method and apparatus for generating a dialogue model |
| CN111159368A (en) * | 2019-12-12 | 2020-05-15 | 华南理工大学 | Reply generation method for personalized dialogue |
| CN111309886A (en) * | 2020-02-18 | 2020-06-19 | 腾讯科技（深圳）有限公司 | Information interaction method and device and computer readable storage medium |
| CN112035628A (en) * | 2020-08-03 | 2020-12-04 | 北京小米松果电子有限公司 | Dialogue data cleaning method, device and storage medium |
| CN112036156A (en) * | 2020-09-25 | 2020-12-04 | 北京小米松果电子有限公司 | Text dialogue method, text dialogue device and storage medium |
| CN112417118A (en) * | 2020-11-19 | 2021-02-26 | 上海交通大学 | Dialog generation method based on marked text and neural network |
| CN112541063A (en) * | 2020-12-08 | 2021-03-23 | 山东师范大学 | Man-machine conversation method and system based on self-learning conversation model |
| WO2021051404A1 (en) * | 2019-09-20 | 2021-03-25 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for auxiliary reply |
| CN112667796A (en) * | 2021-01-05 | 2021-04-16 | 网易（杭州）网络有限公司 | Dialog reply method and device, electronic equipment and readable storage medium |
| CN112668339A (en) * | 2020-12-23 | 2021-04-16 | 北京有竹居网络技术有限公司 | Corpus sample determination method and device, electronic equipment and storage medium |
| CN113051388A (en) * | 2021-04-30 | 2021-06-29 | 中国银行股份有限公司 | Intelligent question and answer method and device, electronic equipment and storage medium |
-
2021
- 2021-07-15 CN CN202110802896.5A patent/CN113378583A/en active Pending
Patent Citations (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120239390A1 (en) * | 2011-03-18 | 2012-09-20 | Kabushiki Kaisha Toshiba | Apparatus and method for supporting reading of document, and computer readable medium |
| US20180061400A1 (en) * | 2016-08-30 | 2018-03-01 | Google Inc. | Using textual input and user state information to generate reply content to present in response to the textual input |
| US20190286996A1 (en) * | 2016-09-05 | 2019-09-19 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Human-machine interactive method and device based on artificial intelligence |
| US20180293483A1 (en) * | 2017-04-11 | 2018-10-11 | Microsoft Technology Licensing, Llc | Creating a Conversational Chat Bot of a Specific Person |
| CN108984655A (en) * | 2018-06-28 | 2018-12-11 | 厦门快商通信息技术有限公司 | A kind of customer service robot intelligent customer service bootstrap technique |
| CN110197279A (en) * | 2019-06-10 | 2019-09-03 | 北京百度网讯科技有限公司 | Transformation model training method, device, equipment and storage medium |
| CN110399474A (en) * | 2019-07-18 | 2019-11-01 | 腾讯科技（深圳）有限公司 | A kind of Intelligent dialogue method, apparatus, equipment and storage medium |
| WO2021051404A1 (en) * | 2019-09-20 | 2021-03-25 | Beijing Didi Infinity Technology And Development Co., Ltd. | Systems and methods for auxiliary reply |
| CN111159368A (en) * | 2019-12-12 | 2020-05-15 | 华南理工大学 | Reply generation method for personalized dialogue |
| CN111143535A (en) * | 2019-12-27 | 2020-05-12 | 北京百度网讯科技有限公司 | Method and apparatus for generating a dialogue model |
| CN111309886A (en) * | 2020-02-18 | 2020-06-19 | 腾讯科技（深圳）有限公司 | Information interaction method and device and computer readable storage medium |
| CN112035628A (en) * | 2020-08-03 | 2020-12-04 | 北京小米松果电子有限公司 | Dialogue data cleaning method, device and storage medium |
| CN112036156A (en) * | 2020-09-25 | 2020-12-04 | 北京小米松果电子有限公司 | Text dialogue method, text dialogue device and storage medium |
| CN112417118A (en) * | 2020-11-19 | 2021-02-26 | 上海交通大学 | Dialog generation method based on marked text and neural network |
| CN112541063A (en) * | 2020-12-08 | 2021-03-23 | 山东师范大学 | Man-machine conversation method and system based on self-learning conversation model |
| CN112668339A (en) * | 2020-12-23 | 2021-04-16 | 北京有竹居网络技术有限公司 | Corpus sample determination method and device, electronic equipment and storage medium |
