CN117910481A - Spoken language dialogue method and device for assisting language learning and dialogue robot - Google Patents
Spoken language dialogue method and device for assisting language learning and dialogue robot Download PDFInfo
- Publication number
- CN117910481A CN117910481A CN202410317012.0A CN202410317012A CN117910481A CN 117910481 A CN117910481 A CN 117910481A CN 202410317012 A CN202410317012 A CN 202410317012A CN 117910481 A CN117910481 A CN 117910481A
- Authority
- CN
- China
- Prior art keywords
- language
- dialogue
- spoken
- agent
- language model
- 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
- 238000000034 method Methods 0.000 title claims abstract description 53
- 235000002198 Annona diversifolia Nutrition 0.000 claims description 11
- 241000282842 Lama glama Species 0.000 claims description 11
- 238000010276 construction Methods 0.000 claims description 9
- 238000013461 design Methods 0.000 claims description 8
- 238000012360 testing method Methods 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 5
- 238000004088 simulation Methods 0.000 claims description 5
- 238000012549 training Methods 0.000 claims description 5
- QVFWZNCVPCJQOP-UHFFFAOYSA-N chloralodol Chemical compound CC(O)(C)CC(C)OC(O)C(Cl)(Cl)Cl QVFWZNCVPCJQOP-UHFFFAOYSA-N 0.000 claims 1
- 230000000694 effects Effects 0.000 abstract description 4
- 238000003058 natural language processing Methods 0.000 abstract description 2
- 238000004891 communication Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000006978 adaptation Effects 0.000 description 3
- 230000009977 dual effect Effects 0.000 description 3
- 230000003993 interaction Effects 0.000 description 3
- 238000004590 computer program Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000013519 translation Methods 0.000 description 2
- 241000282840 Vicugna vicugna Species 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Landscapes
- Electrically Operated Instructional Devices (AREA)
Abstract
The invention discloses a spoken language dialogue method and device for assisting language learning and a dialogue robot, belonging to the technical field of natural language processing, wherein the method comprises the following steps: constructing a spoken dialog data set; constructing a dialogue language model for assisting language learning; performing instruction fine tuning on the dialogue language model by using the spoken dialogue data set to obtain a target language model; constructing an agent framework for a spoken language teaching task; setting role tasks for each agent in the agent framework; and in the spoken dialogue process, generating spoken dialogue text according to the role tasks corresponding to the intelligent agents in the intelligent agent framework and the target language model. The spoken language dialogue scheme for assisting language learning can improve the language learning effect and flexibly use the assisting language learning of various scenes.
Description
Technical Field
The invention relates to the technical field of natural language processing, in particular to a spoken language dialogue method and device for assisting language learning and a dialogue robot.
Background
For language learners, language learning is an exercise of a non-open language dialogue. The spoken language dialogue exercise can help language learners flexibly use the learned grammar and vocabulary knowledge to deepen understanding and memory of the language learners, and meanwhile, the spoken language dialogue exercise can comprehensively improve the listening, speaking, reading, understanding and expression abilities of the language learners, improve the communication technology and expression fluency of the language learners and improve the enthusiasm and self-confidence of the language learners.
The conventional language learning method relies on fixed teaching materials and standard courses, and it is difficult to provide personalized language teaching services according to the language proficiency of a language learner and a language learning goal. Meanwhile, most language learners have difficulty in conditionally obtaining real dialogue experience, and cannot have timely feedback and correction to help the language learners to quickly improve language expression capability. Therefore, the traditional language learning scheme has poor learning effect and strong limitation.
Disclosure of Invention
The embodiment of the invention aims to provide a spoken language dialogue method and device for assisting language learning and a dialogue robot, which can solve the problems of poor language learning effect and strong limitation in the prior art.
