Travel knowledge recommendation method and device based on target language model, and electronic equipment

Abstract

The invention discloses a travel knowledge recommendation method and device based on a target language model and electronic equipment, and relates to the technical field of passenger ticket freight rate search or other related technical fields, and the method comprises the steps: receiving a freight rate search request of a user side, and extracting a key parameter set and a travel knowledge tag set in the freight rate search request, querying the plurality of data sources to obtain multi-dimensional information associated with the destination, and preprocessing the multi-dimensional information to obtain a multi-source retrieval result; arbitrating travel retrieval data of each travel knowledge tag in the multi-source retrieval result by adopting a preset arbitration mechanism to obtain arbitrated structured data; and based on the arbitrated structured data and a preset generation parameter, generating a destination travel knowledge recommendation result through a target language model. The technical problem that travel recommendation information depends on manual operation and is low in accuracy when a passenger ticket freight rate search system introduces a destination in the prior art is solved.

Description

Technical Field

[0001] This invention relates to the field of passenger ticket fare search technology or other related fields. Specifically, it relates to a travel knowledge recommendation method, apparatus, and electronic device based on a target language model. Background Technology

[0002] In the rapidly developing digital age, online travel booking platforms have become indispensable tools for people planning their trips. When users search for flight fares through these platforms, they are not only concerned with the price itself, but also expect to obtain more comprehensive destination information to help them make more informed travel decisions. However, existing ticket fare search systems have significant limitations in providing this kind of comprehensive information. For example, many online travel platforms rely on pre-written destination description pages. Because information updates depend on manual operation, the cycle is long, making it difficult to reflect the latest developments of the destination in a timely manner. For example, newly launched attractions, temporarily closed facilities, or sudden large-scale events are often not included quickly, leading to questions about the timeliness and accuracy of the information. To compensate for the shortcomings of static content, some systems attempt to automatically crawl relevant information from multiple third-party travel websites or vertical platforms and then roughly piece it together for display to users. Although this can increase the amount of information in a short time, it ignores the consistency and contextual relationship of information from different sources, easily leading to information duplication, conflicting descriptions, or fragmentation. For example, different descriptions of the same attraction may cause confusion, or different sources may provide contradictory information about the timing of the same event. This not only affects the overall quality of the information but also weakens users' trust in the platform.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This invention provides a travel knowledge recommendation method, apparatus, and electronic device based on a target language model, to at least solve the technical problem in related technologies where travel recommendation information relies on manual operation and has low accuracy when introducing destinations in passenger ticket fare search systems.

[0005] According to one aspect of the present invention, a travel knowledge recommendation method based on a target language model is provided, comprising: receiving a fare search request from a user, and extracting a set of key parameters and a set of travel knowledge tags from the fare search request, wherein the key parameters in the set of key parameters include at least a destination, and each travel knowledge tag in the set of travel knowledge tags corresponds to a specified type of travel knowledge; querying multiple data sources to obtain multi-dimensional information associated with the destination, and preprocessing the multi-dimensional information to obtain multi-source retrieval results; arbitrating the travel retrieval data of each travel knowledge tag in the multi-source retrieval results using a preset arbitration mechanism to obtain arbitrated structured data; and generating destination travel knowledge recommendation results based on the arbitrated structured data and preset generation parameters through a target language model, wherein the destination travel knowledge recommendation results include at least: travel knowledge data and travel suggestions corresponding to each travel knowledge tag.

[0006] Optionally, before extracting the set of key parameters and the set of travel knowledge tags from the fare search request, the method further includes: receiving the style configuration strategy, character length generation strategy, and topic weight allocation strategy input by the user, wherein the character length generation strategy includes at least a character count limit, and the topic weight allocation strategy includes weight parameters corresponding to each travel knowledge tag; and determining the preset generation parameters based on the style configuration strategy, the character length generation strategy, and the topic weight allocation strategy.

[0007] Optionally, the step of extracting the key parameter set and travel knowledge tag set from the fare search request includes: parsing the fare search request to obtain destination information, travel time parameter information, and user preference information, thereby obtaining the key parameter set; and obtaining all selected travel knowledge tags from the fare search request to obtain the travel knowledge tag set, wherein the types of travel knowledge tags include: climate and weather tags, attraction tags, food tags, activity experience tags, and travel tips tags.

[0008] Optionally, the step of querying multiple data sources to obtain multi-dimensional information related to the destination includes: for the climate and weather tag, querying key weather data from multiple meteorological organizations, climate databases, and environmental monitoring stations over multiple time periods, wherein the key weather data includes at least one of the following: temperature range, precipitation, humidity, UV index, air quality, and clothing recommendations; for the attraction tag, querying key attraction data from scenic area databases, online travel platforms, attraction review platforms, and historical and cultural heritage protection lists, wherein the key attraction data includes at least one of the following: opening hours, ticket prices, tour duration, best time to visit, unique features, and historical and cultural background; for the food tag, querying preset catering platforms, food blogs, local food records, and intangible cultural heritage... The data includes multiple key food items from the Intangible Cultural Heritage List, wherein the key food items include at least one of the following: signature dishes, recommended restaurants, flavor characteristics, price range, preparation techniques, and historical origins; for the activity experience tag, the data includes multiple key activity experience items from cultural activity calendars, tourism experience platforms, festival and celebration information, and local life recommendations, wherein the key activity experience items include at least one of the following: activity time, location information, participation method, experience duration, and price information; for the travel tips tag, the data includes multiple key travel data from traffic management department data, tourism safety announcements, consular reminders, and local practical information, wherein the key travel data includes at least one of the following: transportation guide, safety precautions, document requirements, emergency contact information, and consumption tips.

[0009] Optionally, the step of arbitrating the travel retrieval data of each travel knowledge tag in the multi-source retrieval results using a preset arbitration mechanism to obtain the arbitrated structured data includes: obtaining the basic weight, historical accuracy, timeliness factor, and tag type coefficient corresponding to each travel knowledge tag, wherein the basic weight is determined by weighting based on the weight of each data source; calculating the weighted travel retrieval data of each travel knowledge tag in the multi-source retrieval results using a preset arbitration mechanism based on the basic weight, historical accuracy, timeliness factor, and tag type coefficient corresponding to each travel knowledge tag to obtain the authority corresponding to each travel knowledge tag; and determining the arbitrated structured data based on the authority corresponding to each travel knowledge tag.

[0010] Optionally, the step of generating destination travel knowledge recommendation results through a target language model based on the arbitrated structured data and preset generation parameters includes: for each travel knowledge tag, obtaining the tag generation structure and preset key suggestions corresponding to that travel knowledge tag; generating travel recommendation content corresponding to each travel knowledge tag through the target language model based on the arbitrated structured data, preset generation parameters, the tag generation structure and preset key suggestions corresponding to each travel knowledge tag; adding tag type markers to each travel recommendation content to obtain travel knowledge recommendation results corresponding to each travel knowledge tag; and combining the travel knowledge recommendation results corresponding to each travel knowledge tag to generate the destination travel knowledge recommendation results.

[0011] Optionally, after generating destination travel knowledge recommendation results through a target language model based on the arbitrated structured data and preset generation parameters, the method further includes: collecting feedback information from the user regarding the travel knowledge recommendation results corresponding to each travel knowledge tag in the destination travel knowledge recommendation results; generating a feedback score based on the feedback information; and adjusting the basic weight and tag type coefficient corresponding to each travel knowledge tag by combining the feedback score, learning rate, and historical authority.

[0012] According to another aspect of the present invention, a travel knowledge recommendation device based on a target language model is also provided, comprising: a request parsing unit, configured to receive a fare search request from a user and extract a set of key parameters and a set of travel knowledge tags from the fare search request, wherein the key parameters in the set of key parameters include at least: a destination, and each travel knowledge tag in the set of travel knowledge tags corresponds to a specified type of travel knowledge; a multi-source retrieval unit, configured to query multiple data sources to obtain multi-dimensional information associated with the destination, and preprocess the multi-dimensional information to obtain multi-source retrieval results; a multi-source arbitration unit, configured to arbitrate the travel retrieval data of each travel knowledge tag in the multi-source retrieval results using a preset arbitration mechanism to obtain arbitrated structured data; and a model generation unit, configured to generate destination travel knowledge recommendation results based on the arbitrated structured data and preset generation parameters using a target language model, wherein the destination travel knowledge recommendation results include at least: travel knowledge data and travel suggestions corresponding to each travel knowledge tag.

