Method and computer program product for information presentation of travel products

Abstract

This specification provides one or more embodiments of a method for displaying information about travel products and a computer program product. The method is applied to a travel service platform and includes: determining travel-related information input by a user, and obtaining product recommendation information generated by a product recommendation model based on the travel-related information, wherein the product recommendation information indicates at least one set of travel products and summary descriptive terms for each set of travel products; displaying the summary descriptive terms on a platform client, and displaying corresponding travel products for each triggered summary descriptive term.

Description

Technical Field

[0001] This specification relates to one or more embodiments in the field of travel services, and more particularly to a method for displaying information about travel products and a computer program product. Background Technology

[0002] In related technologies, travel service platforms recommend travel products to users through their platform clients for them to choose from. Currently, there are generally two main logics for travel product recommendations: one is a fixed product stream push based on the user's current location or historical behavior; the other is a simple matching and sorting strategy based on user-input keywords. Taking hotels as an example, after a user logs into the travel service platform, the platform client's homepage will display some recommended popular hotels, or further display hotels that match the user's input of a filter condition (such as destination). Summary of the Invention

[0003] In view of the above, one or more embodiments of this specification provide the following technical solutions: According to a first aspect of one or more embodiments of this specification, a method for displaying information about travel products is proposed, applied to a travel service platform, the method comprising: Determine the travel-related information input by the user, and obtain product recommendation information generated by the product recommendation model based on the travel-related information. The product recommendation information is used to indicate at least one set of travel products and summary descriptive words for each set of travel products. The summary descriptive terms are displayed on the platform client, and corresponding travel products are displayed for the triggered summary descriptive terms.

[0004] According to a second aspect of one or more embodiments of this specification, a computer program product is provided, comprising a computer program / instructions that, when executed by a processor, implement the steps of the method as described in the first aspect.

[0005] As can be seen from the first aspect above, this specification proposes a novel method for displaying travel product information. The travel service platform first determines the travel-related information input by the user and obtains product recommendation information generated by the product recommendation model based on the information. Then, it displays the summary descriptive words indicated by the product recommendation information on the platform client, and further displays the travel products corresponding to the descriptive words when any summary descriptive word is triggered.

[0006] Understandably, this solution uses a pre-trained product recommendation model to generate targeted summary descriptive terms for semantic grouping of travel products. This organizes the disorganized travel products on the platform into clearly defined product groups, enabling structured and intent-driven recommendations based on user-inputted travel-related information. This improves the relevance of recommended content to the intended travel information and enhances the accuracy of product recommendations. Furthermore, by generating targeted summary descriptive terms, typical products can be proactively presented to users, helping them understand options and stimulating their potential interest. This allows for the discovery / mining of users' latent travel needs without requiring precise input. Additionally, since users can clearly express their specific intentions by triggering summary descriptive terms, the system can display corresponding travel products based on the triggering results, achieving accurate recommendations tailored to the user's current intent, resulting in significant improvements. Attached Figure Description

[0007] Figure 1 This is a schematic diagram of the architecture of a travel service platform provided in an exemplary embodiment.

[0008] Figure 2 This is a flowchart illustrating an exemplary embodiment of a method for displaying information about travel products.

[0009] Figure 3 This is a schematic diagram of an information input window provided in an exemplary embodiment.

[0010] Figure 4 This is an exemplary embodiment providing a schematic diagram of the display effect of summary descriptive terms and travel products.

[0011] Figure 5 This is an exemplary embodiment providing another summary description and a schematic diagram illustrating the display effect of travel products.

[0012] Figure 6 This is a schematic diagram of a map floating page provided in an exemplary embodiment.

[0013] Figure 7 This is an exemplary embodiment illustrating the triggering and display effect of recommended prompts.

[0014] Figure 8 This is a schematic diagram of the structure of a device provided in an exemplary embodiment.

[0015] Figure 9 This is a block diagram of an information display device for travel products provided in an exemplary embodiment. Detailed Implementation

[0016] The user 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, displayed data, etc.) involved in this manual, as well as the historical evaluation information described below, are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse. This is hereby stated.

[0017] In related technologies, travel service platforms recommend travel products to users through their platform clients for purchase. Currently, travel product recommendation logic typically falls into two categories: one is a fixed product stream push based on the user's current location or historical behavior; the other is a simple matching and sorting strategy based on user-input keywords. Taking hotels as an example, after a user logs into a travel service platform, the platform client's homepage displays some recommended popular hotels, or further displays hotels that match the user's input of a filter condition (such as destination). The inventors have found that the recommendation logic in these related technologies is relatively fixed and simplistic, making it difficult to meet users' travel needs and failing to fully explore their potential travel demands, thus requiring urgent improvement.

[0018] To address this issue, this specification proposes a novel information display scheme for travel products. A pre-trained product recommendation model generates product recommendations based on user-inputted travel-related information, indicating at least one set of travel products and their summative descriptive terms. Responding to user triggers, the system selectively displays travel products corresponding to the triggered summative descriptive terms. This aims to achieve accurate product recommendations while further stimulating users' potential interest and uncovering / mining their latent travel needs. The following detailed description, in conjunction with accompanying drawings and corresponding embodiments, illustrates this scheme.

[0019] Figure 1 This is a schematic diagram of the architecture of a travel service platform provided in an exemplary embodiment. Figure 1 As shown, the platform may include a server 11 and several electronic devices, such as mobile phones 12-14.

[0020] During platform operation, server 11 can run server-side programs for a certain application to implement the relevant functions of that application. For example, when server 11 runs a travel service program, it can act as the server for that service (hereinafter referred to as the platform server). Server 11 can be a physical server containing an independent host, or it can be a virtual server hosted by a host cluster. This embodiment of the specification does not limit this.

[0021] Any electronic device can run a client-side program for an application to implement the application's related functions. For example, when the electronic device runs a travel service program, it can act as a client for that service (hereinafter referred to as a platform client, etc.). The client-side program for the travel service can be launched and run on the corresponding electronic device. For example, this program can be a native application installed on the electronic device, or it can be a mini-program, quick app, or other similar form. Of course, when using web technologies such as HTML5 or similar technologies, the related functions can be implemented through a page displayed by a browser. Here, the browser can be a standalone browser application or a browser module embedded in some applications. Furthermore, mobile phones are only one type of electronic device that users can use. In fact, users can obviously also use electronic devices such as PCs (Personal Computers), tablets, laptops, PDAs (Personal Digital Assistants), wearable devices (such as smart glasses, smartwatches, etc.), etc., and one or more embodiments in this specification do not limit this.

