CN117573859A - Data processing method, system and equipment for automatically advancing scenario and dialogue - Google Patents
Data processing method, system and equipment for automatically advancing scenario and dialogue Download PDFInfo
- Publication number
- CN117573859A CN117573859A CN202410051347.2A CN202410051347A CN117573859A CN 117573859 A CN117573859 A CN 117573859A CN 202410051347 A CN202410051347 A CN 202410051347A CN 117573859 A CN117573859 A CN 117573859A
- Authority
- CN
- China
- Prior art keywords
- scenario
- data
- dialogue
- speaking
- historical
- 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
- 238000003672 processing method Methods 0.000 title claims abstract description 16
- 238000012545 processing Methods 0.000 claims abstract description 30
- 238000000034 method Methods 0.000 claims abstract description 22
- 238000012549 training Methods 0.000 claims description 72
- 230000019771 cognition Effects 0.000 claims description 33
- 230000001149 cognitive effect Effects 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 5
- 230000000981 bystander Effects 0.000 claims description 4
- 230000003247 decreasing effect Effects 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 6
- 241000282693 Cercopithecidae Species 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 210000000988 bone and bone Anatomy 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 235000014347 soups Nutrition 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 241000282414 Homo sapiens Species 0.000 description 1
- 241001465754 Metazoa Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000008451 emotion Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000007654 immersion Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 206010025482 malaise Diseases 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- 238000005406 washing Methods 0.000 description 1
Classifications
-
- 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
Abstract
The invention relates to the field of data processing, in particular to a data processing method, system and equipment for automatically advancing a scenario and a dialogue, aiming at improving participation will of a user. The method for automatically propelling the scenario and the dialogue provided by the invention comprises the following steps: judging whether the scenario needs to be switched according to the historical data; if so, processing the history data to generate new scenario content and virtual roles speaking or whitish; otherwise, processing the historical data to judge whether the user should speak, and acquiring the user's speaking or generating a new round of virtual character speaking or whitish according to the judging result; updating the history data, repeating the steps of determining if a scenario needs to be switched and generating speech, bystanding and/or scenario content. Wherein the history data includes: historical scenario and historical dialogue. The scenario and the dialogue generated by the invention are influenced by the historical data, so that the continuity of the subsequent scenario is improved, and the participation will of the user is improved.
Description
Technical Field
The invention relates to the field of data processing, in particular to a data processing method, system and equipment for automatically advancing a scenario and a dialogue.
Background
The virtual sitcom can be widely applied to the fields of social activity exercise, children safety education, actor training and the like because the virtual sitcom provides the function of participating in role interaction by the user.
In the natural language processing process, model training is often carried out by utilizing historical data, a large language model can simulate the language capability of human beings through learning massive data, has the capability of understanding and generating natural language, and can be used for various text generation tasks such as machine translation, automatic abstract, dialogue system, role playing and the like.
In a scenario application based on a large language model, multiple rounds of conversations by a user with machine virtual characters may be considered a series of data inputs and outputs. As the number of dialog turns increases, processing such data becomes more challenging because of the need to maintain consistency and correlation of dialogs. However, when the data is mishandled or the model fails to understand the context of the dialog, a dialog may be generated that is not subject-matter-related, which may result in reduced user interest and engagement.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a data processing method, a system and equipment for automatically advancing the scenario and the dialogue, which improves the participation will of the user.
In a first aspect of the present invention, a data processing method for automatically advancing a scenario and a dialogue is provided, the method comprising:
judging whether the scenario needs to be switched according to the historical data;
if the scenario needs to be switched, processing the historical data to generate scenario content of a new round and virtual roles speaking or whitish;
if the scenario does not need to be switched, processing the historical data to judge whether the user should speak, and acquiring the user's speaking or generating a new round of virtual character speaking or bystanding according to the judging result;
updating the history data, and repeatedly executing the steps of judging whether the scenario needs to be switched and generating speaking, bystanding and/or scenario contents;
wherein the history data includes: historical scenario and historical dialogue.