In order to solve the technical problems, the invention provides the following technical scheme:
In one aspect, there is provided a spoken dialog method for assisting language learning, the method comprising:
constructing a spoken dialog data set, wherein each piece of data in the spoken dialog data set contains multiple rounds of dialog text suitable for a language learner;
constructing a dialogue language model for assisting language learning;
Performing instruction fine tuning on the dialogue language model by using the spoken dialogue data set to obtain a target language model, wherein the target language model is used for generating dialogue texts for assisting language learning;
constructing an agent framework for a spoken language teaching task, wherein the agent framework comprises at least two agents;
Setting role tasks for each agent in the agent framework;
And in the spoken dialogue process, generating spoken dialogue text according to the role tasks corresponding to the intelligent agents in the intelligent agent framework and the target language model.
Optionally, the step of constructing the spoken dialog data set includes:
automatically generating each piece of data in the spoken dialog data set by taking the teaching textbook as an original corpus through a large language model;
the teaching textbook comprises a college English teaching textbook and a primary school English teaching textbook.
Optionally, the step of automatically generating each piece of data in the spoken dialog data set by using the large language model and the teaching textbook as the original corpus includes:
extracting dialogue theme from each unit of teaching textbook;
inputting the theme into a first large language model, and generating dialogue data which is related to the theme and accords with the textbook difficulty through a first preset prompt;
and inputting the dialogue data generated by the first large language model into a second large language model, and enabling the dialogue data to be subjected to dialogue through a second preset prompt to generate a multi-round dialogue text.
Optionally, the step of constructing a dialogue language model for assisting language learning includes:
the LLaMA large language model is configured as a self-attention network based decoder model to generate a conversational language model.
Optionally, the step of performing instruction fine tuning on the dialogue language model by using the spoken dialogue data set to obtain a target language model includes:
And based on the spoken dialogue data set, performing low-rank approximation on the weight matrix in the LLaMA large language model by adopting a LoRA method to obtain a target language model.
Optionally, when the agent frame is a dual agent frame, the agent frame includes a coach agent and a teacher agent, or the agent frame includes a teacher agent and a student agent;
The training agent is used for teaching according to the language proficiency and the language learning target of the language learner; the teacher agent is used for carrying out dialogue practice simulation spoken language teaching; the student agent is used to play a language learner.
Optionally, when the agent frame is a multi-agent frame, the agent frame includes: course design module, teaching module, thinking back module and test module, every module in the agent frame corresponds an agent.
In another aspect, there is provided a spoken dialog device for assisting language learning, comprising:
A first construction module for constructing a spoken dialog data set, wherein each piece of data in the spoken dialog data set contains a plurality of rounds of dialog text suitable for a language learner;
The second construction module is used for constructing a dialogue language model for assisting language learning;
the fine tuning module is used for carrying out instruction fine tuning on the dialogue language model by using the spoken dialogue data set to obtain a target language model, wherein the target language model is used for generating dialogue texts for assisting language learning;
The third construction module is used for constructing an agent frame for a spoken language teaching task, wherein the agent frame comprises at least two agents;
the setting module is used for setting role tasks for the intelligent agents in the intelligent agent framework;
And the control module is used for generating a spoken dialogue text according to the role tasks corresponding to the intelligent agents in the intelligent agent framework and the target language model in the spoken dialogue process.
In another aspect, a conversation robot is provided, the conversation robot including a processor, a memory, and a program or instructions stored on the memory and executable on the processor, the program or instructions when executed by the processor implementing any of the steps of the spoken conversation method for assisting language learning described above.
The embodiment of the invention provides a spoken language dialogue scheme for assisting language learning, and a spoken language dialogue data set is constructed; constructing a dialogue language model for assisting language learning; performing instruction fine adjustment by using the spoken dialogue data set dialogue language model to obtain a target language model; constructing an agent framework for a spoken language teaching task; setting role tasks for each agent in the agent framework; in the spoken dialogue process, a spoken dialogue text is generated according to the role task and the target language model corresponding to each agent in the agent framework. On the one hand, the spoken dialogue robot for assisting the language learning can provide personalized learning content and feedback according to the language proficiency, the language learning target and the learning style of the language learner, so that the language learning is more suitable for the level, the interest and the learning requirement of the language learner, and the assisting language learning has stronger flexibility; in the second aspect, the target language module simulates a real and natural dialogue scene, so that rich dialogue experience is provided for a language learner, and the language communication capability of the language learner can be improved; in the third aspect, the target language model can generate natural, real and smooth texts by virtue of strong learning ability, language understanding ability and generating ability, so that a spoken language dialogue is close to a real scene, a language learner is helped to obtain real dialogue experience, and language skills are improved.