[0022] It is understandable that, from a hardware perspective, a travel service platform includes servers (such as the aforementioned server 11) and electronic devices (such as the aforementioned mobile phones 12-14); while from a software perspective, the platform includes the platform server (i.e., the server-side application of the travel service platform) and the platform client (i.e., the client-side application of the travel service platform).

[0023] Travel service platforms are at least used to provide travel product sales services to users (i.e. consumers). Users can browse relevant information about travel products online through the platform and then place orders to purchase travel products. For example, the travel products described in the embodiments of this specification may include, but are not limited to, transportation products (such as air tickets, train tickets, bus tickets / long-distance passenger transport, car rental / self-driving services, airport transfers / private car services, etc.), accommodation products (such as star-rated / chain hotels / homestays / apartment hotels, hot spring inns / treehouses / tent campsites / ancient inns, family-friendly hotels / pet-friendly hotels / accessible rooms, etc.), destination activities products (such as attraction tickets, day trips / multi-day trips, special activities, performance / sports tickets, etc.), packaged travel products (or product packages, such as independent travel packages, group tours, customized tours, themed tours, etc.), visa and insurance products (such as visa application services, travel insurance, etc.), and value-added services (such as Wi-Fi rental / international roaming packages, luggage storage / fast-track customs services, travel guides / itinerary planning tools, etc.). This application does not limit the specific content, form, and quantity of travel products.

[0024] The users described in this specification can be either legally registered users on the travel service platform or unregistered temporary users (or "tourists" on the platform). This specification does not impose any limitations on these categories. Users can access the travel service platform through a platform client running on their electronic devices and interact with the client through specific operations. This allows them to control the platform client to exchange data with the platform server, enabling them to browse product information and even place orders online.

[0025] Furthermore, the network 10 for interaction between electronic devices such as mobile phones 12 and server 11 can be implemented using either wired or wireless networks, depending on the communication methods supported by the respective electronic devices. This specification does not impose any restrictions on this. For example, mobile phones 12 typically only support wireless communication, so they can use wireless networks for communication; while PCs can communicate with server 11 using either wired or wireless methods, which will not be elaborated further.

[0026] Figure 2 This is a flowchart illustrating an exemplary method for displaying information about travel products, which can be applied to the aforementioned travel service platform. For example... Figure 2 As shown, the method includes the following steps 202-204.

[0027] Step 202: Determine the travel-related information input by the user, and obtain product recommendation information generated by the product recommendation model based on the travel-related information. The product recommendation information is used to indicate at least one set of travel products and summary descriptive terms for each set of travel products.

[0028] Users can access travel product sales services offered by the travel service platform through the platform's client. Users can input travel-related information through the client, which then submits this information to the platform's server. The travel-related information described in this specification refers to raw input data that reflects the user's current travel interests and intentions. Essentially, it is the initial expression of the user's potential travel needs. This information can provide a semantic starting point for the product recommendation model, generating product groupings and summary descriptive terms with contextual guidance. Furthermore, the travel-related information may not correspond to specific travel products, but rather be brief, generalized, or even vague query content (such as "a certain city," "cherry blossoms," "family trip," "National Day holiday," etc.), which implies the user's initial interest in one or more travel dimensions such as destination, activity theme, time, target audience, and budget. The travel-related information can take the form of text, images, and / or audio / video.

[0029] In one embodiment, a first input window can be displayed on the platform client, allowing users to input relevant travel-related information according to their needs. For example, the system can determine the user's input travel-related information in response to the user's input in the first input window. In this case, the user directly inputs information in the first input window displayed on the client, which aligns with the user habit of "quick search" (especially in scenarios where the platform client is a mobile device). The input operation and results are relatively intuitive, and the path from "thinking" to "searching" is shorter, which helps improve the search initiation rate and subsequent click conversion.

[0030] For example, in response to a user's input triggering action in the first input window, the first input window can be updated to a second input window; wherein the size of the editable area of ​​the second input window is larger than the size of the editable area of ​​the first input window. Furthermore, in response to the user's information input action in the second input window, the travel-related information input by the user can be determined. It is understood that in a larger editable area, users can input more data and / or more dimensions of travel-related information (such as "taking elderly relatives to City A in October, budget 5000, hoping to stay in a hot spring hotel"), thereby specifying richer initial semantic information to the platform, facilitating the expression of diverse and complex needs. Moreover, the larger second input window can integrate intelligent prompts (such as prompt text such as "you can enter destination, time, companions, interests, etc."), segmented tags, or templates to help users organize their language, reducing the difficulty of expression, especially beneficial for users who are not yet clear about their specific needs to gradually clarify their intentions. More complete and structured user input also helps the product recommendation model more accurately identify multi-dimensional intentions (such as simultaneously focusing on "people + budget + activities"), thereby generating more accurate summary descriptive terms and product groupings. In addition, the input triggering operation can be a regular input triggering action, such as clicking in the input box; or it can be a specific preset operation, such as clicking the "I want to input diverse / multimodal information" button. In this mode, a compact first input window is displayed by default to keep the interface simple; and a larger area is expanded only when the user intends to input more information, thus balancing the design contradiction between "lightweight homepage" and "advanced function usability".

[0031] For example, the information input interface displayed by the platform client is as follows: Figure 3As shown, input box 301 in (a) is the first input window, and input box 302 in (b) is the second input window. In input box 301, users can directly input travel-related information in text form, such as "City A." This method provides a smooth "one-click" experience for users with low cognitive load and high certainty (e.g., "I just want to go to City A"). Alternatively, when a user triggers (e.g., clicks) input box 301, the client can refresh and display input box 302, updating the conventional text input box to a larger smart input box, allowing the user to input more and richer travel-related information. This method provides ample space for expression and guidance for users with high uncertainty who need to explore or combine conditions (e.g., "I want to find a cheap beach that's also good for taking photos").

[0032] In one embodiment, the travel-related information can be used to characterize the user's generalized travel intention, and the product recommendation information can be generated by the product recommendation model based on the generalized travel intention, thus possessing strong targeting. The generalized travel intention can be used to characterize the user's travel intention from at least one of the following dimensions: travel destination, travel time period, travel budget, core travel activities, travel style orientation, travelers, and mode of transportation. Since the user's true travel intention cannot be accurately and uniquely determined based on the travel-related information, the aforementioned generalized travel intention can also be called potential travel intention, which is the user's possible travel intention inferred / identified based on the travel-related information.