[0013] Optionally, the travel knowledge recommendation device based on the target language model further includes: a strategy receiving unit, configured to receive a style configuration strategy, a character length generation strategy, and a topic weight allocation strategy input by the user before extracting the key parameter set and the travel knowledge tag set from the fare search request, wherein the character length generation strategy includes at least a character number limit, and the topic weight allocation strategy includes weight parameters corresponding to each travel knowledge tag; and a generation parameter determining unit, configured to determine the preset generation parameters based on the style configuration strategy, the character length generation strategy, and the topic weight allocation strategy.

[0014] Optionally, the request parsing unit includes: a request parsing module, used to parse the fare search request to obtain destination information, travel time parameter information, and user preference information, and to obtain the key parameter set; and a tag acquisition module, used to acquire all selected travel knowledge tags in the fare search request, and to obtain the travel knowledge tag set, wherein the types of the travel knowledge tags include: climate and weather tags, attraction tags, food tags, activity experience tags, and travel tips tags.

[0015] Optionally, the multi-source retrieval unit includes: a weather tag retrieval module, used to query key weather data from multiple meteorological organizations, climate databases, and environmental monitoring stations for multiple time periods for the climate and weather tag, wherein the key weather data includes at least one of the following: temperature range, precipitation, humidity, UV index, air quality, and clothing recommendations; a scenic spot tag retrieval module, used to query key scenic spot data from scenic spot databases, online travel platforms, scenic spot review platforms, and historical and cultural heritage protection lists for the scenic spot tag, wherein the key scenic spot data includes at least one of the following: opening hours, ticket price, tour duration, best time to visit, special features, and historical and cultural background; and a food tag retrieval module, used to query preset catering platforms, food blogs, local chronicle food records, and intangible cultural heritage lists for the food tag. The system includes several key food data entries, which include at least one of the following: signature dishes, recommended restaurants, flavor characteristics, price range, cooking techniques, and historical origins; an activity experience tag retrieval module, used to query multiple key activity experience data entries from cultural activity calendars, tourism experience platforms, festival celebrations, and local life recommendations for the activity experience tags, where the key activity experience data includes at least one of the following: activity time, location information, participation method, experience duration, and price information; and a travel advisory tag retrieval module, used to query multiple key travel data entries from traffic management department data, tourism safety announcements, consular reminders, and local practical information for the travel advisory tags, where the key travel data includes at least one of the following: traffic guides, safety precautions, document requirements, emergency contact information, and consumption tips.

[0016] Optionally, the multi-source arbitration unit includes: a tag weight acquisition module, used to acquire the basic weight, historical accuracy, timeliness factor, and tag type coefficient corresponding to each travel knowledge tag, wherein the basic weight is determined by weighting based on the weight of each data source; a weighted calculation module, used to perform weighted calculation on the travel retrieval data of each travel knowledge tag in the multi-source retrieval results based on the basic weight, historical accuracy, timeliness factor, and tag type coefficient corresponding to each travel knowledge tag, using a preset arbitration mechanism to obtain the authority corresponding to each travel knowledge tag; and an arbitration data determination module, used to determine the arbitrated structured data based on the authority corresponding to each travel knowledge tag.

[0017] Optionally, the model generation unit includes: a tag generation structure acquisition module, used to acquire the tag generation structure and preset key suggestions corresponding to each travel knowledge tag; a travel recommendation content generation module, used to generate travel recommendation content corresponding to each travel knowledge tag through the target language model based on the arbitrated structured data, preset generation parameters, the tag generation structure and preset key suggestions corresponding to each travel knowledge tag; a tag marking module, used to add tag type markings to each travel recommendation content to obtain travel knowledge recommendation results corresponding to each travel knowledge tag; and a result synthesis module, used to synthesize the travel knowledge recommendation results corresponding to each travel knowledge tag to generate the destination travel knowledge recommendation results.

[0018] Optionally, the travel knowledge recommendation device based on the target language model further includes: a feedback collection unit, used to collect feedback information from the user terminal regarding the travel knowledge recommendation results corresponding to each travel knowledge tag in the destination travel knowledge recommendation results after generating destination travel knowledge recommendation results through the target language model based on the arbitrated structured data and preset generation parameters; and a weight adjustment module, used to generate a feedback score based on the feedback information; and to adjust the basic weight and tag type coefficient corresponding to each travel knowledge tag by comprehensively considering the feedback score, learning rate, and historical authority.

[0019] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to execute the travel knowledge recommendation method based on the target language model described above.

[0020] According to another aspect of the present invention, an electronic device is also provided, including one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the travel knowledge recommendation method based on the target language model described above.

[0021] According to another aspect of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements the steps of the travel knowledge recommendation method based on the target language model described above.

[0022] In this disclosure, a fare search request is received from a user, and a set of key parameters and a set of travel knowledge tags are extracted from the fare search request. The key parameters in the set of key parameters include at least the destination, and each travel knowledge tag in the set of travel knowledge tags corresponds to a specified type of travel knowledge. Multiple data sources are queried to obtain multi-dimensional information related to the destination, and the multi-dimensional information is preprocessed to obtain multi-source retrieval results. A preset arbitration mechanism is used to arbitrate the travel retrieval data of each travel knowledge tag in the multi-source retrieval results to obtain arbitrated structured data. Based on the arbitrated structured data and preset generation parameters, destination travel knowledge recommendation results are generated through a target language model. The destination travel knowledge recommendation results include at least the travel knowledge data and travel suggestions corresponding to each travel knowledge tag.

[0023] This disclosure introduces a multi-source data retrieval and arbitration mechanism, combined with large language model generation technology, to generate travel knowledge that meets users' specific needs and preferences. This improves the accuracy and personalization of travel knowledge recommendation results, effectively integrates information from multiple channels, greatly reduces data conflicts and errors, and ensures the accuracy of travel knowledge recommendation results. This solves the technical problem in related technologies where travel recommendation information relies on manual operation and has low accuracy when introducing destinations in passenger ticket fare search systems. Attached Figure Description

[0024] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0025] Figure 1 This is a flowchart of an optional travel knowledge recommendation method based on a target language model according to an embodiment of the present invention;

[0026] Figure 2 This is a schematic diagram of an optional large-model-based controllable generation and dynamic optimization system for travel knowledge according to an embodiment of the present invention;

[0027] Figure 3 This is a schematic diagram of an optional method for controllable generation and dynamic optimization of travel knowledge based on a large model according to an embodiment of the present invention;

[0028] Figure 4 This is a schematic diagram of an optional travel knowledge recommendation device based on a target language model according to an embodiment of the present invention;

[0029] Figure 5 This is a hardware structure block diagram of an electronic device (or mobile device) that performs a travel knowledge recommendation method based on a target language model according to an embodiment of the present invention. Detailed Implementation

[0030] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0031] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0032] To facilitate understanding of the present invention by those skilled in the art, some terms or nouns involved in the various embodiments of the present invention are explained below:

[0033] The Domestic Ticket Price Search Engine (DTPSE) is an online system for searching and comparing domestic flight fares. It allows users to enter their departure point, destination, and travel date, and the system collects fare information from multiple airlines and other travel service providers to provide the most comprehensive fare comparisons and options.

[0034] Large Language Model (LLM) refers to a machine learning model with a large number of parameters used to understand and generate natural language.

[0035] User-generated content (UGC) refers to content created and published on a platform by users rather than by the platform's administrators. In the travel information field, UGC can include travel diaries, photos, reviews, and ratings.

[0036] A Multi-Source Retrieval Engine (MSRE) is a software component that can access and retrieve data from multiple data sources in parallel, including internal knowledge bases, external APIs, and UGC platforms.

[0037] It should be noted that the travel knowledge recommendation method and apparatus based on the target language model in this disclosure can be used in the field of passenger ticket fare search technology for the case of multi-source trimming and controllable generation of travel knowledge for passenger ticket fare search, and can also be used in any field other than the field of passenger ticket fare search technology for the case of multi-source trimming and controllable generation of travel knowledge for passenger ticket fare search. This disclosure does not limit the application field of the travel knowledge recommendation method and apparatus based on the target language model.

