Method, device and program product for implementing travel assistant based on large language model
Through a travel assistant based on a large language model, the model after fine-tuning of scenic spot data is used to handle user travel matters, the problem of insufficient performance of the NLP model is solved, the travel experience is improved, and high-performance and convenient travel services are provided.
Patent Information
- Application Number
- CN202311271635.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-09-28
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2043-09-28
AI Technical Summary
The existing smart travel assistant is based on the NLP model and has limited performance and cannot effectively assist tourists during travel, resulting in poor travel experience.
A travel assistant based on a large language model is used to fine-tune the scenic spot data through a pre-trained large language model, generate processing results for users' travel matters, and provide one-to-one travel services.
Through the high performance and versatility of the large language model, users' experience before, during and after travel is improved, and more accurate and convenient travel services are provided.
Smart Images

Figure CN117273148B_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of artificial intelligence technology, specifically the field of large language models and intelligent assistant technologies, and in particular to a method and device for implementing a travel assistant based on a large language model, a method and device for determining a target large language model, an electronic device, a storage medium, and a computer program product, which can be applied to a travel assistant scenario. Background Art
[0002] Tourists often need to use smart travel assistants to enjoy the travel process conveniently and easily. At present, smart travel assistants are generally implemented based on NLP (Natural Language Processing) models. However, the performance of NLP models is far inferior to that of LLM (Large Language Model) models. Based on the performance limitations of the NLP model itself, the smart travel assistant does not provide ideal assistance to tourists during their travels. Summary of the invention
[0003] The present disclosure provides a method and device for implementing a travel assistant based on a large language model, a method and device for determining a target large language model, an electronic device, a storage medium, and a computer program product.
[0004] According to a first aspect, a method for implementing a travel assistant based on a large language model is provided, comprising: obtaining user conversation information about travel matters; generating a processing result for the conversation information through a target large language model, wherein the target large language model is obtained by fine-tuning the scenic spot data based on a pre-trained large language model; and feeding back the processing result to the user.
[0005] According to the second aspect, a method for determining a target large language model is provided, comprising: obtaining a training sample set, wherein the training samples in the training sample set include scenic spot data and labels; adopting a machine learning method, taking the scenic spot data as the input of a pre-trained large language model, taking the labels corresponding to the input scenic spot data as the expected output of the pre-trained large language model, and training to obtain the target large language model.
[0006] According to a third aspect, a device for implementing a travel assistant based on a large language model is provided, comprising: an acquisition unit, configured to acquire user conversation information about travel matters; a generation unit, configured to generate a processing result for the conversation information through a target large language model, wherein the target large language model is obtained by fine-tuning scenic spot data based on a pre-trained large language model; and a feedback unit, configured to feed back the processing result to the user.
[0007] According to the fourth aspect, a device for determining a target large language model is provided, comprising: a sample set acquisition unit, configured to acquire a training sample set, wherein the training samples in the training sample set include scenic spot data and labels; a fine-tuning unit, configured to adopt a machine learning method, using the scenic spot data as the input of a pre-trained large language model, and using the labels corresponding to the input scenic spot data as the expected output of the pre-trained large language model, to train the target large language model.
[0008] According to the fifth aspect, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method described in any implementation of the first aspect or the second aspect.
[0009] According to a sixth aspect, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause a computer to execute the method described in any implementation of the first aspect or the second aspect.
[0010] According to a seventh aspect, a computer program product is provided, comprising: a computer program, which implements the method described in any implementation of the first aspect or the second aspect when executed by a processor.
[0011] According to the technology disclosed in the present invention, a method for realizing a travel assistant based on a large language model is provided. The user's conversation information about travel matters is processed based on the large language model fine-tuned by scenic spot data, and the processing result for the conversation information is generated and fed back to the user. With the help of the high performance and strong versatility of the large language model, one-to-one travel services can be provided to the user before, during and after the trip, thereby improving the user's travel experience.
[0012] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS
[0013] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.
[0014] Figure 1 is an exemplary system architecture diagram in which an embodiment of the present disclosure may be applied;
[0015] Figure 2 is a flowchart of an embodiment of a method for implementing a travel assistant based on a large language model according to the present disclosure;
[0016] Figure 3Schematic diagram of the fine-tuning process of the pre-trained large language model according to the present disclosure
[0017] Figure 4 Schematic diagram of the overall system architecture of the method for implementing a travel assistant based on a large language model according to the present disclosure
[0018] Figure 5 A schematic diagram of an application scenario of the method for implementing a travel assistant based on a large language model according to this embodiment;
[0019] Figure 6 is a flowchart of another embodiment of a method for implementing a travel assistant based on a large language model according to the present disclosure;
[0020] Figure 7 is a flowchart of an embodiment of a method for determining a target large language model according to the present disclosure;
[0021] Figure 8 is a structural diagram of an embodiment of a device for implementing a travel assistant based on a large language model according to the present disclosure;
[0022] Fig. 9 is a structural diagram of an embodiment of a device for determining a target large language model according to the present disclosure;
[0023] Fig.10 It is a schematic diagram of the structure of a computer system suitable for implementing the embodiments of the present disclosure. DETAILED DESCRIPTION
[0024] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.
[0025] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.
[0026] Figure 1 An exemplary architecture 100 is shown to which the method and apparatus for implementing a travel assistant based on a large language model and the method and apparatus for determining a target large language model of the present disclosure can be applied.
[0027] like Figure 1As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The communication connection between the terminal devices 101, 102, 103 constitutes a topological network, and the network 104 is used to provide a medium for the communication link between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.
[0028] Terminal devices 101, 102, 103 may be hardware devices or software that support network connection for data interaction and data processing. When terminal devices 101, 102, 103 are hardware, they may be various electronic devices that support network connection, information acquisition, interaction, display, processing and other functions, including but not limited to smart phones, tablet computers, e-book readers, laptop computers and desktop computers, etc. When terminal devices 101, 102, 103 are software, they may be installed in the electronic devices listed above. They may be implemented as multiple software or software modules for providing distributed services, or they may be implemented as a single software or software module. No specific limitation is made here.