[0033] The travel destination is used to indicate where the user wants to travel. The user can enter the name of the destination (e.g., ...). Figure 3 (As shown in the example of "City A"). This type of input directly points to geographic space and is the most basic starting point for users' travel decisions. Based on this, the product recommendation model can generate summary descriptive terms such as "Popular cherry blossom viewing spots in City A" or "Hot hotels in popular shopping districts of Area P in City A".

[0034] The travel time period is used to indicate when a user wants to travel, such as when the user inputs "May Day holiday", "December", "National Day", "this weekend", etc. Time information is used to help the product recommendation model determine the price fluctuations and peak traffic of seasonal activities (such as cherry blossom viewing, skiing, etc.), and then generate summary descriptive terms such as "special offer airfare to city A in December" and "limited-time cherry blossom season itinerary".

[0035] Travel budget indicates how much a user wants to spend to complete a trip (i.e., the user focuses on "how much money to spend"), such as a user inputting "travel to country X for 3,000 yuan" or "package with a budget of no more than 5,000 yuan per person". Budget information determines the recommended product tier and combination strategy. For example, the product recommendation model can generate summary descriptive terms such as "high-value Osaka independent travel" or "family travel under 1,000 yuan".

[0036] The core travel activities indicate what activities users want to participate in during their trip (i.e., what users are interested in "doing"), such as users inputting "cherry blossoms in country X," "cherry blossom viewing," or "hot springs." This activity information shows that users have clear interests and preferences, and the product recommendation model can generate summary descriptive terms such as "popular cherry blossom viewing spots in city A" or "ski resorts in city B."

[0037] Travel style orientation is used to indicate the travel style that a user is interested in (such as how the user cares about "how to play" or "what feeling they want"). For example, if a user enters "I want to have a slow-paced X-culture trip," the style orientation reflects the user's travel preferences and psychological needs. Based on this, the product recommendation model can generate summary descriptive terms with emotional value orientation, such as "a quiet ancient town suitable for relaxing" or "light hiking + hot spring slow life."

[0038] The "travelers" category indicates who / who is in the user's travel plans (e.g., who the user is interested in "traveling with"). For example, a user might input "taking the kids to City A" or "accompanying parents to see XX." The travelers' identities influence accommodation type, transportation arrangements, and activity design. The product recommendation model can then generate customized summary descriptions such as "family-friendly hot spring hotels" or "romantic hot spring packages for couples."

[0039] Transportation mode indicates the user's planned mode of transportation for travel (i.e., how the user is concerned about "how to get there / how to travel / how to return"), such as "direct flights to country X" or "sleeper trains to city A". Transportation preferences directly affect itinerary planning and product matching, and the product recommendation model can generate practical summary descriptions such as "hotels near country X" or "10-minute walk from the high-speed rail station".

[0040] Of course, in addition to the dimensions mentioned above, travel-related information can also be used to describe the user's travel intentions in detail from many other dimensions such as travel style (e.g., independent travel, self-driving, group tours), culture / knowledge (e.g., historical sites, museums, intangible cultural heritage), and food (e.g., night markets, ramen, snacks). These will not be elaborated further.

[0041] Traditional solutions in related technologies often require users to provide relatively clear and specific intent information to trigger effective recommendation results. However, in real-world scenarios, many users are in the early stages of travel decision-making, and their search behavior exhibits highly generalized characteristics—for example, entering only broad terms like "City A" or "cherry blossoms" expresses initial interest in the destination and seasonal activities, rather than a specific product need. Traditional recommendation systems, lacking a deep understanding of users' potential intent, struggle to effectively organize and present information. Faced with a massive amount of uncategorized travel products (such as flights, hotels, local activities, and packages), users often find themselves in a dilemma of "information overload but no way to start," unable to quickly grasp the core activities and product structure of a destination, or efficiently identify options that match their preferences. This solution identifies generalized travel intents from the user's input of travel-related information, helping to fully uncover potential user needs and accurately generate product recommendations matching those needs. This results in more precise and diverse travel product recommendations, improving information discovery efficiency and user experience in scenarios with broad intents.

[0042] In one embodiment, the product recommendation information can be generated through a product recommendation model at an appropriate time. For example, the model can be temporarily invoked and product recommendation information generated based on travel-related information input by the user. After determining the user's generalized travel intent based on travel-related information, various travel products related to the generalized travel intent can be further identified, and prompt words can be constructed based on the corresponding product description information. Then, the prompt words are input into the product recommendation model to obtain the product recommendation information output by the model for the generalized travel intent. The prompt words in this method are dynamically constructed based on the current real-time product pool and the user's specific intent, reflecting the latest inventory, price fluctuations, promotional activities, or seasonal changes (such as "hot spring specials in December"), ensuring that the recommended content is consistent with the current market situation. It also supports long-tail / niche intents. Even if the user inputs a low-frequency or combined general intent (such as "traveling with elderly + budget 4000 + city A + slow pace"), the system can instantly aggregate relevant products and generate appropriate descriptive words, avoiding unresponsiveness due to lack of pre-caching. It fully leverages the semantic understanding, induction, and generation capabilities of the product recommendation model, producing novel, diverse, and context-rich summary descriptive terms (such as "a tea room suitable for quiet sitting during ginkgo season"), effectively enhancing the guidance effect.

[0043] For example, product recommendation information can be generated and cached in advance based on preset or historical travel-related information input by the user. Then, when the user inputs travel-related information in the current session, this information can be directly returned and displayed to the user. After determining the user's generalized travel intent based on the travel-related information, product recommendation information specific to the generalized travel intent can be queried from the product recommendation information cached by the travel service platform for various travel intents. Each cached product recommendation for a travel intent is generated by the product recommendation model in response to a prompt word for that intent, and this prompt word is constructed based on the product description information of various travel products related to that intent. It is understood that while temporarily calling the model and generating product recommendation information involves more model calls, it offers higher timeliness; while pre-calling the model and caching product recommendation information, although slightly less timely, can significantly reduce model overhead and improve the response speed to user requests. Therefore, the above methods can be flexibly selected according to the actual situation. This approach boasts high response speed, enhancing user experience (it returns results in milliseconds without real-time model calls, making it particularly suitable for high-concurrency, low-latency scenarios such as homepages, search dropdowns, and popular destinations). It also significantly reduces computational costs and system load, avoiding large model calls for every request, saving GPU (Graphics Processing Unit) / TPU (Tensor Processing Unit) resources, and improving overall system throughput and stability. Furthermore, it ensures consistent recommendation quality: for high-frequency intents, pre-generated content can be manually reviewed and optimized through multiple rounds to ensure accurate, compliant, and marketing-friendly descriptions (e.g., avoiding generating illegal expressions like "cheapest"). It also naturally supports precise operations for trending products, proactively pre-generating high-quality recommendation content for holidays, major promotions, and featured destinations (e.g., "May Day City A Tour"), improving product recommendation quality.