Preferably, the step of processing the history data to generate a new scenario content and the virtual character speaking or blurriness includes:
inputting the history data into a first scenario engine model to generate a new scenario content;
inputting the role cognition data, the scenario content and the history data into a first cognition model, and outputting the updated role cognition data;
inputting the updated role cognition data, the scenario content and the history data into a first dialogue engine model to generate the virtual role speaking or the bystander;
wherein the role awareness data includes: the virtual character recognizes itself and recognizes each of the other characters.
Preferably, the step of processing the history data to generate a new scenario content and the virtual character speaking or blurriness includes:
inputting the history data and the control items into a second scenario engine model to generate new scenario contents;
inputting the role cognition data, the scenario content and the history data into a first cognition model, and outputting updated role cognition data;
inputting the updated character cognition data, the scenario content and the history data into a first dialogue engine model to generate the virtual character speaking or the bystander;
wherein,
the control items are set by operators in the background, and comprise: continuity with historical scenario, degree of influence and interestingness from previous dialog.
Preferably, the step of processing the history data to determine whether the user should speak, and acquiring the user's speaking or generating a new virtual character speaking or a new bypass according to the determination result includes:
inputting the historical data into a speaking order judging model to judge whether the user should speak or not;
if yes, acquiring the speech of the user;
otherwise, processing the history data to generate a new round of virtual character speaking or white-over;
the speaking order judgment model adopts a large language model and is trained.
Preferably, the step of processing the history data to generate a new round of virtual character speech or whitish includes:
inputting the character cognitive data and the historical data into a second cognitive model, and outputting updated character cognitive data;
and inputting the updated role cognition data and the history data into a second dialogue engine model to generate the virtual role speaking or the whitish.
Preferably, the step of determining whether the scenario needs to be switched according to the history data includes:
inputting the historical data into a scenario switching judgment model, and respectively outputting the probability of scenario switching after 0, 1, 2, 3 and more than 3 times of dialogue;
if the probability of 0-round switching is greater than or equal to the preset probability or the probabilities of 1 round, 2 rounds, 3 rounds and more than 3 rounds of switching are all smaller than the probability of 0 round switching, judging that the scenario needs to be switched; otherwise, judging that the scenario does not need to be switched;
the scenario switching judgment model adopts a large language model and is trained, and the training method comprises the following steps:
constructing a first training set, the first training set comprising: a preset first number of training samples and corresponding switching probability labels; each training sample includes: a scenario segment and corresponding dialog; the handover probability tag includes: the scenario segment switches probability reality values after 0, 1, 2, 3 and more than 3 conversations respectively;
selecting a training sample from a first training set, and inputting the training sample into the scenario switching judgment model to obtain an output result; the output result comprises: the scenario segment switches probability prediction values after 0, 1, 2, 3 and more than 3 conversations respectively;
calculating a first loss function according to the output result and the switching probability label, and adjusting model parameters;
the execution is repeated until the first loss function is no longer decreasing.
Preferably, the first scenario engine model adopts a large language model, and has been trained, and the training method includes:
constructing a second training set, the second training set comprising: a preset second number of training samples; each training sample includes: a scenario segment and corresponding dialogue, and the next scenario segment corresponding to the scenario segment;
training the first scenario engine model based on the second training set.
Preferably, the second scenario engine model adopts a large language model, and has been trained, and the training method includes:
constructing a third training set, the third training set comprising: a preset third number of training samples; each training sample includes: a scenario segment and corresponding dialogue, control item label, and the next scenario segment corresponding to the scenario segment;
training the second scenario engine model based on the third training set;
wherein,
the control item tag includes: the consistency with the historical scenario, the degree of influence from the previous session, and the interestingness.
In a second aspect of the invention, a data processing system for automatically advancing episodes and conversations is presented, the system comprising:
the judging module is used for inputting the historical data into the scenario switching judging model and judging whether the scenario needs to be switched or not;
the first generation module is used for processing the historical data to generate a new scenario content and a virtual character speaking or whitish under the condition that the scenario needs to be switched;
the second generation module is used for processing the historical data to judge whether the user should speak or not under the condition that the scenario does not need to be switched, and acquiring the user's speaking or generating a new round of virtual character speaking or bystanding according to the judging result;
the control module is used for updating the historical data and controlling the judging module and the first generating module/the second generating module to repeatedly execute;
wherein the history data includes: historical scenario and historical dialogue.