Drawings
FIG. 1 is a flow chart of steps of a spoken dialog method for assisting language learning, provided by an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of an agent framework for a spoken language teaching task, according to an embodiment of the present invention;
Fig. 3 is a block diagram of a spoken dialog device for assisting language learning according to an embodiment of the present invention.
Detailed Description
In order to make the technical problems, technical solutions and advantages to be solved more apparent, the following detailed description will be given with reference to the accompanying drawings and specific embodiments.
The spoken dialog scheme for assisting language learning provided by the embodiment of the invention is described in detail below through specific embodiments and application scenarios thereof with reference to the accompanying drawings.
As shown in fig. 1, a spoken dialog method for assisting language learning according to an embodiment of the present invention includes the steps of:
step 101: a spoken dialog data set is constructed.
Wherein each piece of data in the spoken dialog data set contains multiple rounds of dialog text suitable for a language learner. Dialogs with language learners by the spoken dialog robot include text dialogs and spoken dialogs.
Preferably, each of the multiple rounds of dialog text in the dialog data set for the language learner contains 4-6 rounds of dialog. Of course, the number of the dialogues included in a piece of data is not limited to 4-6, and may be 3, 7, etc., and the number of the dialogues specifically included may be flexibly set by those skilled in the art, which is not particularly limited in the embodiment of the present invention.
The spoken dialogue method for assisting language learning provided by the embodiment of the invention can be applied to a dialogue robot or an intelligent module, a computer program corresponding to the spoken dialogue method for assisting language learning is stored in a storage medium of the dialogue robot or the intelligent module, and a processor executes the computer program to assist language learning. The intelligent module may be a spoken dialog device for assisting language learning.
One way to optionally construct a spoken dialog data set may be as follows:
Automatically generating each piece of data in the spoken dialog data set by taking the teaching textbook as an original corpus through a large language model; the teaching textbook comprises a college English teaching textbook and a primary school English teaching textbook.
In the actual implementation process, the original corpus for college English teaching in the spoken dialogue data set is automatically generated by a college language model based on the text book of the New college English visual listening and speaking version 3, and the original corpus for middle and primary school English teaching is automatically generated by the college language model based on the English learning auxiliary report.
In an alternative embodiment, the manner of automatically generating each piece of data in the spoken dialog data set from the teaching textbook to the original corpus by means of the large language model may comprise the following sub-steps:
sub-step 1: extracting dialogue theme from each unit of teaching textbook;
The number of the units contained in the teaching textbook is determined according to the specific content of the teaching textbook, for example, eight units can be contained in the teaching textbook, and then dialogue topics are extracted from the eight units of the teaching textbook.
Sub-step 2: inputting the theme into a first large language model, and generating dialogue data which is related to the theme and accords with the textbook difficulty through a first preset prompt;
Sub-step 3: and inputting dialogue data generated by the first large language model into the second large language model, and enabling the dialogue data to be subjected to dialogue through a second preset prompt to generate multi-round dialogue texts.
The first preset hint may be set "Remember you are a Chinese college student, please give a one-sentence greeting base the following topic" to correspond to a chinese translation as "please greetings for a sentence based on the following subject matter".
The second preset hint may be set to "Remember you are a Chinese college student, please give a one-sentence respond for the following sentences" for chinese translation to "remember that you are a chinese college student, please respond to a sentence below.
In a practical implementation, the large language model may be ChatGPT, chatGPT, an abbreviation of CHAT GENERATIVE PRE-trained Transformer, a chat bot developed at OpenAI.