[0044] For example, some preset travel-related information can be created, or some common and popular travel-related information can be statistically summarized from historical requests received by the platform. Then, the platform server can call the product recommendation model to generate corresponding product recommendation information in batches according to a preset period (such as daily, hourly, etc., which can be flexibly set according to the update cycle of specific products), and cache these travel-related information and corresponding product recommendation information locally on the server (or in a cloud database). Subsequently, when any user input of any travel-related information is received from the platform client at any time, the system can check whether the travel-related information exists in the cached preset / popular travel-related information: if it does not exist, it indicates that the information is relatively novel (the corresponding travel intention is relatively rare / niche), and the product recommendation model can be called to temporarily generate product recommendation information corresponding to the information; conversely, if it exists, it indicates that the information is preset / popular travel-related information (the corresponding travel intention is easily predictable or is a historical hot intention of other users), and the corresponding product recommendation information can be retrieved from the cache and directly returned to the platform client for display. This method can respond to popular and niche travel intentions separately, which helps to improve the accuracy of product recommendation information.

[0045] In one embodiment, the product recommendation model described in this specification can be a generative language model, such as an LLM (Large Language Model) or a VLM (Vision-Language Model). The generative language model can utilize massive amounts of prior textual knowledge to naturally generate fluent, diverse, and contextualized descriptive terms (such as "A City's secret spot with ginkgo leaves covering the path"), and supports unseen intent combinations (such as "with pet + hot spring + budget 5000"), effectively handling long-tail / niche travel intents without retraining.

[0046] After identifying the various travel products related to the generalized travel intention, prompts can be constructed using product description information from at least some of these travel products. The product description information for any travel product may include at least some data from the product details page and / or at least some consumer review data corresponding to historical orders, and this information may be structured data.

[0047] In addition, prompts need to be constructed in a corresponding format according to the specific form of the product recommendation model. For example, when the product recommendation model is LLM, the product description information should only contain text information in order to construct text-formatted prompts; while when the product recommendation model is VLM, the product description information may contain text and / or image information, in which case text and / or image-formatted prompts can be constructed.

[0048] In addition, the product recommendation model can also employ an intent classification / clustering / rule template system. This system consists of an intent recognition module (such as a BERT-based classifier with bidirectional encoder representation, used to determine the generalized travel intent corresponding to travel-related information), a product clustering module (such as K-means, hierarchical clustering algorithms, used to group candidate travel products according to semantics / attributes), and a preset template engine (such as used to generate corresponding summary descriptive terms for dimensions such as "travel destination" and "core travel activities"). The internal logic of each module / engine is clear (not a logical black box), with strong interpretability and controllability, which helps to achieve efficient and accurate control over the travel recommendation logic. Alternatively, the product recommendation model can also employ a knowledge graph-enhanced recommendation model (KG-based Recommender), such as using a travel domain knowledge graph (such as "City A → Related attractions → Region P; Related season → Cherry blossom viewing; Related transportation → Airport A") for reasoning. This approach also has strong interpretability, is suitable for generating reasons for "why recommend," and can also assist in constructing prompt words, which will not be elaborated further.

[0049] Of course, a hybrid architecture can also be used. For example, a non-generative model can be used for initial screening and grouping (such as quickly recalling relevant travel products and clustering them by attribute based on general user intent), while a generative language model can be used to generate summary descriptive words (such as constructing prompt words from clustering results and corresponding product descriptions, and inputting them into an LLM to generate summary descriptive words). For example, the prompt word could be "You are a travel assistant. Please generate an attractive Chinese title for the following hotel group: These hotels are located in the business district of area P in city A, within a ≤5-minute walk of the subway station, priced between 600-900 yuan, most include free breakfast, suitable for independent travelers." Correspondingly, the LLM outputs "High-value independent travel hotels in area P" or "Hotels with breakfast included near the subway in area P," etc. This hybrid architecture of "traditional model data preparation + LLM recommendation result generation" ensures both efficiency and controllability, while also leveraging the linguistic expressiveness of generative models, making it highly practical.

[0050] In one embodiment, when the generalized travel intent is used to represent the user's travel intent from the perspective of travel destination, the prompt words constructed for the generalized travel intent may include product description information of travel products related to the travel destination. For example, if the user inputs "City A" as travel-related information, 100 hotel / flight / entertainment products related to City A can be filtered accordingly. Prompt words can then be constructed using the product description information of at least some of these products. For instance, if there are 400 hotel products, prompt words can be constructed using the product description information of each of these 400 hotel products; alternatively, these 100 hotel products can be categorized (e.g., by price or distance from attractions), and prompt words can be constructed using the product description information of a portion of the products in each categorized product set. This method has a clear data source and strong controllability (the input information comes from real travel products on the platform, which helps ensure the authenticity and redeemability of the recommended content). Moreover, the summary descriptions generated based on product description information can directly reflect the core selling points of existing travel products. Once triggered by the user, they can seamlessly jump to the corresponding product list, simplifying the purchase process and helping to improve recommendation conversion efficiency.

[0051] And / or, the prompts constructed for the generalized travel intention may include, in addition to product descriptions of travel products related to the generalized travel intention (such as product descriptions of at least some travel products related to the generalized travel intention), non-product-related contextual information related to the generalized travel intention. This non-product-related contextual information refers to auxiliary external or environmental information in the travel recommendation scenario that does not directly originate from the attributes or descriptions of the available travel products (such as airfares, hotels, tickets, packages, etc.), but is closely related to the user's travel decision and can enhance their understanding of the destination or travel intention. Essentially, this information supplements and expands the semantics of the travel scenario and can originate from internal or external knowledge systems, real-time environmental dynamics, user group behavior trends, or cultural and geographical common sense, used to answer the "meta-questions" that users often encounter during the decision-making process for the product recommendation model. Examples include: "Is this place worth visiting now?", "What's special about this season?", "Why do people recommend this place?", "Besides accommodation and transportation, what else can I experience?", etc. This type of information does not have direct transactional attributes (i.e., it cannot be "purchased"), but it profoundly influences the user's perception and interest in the destination, ultimately affecting the user's travel decision.