[0029] The server 105 may be a server that provides various services, for example, a background processing server that receives conversation information about travel matters sent by users through the terminal devices 101, 102, and 103, processes the conversation information based on the large language model fine-tuned by the scenic spot data, and generates and feeds back the processing results of the conversation information to the user. As an example, the server 105 may be a cloud server.
[0030] It should be noted that the server can be hardware or software. When the server is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or it can be implemented as a single server. When the server is software, it can be implemented as multiple software or software modules (for example, software or software modules used to provide distributed services), or it can be implemented as a single software or software module. No specific limitation is made here.
[0031] It should also be noted that the method for implementing a travel assistant based on a large language model and the method for determining a target large language model provided in the embodiments of the present disclosure can be executed by a server, or by a terminal device, or by a server and a terminal device in cooperation with each other. Accordingly, the various parts (e.g., various units) included in the device for implementing a travel assistant based on a large language model can all be set in the server, or all be set in the terminal device, or can be set in the server and the terminal device respectively.
[0032] It should be understood that Figure 1The number of terminal devices, networks, and servers in the system is only illustrative. Depending on the implementation needs, there may be any number of terminal devices, networks, and servers. When the electronic device on which the method for implementing a travel assistant based on a large language model and the method for determining a target large language model are running does not need to transmit data with other electronic devices, the system architecture may only include the electronic device (such as a server or terminal device) on which the method for implementing a travel assistant based on a large language model and the method for determining a target large language model are running.
[0033] Please refer to Figure 2 , Figure 2 A flowchart of a method for implementing a travel assistant based on a large language model provided in an embodiment of the present disclosure. In process 200, the following steps are included:
[0034] Step 201, obtaining user's session information related to travel matters.
[0035] In this embodiment, the execution subject of the method for implementing the travel assistant based on the large language model (for example, Figure 1 The terminal device or server in the system can obtain the user's session information about travel items remotely or locally through a wired network connection or a wireless network connection. In a real situation, travel is, for example, traveling or sightseeing.
[0036] The session information can be represented by information carriers such as text, voice, image, video, etc. For example, the session information can be a text in which the user describes his / her travel requirements (such as requirements for food, accommodation, transportation, sightseeing, shopping, entertainment, etc.). For another example, the session information can represent an image of a scenic spot of a specific scenery. Without knowing the travel location in the image, the user can represent the travel location he / she expects to have the same or similar scenery based on the image.
[0037] In some optional implementations of this embodiment, the execution subject may perform step 201 in the following manner: using a multi-round dialogue method to obtain conversation information about travel matters from the user.
[0038] As an example, a question-and-answer approach can be used to conduct multiple rounds of dialogue with the user through the target large language model, and the dialogue content of the multiple rounds of dialogue can be integrated to obtain the user's conversation information about travel matters.
[0039] In this implementation, through multiple rounds of dialogue, the user's conversation information about travel matters can be obtained more thoroughly, which helps the subsequent target large language model to fully understand the user's intentions and requirements.
[0040] In some optional implementations of this embodiment, the above-mentioned execution entity can further determine the session information in the following manner: through a virtual digital human, using a multi-round dialogue method, to obtain the user's session information about travel matters.
[0041] Virtual digital human, also known as virtual image, digital human, etc., is a digital AI (Artificial Intelligence) robot with multiple human characteristics formed by computer means such as deep learning, semantic understanding, and speech synthesis. The character image generation module, speech generation module, animation generation module, audio and video synthesis display module, and interaction module constitute the general system framework of virtual digital human. In short, virtual digital human is a digital virtual character created based on CG (Computer Graphics) technology and artificial intelligence technology.
[0042] In terms of interactive capabilities, digital humans include interactive and non-interactive types. Among them, interactive digital humans, driven by real people or intelligent systems, have human-like movements and perception capabilities and can interact with the real world. This implementation method uses interactive digital humans.
[0043] In terms of application scenarios, digital humans can be divided into identity-based and service-based. The former includes virtual avatars and virtual IP (Intellectual Property) / idols, which are the product of the "advanced" of real-world characters to the virtual world, and have more obvious "personality" attributes; while the latter mainly provides anthropomorphic services such as office and companionship for corporate or individual users, thus replacing some human resources. This implementation method is to use identity-based digital humans.
[0044] The above-mentioned execution entity can provide multiple virtual digital humans to the user, and determine the target virtual digital human preferred by the user based on the user's selection operation; then, conduct multiple rounds of voice conversations with the user through the target virtual digital human; finally, integrate the multiple rounds of voice conversations to determine the user's conversation information about travel matters.
[0045] In this implementation, multiple rounds of dialogue interaction are carried out between the virtual digital human and the user, which improves the interest of the interaction process and improves the efficiency of obtaining conversation information.
[0046] In some optional implementations of this embodiment, the execution subject may further determine the conversation information in the following manner: in a multi-round conversation process between the virtual digital human and the user, iteratively perform the following conversation operations:
[0047] First, receive and recognize the user's voice conversation information.
[0048] Second, the voice conversation information is converted into textual conversation information.
[0049] In this implementation, the execution subject may use speech-to-text technology to convert the speech conversation information into text-based conversation information.
[0050] Third, feedback information of conversation information is generated through the target large language model.
[0051] As an example, first, the execution subject determines the slots in the conversation information through the target large language model. The slot refers to the information or attributes that need to be extracted from the text in order to complete a specific task or answer a specific question. Then, the uncertainty of each slot in the multiple slots is determined. The uncertainty of the slot indicates that there are many options for the answer corresponding to the slot and it cannot be clearly determined. Finally, the target large language model generates feedback information of the conversation information according to the slot with the largest uncertainty.
[0052] Fourth, generate speech feedback information representing the feedback information.
[0053] In this implementation, the above-mentioned execution subject can generate voice feedback information representing the feedback information through text-to-speech technology.
[0054] Fifth, voice feedback information is broadcast to users through virtual digital humans.