Step 102: a dialogue language model for assisting language learning is constructed.
The conversational language model may be generated by:
The LLaMA large language model is configured to generate a dialog language model (transducer) based on a decoder model of a Self-Attention (Self-Attention) network. The dialogue language model is built based on LLaMA large language model. LLaMA is a large language model of Meta open source, is a basic model of a series of open source models, and comprises well-known vicuna series, longChat series and the like which are obtained by fine tuning based on the models. The LLaMA large language model is then an upgraded version of the LLaMA large language model.
In an alternative embodiment, the spoken dialog data set is used to fine-tune the dialog language model in such a way that the target language model is obtained as follows:
Based on the spoken dialogue dataset, a LoRA method is adopted to perform low-rank approximation on the weight matrix in the LLaMA large language model to obtain the target language model.
The Llama2 large language model adopts LoRA (Low-Rank Adaptation) method to carry out fine adjustment so as to improve the efficiency of parameter adjustment and the performance of the model on specific spoken language teaching tasks. The LoRA method realizes effective task-specific adjustment of the model under the condition of not significantly increasing computational burden by performing low-rank approximation on the weight matrix inside the model.
Step 103: and performing instruction fine tuning on the dialogue language model by using the spoken dialogue data set to obtain a target language model.
Wherein the target language model is used to generate dialogue text that assists in language learning.
The instruction fine tuning is specially aimed at the scenes of the oral English teaching, and parameters of the language model are adjusted to optimize oral reaction and guidance for English learners. The target language model obtained by fine adjustment can understand and generate natural language to communicate natural language, and is suitable for the spoken language teaching task and good in performance.
Parameters for fine tuning include, but are not limited to: max_new_ tokens:1024 longest input; temperature: 0.6 temperature; the answer probability returned by top_p is greater than 0.9; top_k 50 returns answer front 50. It should be noted that, the latter three parameters are all related to randomness, and in the actual implementation process, a person skilled in the art can flexibly adjust specific values of the fine tuning parameters according to actual requirements.
Step 104: an agent framework for spoken language teaching tasks is constructed.
Wherein, include at least two agent in the agent frame. A schematic structural diagram of an agent framework suitable for a spoken language teaching task is shown in FIG. 2.
When the agent frame is a dual agent frame, the agent frame may include a coach (Instructor) agent and a teacher (Teacher) agent, or the agent frame may include a teacher (Teacher) agent and a Student (Student) agent.
The training agent is used for teaching and designing according to the language proficiency and the language learning target of the language learner; the teacher intelligent body is used for carrying out dialogue practice simulation spoken language teaching; student agents are used to impersonate language learners.
And all the intelligent agents cooperate to complete the spoken language teaching task through information synchronization. Specifically, the Instructor agent carries out teaching design to instruct Teacher the agent to carry out spoken language teaching, the Student agent carries out dialogue exercise with Teacher agent, the simulation spoken language teaching scene helps Teacher agent to proficiency in teaching process.
When the agent frame is a multi-agent frame, the agent frame includes: course design module, teaching module, thinking back module and test module, every module in the agent frame corresponds an agent. Wherein, course design module can be equivalent to instruction (Instructor) agent, and teaching module can be equivalent to teacher (Teacher) agent.
Step 105: role tasks are set for each agent in the agent framework.
In this step, the promt (the promt may also be referred to as a task or a prompt) of each role played by the agent in the agent framework is designed, and the agent framework is constructed using the promt corresponding to each agent.
The prompt employed by Teacher agent may be "as shown above is a session between a student and a teacher, presenting the subject of the session and advice of your teaching assistant.As an english teacher, you need to ensure that your teaching assistant's advice is followed in the interaction with the student. /(I)In addition, teachers are encouraged to provide positive feedback, gradually guide students, and end up with a profound problem. /(I)Please answer using a piece of text. Your identity is not emphasized. Remember that you are a teacher and only a student.