[0052] For example, the non-product-related contextual information may include geographical and cultural knowledge (such as "City A has 17 World Cultural Heritage sites", "City A is known as the 'Hometown of XX'"), seasonality and festivals (such as "December is the ski season in City A", "The best time to view cherry blossoms in City A is from late March to early April", "City A is hosting the X event", etc.), policies and travel conditions (such as "Citizens of Country A can apply for e-visas to Country B", "Country A offers visa-free entry to tourists from Country A for X years", "City A has a weekend traffic restriction policy based on odd and even license plate numbers", etc.), user behavior and market trends (such as "The search volume for 'City A + family travel' has increased by 150% in the past 7 days", "The collection rate of this destination is higher than the average of similar cities", etc.), weather and natural conditions (such as "The current temperature in City A is 28℃, sunny and suitable for beach activities", "The lavender in City A will enter its peak blooming period in mid-June", etc.), media and word-of-mouth content (such as "Rated by XX as one of the most worthwhile cities to visit in X continent in 2025"), etc., which will not be elaborated further.

[0053] Understandably, the non-product contextual information enhances the scenario awareness of product recommendation models. For example, phrases like "cherry blossom season," "hometown of XX," and "family-friendly city" help the model understand why certain products are worth recommending, thus generating more narrative and emotionally resonant descriptive terms, such as "Winter lighting collection in City A" and "High-value hotels in District P." Furthermore, it helps activate latent demand and improves the efficiency of demand discovery: users may not know that "City A has an X celebration in December," but they may become interested after seeing "X celebration limited-time packages." In this case, the non-product contextual information stimulates demand, solving the problem of "users not knowing what they want." It also improves the timeliness and sensitivity to trending topics in the product recommendation information generated by the model: incorporating real-time events (such as festivals, exhibitions, weather), policy changes, or public opinion trends keeps the recommended content fresh and relevant, avoiding "static product piling up." Furthermore, this solution supports cold start or product-sparse scenarios: for newly launched destinations or those with few products (such as niche islands), even with insufficient product data, attractive guiding words can be generated using external context (such as "hailed as the last hidden gem of XX") to maintain user experience. This helps ensure that the generated XX more closely reflects real travel decision-making logic: users not only consider "what products are available," but also care about "whether this place is worth visiting" and "whether now is a good time to go," and non-product-related contextual information is key to accurately answering these questions.

[0054] Step 204: Display the summary descriptive terms on the platform client and display the corresponding travel products for the triggered summary descriptive terms.

[0055] After receiving product recommendation information from the platform server, the platform client can display this information to the user for viewing. For example, it can display various summary descriptive terms. If a user is interested in a particular descriptive term, they can trigger that term (e.g., by clicking to select it). At this point, the client can further display at least a portion of a group of travel products corresponding to that term, thus achieving product recommendation. It's understandable that "displaying travel products" actually means displaying relevant information about the travel products, such as product tag information.

[0056] In one embodiment, when certain summary descriptive terms are detected as being triggered, it is necessary to first determine the travel products corresponding to those descriptive terms—these are the products to be displayed. Specifically, when determining the travel products to be displayed for a triggered summary descriptive term, if the product recommendation information includes product identifiers for at least one set of travel products and summary descriptive terms for each set, the travel products represented by the product identifiers corresponding to the triggered summary descriptive term can be identified as the travel products to be displayed. For example, if the model directly outputs a mapping relationship of "summary descriptive term A → product ID list," when a user clicks on descriptive term A, the corresponding travel products can be directly retrieved and displayed according to the ID list. This method offers fast response speed and a smoother user experience: no real-time calculation or query condition matching is required; product details can be retrieved in batches by ID, which can be completed in milliseconds, making it particularly suitable for high-concurrency scenarios (such as homepage recommendations and search dropdowns). The recommendation results are highly controllable and auditable: each set of products can be manually verified, sorted, or A / B tested in advance to ensure the quality, price competitiveness, and compliance of the displayed products (e.g., excluding hotels with high negative review rates). Furthermore, it reduces backend query pressure by avoiding the execution of complex filtering logic (such as multi-field combination filtering) with every click, thus reducing the load on the database or search engine. In addition, it supports offline generation and caching optimization: the entire "description + ID list" structure can be generated and cached in advance, which is naturally compatible with the aforementioned caching mechanism and helps to improve the overall throughput of the system.

[0057] Alternatively, if the product recommendation information includes at least one set of filtering rules for travel products and summary descriptive terms for each set of travel products, the platform can also filter travel products to be displayed from the travel products maintained by the travel service platform based on the filtering rules corresponding to the triggered summary descriptive terms. For example, the product recommendation information may contain structured filtering rules such as "summary descriptive term B → {destination=City A, region=Zone P, price ≤1000, includes breakfast, rating ≥4.5}". After the user clicks, the platform filters products that meet the conditions from the full product pool in real time according to the above filtering rules and returns them to the platform client for display. This solution has strong real-time performance and can accurately reflect the latest product status: the displayed content is based on the latest data in the current product library (price, inventory, rating, promotional tags, etc.), which can avoid displaying products that have been discontinued or increased in price, thus helping to improve credibility and conversion rate. It is highly flexible and can adapt to dynamic market changes: there is no need to regenerate complete product recommendation information corresponding to travel-related information. As long as the product attributes are updated, new products can be automatically included in the display scope (e.g., if a newly opened hotel meets the rule of "Zone P + includes breakfast", it can be immediately filtered out and made visible to users). Furthermore, this solution aligns better with the "intent-attribute" mapping logic: the filtering rules are essentially a structured expression of generalized travel intentions (e.g., "family travel" → {families with children, rooms ≥ 2 beds}), with clear logic that is easy to explain and debug. It also facilitates integration with real-time personalization logic: user profiles can be overlaid on basic filtering rules (e.g., "increasing the weight of the 'special offer' tag for price-sensitive users"), which helps achieve lightweight personalized recommendations.