[0055] In this implementation, a specific implementation of multi-round dialogue is provided. During the multi-round dialogue process, the target large language model generates feedback information in a targeted manner based on the user's conversation information in this round, and reports it to the user through a virtual digital human, thereby further improving the efficiency of obtaining conversation information while enhancing the fun of the interaction process.
[0056] In some optional implementations of this embodiment, the above-mentioned execution subject can perform the above-mentioned third step in the following manner: first, determine the target task corresponding to the session information; then, based on the session information carrying the prompt guidance information corresponding to the target task, generate feedback information of the session information through the target large language model.
[0057] The target task may be any task related to travel matters, including but not limited to travel itinerary planning, scenic spot explanation, knowledge quiz, small talk conversation, etc. The above-mentioned execution subject may parse the conversation and determine the target task corresponding to the conversation information.
[0058] After determining the target task corresponding to the session information, the large language model is called by carrying prompt information (prompt guidance information) on the session information to generate feedback information of the session information.
[0059] In this implementation, the target large language model processes the conversation information carrying the prompt guidance information corresponding to the target task to generate feedback information, which helps the target large language model to understand the conversation information more thoroughly and improves the accuracy of the feedback information.
[0060] In some optional implementations of this embodiment, the above-mentioned execution subject may generate feedback information in the following manner: generate feedback information of the session information according to the user's preference information on travel through a target large language model.
[0061] As an example, the above-mentioned execution entity can form a more targeted travel recommendation response information service through the target large language model based on some of the user's proprietary preference prompt words in dining, shopping, travel methods, cultural hobbies, etc., to guide the user's reservation operations and promote the merchant's information promotion.
[0062] In this implementation, the feedback information is generated through the target large language model by combining the conversation information carrying the prompt guidance information corresponding to the target task and the user's travel preference information, which further improves the fit between the feedback information and the user and improves the interaction efficiency.
[0063] Step 202: Generate a processing result for the conversation information through the target large language model.
[0064] In this embodiment, the execution subject can generate a processing result for the conversation information through a target large language model, wherein the target large language model is obtained by fine-tuning the scenic spot data based on the pre-trained large language model.
[0065] A large language model refers to a model that can adapt to a series of downstream tasks after being trained on large-scale and broad data. Large language models have the two attributes of "large scale" and "pre-training", which can greatly improve the generalization, versatility and practicality of artificial intelligence. In practical applications, large models have demonstrated their strong strength in natural language processing, image recognition, speech recognition and other fields.
[0066] Large language models have the following advantages:
[0067] 1. Powerful generalization ability: The large language model can extract the patterns and features of the data through pre-training of a large amount of data, thus having stronger generalization ability and being able to adapt to a variety of tasks.
[0068] 2. High performance: Large models are trained using high-performance GPU (graphics processing unit) clusters, which can accelerate the model training process and improve model performance.
[0069] 3. Strong versatility: Large models can be applied to a variety of fields and tasks and have strong versatility.
[0070] Continue to refer Figure 3 , showing a schematic diagram of a fine-tuning process 300 of a pre-trained large language model. In this embodiment, the above-mentioned execution subject can obtain the target large language model in the following manner:
[0071] First, obtain scenic spot data 301. The scenic spot data is data related to the scenic spot, including but not limited to introduction data of the scenic spot, relevant field knowledge, number of tourists, meteorological data, etc. Then, training samples can be obtained based on the annotation of the scenic spot data by the operator. For different tasks performed subsequently by the fine-tuned large language model, the scenic spot data can be annotated differently to obtain different training samples. Finally, the pre-trained large language model 302 is fine-tuned through the generated training samples to obtain the fine-tuned target large language model 303. In addition, according to the requirements of the large language model for different specific tasks, the structure of the pre-trained model can also be fine-tuned.
[0072] In this embodiment, the execution subject can input the conversation information about travel matters into the target large language model, and generate processing results for the conversation information through the target large language model. The target large language model can generate different processing results for conversation information with different requirements represented by different semantics.
[0073] In some optional implementations of this embodiment, the above-mentioned execution entity can perform the above-mentioned step 202 in the following manner: in response to determining that the session information represents the travel itinerary requirements of the user before the trip, generating travel itinerary planning information that meets the travel itinerary requirements through the target large language model.
[0074] For example, the dialogue information is "Please prepare a four-day and three-night travel guide for three people in Guangxi. The itinerary needs to include Yangshuo and Guilin, and give a suitable round-trip air ticket from Beijing to Guangxi, and the air ticket price should not exceed 1,500 yuan." The above dialogue information is input into the target large language model, and the target large language model generates travel itinerary planning information that meets the travel itinerary requirements.
[0075] In this implementation, based on the high performance and strong generalization of the target large language model, more reasonable travel itinerary planning information that meets user requirements can be generated, which helps to improve the user's travel experience.
[0076] In some optional implementations of this embodiment, the above-mentioned execution entity can also perform the above-mentioned step 202 in the following manner: first, determine the travel reservation service required in the travel itinerary according to the travel itinerary planning information; then, complete the reservation operation of the travel reservation service through the target large language model.
[0077] Travel reservation services include but are not limited to hotel, restaurant, train ticket, bus ticket, air ticket and other reservation services.
[0078] After determining the travel itinerary planning information, the above-mentioned execution entity can clarify each time point in the travel itinerary; and then determine the scheduled time of the travel reservation service according to the time point, so as to complete the reservation operation of the travel reservation service through the target large language model.
[0079] For example, the above-mentioned execution entity can push the booking interface of the travel reservation service to the user so that the user can complete the reservation operation; for another example, the above-mentioned execution entity can determine the travel reservation service adapted to the user through the target large language model based on the user's consumption habits and personal preferences, and then directly complete the reservation operations such as verification and payment.
[0080] In this implementation, the user's travel itinerary planning information is determined through the target large language model, and the travel reservation services involved in the user's travel itinerary are further booked, thereby avoiding the user's search, aggregation and other operations for travel reservation services, improving the user's operational convenience, and enhancing the user's travel experience.