[0038] It should be noted that the information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, and displayed data) collected in this public disclosure are information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of related data all comply with the relevant laws, regulations, and standards of the relevant regions, necessary confidentiality measures have been taken, and they do not violate public order and good morals. Corresponding operation entry points are provided for users to choose to authorize or refuse. For example, this system has interfaces with relevant users or organizations. Before obtaining relevant information, a request to obtain the information needs to be sent to the aforementioned user or organization through the interface, and the relevant information is obtained only after receiving consent from the aforementioned user or organization.

[0039] It should be noted that in this disclosure, customer information is collected and analyzed, and users are provided with corresponding operation entry points to choose whether to agree to or reject the automated decision results; if the user chooses to reject, the process will proceed to the expert decision-making process.

[0040] The following embodiments of the present invention can be applied to various travel knowledge recommendation systems / applications / devices based on target language models. The present invention is applicable to online travel service scenarios, particularly in domestic airfare search applications. Specifically, when users search for flight fares through various online travel booking platforms, the present invention can provide deeply integrated destination travel knowledge, including but not limited to climate conditions, attraction descriptions, food recommendations, activity experiences, and travel safety tips.

[0041] This invention significantly improves the accuracy and personalization of travel knowledge by introducing a multi-source data retrieval and arbitration mechanism, a controllable large language model generation technology, and a closed-loop feedback optimization system. By dynamically adjusting the authority of data sources and considering historical accuracy and timeliness, this invention effectively filters and integrates information from multiple channels, greatly reducing data conflicts and errors, and ensuring high precision in travel knowledge. Based on specific tags in user query requests and real-time feedback, this invention can generate travel knowledge that meets specific user needs and preferences, improving information relevance and user experience, and satisfying the needs of personalized travel planning.

[0042] This invention continuously adjusts the system's behavior by collecting explicit and implicit user feedback, optimizing data source selection, generation strategies, and arbitration rules, forming a self-evolving and self-correcting dynamic optimization loop, thereby maintaining high service quality and applicability in the long term.

[0043] The present invention will now be described in detail with reference to various embodiments.

[0044] Example 1

[0045] According to an embodiment of the present invention, an embodiment of a travel knowledge recommendation method based on a target language model is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0046] Figure 1 This is a flowchart of an optional travel knowledge recommendation method based on a target language model according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0047] Optionally, before extracting the set of key parameters and the set of travel knowledge tags from the fare search request, the method further includes: receiving user input of a style configuration strategy, a character length generation strategy, and a topic weight allocation strategy. The character length generation strategy includes at least a character limit, and the topic weight allocation strategy includes weight parameters corresponding to each travel knowledge tag. Based on the style configuration strategy, the character length generation strategy, and the topic weight allocation strategy, preset generation parameters are determined.

[0048] This embodiment designs a flexible system initialization method. Before extracting the set of key parameters and the set of travel knowledge tags from the fare search request, it can receive various strategy settings input by the user, which helps to adjust and optimize the format and focus of the system-generated content to better meet the user's personalized needs.

[0049] In this embodiment, the style configuration strategy allows users to choose the style of the generated content. For example, the "Concise Guide" style tends to provide a concise overview, suitable for quick browsing; while the "Detailed Description" style focuses on providing detailed background information and in-depth analysis, suitable for in-depth research. This style configuration strategy can be adjusted according to user preferences or specific scenario requirements to ensure that the generated content matches the user's expectations. The character length generation strategy includes at least a character limit, which helps control the length of the generated content, ensuring that information is conveyed accurately and efficiently. The character limit can be flexibly adjusted according to user reading habits, platform display limitations, and content complexity. For example, for short trips or scenarios requiring quick information viewing, a shorter character limit can be set to generate more concise text. Regarding the topic weight allocation strategy, users can customize the weight parameters corresponding to each travel knowledge tag, which can be dynamically adjusted based on the user's interests or the characteristics of the destination. For example, if a user is particularly interested in food, they can increase the weight of food recommendations in the search request. The system will then generate more food-related information based on this preference, helping the system more accurately capture user interests and provide highly customized content presentation.

[0050] Based on the input of the above-mentioned style configuration strategy, character length generation strategy, and topic weight allocation strategy, this embodiment can determine a set of preset generation parameters. These parameters may include, but are not limited to, text style, length limit, and priority of travel knowledge tags. In this way, the system can more intelligently generate travel knowledge that not only meets user preferences but also complies with platform rules and display restrictions, further improving the applicability of information and user satisfaction.

[0051] Step S101: Receive the fare search request from the user and extract the key parameter set and travel knowledge tag set from the fare search request. The key parameters in the key parameter set include at least the destination, and each travel knowledge tag in the travel knowledge tag set corresponds to a specified type of travel knowledge.

[0052] Optionally, the steps of extracting the key parameter set and travel knowledge tag set from the fare search request include: parsing the fare search request to obtain destination information, travel time parameter information, and user preference information, and obtaining the key parameter set; obtaining all selected travel knowledge tags in the fare search request to obtain the travel knowledge tag set, wherein the types of travel knowledge tags include: climate and weather tags, attraction tags, food tags, activity experience tags, and travel tips tags.

[0053] Destination information forms the basis for generating destination travel knowledge, while travel time parameters help filter destination information related to a specific time point, such as seasonal activities or weather conditions. User preference information can be explicitly specified by the user during the search or automatically inferred by the system based on the user's past behavior; it helps the system generate travel knowledge that is more closely aligned with the user's interests.

[0054] This embodiment can accurately understand the user's query intent, not only limited to flight fares, but also covering diverse needs for destination knowledge. This helps the system provide richer, more detailed, and personalized travel information, enhancing the user's experience and decision support when planning their trip. For example, when a user searches for flights to a certain destination in September, the system can simultaneously notice that the user may be interested in climate information, attraction descriptions, and travel tips, and then proactively retrieve and generate relevant knowledge to provide the user with comprehensive guidance for their trip.

[0055] Step S102: Query multiple data sources to obtain multi-dimensional information about the associated destinations, and preprocess the multi-dimensional information to obtain multi-source retrieval results.

[0056] In this embodiment, after receiving a fare search request from the user and parsing out the destination and travel knowledge tags, the system can start the query process. For different types of travel knowledge tags, it can retrieve multi-dimensional information about related destinations from multiple data sources. This information is not limited to basic flight fares and schedules, but also covers a wealth of content such as weather, attractions, food, activities, and travel tips.

[0057] Optionally, the step of querying multiple data sources to obtain multi-dimensional information about the associated destination includes: for the climate and weather tag, querying key weather data from multiple meteorological organizations, climate databases, and environmental monitoring stations for multiple time periods, wherein the key weather data includes at least one of the following: temperature range, precipitation, humidity, UV index, air quality, and clothing recommendations; for the attraction tag, querying key attraction data from scenic area databases, online travel platforms, attraction review platforms, and historical and cultural heritage protection lists, wherein the key attraction data includes at least one of the following: opening hours, ticket prices, tour duration, best time to visit, special features, and historical and cultural background; for the food tag, querying preset catering platforms, food blogs, local food records, and intangible cultural heritage... The data includes several key culinary information from the cultural heritage list, including at least one of the following: signature dishes, recommended restaurants, flavor characteristics, price range, preparation techniques, and historical origins. For the activity experience tag, the data includes several key activity experience information from cultural activity calendars, tourism experience platforms, festival information, and local lifestyle recommendations, including at least one of the following: activity time, location information, participation methods, experience duration, and price information. For the travel advisory tag, the data includes several key travel information from traffic management departments, tourism safety notices, consular reminders, and local practical information, including at least one of the following: transportation guides, safety precautions, document requirements, emergency contact information, and consumption tips.

[0058] In handling climate and weather tags, this embodiment can access multiple meteorological organizations, climate databases, and environmental monitoring stations in parallel to collect key weather data over multiple time periods. This data may include, but is not limited to, temperature range, precipitation, humidity, UV index, air quality, and even clothing recommendations. By integrating this data, the system can provide a comprehensive climate overview, helping users make more appropriate travel preparations and clothing choices. For attraction tags, this embodiment will retrieve multiple key attraction data from scenic area databases, online travel platforms, attraction review platforms, and historical and cultural heritage protection lists, such as opening hours, ticket prices, tour duration, and best visiting times. It also includes the attraction's unique features and related cultural and historical background information, helping users understand the attraction comprehensively, plan reasonable travel time, and even inspire a deeper understanding and exploration of the destination's culture.