| CN112667796A (en) * | 2021-01-05 | 2021-04-16 | 网易（杭州）网络有限公司 | Dialog reply method and device, electronic equipment and readable storage medium |
| CN113051388A (en) * | 2021-04-30 | 2021-06-29 | 中国银行股份有限公司 | Intelligent question and answer method and device, electronic equipment and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 王乾铭;李吟;: "基于深度学习的个性化聊天机器人研究", 计算机技术与发展, no. 04, 30 April 2020 (2020-04-30) * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114247143A (en) * | 2021-12-21 | 2022-03-29 | 北京蔚领时代科技有限公司 | Digital human interaction method, device, equipment and storage medium based on cloud server |
| CN114490967A (en) * | 2021-12-28 | 2022-05-13 | 北京百度网讯科技有限公司 | Training method of dialogue model, dialogue method and device of dialogue robot and electronic equipment |
| CN114490967B (en) * | 2021-12-28 | 2023-10-31 | 北京百度网讯科技有限公司 | Training method of dialogue model, dialogue method and device of dialogue robot and electronic equipment |
| WO2024066920A1 (en) * | 2022-09-30 | 2024-04-04 | 腾讯科技（深圳）有限公司 | Processing method and apparatus for dialogue in virtual scene, and electronic device, computer program product and computer storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109522419B (en) | Session information completion method and device | |
| CN113378583A (en) | Dialogue reply method and device, dialogue model training method and device, and storage medium | |
| CN110602516A (en) | Information interaction method and device based on live video and electronic equipment | |
| CN110175223A (en) | A kind of method and device that problem of implementation generates | |
| CN109460503B (en) | Answer input method, answer input device, storage medium and electronic equipment | |
| CN104035995A (en) | Method and device for generating group tags | |
| CN113689530B (en) | Method and device for driving digital person and electronic equipment | |
| CN110929176A (en) | Information recommendation method and device and electronic equipment | |
| CN112151041B (en) | Recording method, device, equipment and storage medium based on recorder program | |
| CN115840841A (en) | Multi-modal dialog method, device, equipment and storage medium | |
| CN113177419B (en) | Text rewriting method and device, storage medium and electronic equipment | |
| CN113656557A (en) | Message reply method, device, storage medium and electronic equipment | |
| CN111723606A (en) | Data processing method and device and data processing device | |
| CN111739535A (en) | Voice recognition method and device and electronic equipment | |
| CN109165285A (en) | Handle the method, apparatus and storage medium of multi-medium data | |
| CN116737883A (en) | Man-machine interaction method, device, equipment and storage medium | |
| CN113542797A (en) | Interaction method and device in video playing and computer readable storage medium | |
| CN112988956A (en) | Method and device for automatically generating conversation and method and device for detecting information recommendation effect | |
| CN113923517B (en) | Background music generation method and device and electronic equipment | |
| CN110162710A (en) | Information recommendation method and device under input scene | |
| CN112820265B (en) | Speech synthesis model training method and related device | |
| CN111160051B (en) | Data processing method, device, electronic equipment and storage medium | |
| CN113901832A (en) | Man-machine conversation method, device, storage medium and electronic equipment | |
| CN114579702A (en) | Message sending method, device, terminal and storage medium for preventing misoperation | |
| CN113420553A (en) | Text generation method and device, storage medium and electronic equipment |
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 |