The sample employed by the Student agent may be "as shown above is a session between a Student and a teacher. Bearing in mind that your character is a Chinese student who is learning English. You may make some grammatical errors and use only some simple sentences.
The prompt employed by Teacher agent may be "as shown above is a session between a student and a teacher. As an english teacher, you should provide positive feedback, guide the student step by step, and end up with a profound question.Please answer using a piece of text. Your identity is not emphasized. Remember that you are a teacher and only a student.
Step 106: in the spoken dialogue process, a spoken dialogue text is generated according to the role task and the target language model corresponding to each agent in the agent framework.
The intelligent body framework after configuration comprises demand analysis, teaching design and teaching implementation, a set of spoken language teaching flow with definite targets and high system efficiency is formed, and personalized spoken language teaching service can be realized according to different language learning demands of language learners. The method can effectively solve the problems that the prior art can not simulate natural and real dialogue and can not realize personalized spoken language teaching service according to different language learning requirements of language learners during auxiliary language learning.
Taking the intelligent body frame as an example, when auxiliary language learning is carried out, a teaching plan can be formulated in advance for a teacher in the course design module, the teaching module can carry out auxiliary teaching and dialogue interaction with a language learner according to the formulated teaching plan, the thinking returning module evaluates the performances of the language learner and the spoken language dialogue robot in the dialogue interaction and carries out thinking returning on the defects of the dialogue robot, and the test module gives test questions according to the level of the language learner to evaluate whether the language learner grasps the learned knowledge. In the embodiment of the invention, the spoken language dialogue robot is taken as an example for assisting English teaching, and the spoken language dialogue robot can be also suitable for Chinese teaching scenes.
The spoken language dialogue method for assisting language learning provided by the embodiment of the invention constructs a spoken language dialogue data set; constructing a dialogue language model for assisting language learning; performing instruction fine adjustment by using the spoken dialogue data set dialogue language model to obtain a target language model; constructing an agent framework for a spoken language teaching task; setting role tasks for each agent in the agent framework; in the spoken dialogue process, a spoken dialogue text is generated according to the role task and the target language model corresponding to each agent in the agent framework. On the one hand, the spoken dialogue method for assisting the language learning can provide personalized learning content and feedback according to the language proficiency, the language learning target and the learning style of the language learner, so that the language learning is more suitable for the level, the interest and the learning requirement of the language learner, and the assisting language learning has stronger flexibility; in the second aspect, the target language module simulates a real and natural dialogue scene, so that rich dialogue experience is provided for a language learner, and the language communication capability of the language learner can be improved; in the third aspect, the target language model can generate natural, real and smooth texts by virtue of strong learning ability, language understanding ability and generating ability, so that a spoken language dialogue is close to a real scene, a language learner is helped to obtain real dialogue experience, and language skills are improved.
Fig. 3 is a block diagram of a spoken dialog device for assisting language learning, in which an embodiment of the invention is implemented.
The spoken language dialogue device for assisting language learning provided by the embodiment of the invention comprises the following functional modules:
A first construction module 301, configured to construct a spoken dialog data set, where each piece of data in the spoken dialog data set includes a plurality of rounds of dialog text suitable for a language learner;
A second construction module 302, configured to construct a dialogue language model for assisting language learning;
A fine tuning module 303, configured to perform instruction fine tuning on the dialogue language model using the spoken dialogue data set to obtain a target language model, where the target language model is used to generate dialogue text for assisting language learning;
A third construction module 304, configured to construct an agent framework for a spoken language teaching task, where the agent framework includes at least two agents;
a setting module 305, configured to set role tasks for each agent in the agent framework;
and the control module 306 is configured to generate a spoken dialogue text according to the role task corresponding to each agent in the agent framework and the target language model in the spoken dialogue process.
Optionally, the first building module is specifically configured to: automatically generating each piece of data in the spoken dialog data set by taking the teaching textbook as an original corpus through a large language model;
the teaching textbook comprises a college English teaching textbook and a primary school English teaching textbook.