[0059] Furthermore, in processing food tags, this embodiment extracts various key food data by querying preset catering platforms, food blogs, local chronicles of food records, and intangible cultural heritage lists. This data includes featured dishes, recommended restaurants, flavor characteristics, price ranges, preparation techniques, and historical origins. This not only provides users with practical catering information but also enhances the fun and cultural experience of travel, allowing users to appreciate local characteristics and historical stories while enjoying authentic cuisine. When processing activity experience tags, this embodiment searches cultural activity calendars, tourism experience platforms, festival information, and local lifestyle recommendations to obtain a series of key activity experience data, including activity time, location information, participation methods, experience duration, and price information. This helps users enrich their itinerary, participate in local activities, and improve the overall travel experience and memorability. For travel advisory tags, this embodiment collects key travel data from data sources such as traffic management departments, travel safety notices, consular reminders, and local practical information. This data includes transportation guides, safety tips, document requirements, emergency contact information, and consumption tips. This helps users avoid common travel problems, ensure personal safety, and make reasonable financial plans and budgets.

[0060] Step S103: Use a preset arbitration mechanism to arbitrate the travel retrieval data of each travel knowledge tag in the multi-source retrieval results to obtain the arbitrated structured data.

[0061] In this embodiment, arbitration analysis is performed on the travel search data under each type of travel knowledge tag, and finally, structured data after arbitration is generated. This helps to filter out the most accurate and relevant information, reduce data conflicts, and improve the quality and reliability of travel knowledge.

[0062] Optionally, the step of arbitrating the travel retrieval data for each travel knowledge tag in the multi-source retrieval results using a preset arbitration mechanism to obtain the arbitrated structured data includes: obtaining the basic weight, historical accuracy, timeliness factor, and tag type coefficient corresponding to each travel knowledge tag, wherein the basic weight is determined by weighting based on the weight of each data source; based on the basic weight, historical accuracy, timeliness factor, and tag type coefficient corresponding to each travel knowledge tag, performing weighted calculation on the travel retrieval data for each travel knowledge tag in the multi-source retrieval results using a preset arbitration mechanism to obtain the authority corresponding to each travel knowledge tag; and determining the arbitrated structured data based on the authority corresponding to each travel knowledge tag.

[0063] Among them, the basic weight reflects the inherent credibility of the data source, the historical accuracy quantifies the accuracy of the information from the data source over a period of time, the timeliness factor measures the freshness of the information, and the label type coefficient fine-tunes the weights according to the category of travel knowledge to ensure that the information under each label receives appropriate attention.

[0064] Based on the parameters obtained above, a pre-defined arbitration mechanism is used to weight the travel search data under each travel knowledge tag, deriving the authority level for each tag. The authority level calculation formula comprehensively considers the weight of the data source, historical performance, and real-time nature, providing a quantitative evaluation of the information and helping the system make informed choices when faced with conflicts between different data sources. After calculating the authority level, the system can determine the arbitrated structured data based on the authority level of each travel knowledge tag, ensuring that the final output travel knowledge is both accurate and authoritative. Furthermore, the structured data format facilitates the subsequent controllable generation module, enabling more precise control over the style, length, and focus of the generated content.

[0065] The arbitration mechanism in this embodiment not only helps the system fairly and scientifically compare and select the most reliable information when faced with conflicts from different data sources, but also dynamically adjusts weights based on real-time data performance. This allows the system to adapt to changes in data sources and maintain the accuracy and timeliness of information. For example, for climate and weather tags, if data provided by a meteorological organization has a high accuracy rate over the past month, the combination of its base weight and historical accuracy will increase its authority, making it more likely to be prioritized in the arbitration process. Similarly, for attraction tags, if a tourist attraction's database is frequently updated, the timeliness factor can reflect the freshness of its data, thus obtaining a higher authority score in arbitration.

[0066] By employing a method of authority calculation and structured data determination, this embodiment can ensure that the generated travel knowledge not only covers various types of information that users want to know, but also guarantees the quality of information from the source. This helps to increase users' trust in the travel knowledge generated by the system, while also improving the efficiency of travel planning and the accuracy of decision-making.

[0067] Step S104: Based on the arbitrated structured data and preset generation parameters, generate destination travel knowledge recommendation results through the target language model. The destination travel knowledge recommendation results include at least: travel knowledge data and travel suggestions corresponding to each travel knowledge tag.

[0068] In this embodiment, destination travel knowledge recommendation results can be generated through the target language model. These results not only cover the travel knowledge data corresponding to each travel knowledge tag, but also incorporate targeted travel suggestions, helping users obtain comprehensive and specific travel information.

[0069] Optionally, the step of generating destination travel knowledge recommendation results using a target language model based on the arbitrated structured data and preset generation parameters includes: for each travel knowledge tag, obtaining the tag generation structure and preset key suggestions corresponding to that travel knowledge tag; generating travel recommendation content corresponding to each travel knowledge tag using a target language model based on the arbitrated structured data, preset generation parameters, the tag generation structure and preset key suggestions corresponding to each travel knowledge tag; adding tag type markers to each travel recommendation content to obtain travel knowledge recommendation results corresponding to each travel knowledge tag; and combining the travel knowledge recommendation results corresponding to each travel knowledge tag to generate destination travel knowledge recommendation results.

[0070] For each travel knowledge tag, this embodiment can obtain a specific tag generation structure. For example, the generation structure of the climate and weather tag may involve the weather conditions of the day, the forecast for the next few days, and corresponding clothing suggestions. At the same time, the system can also read preset key suggestions, such as suggestions that emphasize local specialties and the average cost per person in food recommendations.

[0071] Utilizing the aforementioned tag generation structure and preset key suggestions, combined with the arbitrated structured data and preset generation parameters, the system generates travel recommendation content corresponding to each travel knowledge tag through a target language model. This fully leverages the natural language generation capabilities of the large model to provide personalized recommendations that are both factually accurate and meet user reading preferences and information needs. It's worth noting that the system can also add specific typographical markers to each part of the generated travel recommendations, facilitating hierarchical source display on the front end. For example, for climate data, markers can be embedded in the generated text to indicate that the data originates from a meteorological organization and to assess its real-time performance and reliability.

[0072] This embodiment, by combining target language models and structured data, not only generates reliable, data-driven information but also provides personalized and contextualized travel recommendations based on the characteristics of different travel knowledge tags and user needs. This helps improve the applicability and appeal of travel knowledge, making users feel more satisfied and convenient when obtaining destination information. Simultaneously, by adding typified tags, the system can improve information transparency, making it easier for users to understand the data source and quality, and enhancing user trust in the content generated by the system.

[0073] Optionally, after generating destination travel knowledge recommendation results through the target language model based on the arbitrated structured data and preset generation parameters, the method further includes: collecting user feedback on the travel knowledge recommendation results corresponding to each travel knowledge tag in the destination travel knowledge recommendation results; generating feedback scores based on the feedback information; and adjusting the basic weights and tag type coefficients corresponding to each travel knowledge tag by comprehensively considering the feedback scores, learning rate, and historical authority.

[0074] Optionally, after generating destination travel knowledge recommendations, this embodiment can further collect user feedback on these recommendations. This step focuses on user acceptance and satisfaction with the recommended content under each travel knowledge tag, including but not limited to user likes, dislikes, corrections, and frequent browsing or sharing of certain knowledge points. Collecting feedback helps the system understand which travel knowledge tags are most popular with users and which information needs further improvement, providing practical data support for subsequent optimization efforts.

[0075] This embodiment generates feedback scores based on collected user feedback information. The scoring mechanism comprehensively considers both explicit and implicit user feedback. It can be a direct score of recommended content under a single travel knowledge tag, or an indirect score inferred from user actions such as copying, sharing, or re-clicking. Generating feedback scores helps provide an objective performance evaluation of the data source under each travel knowledge tag, providing a basis for subsequent weight adjustments.