[0081] In some optional implementations of this embodiment, the above-mentioned execution entity can perform the reservation operation of the travel reservation service in the following manner: first, determine the target application that supports the travel reservation service; then, through the target large language model, call the third-party plug-in corresponding to the target application to complete the reservation operation of the travel reservation service.
[0082] The target application supporting the travel reservation service is an application that can provide reservation operations for the travel reservation service. For example, the target application is an application such as Ctrip and Alitrip.
[0083] For each target application, a corresponding third-party plug-in can be set. After the target application is determined, the third-party plug-in corresponding to the target application can be called through the target large language model to complete the reservation operation of the travel reservation service.
[0084] In this implementation, a specific implementation of a predetermined operation is provided, which helps to improve the execution efficiency of the predetermined operation.
[0085] In some optional implementations of this embodiment, the above-mentioned execution entity can perform the above-mentioned step 202 in the following manner: in response to determining that the session information represents a recommendation request for travel experience items of the user during the trip, generate travel experience item data that meets the recommendation request through the target large language model.
[0086] Among them, travel experience projects may include projects involving travel routes, food, accommodation, etc.
[0087] In this implementation, by means of the high performance and strong generalization of the target large language model, travel experience project data that meets the recommendation request can be generated quickly and accurately.
[0088] In some optional implementations of this embodiment, the above-mentioned execution entity can perform the above-mentioned step 202 in the following manner: in response to a request to record that the session information represents the user's travel experience after the trip, travel experience data that meets the recording request is generated through the target large language model.
[0089] Travel experience record requests include requests to generate travel notes and merchant reviews. By leveraging the AIGC (Artificial Intelligence Generated Content) generation capabilities of the large language model, travel notes and merchant reviews are automatically generated based on the user's itinerary data.
[0090] In this implementation, by means of the high performance and strong generalization of the target large language model, travel experience data that meets the recording request can be generated quickly and accurately.
[0091] In some optional implementations of this embodiment, the above-mentioned execution entity can generate travel experience data in the following manner: based on the user's itinerary data and personal style data, generate travel experience data that meets the recording request and conforms to the personal style represented by the personal style data through the target large language model.
[0092] As an example, the execution subject may obtain and analyze the user's relevant data to determine the user's personal style data. The user's relevant data may be, for example, travel notes written by the user himself or a merchant review data. Then, the user's itinerary data and personal style data are input into the target large language model to generate travel experience data that meets the recording request and conforms to the personal style represented by the personal style data.
[0093] In this implementation, the user's itinerary data and personal style data are combined to generate travel experience data. With the help of the high performance and strong generalization of the target large language model, the adaptability of the travel experience data to the user is further improved.
[0094] Step 203: Feedback the processing result to the user.
[0095] In this embodiment, the above execution entity can feed back the processing result to the user.
[0096] In this embodiment, the above-mentioned execution entity can feed back the processing results to the terminal device corresponding to the user, and display the processing results to the user through a virtual digital human.
[0097] In some optional implementations of this embodiment, the execution subject may perform step 203 in the following manner: providing the user with feedback on travel experience project data through a virtual digital human.
[0098] As an example, the above-mentioned execution entity can generate audio and video data representing travel project experience data, and display and broadcast the travel experience project data to users through a virtual digital person.
[0099] In this implementation, travel experience project data is fed back to users through virtual digital humans, and based on the interaction between the virtual digital humans and users, the user experience during the travel process is improved.
[0100] In some optional implementations of this embodiment, the above-mentioned execution subject may perform the process of feeding back travel experience project data to the user through the virtual digital human in the following manner:
[0101] First, according to the travel experience project data, the action and emotion data of the virtual digital person in the process of explaining the travel experience project data are determined. Then, in the process of using the virtual digital person to feedback the travel experience project data to the user, the action and emotion data are used to guide the operation of the virtual digital person.
[0102] As an example, the above-mentioned execution entity can parse the travel experience project data to determine the historical and cultural background, architectural style and design concept, related myths and legends, natural landscape and ecological environment, social and cultural significance of the travel experience project represented by the travel experience project data, and thus determine the action and emotional data of the virtual digital person in the process of explaining the travel experience project data based on the above-mentioned relevant data, so that the virtual digital person can vividly feedback the travel experience project data to the user, help the user better understand and feel the charm of the travel experience project, make the user immersive, and enhance the user's experience.
[0103] In some optional implementations of this embodiment, the above-mentioned execution subject may perform the process of feeding back travel experience project data to the user through the virtual digital human in the following manner:
[0104] First, determine the scenic spot IP of the travel project corresponding to the travel experience project data; then, feedback the travel experience project data to the user through a virtual digital person representing the scenic spot IP.
[0105] The IP of a scenic spot is actually an element or symbol that is simple, distinct and distinctive, can attract tourists, give the destination personality and vitality, and can represent the destination. It can be a reproduction of a story, a film, a game or a legend, or an architectural element, or a person, animal, cartoon, etc.
[0106] As an example, when the travel project is Mount Emei, the virtual digital human representing the IP of the scenic spot can be a three-dimensional digital human image of the Monkey King; when the travel project is Yueyang Tower, the virtual digital human representing the IP of the scenic spot can be a three-dimensional digital human image of poets such as Li Bai and Du Fu.
[0107] By using a virtual digital person who represents the IP of a scenic spot to provide users with feedback on travel experience project data, the image of the virtual digital person is more compatible with the travel experience project, which helps to further improve the user's experience during travel.
[0108] Continue to refer Figure 4 , is a schematic diagram of an overall system architecture 400 of a method for implementing a travel assistant based on a large language model according to this embodiment. The overall system architecture 400 includes an infrastructure layer 401, a large model layer 402, a framework layer 403, an AI application model layer 404, a central control application layer 405, and an application layer 406.
[0109] The infrastructure layer refers to the basic equipment required for the operation of large language models, involving basic equipment in terms of GPU computing power, RDMA (Remote Direct Memory Access) high-speed network, data storage, system scheduling, etc.