In a third aspect the invention proposes a computer readable storage device storing a computer program capable of being loaded by a processor and executing a method as described above.
The invention has the following beneficial effects:
the method provided by the invention can automatically advance the development of the scenario and the dialogue among the roles by utilizing the history data, ensure that the newly generated scenario, role speaking and bystanding keep internal logic continuity with the previous, and effectively improve the participation degree and immersion feeling of the user. After the new scenario is generated, the system preferentially generates the speech or the side of the virtual character instead of waiting for user input, so that experience interruption or reduction caused by uncertain how the user responds is avoided.
The invention also introduces the cognitive data of the roles into dialogue generation, so that the generated speaking and the generated paralogue are closely connected with the psychological activities of the roles, and the roles are endowed with richer emotion.
The invention also introduces the control items into the generation of the scenario, and can effectively guide the development direction of the subsequent scenario, which ensures the continuity and the interestingness of the scenario and avoids the occurrence of the phenomena of dislocation, fun and the like of the scenario before and after the scenario.
In a word, the invention obviously improves the scenario generation quality and the smoothness of user experience through a series of fine data processing and analysis technologies, and effectively improves the participation will of the user.
Drawings
FIG. 1 is a schematic diagram of the main steps of an embodiment of a data processing method for self-propelled scenario and dialogue of the present invention;
FIG. 2 is a schematic diagram of main steps of training a scenario switching judgment model according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of the principal components of an embodiment of the data processing system of the present invention for self-propelled storyline and conversation.
Detailed Description
Preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are merely for explaining the technical principles of the present invention, and are not intended to limit the scope of the present invention.
For the purposes of making the objects, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions of the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments of the present application without making any inventive effort, are intended to be within the scope of the present invention.
It should be noted that in the description of the present invention, the terms "first," "second," and the like are merely used for convenience of description and are not to be construed as limiting the invention as to the relative importance of the device, element or parameter being described or implied. In addition, the term "and/or" in the present invention is merely an association relationship describing the association object, and indicates that three relationships may exist, for example, a and/or B may indicate: a exists alone, A and B exist together, and B exists alone. In this context, unless otherwise specified, the term "/" generally indicates that the associated object is an "or" relationship.
In the present invention, the virtual characters refer to all the characters driven by the computer program except the roles played by the user in the scenario, and the speaking, actions, psychology and the like of the characters are generated by the computer program.
Fig. 1 is a schematic diagram of main steps of an embodiment of a data processing method for self-propulsion scenario and dialogue of the present invention. As shown in fig. 1, the method of the present embodiment includes steps a10-a40:
step A10, judging whether the scenario needs to be switched according to the historical data; if so, go to step A20; otherwise, go to step a30.
Wherein the history data includes: historical scenario and historical dialogue.
This step may specifically comprise steps a11-a12:
and step A11, inputting the historical data into a scenario switching judgment model, and respectively outputting the probabilities of scenario switching after 0 rounds of dialogue, 1 round of dialogue, 2 rounds of dialogue, 3 rounds of dialogue and more than 3 rounds of dialogue.
Step A12, if the probability of 0 round of switching is greater than or equal to the preset probability (such as 80%), or the probabilities of 1 round, 2 rounds, 3 rounds and more than 3 rounds of switching are all smaller than the probability of 0 round of switching, judging that the scenario needs to be switched; otherwise, judging that the scenario does not need to be switched.
For example, model output: there is also a probability of more than 3 rounds of switching of 2%, there is also a probability of 3 rounds of switching of 8%, there is also a probability of 2 rounds of switching of 10%, there is also a probability of 1 round of switching of 50%, and there is also a probability of 0 round of switching of 30%. At this time, it is determined that the scenario does not need to be switched.
The scenario switching judgment model adopts a large language model and is trained.