[0076] To continuously improve the quality and relevance of destination travel knowledge recommendations, this embodiment comprehensively considers feedback ratings, preset learning rates, and historical authority to dynamically adjust the base weight and tag type coefficient for each travel knowledge tag. This adjustment mechanism automatically learns from real user feedback, optimizing the data source selection strategy over time. For example, if a user provides poor feedback on information from a data source under the attraction tag, the system will reduce the base weight of that data source, decreasing its proportion in future searches, while simultaneously increasing the weight of another data source to seek higher-quality information. Similarly, if a user shows high interest in information from travel tips tags, the system can increase the importance of this information, adjusting the corresponding tag type coefficient to give it a more prominent position in future knowledge generation.

[0077] Through the above steps, fare search requests from users can be received, and a set of key parameters and a set of travel knowledge tags can be extracted from the fare search requests. The key parameters in the key parameter set include at least the destination, and each travel knowledge tag in the travel knowledge tag set corresponds to a specific type of travel knowledge. Multiple data sources are queried to obtain multi-dimensional information about the associated destinations, and this multi-dimensional information is preprocessed to obtain multi-source search results. A preset arbitration mechanism is used to arbitrate the travel search data for each travel knowledge tag in the multi-source search results, resulting in arbitrated structured data. Based on the arbitrated structured data and preset generation parameters, destination travel knowledge recommendation results are generated through a target language model. These destination travel knowledge recommendation results include at least the travel knowledge data and travel suggestions corresponding to each travel knowledge tag. In this embodiment, by introducing a multi-source data retrieval and arbitration mechanism, combined with large language model generation technology, travel knowledge that meets the specific needs and preferences of users is generated, improving the accuracy and personalization of travel knowledge recommendation results. This effectively integrates information from multiple channels, greatly reduces data conflicts and errors, and ensures the accuracy of travel knowledge recommendation results. Thus, it solves the technical problem in related technologies where travel recommendation information relies on manual operation and has low accuracy when introducing destinations in passenger ticket fare search systems.

[0078] The following describes in detail another optional implementation method.

[0079] This invention provides a controllable generation and dynamic optimization system for travel knowledge based on a large model. First, based on a user's request to query destination fares, the system automatically triggers a destination knowledge generation process. The system ensures the accuracy of information through multi-source data retrieval and arbitration mechanisms, and continuously optimizes the generation effect based on user feedback.

[0080] Figure 2 This is a schematic diagram of an optional large-model-based controllable generation and dynamic optimization system for travel knowledge according to an embodiment of the present invention, such as... Figure 2 As shown, the system comprises six parts: a request parsing module, a multi-source retrieval engine, a multi-source arbitration module, a controllable generation module, a feedback processing module, and a storage medium. Each part will be described in detail below.

[0081] The request parsing module is responsible for receiving user fare search requests, extracting key parameters such as destination and time from the requests, and triggering the destination knowledge generation process.

[0082] Multi-source retrieval engine: Responsible for parallel querying of internal knowledge base and external data sources to obtain multi-dimensional information related to the destination (including climate, attractions, food, etc.), and performing preliminary cleaning and formatting of data from different sources.

[0083] Multi-source arbitration module: responsible for conflict detection and authority assessment of multi-source retrieval results, and uses dynamic weight calculation and multi-source voting mechanism to select the optimal result to ensure the accuracy and reliability of the output information.

[0084] Controllable generation module: Responsible for receiving the structured data after arbitration, and generating natural language descriptions that meet the requirements through a large model based on preset generation parameters (style, length, key themes, etc.).

[0085] Feedback processing module: responsible for collecting explicit user feedback (likes / dislikes) and implicit behaviors (copying / sharing), analyzing the quality of generated content, and dynamically adjusting arbitration weights and generation strategies.

[0086] Storage medium: The physical device for data storage, used to store multi-source data interface configurations, authority weight tables, user feedback data, generation history, etc. Arbitration weights and user preferences are stored in key-value pairs, and generated content is stored in text format.

[0087] Figure 3 This is a schematic diagram of an optional method for controllable generation and dynamic optimization of travel knowledge based on a large model according to an embodiment of the present invention, such as... Figure 3 As shown, it includes the following steps:

[0088] Step 1: Parameter adjustment.

[0089] The generated parameters are set according to the pre-configured strategy, including the style of the generated content (such as "concise guide" or "detailed description"), target length (character limit), the themes to be highlighted (such as the weighting of "climate", "food" and "attractions"), and security filtering rules.

[0090] Step 2: Parse the request.

[0091] Receive the user's destination query request, parse the request parameters, and extract core information: destination name, travel time, and potential preferences (if any) obtained through the user's history or real-time interactions. If no clear destination information is available, terminate the subsequent process.

[0092] Step 3: Multi-source data retrieval and arbitration.

[0093] Parallel retrieval: Based on destination parameters, query internal knowledge bases (such as attractions and policies), external authoritative website APIs, and verified UGC platform content in parallel.

[0094] Multi-source arbitration and verification: Consistency verification of search results. If information from multiple data sources conflicts, an arbitration mechanism is initiated. Based on the preset authority weight of the data sources (e.g., institutional data sources have a higher weight than commercial platforms), historical accuracy, and data timeliness, a comprehensive weight is calculated, and the data result with the highest weight is selected. Optionally, the arbitration result is temporarily stored in a predetermined database in a structured data format.

[0095] Step 4: Parameterized generation control.

[0096] The large language model is invoked to combine the arbitrated structured data with the generation parameters set in step one to generate a natural language description that meets the requirements. The generation process is constrained by parameters to ensure that the content style, length, and focus are consistent with the presets.

[0097] Step 5: Embedded tag injection and result output, including source tagging and travel knowledge recommendation results output.

[0098] Hidden source tags are embedded in the generated natural language text for key factual points (such as temperature data and the opening status of tourist attractions), recording the data source ID, authority score, and update timestamp. The front end parses and renders the data based on the tags, providing users with a tiered source display (such as hover tooltip).

[0099] Step Six: Feedback Collection and Dynamic Optimization.

[0100] Collect feedback: Collect explicit feedback (such as likes, dislikes, and error corrections) and implicit feedback (such as copying and sharing of certain types of information and subsequent search keywords) from users in real time.

[0101] Analysis and optimization: Dynamically adjust system behavior based on feedback data.

[0102] Search strategy: If the accuracy of a data source continues to decline, reduce its authority weight.

[0103] Generation parameters: If users frequently interact with a certain type of content, the generation weight of that type of topic will be increased accordingly.

[0104] Arbitration Rules: Optimize the calculation formulas for arbitration thresholds and weights.

[0105] The adjusted strategy takes effect immediately and is used for subsequent request processing, forming a closed loop of continuous optimization.

[0106] Specifically, the implementation details of this embodiment in the process of performing multi-source arbitration and controllable method generation of travel knowledge for fare search may include:

[0107] 1. Request parsing and tag recognition.

[0108] When a user initiates a fare search request, the system automatically parses the request parameters and identifies the travel knowledge tags contained within. Each tag corresponds to a specific type of travel knowledge:

[0109] Dst01: Climate and weather information (such as temperature, precipitation, and seasonal characteristics);

[0110] Dst03: Attraction information (such as attraction introduction, opening hours, and ticket information);

[0111] Dst04: Food information (such as specialty dishes, restaurant recommendations, food culture);

[0112] Dst08: Activity Experiences (such as local activities, experiences, and entertainment recommendations);

[0113] Dst09: Travel tips (such as transportation guides, precautions, and useful information).

[0114] Example:

[0115] When a user request contains "Dst01, Dst03, Dst04", the system will generate three types of travel information in parallel: climate information, attraction introductions, and food recommendations.

[0116] 2. Implementation of multi-source data retrieval and processing.

[0117] The system employs differentiated data retrieval and processing strategies for the five categories of travel knowledge tags:

[0118] Dst01 (Weather):

[0119] Data sources: Meteorological Organization API, Meteorological Organization data, historical climate database, and real-time data from environmental monitoring stations;

[0120] Key data: temperature range, precipitation, humidity, UV index, air quality, and clothing recommendations;

[0121] Update frequency: Real-time data (updated every 30 minutes), seasonal data (updated annually);

[0122] Key processing areas: numerical accuracy verification, extreme weather early warning, and seasonal feature extraction;

[0123] Example: Searching for "Beijing September climate" yields data such as temperature 25-28℃, precipitation 60mm, and humidity 65%.