[0110] The large model layer corresponds to the pre-trained large language model, such as the "Wenxin Yiyan" large language model, which is an industrial-level knowledge-enhanced large model trained based on large-scale industrial data and knowledge graphs.
[0111] The framework layer is the framework layer above the big language model, which is the big language model development and service operation platform. Using this framework, the annotation data of specific travel attractions can be used to generate a travel assistant AI model, which is the target big language model, through model tuning, prompt engineering and other operations.
[0112] The AI application model layer corresponds to the fine-tuned large language model in each business field, including the travel assistant AI model and the digital human AI model. Among them, the travel assistant AI model is an industry model generated after fine-tuning based on the pre-trained large language model and scenic spot annotation data.
[0113] The central control application layer includes the travel assistant central control and the digital human central control. Among them, the travel assistant central control provides capabilities such as task dialogue, knowledge questions and answers, assistant skills, and third-party plug-ins. The digital human central control provides capabilities such as speech recognition, speech synthesis, graphics rendering, and video drive.
[0114] Specifically, the capabilities of the travel assistant central control are described as follows:
[0115] Task dialogue: The travel assistant central control has a multi-round dialogue function. For example, the user asks the travel assistant, "Please prepare a four-day and three-night travel guide in Guangxi for three people. The itinerary needs to include Yangshuo and Guilin, and provide suitable round-trip air tickets from Beijing to Guangxi. The air ticket price should not exceed 1,500 yuan." The travel assistant generates travel itinerary planning information through the travel assistant central control.
[0116] Knowledge Q&A: The travel assistant central control has a knowledge Q&A function. For example, when a user asks the travel assistant, "Please tell me the history of the Hall of Supreme Harmony in the Forbidden City," the travel assistant AI model returns the introduction data of the Hall of Supreme Harmony in the Forbidden City to the travel assistant central control. The digital human central control uses speech synthesis technology to generate an introductory voice for the text information returned by the travel assistant central control, and returns the introductory voice to the application layer for voice broadcast based on the virtual digital human.
[0117] Assistant skills: For example, questions and answers about general knowledge, such as today's weather in Guangxi; questions and answers about industry skills, such as an introduction to the Longji Terraces in Guangxi; character role-playing, such as providing me with tour guide services as a Guangxi tour guide, etc.
[0118] Third-party plug-in: a target application that supports travel booking services. For example, after determining the travel itinerary planning information, the user needs to purchase an air ticket and directly calls the third-party API (Application Program Interface) corresponding to the target application that supports the air ticket booking service through the central control application layer.
[0119] The capabilities of the Digital Human Central Control are described as follows:
[0120] Speech recognition: It can convert binary speech received by terminal applet or application into text. For example, when a user asks a query to the travel assistant through voice, the digital human central control receives the speech file and converts it into text.
[0121] Speech synthesis: For text data, the Digital Human Central Control provides speech synthesis function to convert text data into voice or video data.
[0122] Graphics rendering: used to render the image of virtual digital people.
[0123] Continue to see Figure 5 , Figure 5 FIG. 5 is a schematic diagram 500 of an application scenario of the method for implementing a travel assistant based on a large language model according to the present embodiment. Figure 5 In the application scenario, the user can conduct voice and text conversations through the travel assistant on the terminal device 501, so that the travel assistant can make travel itinerary planning through multiple rounds of voice conversations. First, after the user obtains the user's voice conversation through the travel assistant deployed on the terminal, the voice conversation data is sent to the digital human central control 502; then, the digital human central control calls the model with voice-to-text function to convert the voice conversation data into text data; then, after the digital human central control obtains the converted text data, it sends it to the travel assistant central control 503; then, the travel assistant central control distributes the text data to the corresponding task dialogue module; then, the task dialogue module calls the target language model (travel assistant AI big model) 504 based on the text data carrying prompt information; then, the target big language model generates the processing result (travel itinerary planning information) and returns it to the task dialogue module; finally, the travel assistant central control sends the processing result back to the digital human central control to feedback the processing result to the user through the virtual digital human. In addition, the travel assistant central control will call the third-party reservation service to complete the reservation operation of the travel reservation service.
[0124] In this embodiment, a method for implementing a travel assistant based on a large language model is provided. The user's conversation information about travel matters is processed based on the large language model that has been fine-tuned with scenic spot data, and the processing results for the conversation information are generated and fed back to the user. With the help of the high performance and strong versatility of the large language model, one-to-one travel services can be provided to the user before, during and after the trip, thereby improving the user's travel experience.
[0125] Continue to refer Figure 6 , shows a schematic process 600 of another embodiment of a method for implementing a travel assistant based on a large language model according to the present disclosure. In process 600, the following steps are included:
[0126] Step 601, during the multi-round dialogue between the virtual digital human and the user, the following dialogue operations are iteratively performed:
[0127] Step 6011, receiving and identifying the user's voice conversation information.
[0128] Step 6012: Convert the voice conversation information into textual conversation information.
[0129] Step 6013, determine the target task corresponding to the session information.
[0130] Step 6014, based on the conversation information carrying the prompt guidance information corresponding to the target task, feedback information of the conversation information is generated through the target large language model according to the user's preference information in travel.
[0131] Step 6015, generating voice feedback information representing the feedback information.
[0132] Step 6016: broadcast voice feedback information to the user through the virtual digital human.
[0133] Step 602 , in response to determining that the conversation information represents the travel itinerary requirements of the user before traveling, generating travel itinerary planning information that meets the travel itinerary requirements through the target large language model.
[0134] Step 603, determining the travel reservation services required in the travel itinerary according to the travel itinerary planning information.
[0135] Step 604: Complete the reservation operation of the travel reservation service through the target large language model.
[0136] Step 605 , in response to determining that the session information represents a recommendation request for a travel experience item of the user during travel, generating travel experience item data satisfying the recommendation request through a target large language model.
[0137] Step 606, in response to determining that the session information represents a record request of the user's travel experience after the trip, based on the user's itinerary data and personal style data, travel experience data that meets the record request and conforms to the personal style represented by the personal style data is generated through the target large language model.
[0138] Step 607: Feedback the processing result to the user through the virtual digital human.