Step A20, processing the history data to generate a new scenario content and a virtual character speaking or whitish, and turning to step A40. This step may specifically comprise steps a21-a23:
and step A21, inputting the history data into a first scenario engine model to generate scenario contents of a new round.
Step A22, inputting the role cognition data, the scenario content and the history data into a first cognition model, and outputting updated role cognition data;
step A23, inputting the updated role cognition data, scenario content and history data into a first dialogue engine model to generate virtual role speaking or whitish;
wherein the character recognition data includes: the virtual character recognizes itself and recognizes each of the other characters; the first scenario engine model, the first cognitive model and the first dialogue engine model all adopt large language models and are trained.
In this embodiment, after a new scenario is generated, a virtual character is first generated to speak or talk instead of making the user speak first, so that the user is prevented from knowing what is said at a time, and the user experience is reduced.
And step A30, processing the historical data to judge whether the user should speak, and acquiring the user's speaking or generating a new round of virtual character speaking or whitish according to the judging result.
This step may specifically comprise steps a31-a33:
step A31, inputting the history data into a speaking order judging model to judge whether the user should speak; if yes, go to step a32; otherwise, go to step a33.
The speaking order judging model adopts a large language model and is trained.
Step a32, the speech of the user is acquired, and the process goes to step a40.
And step A33, generating a new round of virtual character speaking or whitish according to the historical data.
The step A33 may specifically include steps A331-A332:
and step A331, inputting the character cognition data and the history data into a second cognition model, and outputting updated character cognition data.
And step A332, inputting the updated role cognition data and history data into the second dialogue engine model to generate virtual role speaking or white-out.
Wherein the second cognitive model and the second dialog engine model both employ a large language model and have been trained.
Step A40, updating the history data, and turning to step A10.
If new scenario content is generated, updating the historical scenario accordingly; if the user's speech, avatar speech, or bystanding is acquired, the historical dialog is updated accordingly.
In a preferred embodiment, control items may be added as new episodes of the episode are generated, directing the episode toward a particular direction. In this embodiment, step A20 may specifically include steps A25-A27:
and step A25, inputting the history data and the control items into a second scenario engine model to generate scenario contents of a new round.
And step A26, inputting the role cognition data, the scenario content and the history data into the first cognition model, and outputting updated role cognition data.
And step A27, inputting the updated role cognition data, scenario content and history data into the first dialogue engine model to generate virtual role speaking or whitish.
Wherein, control project can be set up in the background by the operator, include: continuity with history scenario (0-100 points), degree of influence of previous dialog (0-100 points), interestingness (0-100 points) and the like; the second scenario engine model, the first cognitive model and the first dialogue engine model all adopt large language models and are trained.
For example, in a western tour, the history scenario is that a grand monkey has beaten white bone essence, the history dialogue is that tangheng is blame for grand monkey that should not kill innocent, and grand monkey is not believed while explained.
When the next scenario is generated, if no control item is set, an abrupt scenario may be generated, for example: the white bone essence opens a crock Shang Fandian, and the spareribs soup is delicious, cheap and sufficient in price, so that the spareribs essence sounds the famous air, and the spareribs soup is happy from now on.
If the continuity of the control project with the history scenario is set to 100 points, the influence degree of the previous dialog is 100 points, the interestingness is 50 points, the following scenario can be generated: the eight rings of the pig come out of the round place, the grand monkey answers the face of the heart-washing innovation, the tangheng is happy and allowed, and the teachers and the apprentices are good.
If 100 points of continuity with the historical scenario in the control project are set, the influence degree of the previous dialog is 100 points, and the interestingness is 90 points, the following more interesting scenario can be generated: under the sun's anger, the back flower, fruit and mountain go on the leisure day, and the monk continues to suffer from the trouble.
In the embodiment of the invention, the role cognition data comprises two variables: variable 1 is the knowledge of itself by the avatar. For example, I are a gentle person, I have a patience with others, I like small animals, have friends to find I to play with a puppy, I have a happiness, etc. Variable 2 is a list that is the knowledge of each of the other roles by the virtual role. For example, character a's knowledge of character b: character b likes me well, but i dislikes character b because he is not careful about others and will appear clearly. Cognition of character c to character d: although not seen, hearing the first says that he is a boring person.