[0058] In one embodiment, after identifying a set of travel products corresponding to the triggered summary descriptive term, the platform client can display at least a portion of these products. For example, after identifying (e.g., by querying cache or filtering) a set of travel products, the platform client can batch-send the relevant information of this set of products to the client, which will then display them according to preset rules (e.g., in list order). For instance, the travel products corresponding to the triggered summary descriptive term can be centrally displayed in the same display area of ​​the platform client. This method directly displays related hotels, tickets, day trip packages, etc., corresponding to the triggered summary descriptive term in a unified list or waterfall layout on the page. Users can see the recommended results all at once without additional clicks or switching tags, making it suitable for quick browsing and horizontal comparison, and helping to reduce cognitive load. Furthermore, since each displayed travel product corresponds to the same summary descriptive term, and the travel scenario expressed (e.g., "cherry blossom viewing") is presented centrally, it strengthens the overall feel of the theme, highlights the direct mapping between "intent and result," enhances guidance consistency, and effectively avoids information and cognitive disconnect. In addition, it facilitates unified sorting and mixed sorting optimization of the algorithm: the system can merge and sort different categories of products based on a unified goal (such as overall conversion rate, user preference) (such as inserting high-conversion tickets between hotels), which helps to improve the overall business effect.

[0059] Alternatively, the platform client can display travel products belonging to different product sets within a set of travel products corresponding to the triggered summary descriptive term in different display areas. These travel products are divided into multiple product sets according to at least one of the following dimensions: product type, product price, or the geographical region of the product delivery location. In this approach, after a user triggers a summary descriptive term, the page is divided into multiple display areas, each displaying individual travel products within the corresponding product set. For example, when dividing product sets by product type: the "Hotels" area displays hotels near area P, the "Transportation" area displays discounted buses / shuttle services from the airport to the city center in city A, and the "Entertainment" area displays food vouchers and attraction tickets in city A. This zoning display method improves the structure of the displayed information and helps reduce the user's filtering costs: through explicit categorization, it helps users quickly focus on product categories they care about (e.g., only viewing "Hotels"), budget ranges (e.g., "under ¥800"), or geographical locations (e.g., "near scenic spot X"), effectively improving decision-making efficiency. Furthermore, it supports multi-category collaborative recommendations to build complete travel plans: Travel decisions typically involve combinations of "flight + hotel + attraction + dining," and zoning displays naturally guide users to discover related products in one stop (a group of products corresponding to the same triggered descriptive term may include multiple different types of products), which helps promote cross-category add-ons (such as ordering airport transfer services after booking a hotel). In addition, zoning displays can also improve the experience of long lists and avoid information overload: when there are many recommended products, zoning can effectively alleviate visual fatigue, allowing users to expand on the modules they need, increasing dwell time and satisfaction.

[0060] like Figure 4 and Figure 5 As shown, if the user enters travel-related information in the first input box, such as the travel destination "City A," the platform determines (either by temporarily calling the model or querying the cache) various summary descriptive terms, such as "Top 20 Popular Attractions in City A," "Special Offer Flights to City A in December," and "Hotels in Popular Shopping Districts of City A." If the user does not trigger any descriptive term, the platform can default to displaying the travel product corresponding to the first descriptive term, or default to displaying a preset number (e.g., one) of travel products corresponding to each descriptive term; further details are omitted. like Figure 4As shown, if a user clicks on the descriptive phrase "Hotels in Popular Business Districts of City A," some or all of the hotels in a group (assuming a total of 70 hotels) corresponding to that descriptive phrase will be displayed in the area below the descriptive phrase's display location. For example, the first four hotels, "Hotels 11 in City A" to "Hotels 44 in City A," will be displayed sequentially, while the rest will be collapsed, and a control that can be triggered to expand them, such as "View All 70 Hotels," will be displayed. Of course, if the user scrolls down to the bottom of the interface and continues to scroll down to trigger a bottom refresh, the control will automatically open and further display all hotels as needed.

[0061] like Figure 5 As shown, if a user clicks on the descriptive phrase "Hotels in Popular Business Districts of City A," all 70 hotels corresponding to this descriptive phrase can be grouped by their respective regions, such as over 30 hotels in Zone P, 25 hotels in Zone Q, etc. The hotels in each region are then displayed separately. For example, in the display area corresponding to "Zone P - City Center Shopping and Dining Paradise," the top three hotels out of the 30+ hotels in Zone P (i.e., Hotels 11 to 33 in City A) are displayed, the rest are collapsed, and a control that can be triggered to expand "View All 30+ Hotels" is displayed. In the display area corresponding to "Zone Q - Modern Urban Hub," the top three hotels out of the 25 hotels in Zone Q (i.e., Hotels 44 to 66 in City A) are displayed, the rest are collapsed, and a control that can be triggered to expand "View All 25 Hotels" is displayed, and so on. Further details are omitted.

[0062] As mentioned earlier, showcasing travel products essentially means showcasing the product label information of those products. For example... Figure 4 As shown, the product label information for any hotel may include text information (such as hotel name, class, English name, quantity sold, price, promotional information, and ancillary services (such as free parking, free airport pick-up / drop-off, vending machines), etc.) and / or image information (such as introductory videos, hotel photos, etc.). Among them, the image information for hotels 11 to 33 in City A is an introductory video (the video playback button 401 in the lower right corner is visible; clicking it will play the video, or it can play by default); the image information for hotel 44 in City A is a hotel photo 402 (there is no video playback button in the lower right corner, indicating that it is a static image). Figure 5 Similarly, I will not elaborate further.

[0063] In addition, the display area corresponding to each travel product can also display a map / navigation sign for the product delivery location, such as... Figure 5 The map / navigation icon 501 is displayed in the upper right corner of Hotel 11 in City A. Clicking this control will trigger a display. Figure 6 The floating map window shown allows users to view the hotel's exact location and surrounding environment. Pulling up the top edge of the floating window triggers full-screen mode, converting it into a full-screen map page; further details are omitted.

[0064] In one embodiment, the product recommendation information generated by the product recommendation model may further include the feature description information of the travel products. Therefore, when displaying travel products corresponding to triggered summary descriptive terms, the feature description information can also be displayed. For example, for each travel product corresponding to a triggered summary descriptive term, the feature description information of that travel product can be displayed separately in the display area corresponding to each travel product. The feature description information of any travel product can also be referred to as the product's "core selling point information," used to emphasize the product's features / characteristics / user experience in at least one aspect. Through this solution, users can directly and accurately learn about product features from the list without having to click into the product details page, significantly shortening the decision-making path and helping to improve click-through conversion rates.