[0139] It can be seen from this embodiment that Figure 2 Compared with the corresponding embodiments, process 600 of the method for realizing a travel assistant based on a large language model in this embodiment specifically illustrates a process of obtaining conversation information through multiple rounds of dialogue based on a virtual digital human, a process of generating different processing results for different conversation information by a target large language model, and a process of feeding back processing results through a virtual digital human. With the help of the high performance and strong versatility of the large language model, one-to-one travel services can be provided to users before, during and after a trip, thereby improving the user's travel experience.
[0140] Please refer to Figure 7 , Figure 7 A flowchart of a method for determining a target large language model provided by an embodiment of the present disclosure. In process 700, the following steps are included:
[0141] Step 701: Obtain a training sample set.
[0142] In this embodiment, the execution subject of the method for determining the target large language model (for example, Figure 1 The terminal device or server in the network can obtain the training sample set remotely or locally through a wired network connection or a wireless network connection.
[0143] The training samples in the training sample set include scenic spot data and labels. The scenic spot data refers to the data related to the scenic spot, including but not limited to the introduction data of the scenic spot, related domain knowledge, the number of tourists, and meteorological data. Figure 3 ,The manual labeling method can be used to obtain labels and generate ,training sample sets based on the annotation of scenic spot data.
[0144] For different tasks performed subsequently by the fine-tuned large language model, the scenic spot data can be labeled differently to obtain different training samples.
[0145] Step 702, adopting a machine learning method, using the scenic spot data as the input of the pre-trained large language model, using the label corresponding to the input scenic spot data as the expected output of the pre-trained large language model, and training to obtain a target large language model.
[0146] In this embodiment, the above-mentioned execution entity can adopt a machine learning method, use the scenic spot data as the input of the pre-trained large language model, and use the label corresponding to the input scenic spot data as the expected output of the pre-trained large language model to train and obtain the target large language model.
[0147] As an example, the above execution entity can iteratively perform the following fine-tuning operations: first, select an untrained target training sample from the training sample set; then, input the scenic spot data in the target training sample into the current large language model to obtain the actual output; then, calculate the loss between the actual output and the label in the target training sample; finally, update the current large language model according to the loss.
[0148] By iteratively performing the above-mentioned fine-tuning operation, in response to reaching a preset end condition, a target large language model is obtained, wherein the preset end condition is, for example, that the number of training times exceeds a preset number threshold, the training loss tends to converge, and the training time exceeds a preset time threshold.
[0149] In this embodiment, a method for determining a target large language model is provided, and a high-performance travel assistant model can be obtained by fine-tuning the pre-trained large language model with scenic spot data.
[0150] Continue to refer Figure 8 As an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of a device for implementing a travel assistant based on a large language model. Figure 2Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.
[0151] like Figure 8 As shown, the device 800 for implementing a travel assistant based on a large language model includes: an acquisition unit 801, configured to acquire user conversation information about travel matters; a generation unit 802, configured to generate a processing result for the conversation information through a target large language model, wherein the target large language model is obtained by fine-tuning the scenic spot data based on a pre-trained large language model; and a feedback unit 803, configured to feed back the processing result to the user.
[0152] In some optional implementations of this embodiment, the acquisition unit 801 is further configured to: acquire the user's conversation information about travel matters by adopting a multi-round dialogue method.
[0153] In some optional implementations of this embodiment, the acquisition unit 801 is further configured to: acquire the user's conversation information about travel matters through a virtual digital human in a multi-round dialogue manner.
[0154] In some optional implementations of this embodiment, the acquisition unit 801 is further configured to: during multiple rounds of dialogue between the virtual digital human and the user, iteratively perform the following conversation operations: receive and recognize the user's voice conversation information; convert the voice conversation information into text-based conversation information; generate feedback information of the conversation information through the target large language model; generate voice feedback information representing the feedback information; and broadcast the voice feedback information to the user through the virtual digital human.
[0155] In some optional implementations of this embodiment, the acquisition unit is further configured to: determine the target task corresponding to the session information; and generate feedback information of the session information through the target large language model based on the session information carrying the prompt guidance information corresponding to the target task.
[0156] In some optional implementations of this embodiment, the acquisition unit 801 is further configured to: generate feedback information of the conversation information according to the user's travel preference information through the target large language model.
[0157] In some optional implementations of this embodiment, the above-mentioned generation unit 802 is further configured to: in response to determining that the conversation information represents the travel itinerary requirements of the user before the trip, generate travel itinerary planning information that meets the travel itinerary requirements through the target large language model.
[0158] In some optional implementations of this embodiment, the above-mentioned generation unit 802 is further configured to: determine the travel reservation service required in the travel itinerary according to the travel itinerary planning information; and complete the reservation operation of the travel reservation service through the target large language model.
[0159] In some optional implementations of this embodiment, the above-mentioned generation unit 802 is further configured to: determine a target application that supports the travel reservation service; and call a third-party plug-in corresponding to the target application through a target large language model to complete the reservation operation of the travel reservation service.
[0160] In some optional implementations of this embodiment, the above-mentioned generation unit 802 is further configured to: in response to determining that the session information represents a recommendation request for a travel experience item of the user during the trip, generate travel experience item data that meets the recommendation request through the target large language model.
[0161] In some optional implementations of this embodiment, the feedback unit 803 is further configured to: provide the user with feedback on travel experience project data through a virtual digital human.
[0162] In some optional implementations of this embodiment, the feedback unit 803 is further configured to: determine the action and emotion data of the virtual digital person in the process of explaining the travel experience project data based on the travel experience project data; and guide the operation of the virtual digital person through the action and emotion data in the process of feeding back the travel experience project data to the user through the virtual digital person.
[0163] In some optional implementations of this embodiment, the feedback unit 803 is further configured to: determine the scenic spot IP of the travel project corresponding to the travel experience project data; and feedback the travel experience project data to the user through a virtual digital person representing the scenic spot IP.