When the first dialog engine model and the second dialog engine model generate the speech of the virtual character, the two variables are input so as to influence the generation of the speech. When a virtual character cuts into a new scene or hears other characters to speak, the two variables are updated by using the first cognitive model or the second cognitive model, namely, the virtual character is influenced by the outside, and the cognition is changed.
Fig. 2 is a schematic diagram of main steps of training a scenario switching judgment model according to an embodiment of the present invention. As shown in fig. 2, the method for training the scenario switching judgment model in the present embodiment includes steps B10-B40:
and step B10, constructing a first training set.
Wherein the first training set comprises: a preset first number of training samples and corresponding switching probability labels; each training sample includes: a scenario segment and corresponding dialog; the handover probability label includes: the scenario segment switches probability realism values after 0, 1, 2, 3 and more than 3 conversations, respectively.
And step B20, selecting a training sample from the first training set, and inputting the training sample into the scenario switching judgment model to obtain a judgment result.
Wherein, the judging result comprises: probability predictors for scenario segments to switch after 0, 1, 2, 3, and more than 3 conversations, respectively.
And step B30, calculating a first loss function according to the judging result and the switching probability label, and adjusting model parameters.
Step B40, go to step B20 until training is completed when the first loss function is no longer decreasing.
For example, the switch probability label for training sample 1 is:
there are also 100% of more than 3 rounds of switching, 0% of 2 rounds of switching, 0% of 1 round of switching, and 0% of 0 round of switching.
The judgment result of the model output is:
there is also 80% of the probability of more than 3 rounds of switching, 10% of the probability of 3 rounds of switching, 5% of the probability of 2 rounds of switching, 4% of the probability of 1 round of switching, and 1% of the probability of 0 round of switching.
In this embodiment, the first scenario engine model adopts a large language model and has been trained, and the training method includes steps C10-C20:
and step C10, constructing a second training set.
Wherein the second training set comprises: a preset second number of training samples. Each training sample includes: a scenario segment and corresponding dialog, and a next scenario segment corresponding to the scenario segment.
And step C20, training the first scenario engine model based on the second training set.
In this embodiment, the second scenario engine model adopts a large language model, and has been trained, and the training method includes steps D10-D20:
and D10, constructing a third training set.
Wherein the third training set comprises: a preset third number of training samples. Each training sample includes: a scenario segment and corresponding dialogue, control item label, and the next scenario segment corresponding to the scenario segment; the control item tag includes: continuity with history scenario (0-100 points), degree of influence of previous dialog (0-100 points), interestingness (0-100 points), etc.
And step D20, training the second scenario engine model based on the third training set.
In the embodiment of the invention, besides the history dialogue, the scenario is input in the first dialogue engine model and the second dialogue engine model, so that the scenario can influence the role dialogue.
For example: if the previous scenario mentioned that the role played by the user just has grown a big sickness and the woman's principal angle is already known, then the newly generated woman's principal angle speaking may be "feel good in bar today? ", but not" do you go to attend with marathon race the following day? ".
FIG. 3 is a schematic diagram of the principal components of an embodiment of the data processing system of the present invention for self-propelled storyline and conversation. As shown in fig. 3, the system of the present embodiment includes: a judging module 10, a first generating module 20, a second generating module 30, and a control module 40.
The judging module 10 is used for inputting the historical data into a scenario switching judging model to judge whether the scenario needs to be switched; the first generation module 20 is configured to process the history data to generate a new scenario content and a virtual character speaking or whitish in case that scenario needs to be switched; the second generating module 30 is configured to process the history data to determine whether the user should speak, and obtain the user's speaking or generate a new virtual character speaking or a new side according to the determination result, if the scenario does not need to be switched; the control module 40 is used for updating the history data, and controlling the judging module and the first generating module 20/the second generating module 30 to repeatedly execute; the history data includes: historical scenario and historical dialogue.
Although the steps are described in the above-described sequential order in the above-described embodiments, it will be appreciated by those skilled in the art that in order to achieve the effects of the present embodiments, the steps need not be performed in such order, and may be performed simultaneously (in parallel) or in reverse order, and such simple variations are within the scope of the present invention.