[0065] As mentioned earlier, product recommendation information may indicate multiple groups of travel products, each with a corresponding summary descriptive term. When product recommendation information indicates multiple groups of travel products, each product in any two groups may be completely different, meaning that no two groups have the same product; or, products in two groups may have at least one in common, meaning that at least one product belongs to multiple travel product groups simultaneously.

[0066] Correspondingly, the feature description information for any travel product can be a fixed set of information. Regardless of which group the travel product belongs to, this feature description information will be displayed in the product's display area after the summary description of that group is triggered. Alternatively, any travel product may have multiple feature descriptions, where different feature descriptions describe different product features (i.e., different semantics). In this case, the feature description information for a group of travel products corresponding to any summary description can include dynamic information, which can be matched with the core focus dimension of the summative description. Thus, for this travel product, different feature descriptions will be displayed after the summary descriptions corresponding to different product groups are triggered, thereby presenting different features of the same product to the user under different summary descriptions. For example, for the same hotel, the feature under the description "A core business district" is related to shopping (because the triggering of this description indicates that the user is more concerned about the convenience of shopping at this time); the feature under the description "Top 10 hotels near S temple" is related to location / transportation (because the triggering of this description indicates that the user is more concerned about the convenience of location / transportation at this time). For example, if a user clicks on "Special offer airfares to City A in December", it indicates that the user's potential travel intention is more concerned with airfare prices; if a user clicks on "Hotels in popular business districts of City A", it indicates that the user's potential travel intention is more concerned with business districts and shopping convenience; if a user clicks on "Popular cherry blossom viewing spots in City A", it indicates that the user's search intention is more concerned with natural landscapes (cherry blossom scenery), etc.

[0067] like Figure 5 As shown, the display areas for each travel product showcase the product's feature descriptions (502). When a user triggers the description "Hotels in Popular Shopping Districts of City A," the feature description for Hotel 11 in City A is "Located in the heart of District P, featuring trendy design, with views of the bustling streetscape from the window, immersing you in urban vitality"; the feature description for Hotel 22 in City A is "Located in the heart of District P, allowing you to immediately unload your shopping spoils after your shopping trip, perfectly suited for a shopping spree," etc., which will not be elaborated further.

[0068] like Figure 7 As shown in (a), when a user triggers the descriptive phrase "Hotels in popular business districts of City A," the characteristic description information 701 for Hotel 11 in City A is "Located in the core of District P, with a trendy design, and a view of the bustling street scene from the window, immersing you in urban vitality"—emphasizing the hotel's commercial environment. This is because the user triggering "Hotels in popular business districts of City A" indicates that the user is more concerned about the hotel's commercial environment. And as... Figure 7 As shown in (b), when a user triggers the description of "Hotel Recommendation with 0 Distance from Subway", the feature description information 702 of Hotel 11 in City A is "1.9km from the business district center, directly connected to Station X, easy and convenient to XXXX, suitable for both business and leisure" - focusing on the convenience of the hotel's transportation. This is because the user triggering "Hotel Recommendation with 0 Distance from Subway" indicates that the user is more concerned about the hotel's transportation.

[0069] This solution is based on the principle / fact that "triggered summary descriptive words can reflect a user's specific travel intentions." It proposes to display targeted descriptions of travel products based on these triggered summary descriptive words, ensuring users can directly access key information strongly relevant to their intentions from the list. This creates a semantic loop from summary descriptive words to product features, strengthening users' perception that the platform "understands their needs," thereby enhancing user trust in the platform and products and improving experience consistency. Furthermore, because the information is more accurate, it reduces "accidental clicks and back" actions, improving page dwell time and overall interaction efficiency, while reducing server-side request pressure on detail pages. When the same product corresponds to multiple feature descriptions, targeted display enables intelligent "one product, multiple facets," maximizing product exposure value while ensuring each display highly aligns with the current user's intent. This solution effectively solves the problem of users "not seeing clearly and not selecting accurately" at the general intent stage, and significantly improves the relevance, persuasiveness, and conversion efficiency of the recommendation system through precise, dynamic, and contextualized information presentation.

[0070] In addition, more travel products can be recommended, along with corresponding product descriptions and feature descriptions. For example... Figure 7As shown in (a), the description of Hotel X in City A is "the hotel is in a very quiet location", and the description of Hotel Y in City A is "it is very convenient to take the subway".

[0071] In one embodiment, when a summary descriptor is displayed in the first display area of ​​the platform client, for any triggered summary descriptor, at least one recommended descriptor associated with that summary descriptor can also be displayed in the second display area of ​​the platform client; furthermore, when any recommended descriptor is triggered, that recommended descriptor can be added to the first display area. This solution proposes a dynamic guidance mechanism with dual-area linkage, guiding users to browse more travel products that better match their travel intentions through changes in recommended prompts and interface display content. Figure 7 As shown in (a), the summary description is displayed in the first display area 703 at the top of the interface. When the user scrolls down, the hotels corresponding to the selected description "Hotels in Popular Business Districts of City A" move up and off the screen. At this time, several recommended descriptions associated with the selected description "Hotels in Popular Business Districts of City A" are further displayed below each hotel (such as "Recommended hotels 0 distance from the subway", "Hotels offering flight transfer service", "Hotels with complete children's facilities", and "Recommended five-star hotel packages with great value"). Furthermore, if the user selects the recommended description "Recommended hotels 0 distance from the subway", the client will add the recommended suggestion to the first display area 703 to replace the original summary descriptions. Correspondingly, the area below the first display area 703 will display the hotels corresponding to the selected "Recommended hotels 0 distance from the subway", such as hotels 11 to 22 and 77 in City A belonging to area P, and hotel 44 in City A belonging to area Q, etc., which will not be described in detail.