Optionally, the first building module includes:
The first sub-module is used for extracting dialogue topics from all units of the teaching textbook;
The second sub-module is used for inputting the theme into the first large language model, and generating dialogue data which is related to the theme and accords with the textbook difficulty through a first preset prompt;
and the third sub-module is used for inputting the dialogue data generated by the first large language model into the second large language model, and making the dialogue through a second preset prompt to generate a plurality of rounds of dialogue texts.
Optionally, the dialog language model is generated by:
the LLaMA large language model is configured to generate a conversational language model based on the decoder model of the self-attention network.
Optionally, the fine tuning module is specifically configured to perform low-rank approximation on the weight matrix in the LLaMA2 large language model by using a LoRA method based on the spoken dialogue dataset, so as to obtain a target language model.
Optionally, when the agent frame is a dual agent frame, the agent frame includes a coach agent and a teacher agent, or the agent frame includes a teacher agent and a student agent;
The training agent is used for teaching according to the language proficiency and the language learning target of the language learner; the teacher agent is used for carrying out dialogue practice simulation spoken language teaching; the student agent is used to play a language learner.
Optionally, when the agent frame is a multi-agent frame, the agent frame includes: course design module, teaching module, thinking back module and test module, every module in the agent frame corresponds an agent.
According to the spoken dialogue device for assisting language learning, provided by the embodiment of the invention, on one hand, the spoken dialogue device for assisting language learning can provide personalized learning content and feedback according to the language proficiency, the language learning target and the learning style of a language learner, so that the language learning is more suitable for the level, the interest and the learning requirement of the language learner, and the assisting language learning flexibility is higher; in the second aspect, the target language module simulates a real and natural dialogue scene, so that rich dialogue experience is provided for a language learner, and the language communication capability of the language learner can be improved; in the third aspect, the target language model can generate natural, real and smooth texts by virtue of strong learning ability, language understanding ability and generating ability, so that a spoken language dialogue is close to a real scene, a language learner is helped to obtain real dialogue experience, and language skills are improved.
The spoken dialog device for assisting language learning shown in fig. 3 provided by the embodiment of the present invention can implement each process implemented by the method embodiment of fig. 1, and in order to avoid repetition, a description is omitted here.
Optionally, the embodiment of the present invention further provides a conversation robot, including a processor, a memory, and a program or an instruction stored in the memory and capable of running on the processor, where the program or the instruction when executed by the processor implements each process executed by the spoken dialogue device for assisting language learning, and the process can achieve the same technical effect, so that repetition is avoided, and details are not repeated here.
Wherein, the processor is the processor in the dialogue robot described in the above embodiment. The readable storage medium includes a computer readable storage medium such as a read-only memory (ROM), a random access memory (Random Access Memory, RAM), a magnetic disk or an optical disk, and the like.
The conversation robot may be a device having an operating system. The operating system may be an Android operating system, an iOS operating system, or other possible operating systems, and the embodiment of the present invention is not limited specifically.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
While the foregoing is directed to the preferred embodiments of the present invention, it will be appreciated by those skilled in the art that various modifications and adaptations can be made without departing from the principles of the present invention, and such modifications and adaptations are intended to be comprehended within the scope of the present invention.
Claims (9)
1. A spoken dialog method for assisting language learning, comprising:
constructing a spoken dialog data set, wherein each piece of data in the spoken dialog data set contains multiple rounds of dialog text suitable for a language learner;
constructing a dialogue language model for assisting language learning;
Performing instruction fine tuning on the dialogue language model by using the spoken dialogue data set to obtain a target language model, wherein the target language model is used for generating dialogue texts for assisting language learning;
constructing an agent framework for a spoken language teaching task, wherein the agent framework comprises at least two agents;
Setting role tasks for each agent in the agent framework;
And in the spoken dialogue process, generating spoken dialogue text according to the role tasks corresponding to the intelligent agents in the intelligent agent framework and the target language model.