Further, the invention also provides an embodiment of the computer readable storage device. The storage device of the present embodiment stores therein a computer program that can be loaded by a processor and execute the method as described above.
The computer readable storage device may include: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
Those of skill in the art will appreciate that the various illustrative method steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative components and steps have been described above generally in terms of functionality in order to clearly illustrate the interchangeability of electronic hardware and software. Whether such functionality is implemented as electronic hardware or software depends upon the particular application and design constraints imposed on the solution. Those skilled in the art may implement the described functionality using different approaches for each particular application, but such implementation is not intended to be limiting.
Thus far, the technical solution of the present invention has been described in connection with the preferred embodiments shown in the drawings. However, it will be readily appreciated by those skilled in the art that the scope of the invention is obviously not limited to these specific embodiments. Equivalent modifications and substitutions for related technical features may be made by those skilled in the art without departing from the principles of the present invention, and such modifications and substitutions will be within the scope of the present invention.
Claims (10)
1. A data processing method for automatically advancing a scenario and a dialogue, the method comprising:
judging whether the scenario needs to be switched according to the historical data;
if the scenario needs to be switched, processing the historical data to generate scenario content of a new round and virtual roles speaking or whitish;
if the scenario does not need to be switched, processing the historical data to judge whether the user should speak, and acquiring the user's speaking or generating a new round of virtual character speaking or bystanding according to the judging result;
updating the history data, and repeatedly executing the steps of judging whether the scenario needs to be switched and generating speaking, bystanding and/or scenario contents;
wherein the history data includes: historical scenario and historical dialogue.
2. A data processing method for automatically advancing a scenario and dialogue according to claim 1, wherein the step of processing the history data to generate a new scenario content and a virtual character speaking or blurriness includes:
inputting the history data into a first scenario engine model to generate a new scenario content;
inputting the role cognition data, the scenario content and the history data into a first cognition model, and outputting updated role cognition data;
inputting the updated character cognition data, the scenario content and the history data into a first dialogue engine model to generate the virtual character speaking or the bystander;
wherein the role awareness data includes: the virtual character recognizes itself and recognizes each of the other characters.
3. A data processing method for automatically advancing a scenario and dialogue according to claim 1, wherein the step of processing the history data to generate a new scenario content and a virtual character speaking or blurriness includes:
inputting the history data and the control items into a second scenario engine model to generate new scenario contents;
inputting the role cognition data, the scenario content and the history data into a first cognition model, and outputting updated role cognition data;
inputting the updated character cognition data, the scenario content and the history data into a first dialogue engine model to generate the virtual character speaking or the bystander;
wherein,
the control items are set by operators in the background, and comprise: continuity with historical scenario, degree of influence and interestingness from previous dialog.
4. The data processing method of self-propelled scenario and dialogue according to claim 1, wherein the step of processing the history data to determine whether or not the user should speak, and acquiring the user's speaking or generating a new round of virtual character speaking or bystanding according to the determination result comprises:
inputting the historical data into a speaking order judging model to judge whether the user should speak or not;
if yes, acquiring the speech of the user;
otherwise, processing the history data to generate a new round of virtual character speaking or white-over;
the speaking order judgment model adopts a large language model and is trained.
5. A data processing method for self-propelled scenario and dialogue according to claim 4, wherein the step of processing said history data to generate a new round of virtual character utterances or whitish "comprises:
inputting the character cognitive data and the historical data into a second cognitive model, and outputting updated character cognitive data;
and inputting the updated role cognition data and the history data into a second dialogue engine model to generate the virtual role speaking or the whitish.