[0164] In some optional implementations of this embodiment, the above-mentioned generation unit 802 is further configured to: in response to a record request that determines that the session information represents the user's travel experience after the trip, generate travel experience data that meets the record request through the target large language model.
[0165] In some optional implementations of this embodiment, the above-mentioned generation unit 802 is further configured to: generate travel experience data that meets the recording request and conforms to the personal style represented by the personal style data through the target large language model based on the user's itinerary data and personal style data.
[0166] In this embodiment, a device for implementing a travel assistant based on a large language model is provided. A device for implementing a travel assistant based on a large language model is provided. The user's conversation information about travel matters is processed based on the large language model that has been fine-tuned with scenic spot data, and the processing result of the conversation information is generated and fed back to the user. With the help of the high performance and strong versatility of the large language model, one-to-one travel services can be provided to the user before, during and after the trip, thereby improving the user's travel experience.
[0167] Continue to refer Fig. 9 As an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of a device for determining a target large language model. Figure 7 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.
[0168] like Fig. 9 As shown, the device 900 for determining the target large language model includes: a sample set acquisition unit 901, configured to acquire a training sample set, wherein the training samples in the training sample set include scenic spot data and labels; a fine-tuning unit 902, configured to adopt a machine learning method, take the scenic spot data as the input of the pre-trained large language model, take the label corresponding to the input scenic spot data as the expected output of the pre-trained large language model, and train to obtain the target large language model.
[0169] In this embodiment, a device for determining a target large language model is provided, and a high-performance travel assistant model can be obtained by fine-tuning the pre-trained large language model with scenic spot data.
[0170] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can implement the method for realizing a travel assistant based on a large language model and the method for determining a target large language model described in any of the above embodiments when executing.
[0171] According to an embodiment of the present disclosure, the present disclosure also provides a readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to implement the method of realizing a travel assistant based on a large language model and the method of determining a target large language model described in any of the above embodiments when executed.
[0172] An embodiment of the present disclosure provides a computer program product, which, when executed by a processor, can implement the method for implementing a travel assistant based on a large language model and the method for determining a target large language model described in any of the above embodiments.
[0173] Fig.10 A schematic block diagram of an example electronic device 1000 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.
[0174] like Fig.10 As shown, the device 1000 includes a computing unit 1001, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1002 or a computer program loaded from a storage unit 1008 into a random access memory (RAM) 1003. In the RAM 1003, various programs and data required for the operation of the device 1000 can also be stored. The computing unit 1001, the ROM 1002, and the RAM 1003 are connected to each other via a bus 1004. An input / output (I / O) interface 1005 is also connected to the bus 1004.
[0175] A number of components in the device 1000 are connected to the I / O interface 1005, including: an input unit 1006, such as a keyboard, a mouse, etc.; an output unit 1007, such as various types of displays, speakers, etc.; a storage unit 1008, such as a disk, an optical disk, etc.; and a communication unit 1009, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1009 allows the device 1000 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.
[0176] The computing unit 1001 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 1001 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 1001 performs the various methods and processes described above, such as a method for implementing a travel assistant based on a large language model and a method for determining a target large language model. For example, in some embodiments, a method for implementing a travel assistant based on a large language model and a method for determining a target large language model may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 1008. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 1000 via the ROM 1002 and / or the communication unit 1009. When the computer program is loaded into RAM 1003 and executed by the computing unit 1001, one or more steps of the method for implementing a travel assistant based on a large language model and the method for determining a target large language model described above may be performed. Alternatively, in other embodiments, the computing unit 1001 may be configured to execute the method for implementing a travel assistant based on a large language model and the method for determining a target large language model in any other appropriate manner (e.g., by means of firmware).
[0177] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.
[0178] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.
[0179] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
[0180] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).
[0181] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.
[0182] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and virtual private servers (VPS) services; it may also be a server for a distributed system, or a server combined with a blockchain.
[0183] According to the technical solution of the embodiment of the present disclosure, a method for realizing a travel assistant based on a large language model is provided. The user's conversation information about travel matters is processed based on the large language model fine-tuned by scenic spot data, and the processing result for the conversation information is generated and fed back to the user. With the help of the high performance and strong versatility of the large language model, one-to-one travel services can be provided to the user before, during and after the trip, thereby improving the user's travel experience.
[0184] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution provided by this disclosure can be achieved, and this document does not limit this.
[0185] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.
Claims
1. A method for implementing a travel assistant based on a large language model, comprising: Get the user's session information about travel matters; Generate a processing result for the session information through a target large language model, wherein the target large language model is obtained by fine-tuning the scenic spot data based on a pre-trained large language model, the session information represents a recommendation request for a travel experience item during the travel of the user, and the processing result represents travel experience item data that meets the recommendation request; Feedback of the processing result to the user includes: determining the scenic spot IP of the travel project corresponding to the travel experience project data; and feeding back the travel experience project data to the user through a virtual digital person representing the scenic spot IP.
2. The method according to claim 1, wherein: The obtaining of the user's session information related to travel matters includes: A multi-round dialogue method is adopted to obtain the user's conversation information related to travel matters.
3. The method according to claim 2, wherein: The method of using a multi-round dialogue method to obtain the user's conversation information about travel matters includes: Through a virtual digital human, a multi-round dialogue method is adopted to obtain the user's conversation information about travel matters.
4. The method according to claim 3, wherein: The method of obtaining the user's conversation information about travel matters by using a virtual digital person in a multi-round dialogue manner includes: During multiple rounds of conversations between the virtual digital human and the user, the following conversation operations are iteratively performed: Receiving and identifying the user's voice conversation information; Converting the voice conversation information into textual conversation information; generating feedback information of the conversation information by using the target large language model; generating speech feedback information representing the feedback information; The voice feedback information is broadcasted to the user through the virtual digital human.
5. The method according to claim 4, wherein: The generating the feedback information of the conversation information by using the target large language model includes: Determining a target task corresponding to the session information; Based on the conversation information carrying the prompt guidance information corresponding to the target task, feedback information of the conversation information is generated through the target large language model.