[0124] Dst03 (Attraction Information):

[0125] Data sources: A-level scenic area database, online travel platform attraction data, user reviews and ratings, and cultural heritage protection list;

[0126] Key data: opening hours, ticket prices, tour duration, best time to visit, highlights, historical and cultural background;

[0127] Key areas of focus: verification of open status, analysis of visitor experience, and exploration of cultural value;

[0128] Example: Search for " The museum is open from 8:30 to 17:00 and offers admission tickets. The tour costs approximately 300 yuan and is recommended to last 3-4 hours.

[0129] Dst04 (Food Recommendation):

[0130] Data sources: food platforms, food blogs, local chronicles of food, and intangible cultural heritage lists;

[0131] Key data: Featured dishes, recommended restaurants, flavor characteristics, price range, cooking techniques, and historical background;

[0132] Key processing areas: aggregation of word-of-mouth reviews, identification of local characteristics, and assessment of consumption levels;

[0133] Example: Searching for "Peking duck" yields... Germany Recommended time-honored brands such as Fang, with an average cost of 150-300 yuan per person.

[0134] Dst08 (Event Experience):

[0135] Data sources: cultural event calendars, tourism experience platforms, festival and celebration information, and local lifestyle recommendations;

[0136] Key data: event time, location, participation method, experience duration, target audience, and pricing information;

[0137] Key areas of focus: verifying time validity, assessing experience value, and analyzing ease of participation;

[0138] Example: Searching for "Beijing September events" yields... celebration, Experience programs such as performances and alleyway tours.

[0139] Dst09 (Travel Advisory):

[0140] Data sources: Traffic management department data, travel safety notices, consular reminders, and local practical information;

[0141] Key data: Traffic guide, safety precautions, document requirements, emergency contact information, and consumer tips;

[0142] Key areas of focus: security assessment, usability verification, and timeliness assurance;

[0143] Example: Search for " "Travel Tips" provides information on subway operating hours, taxi fare rates, and travel warnings.

[0144] 3. Implementation of a multi-source arbitration mechanism.

[0145] For each data type with a unified arbitration mechanism, but with differentiated weight configurations:

[0146] Formula for calculating authority:

[0147] Authority = Basic weight × Historical accuracy × Timeliness factor × Type coefficient;

[0148] in:

[0149] Basic weights: inherent weights of data sources (institutional data 0.95, commercial data 0.80, UGC data 0.60).

[0150] Historical accuracy: Accuracy of data over the past 30 days (number of correct answers / total number of correct answers);

[0151] Timeliness factor: 1 / (1+0.2×data delay in hours);

[0152] Type coefficient: Adjusted according to different label types (Dst01=1.2, Dst03=1.0, Dst04=0.9, Dst08=1.1, Dst09=1.3).

[0153] Collision detection threshold:

[0154] Different conflict determination thresholds are used for different types of data:

[0155] Numerical data (temperature, price): relative difference >5%;

[0156] Text-based data (descriptions, suggestions): semantic similarity < 0.75;

[0157] Stateful data (open / closed): direct comparison for consistency;

[0158] Severe conflict: Triggers manual review.

[0159] Example:

[0160] When multiple sources provide different temperature data:

[0161] Meteorological organization: 25.3℃ (authority rating 0.95);

[0162] Commercial platform: 26.0℃ (authority level 0.80);

[0163] UGC content: 24.5℃ (authority level 0.60);

[0164] Weighted average temperature = (25.3×0.95+26.0×0.80+24.5×0.60) / (0.95+0.80+0.60) = 25.6℃.

[0165] 4. Controllable content generation and implementation.

[0166] For each tag type, a differentiated generation strategy is adopted:

[0167] Dst01 (climate) generation:

[0168] Structure: Current situation description + trend prediction + action recommendations;

[0169] Key points: Data accuracy, practical suggestions;

[0170] Example: "The average temperature in Beijing in September is 25℃, with a 30% chance of precipitation. It is recommended to bring rain gear."

[0171] Dst03 (Scenic Spot) Generation:

[0172] Structure: Featured introduction + Practical information + Tour suggestions;

[0173] Key points: cultural connotations, practical information;

[0174] Example: Open from 8:30 to 17:00, a visit of 3-4 hours is recommended.

[0175] Dst04 (Food) generated:

[0176] Structure: Featured Recommendations + Restaurant Information + Consumption Guide;

[0177] Key factors: reputation and consumer spending levels;

[0178] Example: Peking duck "The average German spends 200-300 yuan."

[0179] Dst08 (Activity) Generation:

[0180] Structure: Activity introduction + participation information + experiential value;

[0181] Key points: Participation methods and experience;

[0182] Example: "The Mid-Autumn Festival celebration requires advance reservations and is suitable for family participation."

[0183] Dst09 (prompt) generated:

[0184] Structure: Important Notice + Practical Information + Emergency Guidelines;

[0185] Key points: Safety and usability;

[0186] Example: "It is recommended to take the subway and keep your personal belongings safe."

[0187] 5. Embedded tagging and tracing.

[0188] Add typed tags to each generated content:

[0189]

[0190] data-source="weather_gov";

[0191] data-confidence="0.95">;

[0192] The average temperature in Beijing in September is 25℃.

[0193] Differentiated display strategy:

[0194] Dst01: Displays data update time and quality score;

[0195] Dst03: Displays image links and user ratings;

[0196] Dst04: Provides restaurant location and average cost per person;

[0197] Dst08: Includes event time and reservation methods;

[0198] Dst09: Emphasize safety precautions.

[0199] 6. Dynamic optimization and feedback mechanism.

[0200] Establish an independent optimization system for each tag type:

[0201] Feedback collection:

[0202] Explicit feedback: user likes, dislikes, and error corrections;

[0203] Implicit feedback: reading time, sharing behavior, subsequent searches;

[0204] Behavioral feedback: Actual behavioral conversion based on generated content;

[0205] Authority Adjustment Formula:

[0206] New authority level = old authority level × (1-α) + feedback score × α;

[0207] Where α=0.1 is the learning rate, and the feedback score is calculated based on user feedback.

[0208] Type-based optimization strategy:​

[0209] Dst01 feedback: It affects the weight of meteorological data sources and adjusts the climate prediction algorithm;

[0210] Dst03 feedback: Optimized attraction recommendation strategy, updated opening time verification;

[0211] Dst04 feedback: Adjusted restaurant recommendation weights and updated price information;

[0212] Dst08 Feedback: Optimized the timeliness of activity recommendations and updated the participation method;

[0213] Dst09 feedback: Strengthen security information verification and update emergency contact information.

[0214] 7. Complete processing example.

[0215] Extract user request: {"Destination":["BJS Beijing"],"Dst00":["Dst01 Climate","Dst03 Attractions","Dst09 Travel Tips"],"Date":"23SEP25"}.

[0216] Processing procedure:

[0217] Parse the request to identify the weather, attraction, and travel tips that need to be generated;

[0218] Parallel retrieval of three types of data:

[0219] Climate data: temperature, precipitation, air quality;

[0220] Attraction data: opening hours, ticket prices;

[0221] Travel data: traffic information, safety tips;

[0222] Arbitration is performed for each type of data:

[0223] Climate: Weighted average temperature 25.6℃;

[0224] Attractions: Choose the opening hours with the highest level of authority;

[0225] Travel: Verify traffic safety information;

[0226] Generate three types of content, add type tags, and generate the final response according to the configuration file parameters.

[0227] The above implementation methods allow for information retrieval from multiple authoritative data sources, and a dynamic authority arbitration mechanism effectively resolves information conflicts and outdated issues. By comprehensively evaluating basic weights, historical accuracy, timeliness factors, and type coefficients, the most reliable data source is selected, ensuring that the destination-related knowledge received by users is authentic, accurate, and authoritative. This reduces travel inconvenience caused by incorrect or incomplete information, and enhances user trust and satisfaction.