6. The data processing method of self-propelled scenario and dialogue according to claim 1, wherein the step of judging whether the scenario needs to be switched according to the history data comprises:
inputting the historical data into a scenario switching judgment model, and respectively outputting the probability of scenario switching after 0, 1, 2, 3 and more than 3 times of dialogue;
if the probability of 0-round switching is greater than or equal to the preset probability or the probabilities of 1 round, 2 rounds, 3 rounds and more than 3 rounds of switching are all smaller than the probability of 0 round switching, judging that the scenario needs to be switched; otherwise, judging that the scenario does not need to be switched;
the scenario switching judgment model adopts a large language model and is trained, and the training method comprises the following steps:
constructing a first training set, the first training set comprising: a preset first number of training samples and corresponding switching probability labels; each training sample includes: a scenario segment and corresponding dialog; the handover probability tag includes: the scenario segment switches probability reality values after 0, 1, 2, 3 and more than 3 conversations respectively;
selecting a training sample from a first training set, and inputting the training sample into the scenario switching judgment model to obtain an output result; the output result comprises: the scenario segment switches probability prediction values after 0, 1, 2, 3 and more than 3 conversations respectively;
calculating a first loss function according to the output result and the switching probability label, and adjusting model parameters;
the execution is repeated until the first loss function is no longer decreasing.
7. The data processing method of self-propelled scenario and dialogue of claim 2, wherein the first scenario engine model employs a large language model and has been trained, the training method comprising:
constructing a second training set, the second training set comprising: a preset second number of training samples; each training sample includes: a scenario segment and corresponding dialogue, and the next scenario segment corresponding to the scenario segment;
training the first scenario engine model based on the second training set.
8. A data processing method for automatically advancing a scenario and dialogue according to claim 3, wherein the second scenario engine model adopts a large language model and has been trained, the training method comprising:
constructing a third training set, the third training set comprising: a preset third number of training samples; each training sample includes: a scenario segment and corresponding dialogue, control item label, and the next scenario segment corresponding to the scenario segment;
training the second scenario engine model based on the third training set;
wherein,
the control item tag includes: the consistency with the historical scenario, the degree of influence from the previous session, and the interestingness.
9. A data processing system for automatically advancing a scenario and dialogue, the system comprising:
the judging module is used for inputting the historical data into the scenario switching judging model and judging whether the scenario needs to be switched or not;
the first generation module is used for processing the historical data to generate a new scenario content and a virtual character speaking or whitish under the condition that the scenario needs to be switched;
the second generation module is used for processing the historical data to judge whether the user should speak or not under the condition that the scenario does not need to be switched, and acquiring the user's speaking or generating a new round of virtual character speaking or bystanding according to the judging result;
the control module is used for updating the historical data and controlling the judging module and the first generating module/the second generating module to repeatedly execute;
wherein the history data includes: historical scenario and historical dialogue.
10. A computer readable storage device storing a computer program capable of being loaded by a processor and performing the method according to any of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410051347.2A CN117573859A (en) | 2024-01-15 | 2024-01-15 | Data processing method, system and equipment for automatically advancing scenario and dialogue |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410051347.2A CN117573859A (en) | 2024-01-15 | 2024-01-15 | Data processing method, system and equipment for automatically advancing scenario and dialogue |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117573859A true CN117573859A (en) | 2024-02-20 |
Family
ID=89864592
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410051347.2A Pending CN117573859A (en) | 2024-01-15 | 2024-01-15 | Data processing method, system and equipment for automatically advancing scenario and dialogue |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117573859A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120190456A1 (en) * | 2011-01-21 | 2012-07-26 | Rogers Henk B | Systems and methods for providing an interactive multiplayer story |