2. The method of claim 1, wherein the step of constructing a spoken dialog data set comprises:
automatically generating each piece of data in the spoken dialog data set by taking the teaching textbook as an original corpus through a large language model;
the teaching textbook comprises a college English teaching textbook and a primary school English teaching textbook.
3. The method of claim 2, wherein the step of automatically generating each piece of data in the spoken dialog data set from the teaching textbook as the original corpus by the large language model comprises:
extracting dialogue theme from each unit of teaching textbook;
inputting the theme into a first large language model, and generating dialogue data which is related to the theme and accords with the textbook difficulty through a first preset prompt;
and inputting the dialogue data generated by the first large language model into a second large language model, and enabling the dialogue data to be subjected to dialogue through a second preset prompt to generate a multi-round dialogue text.
4. The method of claim 1, wherein the step of constructing a dialogue language model for assisting language learning comprises:
the LLaMA large language model is configured as a self-attention network based decoder model to generate a conversational language model.
5. The method of claim 4, wherein the step of using the spoken dialog data set to fine tune the dialog language model to obtain a target language model, comprises:
And based on the spoken dialogue data set, performing low-rank approximation on the weight matrix in the LLaMA large language model by adopting a LoRA method to obtain a target language model.
6. The method according to claim 1, characterized in that:
When the intelligent body frame is a double intelligent body frame, the intelligent body frame comprises a training intelligent body and a teacher intelligent body, or the intelligent body frame comprises a teacher intelligent body and a student intelligent body;
The training agent is used for teaching according to the language proficiency and the language learning target of the language learner; the teacher agent is used for carrying out dialogue practice simulation spoken language teaching; the student agent is used to play a language learner.
7. The method of claim 1, wherein when the agent frame is a multi-agent frame, the agent frame comprises: course design module, teaching module, thinking back module and test module, every module in the agent frame corresponds an agent.
8. A spoken dialog device for assisting language learning, comprising:
A first construction module for constructing a spoken dialog data set, wherein each piece of data in the spoken dialog data set contains a plurality of rounds of dialog text suitable for a language learner;
The second construction module is used for constructing a dialogue language model for assisting language learning;
the fine tuning module is used for carrying out instruction fine tuning on the dialogue language model by using the spoken dialogue data set to obtain a target language model, wherein the target language model is used for generating dialogue texts for assisting language learning;
The third construction module is used for constructing an agent frame for a spoken language teaching task, wherein the agent frame comprises at least two agents;
the setting module is used for setting role tasks for the intelligent agents in the intelligent agent framework;
And the control module is used for generating a spoken dialogue text according to the role tasks corresponding to the intelligent agents in the intelligent agent framework and the target language model in the spoken dialogue process.
9. A conversation robot comprising a processor, a memory, and a program or instruction stored on the memory and executable on the processor, which when executed by the processor, implements the spoken conversation method for assisting language learning of any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410317012.0A CN117910481A (en) | 2024-03-20 | 2024-03-20 | Spoken language dialogue method and device for assisting language learning and dialogue robot |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410317012.0A CN117910481A (en) | 2024-03-20 | 2024-03-20 | Spoken language dialogue method and device for assisting language learning and dialogue robot |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117910481A true CN117910481A (en) | 2024-04-19 |