6. The method according to claim 5, wherein: The generating the feedback information of the conversation information by using the target large language model includes: Feedback information of the conversation information is generated through the target large language model according to the user's travel preference information.
7. The method according to claim 1, wherein: The generating a processing result for the session information by using the target large language model includes: In response to determining that the conversation information represents a travel itinerary requirement of the user before traveling, travel itinerary planning information that meets the travel itinerary requirement is generated through the target large language model.
8. The method according to claim 7, wherein: The generating of the processing result for the session information by using the target large language model further includes: Determining the travel reservation services required in the travel itinerary according to the travel itinerary planning information; The reservation operation of the travel reservation service is completed through the target large language model.
9. The method according to claim 8, wherein: The completing the reservation operation of the travel reservation service by using the target large language model includes: Determining target applications that support the travel reservation service; Through the target large language model, a third-party plug-in corresponding to the target application is called to complete the reservation operation of the travel reservation service.
10. The method according to claim 1, wherein: Feedback of the travel experience project data to the user through a virtual digital human includes: Determining, according to the travel experience project data, action and emotion data of the virtual digital person in the process of explaining the travel experience project data; In the process of feeding back the travel experience project data to the user through the virtual digital human, the operation of the virtual digital human is guided by the action emotion data.
11. The method according to claim 1, wherein: The generating a processing result for the session information by using the target large language model includes: In response to determining that the session information represents a request to record the travel experience of the user after the trip, travel experience data satisfying the record request is generated through the target large language model.
12. The method according to claim 11, wherein generating travel experience data satisfying the recording request through the target large language model comprises: According to the itinerary data and personal style data of the user, travel experience data satisfying the recording request and conforming to the personal style represented by the personal style data is generated through the target large language model.
13. A method for determining a target large language model, comprising: Acquire a training sample set, wherein the training samples in the training sample set include scenic spot data and labels; A machine learning method is adopted, with scenic spot data as the input of a pre-trained large language model, and the label corresponding to the input scenic spot data as the expected output of the pre-trained large language model, to train a target large language model, wherein the target large language model is used to determine travel experience project data that satisfies the user's recommendation request for travel experience projects during travel, so as to feed back the travel experience project data to the user through a virtual digital person representing the scenic spot IP of the travel project corresponding to the travel experience project data.
14. A device for implementing a travel assistant based on a large language model, comprising: An acquisition unit, configured to acquire user's session information related to travel matters; a generating unit configured to generate a processing result for the session information through a target large language model, wherein the target large language model is obtained by fine-tuning the scenic spot data based on a pre-trained large language model, the session information represents a recommendation request for a travel experience item during the travel of the user, and the processing result represents travel experience item data that satisfies the recommendation request; A feedback unit is configured to feed back the processing result to the user, including: determining the scenic spot IP of the travel project corresponding to the travel experience project data; and feeding back the travel experience project data to the user through a virtual digital person representing the scenic spot IP.
15. The device according to claim 14, wherein: The acquisition unit is further configured to: A multi-round dialogue method is adopted to obtain the user's conversation information related to travel matters.
16. The device according to claim 15, wherein: The acquisition unit is further configured to: Through a virtual digital human, a multi-round dialogue method is adopted to obtain the user's conversation information about travel matters.
17. The device according to claim 16, wherein: The acquisition unit is further configured to: During multiple rounds of conversations between the virtual digital human and the user, the following conversation operations are iteratively performed: receiving and recognizing the user's voice conversation information; converting the voice conversation information into textual conversation information; generating feedback information of the conversation information through the target large language model; generating voice feedback information representing the feedback information; and broadcasting the voice feedback information to the user through the virtual digital human.
18. The device according to claim 17, wherein: The acquisition unit is further configured to: Determine a target task corresponding to the session information; and generate feedback information of the session information through the target large language model based on the session information carrying prompt guidance information corresponding to the target task.
19. The device according to claim 18, wherein: The acquisition unit is further configured to: Feedback information of the conversation information is generated through the target large language model according to the user's travel preference information.
20. The device according to claim 14, wherein: The generating unit is further configured to: In response to determining that the conversation information represents a travel itinerary requirement of the user before traveling, travel itinerary planning information that meets the travel itinerary requirement is generated through the target large language model.
21. The device according to claim 20, wherein: The generating unit is further configured to: Determine the travel reservation service required in the travel itinerary according to the travel itinerary planning information; and complete the reservation operation of the travel reservation service through the target large language model.
22. The device according to claim 21, wherein The generating unit is further configured to: Determine a target application that supports the travel reservation service; and call a third-party plug-in corresponding to the target application through the target large language model to complete a reservation operation of the travel reservation service.
23. The device according to claim 14, wherein: The feedback unit is further configured to: According to the travel experience project data, the action and emotion data of the virtual digital person in the process of explaining the travel experience project data are determined; in the process of feeding back the travel experience project data to the user through the virtual digital person, the operation of the virtual digital person is guided by the action and emotion data.
24. The device according to claim 14, wherein: The generating unit is further configured to: In response to determining that the session information represents a request to record the travel experience of the user after the trip, travel experience data satisfying the record request is generated through the target large language model.
25. The apparatus according to claim 24, wherein the generating unit is further configured to: According to the itinerary data and personal style data of the user, travel experience data satisfying the recording request and conforming to the personal style represented by the personal style data is generated through the target large language model.
26. A device for determining a target large language model, comprising: A sample set acquisition unit is configured to acquire a training sample set, wherein the training samples in the training sample set include scenic spot data and labels; The fine-tuning unit is configured to adopt a machine learning method, use the scenic spot data as the input of the pre-trained large language model, and use the label corresponding to the input scenic spot data as the expected output of the pre-trained large language model to train a target large language model, wherein the target large language model is used to determine the travel experience project data that satisfies the user's recommendation request for travel experience projects during the trip, so as to feedback the travel experience project data to the user through a virtual digital person representing the scenic spot IP of the travel project corresponding to the travel experience project data.
27. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 13.
28. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 13.
29. A computer program product comprising: A computer program which, when executed by a processor, implements the method according to any one of claims 1 to 13.