| CN112843724A (en) * | 2021-01-18 | 2021-05-28 | 浙江大学 | Game scenario display control method and device, electronic equipment and storage medium |
| CN113391865A (en) * | 2021-06-10 | 2021-09-14 | 广州繁星互娱信息科技有限公司 | Plot chatting method, plot chatting device, plot chatting terminal, plot chatting server and storage medium |
| WO2022180704A1 (en) * | 2021-02-24 | 2022-09-01 | 日本電信電話株式会社 | Display control device, display control method, and display control program |
| CN115779446A (en) * | 2023-01-29 | 2023-03-14 | 深圳市人马互动科技有限公司 | Data processing method based on game task and related product |
| CN116340752A (en) * | 2023-02-23 | 2023-06-27 | 中国人民大学 | Predictive analysis result-oriented data story generation method and system |
| CN116637375A (en) * | 2023-01-30 | 2023-08-25 | 深圳市人马互动科技有限公司 | Intelligent scenario generation method and device |
-
2024
- 2024-01-15 CN CN202410051347.2A patent/CN117573859A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120190456A1 (en) * | 2011-01-21 | 2012-07-26 | Rogers Henk B | Systems and methods for providing an interactive multiplayer story |
| CN112843724A (en) * | 2021-01-18 | 2021-05-28 | 浙江大学 | Game scenario display control method and device, electronic equipment and storage medium |
| WO2022180704A1 (en) * | 2021-02-24 | 2022-09-01 | 日本電信電話株式会社 | Display control device, display control method, and display control program |
| CN113391865A (en) * | 2021-06-10 | 2021-09-14 | 广州繁星互娱信息科技有限公司 | Plot chatting method, plot chatting device, plot chatting terminal, plot chatting server and storage medium |
| CN115779446A (en) * | 2023-01-29 | 2023-03-14 | 深圳市人马互动科技有限公司 | Data processing method based on game task and related product |
| CN116637375A (en) * | 2023-01-30 | 2023-08-25 | 深圳市人马互动科技有限公司 | Intelligent scenario generation method and device |
| CN116340752A (en) * | 2023-02-23 | 2023-06-27 | 中国人民大学 | Predictive analysis result-oriented data story generation method and system |
Non-Patent Citations (2)
| Title |
|---|
| RUNCONG ZHAO等: "NarrativePlay:Interactive Narrative Understanding", 《ARXIV》, 2 October 2023 (2023-10-02), pages 1 - 12 * |
| 何陆潇涵: "单源响应式框架下安卓交互式视频制作平台的设计与实现", 《中国优秀硕士学位论文全文数据库信息科技辑》, no. 04, 15 April 2022 (2022-04-15), pages 138 - 902 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sterelny | Language, gesture, skill: the co-evolutionary foundations of language | |
| WO2021169431A1 (en) | Interaction method and apparatus, and electronic device and storage medium | |
| US6526395B1 (en) | Application of personality models and interaction with synthetic characters in a computing system | |
| JP6719747B2 (en) | Interactive method, interactive system, interactive device, and program | |
| Yap et al. | Individual differences in the joint effects of semantic priming and word frequency revealed by RT distributional analyses: The role of lexical integrity | |
| CN112818107B (en) | Conversation robot for daily life and chat method thereof | |
| CN116009748B (en) | Picture information interaction method and device in children interaction story | |
| US11232789B2 (en) | Dialogue establishing utterances without content words | |
| CN105798918A (en) | Interactive method and device for intelligent robot | |
| JP6699010B2 (en) | Dialogue method, dialogue system, dialogue device, and program | |
| CN115356953B (en) | Virtual robot decision method, system and electronic equipment | |
| CN112530218A (en) | Many-to-one accompanying intelligent teaching system and teaching method | |
| CN115348458A (en) | Virtual live broadcast control method and system | |
| CN113617036A (en) | Game dialogue processing method, device, equipment and storage medium | |
| CN117573859A (en) | Data processing method, system and equipment for automatically advancing scenario and dialogue | |
| CN115293132B (en) | Dialog of virtual scenes a treatment method device, electronic apparatus, and storage medium | |
| CN116975214A (en) | Text generation method, device, storage medium and computer equipment | |
| CN116312456A (en) | Voice dialogue script generation method and device and electronic equipment | |
| CN114422844B (en) | Barrage material generation method, recommendation method, device, equipment, medium and product | |
| CN110111793A (en) | Processing method, device, storage medium and the electronic device of audio-frequency information | |
| CN115424605A (en) | Speech synthesis method, speech synthesis device, electronic equipment and computer-readable storage medium | |
| KR102073386B1 (en) | Game device and method for playing game using the game device | |
| Hasani et al. | Immersive experience with non-player characters dynamic dialogue | |
| CN116226411B (en) | Interactive information processing method and device for interactive project based on animation | |
| CN112927692A (en) | Automatic language interaction method, device, equipment and medium |
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 |