Family
ID=90692642
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410317012.0A Pending CN117910481A (en) | 2024-03-20 | 2024-03-20 | Spoken language dialogue method and device for assisting language learning and dialogue robot |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117910481A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050053900A1 (en) * | 2003-09-05 | 2005-03-10 | Steven Kaufmann | Method of teaching a foreign language to a student providing measurement in a context based learning system |
| CN114595923A (en) * | 2022-01-11 | 2022-06-07 | 电子科技大学 | Group teaching recommendation system based on deep reinforcement learning |
| CN116993544A (en) * | 2023-04-28 | 2023-11-03 | 新大陆数字技术股份有限公司 | LLM-based auxiliary teaching method |
| CN117290488A (en) * | 2023-10-31 | 2023-12-26 | 安徽十锎信息科技有限公司 | Man-machine interaction method and device based on large model, electronic equipment and storage medium |
| CN117350325A (en) * | 2023-10-18 | 2024-01-05 | 华东师范大学 | Universal and configurable multi-agent interaction framework |
-
2024
- 2024-03-20 CN CN202410317012.0A patent/CN117910481A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050053900A1 (en) * | 2003-09-05 | 2005-03-10 | Steven Kaufmann | Method of teaching a foreign language to a student providing measurement in a context based learning system |
| CN114595923A (en) * | 2022-01-11 | 2022-06-07 | 电子科技大学 | Group teaching recommendation system based on deep reinforcement learning |
| CN116993544A (en) * | 2023-04-28 | 2023-11-03 | 新大陆数字技术股份有限公司 | LLM-based auxiliary teaching method |
| CN117350325A (en) * | 2023-10-18 | 2024-01-05 | 华东师范大学 | Universal and configurable multi-agent interaction framework |
| CN117290488A (en) * | 2023-10-31 | 2023-12-26 | 安徽十锎信息科技有限公司 | Man-machine interaction method and device based on large model, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yurko et al. | Learning English grammar online: the main resources | |
| JP2018180299A (en) | Word learning training system | |
| Anggraini | IMPROVING STUDENTS’PRONUNCIATION SKILL USING ELSA SPEAK APPLICATION | |
| Al-Jarf | Grammar Podcasts for ESL College Students. | |
| Suseno | Using the Podcast Contents to Help EFL Students in Improving Their Grammar and Pronunciation | |
| Suwarno et al. | EXPLORING TEACHER'S IMPLEMENTATION OF AUDIO LINGUAL METHOD, CHALLENGES, AND TECHNIQUES FOR IMPROVING STUDENT'S VOCABULARY MASTERY | |
| Kennedy | An exploration of when and how New Zealand secondary school teachers introduce Chinese characters to beginners | |
| Jia et al. | Motivate the learners to practice English through playing with chatbot CSIEC | |
| Santoso et al. | Prototype development and usability evaluation of a mobile-based Arabic language learning application | |
| CN117910481A (en) | Spoken language dialogue method and device for assisting language learning and dialogue robot | |
| KR102395702B1 (en) | Method for providing english education service using step-by-step expanding sentence structure unit | |
| Marchenko et al. | Electronic resources for teaching Ukrainian as a second language | |
| Satriani et al. | English Training Using Birkenbihl Method to Improve Receptive Skills of Grade VII Students at MTs Putri As' adiyah | |
| Strik et al. | Development and Integration of Speech technology into COurseware for language learning: the DISCO project | |
| Shintasiwi et al. | Analysis of English-Speaking Performance for the Effective Language Learning in Digital Era | |
| AlSaleem | Computers in teaching English language | |
| Watanabe et al. | Investigating the Effect of Chinese Pronunciation Teaching Materials Using Speech Recognition and Synthesis Functions. | |
| JP2002278441A (en) | Foreign language conversation teaching material for private lessen | |
| Seneff et al. | Spoken dialogue systems for language learning | |
| Syahputri et al. | Instructional Methods in Speaking Classes: Reviewing Literature | |
| Lemnaru | ROLE OF ARTIFICIAL INTELLIGENCE IN LEARNING ROMANIAN AS A SECOND LANGUAGE FOR INTERNATIONAL STUDENTS | |
| Gulamovna | THE IMPORTANCE OF INTEGRATED SKILLS IN TEACHING ENGLISH FOR EFL LEARNERS | |
| Sun | How to teach phonics to EFL students in primary schools in China | |
| 中西弘 | Effects of Prosody Shadowing on Japanese EFL Learners’ Processing of Object Relative Clauses in English | |
| Arapbayeva et al. | AUDIO-LINGUAL METHOD AND TEACHER AND STUDENT ROLES |
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 |