[0228] The controllable content generation module enables the system to generate travel knowledge that is both compliant and personalized for each travel knowledge tag based on user-defined parameters. This includes not only controlling the level of detail but also matching style and theme, ensuring that the output content meets the user's specific information needs while maintaining a style consistent with the platform's positioning. Furthermore, by dynamically adjusting the generation strategy, the system can continuously optimize the quality of generated content, enhancing the practicality and engaging nature of the travel knowledge.

[0229] By collecting user feedback on generated content in real time, this implementation method can dynamically adjust the authority weight of the data source and the generation strategy, forming a self-learning and optimization capability. This ensures that the system can continuously evolve based on users' real needs and experiences, maintain its information quality and service level in the long term, reduce manual maintenance costs, and improve the long-term effectiveness and competitiveness of the service.

[0230] The technical solution of this invention helps travelers quickly obtain accurate, comprehensive, and personalized destination knowledge when planning their trips, covering multi-dimensional information such as climate, attractions, food, activities, and travel tips. This not only saves travelers time in information gathering but also improves the accuracy of their decisions, increasing their anticipation and satisfaction with their trip. The optimized travel knowledge generation can better meet the specific needs of different user groups, such as families, backpackers, or business travelers, thereby enhancing the overall travel planning experience.

[0231] The following is a detailed description with reference to another embodiment.

[0232] Example 2

[0233] The travel knowledge recommendation device based on the target language model provided in this embodiment includes multiple implementation units, each of which corresponds to a specific implementation step in Embodiment 1 above.

[0234] Figure 4 This is a schematic diagram of an optional travel knowledge recommendation device based on a target language model according to an embodiment of the present invention, such as... Figure 4 As shown, the travel knowledge recommendation device based on the target language model may include: a request parsing unit 41, a multi-source retrieval unit 42, a multi-source arbitration unit 43, and a model generation unit 44.

[0235] The request parsing unit 41 is used to receive the fare search request from the user and extract the key parameter set and the travel knowledge tag set from the fare search request. The key parameters in the key parameter set include at least the destination, and each travel knowledge tag in the travel knowledge tag set corresponds to a specified type of travel knowledge.

[0236] The multi-source retrieval unit 42 is used to query multiple data sources, obtain multi-dimensional information about the associated destination, and preprocess the multi-dimensional information to obtain multi-source retrieval results.

[0237] The multi-source arbitration unit 43 is used to arbitrate the travel retrieval data of each travel knowledge tag in the multi-source retrieval results using a preset arbitration mechanism, so as to obtain the arbitrated structured data.

[0238] The model generation unit 44 is used to generate destination travel knowledge recommendation results based on the arbitrated structured data and preset generation parameters through the target language model. The destination travel knowledge recommendation results include at least: travel knowledge data and travel suggestions corresponding to each travel knowledge tag.

[0239] The aforementioned travel knowledge recommendation device based on the target language model can receive fare search requests from users through the request parsing unit 41, and extract the key parameter set and travel knowledge tag set from the fare search request. The key parameters in the key parameter set include at least the destination. Each travel knowledge tag in the travel knowledge tag set corresponds to a specified type of travel knowledge. The multi-source retrieval unit 42 queries multiple data sources to obtain multi-dimensional information related to the destination, and preprocesses the multi-dimensional information to obtain multi-source retrieval results. The multi-source arbitration unit 43 uses a preset arbitration mechanism to arbitrate the travel retrieval data of each travel knowledge tag in the multi-source retrieval results to obtain arbitrated structured data. The model generation unit 44 generates destination travel knowledge recommendation results based on the arbitrated structured data and preset generation parameters through the target language model. The destination travel knowledge recommendation results include at least the travel knowledge data and travel suggestions corresponding to each travel knowledge tag. In this embodiment, by introducing a multi-source data retrieval and arbitration mechanism, combined with large language model generation technology, travel knowledge that meets the specific needs and preferences of users is generated, improving the accuracy and personalization of travel knowledge recommendation results. This effectively integrates information from multiple channels, greatly reduces data conflicts and errors, and ensures the accuracy of travel knowledge recommendation results. Thus, it solves the technical problem in related technologies where travel recommendation information relies on manual operation and has low accuracy when introducing destinations in passenger ticket fare search systems.

[0240] Optionally, the travel knowledge recommendation device based on the target language model further includes: a strategy receiving unit, used to receive a style configuration strategy, a character length generation strategy, and a topic weight allocation strategy input by the user before extracting the key parameter set and the travel knowledge tag set from the fare search request, wherein the character length generation strategy includes at least a character number limit, and the topic weight allocation strategy includes weight parameters corresponding to each travel knowledge tag; and a generation parameter determining unit, used to determine preset generation parameters based on the style configuration strategy, the character length generation strategy, and the topic weight allocation strategy.

[0241] Optionally, the request parsing unit includes: a request parsing module, used to parse the fare search request to obtain destination information, travel time parameter information, and user preference information, and to obtain a set of key parameters; and a tag acquisition module, used to acquire all selected travel knowledge tags in the fare search request, and to obtain a set of travel knowledge tags, wherein the types of travel knowledge tags include: climate and weather tags, attraction tags, food tags, activity experience tags, and travel tips tags.

[0242] Optionally, the multi-source retrieval unit includes: a weather tag retrieval module, used to query key weather data from multiple meteorological organizations, climate databases, and environmental monitoring stations for multiple time periods for climate and weather tags, wherein the key weather data includes at least one of the following: temperature range, precipitation, humidity, UV index, air quality, and clothing recommendations; a scenic spot tag retrieval module, used to query key scenic spot data from scenic spot databases, online travel platforms, scenic spot review platforms, and historical and cultural heritage protection lists for scenic spot tags, wherein the key scenic spot data includes at least one of the following: opening hours, ticket prices, tour duration, best time to visit, special features, and historical and cultural background; and a food tag retrieval module, used to query preset catering platforms, food blogs, local chronicle food records, and intangible cultural heritage lists for food tags. The system retrieves multiple key food data points, including at least one of the following: signature dishes, recommended restaurants, flavor characteristics, price range, cooking techniques, and historical origins. The activity experience tag retrieval module is used to query multiple key activity experience data points from cultural activity calendars, tourism experience platforms, festival and celebration information, and local life recommendations, including at least one of the following: activity time, location information, participation method, experience duration, and price information. The travel advisory tag retrieval module is used to query multiple key travel data points from traffic management department data, tourism safety notices, consular reminders, and local practical information, including at least one of the following: traffic guides, safety precautions, document requirements, emergency contact information, and consumption tips.

[0243] Optionally, the multi-source arbitration unit includes: a tag weight acquisition module, used to acquire the basic weight, historical accuracy, timeliness factor, and tag type coefficient corresponding to each travel knowledge tag, wherein the basic weight is determined by weighting based on the weight of each data source; a weighted calculation module, used to perform weighted calculation on the travel retrieval data of each travel knowledge tag in the multi-source retrieval results based on the basic weight, historical accuracy, timeliness factor, and tag type coefficient corresponding to each travel knowledge tag, using a preset arbitration mechanism to obtain the authority corresponding to each travel knowledge tag; and an arbitration data determination module, used to determine the structured data after arbitration based on the authority corresponding to each travel knowledge tag.

[0244] Optionally, the model generation unit includes: a tag generation structure acquisition module, used to acquire the tag generation structure and preset key suggestions corresponding to each travel knowledge tag; a travel recommendation content generation module, used to generate travel recommendation content corresponding to each travel knowledge tag through a target language model based on the arbitrated structured data, preset generation parameters, the tag generation structure and preset key suggestions corresponding to each travel knowledge tag; a tag marking module, used to add tag type tags to each travel recommendation content to obtain the travel knowledge recommendation result corresponding to each travel knowledge tag; and a result synthesis module, used to synthesize the travel knowledge recommendation results corresponding to each travel knowledge tag to generate destination travel knowledge recommendation results.

[0245] Optionally, the travel knowledge recommendation device based on the target language model further includes: a feedback collection unit, used to collect user feedback on the travel knowledge recommendation results corresponding to each travel knowledge tag in the destination travel knowledge recommendation results after generating destination travel knowledge recommendation results through the target language model based on the arbitrated structured data and preset generation parameters; a weight adjustment module, used to generate a feedback score based on the feedback information; and to adjust the basic weight and tag type coefficient corresponding to each travel knowledge tag by comprehensively considering the feedback score, learning rate, and historical authority.

[0246] The aforementioned travel knowledge recommendation device based on the target language model may further include a processor and a memory. The aforementioned request parsing unit 41, multi-source retrieval unit 42, multi-source arbitration unit 43, model generation unit 44, etc., are all stored in the memory as program units, and the processor executes the aforementioned program units stored in the memory to realize the corresponding functions.

[0247] The aforementioned processor contains a kernel, which retrieves the corresponding program units from memory. One or more kernels can be configured, and adjusting kernel parameters can implement multi-source arbitration of travel knowledge and recommendation of travel knowledge results for fare search.

[0248] The aforementioned memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0249] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored computer program, wherein, when the computer program is running, it controls the device where the computer-readable storage medium is located to execute the travel knowledge recommendation method based on any one of the above embodiments.

[0250] According to another aspect of the present invention, an electronic device is also provided, including one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the travel knowledge recommendation method based on the target language model as described in any of the first embodiments above.

[0251] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the travel knowledge recommendation method based on a target language model as described in various embodiments of this application.

[0252] This application also provides a computer program product, including a non-volatile computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the travel knowledge recommendation method based on a target language model as described in various embodiments of this application.

[0253] Figure 5 This is a hardware structure block diagram of an electronic device (or mobile device) that performs a travel knowledge recommendation method based on a target language model according to an embodiment of the present invention. Figure 5 As shown, an electronic device may include one or more ( Figure 5 The processor 502 (which may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and a memory 504 for storing data may also be included. In addition, it may include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, a keyboard, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 5 The structure shown is for illustrative purposes only and does not limit the structure of the electronic device described above. For example, the electronic device may also include components that are more... Figure 5The more or fewer components shown, or having the same Figure 5 The different configurations shown.

[0254] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0255] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0256] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0257] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0258] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0259] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0260] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A travel knowledge recommendation method based on a target language model, characterized in that, include: The system receives a fare search request from a user and extracts a set of key parameters and a set of travel knowledge tags from the fare search request. The key parameters in the set of key parameters include at least the destination, and each travel knowledge tag in the set of travel knowledge tags corresponds to a specified type of travel knowledge. Multiple data sources are queried to obtain multi-dimensional information related to the destination, and the multi-dimensional information is preprocessed to obtain multi-source retrieval results; A preset arbitration mechanism is used to arbitrate the travel retrieval data of each travel knowledge tag in the multi-source retrieval results, resulting in arbitrated structured data; Based on the arbitrated structured data and preset generation parameters, destination travel knowledge recommendation results are generated through the target language model. The destination travel knowledge recommendation results include at least: travel knowledge data and travel suggestions corresponding to each travel knowledge tag.

2. The travel knowledge recommendation method according to claim 1, characterized in that, Before extracting the set of key parameters and the set of travel knowledge tags from the fare search request, the process also includes: The system receives style configuration strategy, character length generation strategy, and topic weight allocation strategy input from the user terminal. The character length generation strategy includes at least a character count limit, and the topic weight allocation strategy includes weight parameters corresponding to each travel knowledge tag. Based on the style configuration strategy, the character length generation strategy, and the topic weight allocation strategy, the preset generation parameters are determined.

3. The travel knowledge recommendation method according to claim 1, characterized in that, The steps for extracting the set of key parameters and the set of travel knowledge tags from the fare search request include: The fare search request is parsed to obtain destination information, travel time parameters, and user preference information, thus obtaining the set of key parameters; Obtain all selected travel knowledge tags from the fare search request to obtain the travel knowledge tag set, wherein the types of travel knowledge tags include: climate and weather tags, attraction tags, food tags, activity experience tags, and travel tips tags.

4. The travel knowledge recommendation method according to claim 3, characterized in that, The steps of querying multiple data sources to obtain multi-dimensional information related to the destination include: For the climate and weather tag, query key weather data from multiple meteorological organizations, climate databases, and environmental monitoring stations for multiple time periods. The key weather data includes at least one of the following: temperature range, precipitation, humidity, UV index, air quality, and clothing recommendations. For the aforementioned attraction tags, query multiple key attraction data from scenic area databases, online travel platforms, attraction review platforms, and historical and cultural heritage protection lists. The key attraction data includes at least one of the following: opening hours, ticket price, tour duration, best time to visit, special features, and historical and cultural background. For the aforementioned food tags, multiple key food data are queried from preset catering platforms, food blogs, local food records, and intangible cultural heritage lists. The key food data includes at least one of the following: signature dishes, recommended restaurants, flavor characteristics, price range, preparation techniques, and historical origins. For the aforementioned activity experience tag, query multiple key activity experience data from cultural activity calendars, tourism experience platforms, festival celebration information, and local life recommendations. The key activity experience data includes at least one of the following: activity time, location information, participation method, experience duration, and price information. For the aforementioned travel advisory label, multiple key travel data are queried from traffic management department data, travel safety notices, consular reminders, and local practical information. Among these, the key travel data includes at least one of the following: traffic guide, safety precautions, document requirements, emergency contact information, and consumption tips.

5. The travel knowledge recommendation method according to claim 1, characterized in that, The steps of arbitrating the travel retrieval data for each travel knowledge tag in the multi-source retrieval results using a preset arbitration mechanism to obtain arbitrated structured data include: Obtain the basic weight, historical accuracy, timeliness factor, and tag type coefficient corresponding to each travel knowledge tag, wherein the basic weight is determined by weighting the weights of each data source; Based on the basic weight, historical accuracy, timeliness factor and tag type coefficient corresponding to each travel knowledge tag, a preset arbitration mechanism is used to perform weighted calculation on the travel retrieval data of each travel knowledge tag in the multi-source retrieval results to obtain the authority corresponding to each travel knowledge tag. Based on the authority level corresponding to each travel knowledge tag, the structured data after arbitration is determined.

6. The travel knowledge recommendation method according to claim 1, characterized in that, Based on the arbitrated structured data and preset generation parameters, the steps for generating destination travel knowledge recommendation results using a target language model include: For each travel knowledge tag, obtain the tag generation structure and preset key suggestions corresponding to that travel knowledge tag; Based on the arbitrated structured data, preset generation parameters, the tag generation structure corresponding to each travel knowledge tag, and preset key suggestions, travel recommendation content corresponding to each travel knowledge tag is generated through the target language model; Add a typified tag to each of the travel recommendation contents to obtain the travel knowledge recommendation result corresponding to each travel knowledge tag; By combining the travel knowledge recommendation results corresponding to each of the aforementioned travel knowledge tags, the destination travel knowledge recommendation results are generated.

7. The travel knowledge recommendation method according to claim 1, characterized in that, After generating destination travel knowledge recommendation results using the target language model based on the arbitrated structured data and preset generation parameters, the process further includes: Collect feedback information from the user terminal regarding the travel knowledge recommendation results corresponding to each travel knowledge tag in the destination travel knowledge recommendation results; A feedback score is generated based on the feedback information; Based on the feedback ratings, learning rates, and historical authority, the base weights and tag type coefficients corresponding to each travel knowledge tag are adjusted.

8. A travel knowledge recommendation device based on a target language model, characterized in that, include: The request parsing unit is used to receive a fare search request from a user and extract a set of key parameters and a set of travel knowledge tags from the fare search request. The key parameters in the set of key parameters include at least the destination, and each travel knowledge tag in the set of travel knowledge tags corresponds to a specified type of travel knowledge. The multi-source retrieval unit is used to query multiple data sources to obtain multi-dimensional information related to the destination, and to preprocess the multi-dimensional information to obtain multi-source retrieval results. The multi-source arbitration unit is used to arbitrate the travel retrieval data of each travel knowledge tag in the multi-source retrieval results using a preset arbitration mechanism, so as to obtain the arbitrated structured data. The model generation unit is used to generate destination travel knowledge recommendation results based on the arbitrated structured data and preset generation parameters through a target language model. The destination travel knowledge recommendation results include at least: travel knowledge data and travel suggestions corresponding to each travel knowledge tag.

9. An electronic device, characterized in that, The system includes one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the travel knowledge recommendation method based on the target language model as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the travel knowledge recommendation method based on the target language model as described in any one of claims 1 to 7.

Images