[0072] This solution supports the progressive expression of multi-dimensional intentions, lowering the input threshold for users: initially, a user may only have a vague intention (e.g., "City A"), which the system can then refine using summarizing descriptive terms (e.g., "Hotels in popular shopping areas of City A"). After clicking, relevant dimensions are recommended (e.g., "Includes hot springs," "Under ¥500," "Family-friendly"). Users don't need to actively input "budget" or "target demographic"; they can simply select to add intentions. Furthermore, it allows for dynamic combination and flexible adjustment of intentions, enabling users to gradually build complex intentions. It also supports changing the focus of intentions at any time (if not interested in "cherry blossom viewing," users can click on the new recommended term "food guide," and the system will automatically switch the context). It supports exploratory decision-making, allowing users to continuously refine and focus their needs while browsing, rather than accurately expressing all conditions at once, thus more accurately and comprehensively uncovering users' potential travel intentions. In addition, the first display area is used as a "dashboard of current intent", clearly showing the dimensions selected by the user. Moreover, the user can delete or replace a descriptive word at any time, and the system responds in real time, forming a transparent and controllable interactive loop. This visualizes the intent state, which helps to enhance the user's understanding and trust in the recommendation logic and improve the user's sense of control, avoiding the confusion or insecurity caused by "black box recommendation".

[0073] In summary, this solution can achieve the following technical effects: Breaking the limitations of a "fixed product flow": replacing indiscriminate listing with semantic grouping. The platform first obtains "product recommendation information" generated by a product recommendation model. This information not only indicates travel products but also includes "summary descriptive terms" for each group of products. These descriptive terms are essentially semantic clustering and intent labeling of the product set, organizing the originally chaotic product flow into categorized units with clear themes and user value orientations. This allows the platform to present clearly structured and intent-driven product groupings even when users only input broad information (such as "City A"), significantly improving information readability and relevance.

[0074] Beyond simple keyword matching: Guiding intent and stimulating demand through summarizing descriptive terms. These summarizing descriptive terms don't come directly from user input, but are high-level semantic expressions intelligently generated by the product recommendation model based on travel-related information input by the user (which might only be "City A"). This means the system no longer passively waits for users to refine their needs, but proactively converts underlying product data into travel scenario language that users can understand. This helps users quickly understand "what are the typical activities or product types" (satisfying "categorization" needs), and stimulates users' unexpressed interests through diverse descriptive terms (such as "cherry blossom viewing," "food," and "discounted airfares"), achieving the "discovery" and "mining" of interests and needs. For example, when a user searches for "City A," the platform might display descriptive terms such as "guesthouses and inns around X Mountain Bamboo Forest," "one-day tour packages to X scenic area," and "autumn limited-time red leaf train experience." Even if the user isn't initially aware of these options, they can form new travel intentions while browsing.

[0075] Constructing a "guide-feedback" closed loop: This involves interactively guiding users to accurately and conveniently express their needs. By displaying summary descriptive terms on the platform client and showcasing corresponding travel products for triggered terms, a lightweight yet efficient user intent confirmation mechanism is formed: a user's click on a descriptive term is itself explicit feedback on their current interest dimension (such as price sensitivity, preferred shopping areas, interest in seasonal activities, etc.). The platform can dynamically adjust subsequent recommendations accordingly, achieving a smooth transition from broad intent to refined intent. This not only improves conversion efficiency but also enables the system to continuously uncover and refine potential user needs.

[0076] As can be seen, through the innovative architecture of "product recommendation model generating summary descriptive terms + interactive layered display," this solution effectively solves the problems of rigid recommendation logic, inability to respond to general intentions, and difficulty in stimulating potential needs in related technologies. This solution shifts travel product recommendation from "product-centric" to "user cognition / intent-centric," satisfying users' needs for information classification and scenario understanding in the early stages of decision-making, while proactively uncovering and shaping their deep travel intentions through guided interaction, thereby significantly improving the user experience and the intelligence level of platform services.

[0077] Figure 8 This is a schematic structural diagram of a device provided in an exemplary embodiment. Please refer to... Figure 8At the hardware level, the device includes a processor 802, an internal bus 804, a network interface 806, memory 808, and non-volatile memory 810, and may also include other hardware required for its functions. One or more embodiments of this specification can be implemented in software, for example, the processor 802 reads the corresponding computer program from the non-volatile memory 810 into memory 808 and then runs it. Of course, besides software implementation, one or more embodiments of this specification do not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution entity of the following processing flow is not limited to individual logic units, but can also be hardware or logic devices.

[0078] Please refer to Figure 9 Information display devices for travel products can be applied to, for example... Figure 9 The device shown implements the technical solution of this specification. Specifically, the device is applied to a travel service platform and includes: The acquisition unit 901 is used to determine the travel-related information input by the user and acquire product recommendation information generated by the product recommendation model for the travel-related information. The product recommendation information is used to indicate at least one set of travel products and summary descriptive words for each set of travel products. Display unit 902 is used to display the summary descriptive terms on the platform client and to display the corresponding travel products for the triggered summary descriptive terms.

[0079] Optionally, the travel-related information is used to characterize the user's generalized travel intention, and the product recommendation information is generated by the product recommendation model based on the generalized travel intention; The generalized travel intent is used to characterize the user's travel intent from at least one of the following dimensions: travel destination, travel time period, travel budget, core travel activities, travel style orientation, travelers, and mode of transportation.

[0080] Optionally, the acquisition unit 901 is specifically used for: Identify various travel products related to the generalized travel intention, and construct prompt words based on the corresponding product description information; input the prompt words into the product recommendation model to obtain product recommendation information for the generalized travel intention output by the model; or, From the product recommendation information cached by the travel service platform for various travel intentions, product recommendation information for the generalized travel intention is queried; wherein, the product recommendation information for each cached travel intention is generated by the product recommendation model in response to the prompt words for that travel intention, and the prompt words are constructed based on the product description information of various travel products related to that travel intention.

[0081] Optional, When the generalized travel intent is used to represent the user's travel intent from the perspective of travel destination, the prompts constructed for the generalized travel intent include product description information of travel products related to the travel destination; and / or, The prompts constructed for the generalized travel intent include product descriptions of travel products related to the generalized travel intent and non-product-related contextual information related to the generalized travel intent.

[0082] Optionally, the display unit 902 is specifically used for: When the product recommendation information includes product identifiers for at least one set of travel products and summary descriptions for each set of travel products, the travel product represented by the product identifier corresponding to the triggered summary description is determined as the travel product to be displayed; or, When the product recommendation information includes at least one set of filtering rules for travel products and summary descriptive terms for each set of travel products, the travel products to be displayed are selected from the travel products maintained by the travel service platform according to the filtering rules corresponding to the triggered summary descriptive terms.

[0083] Optionally, the display unit 902 is specifically used for: In the same display area of ​​the platform client, travel products corresponding to the triggered summary descriptive terms are displayed in a centralized manner; or, In different display areas of the platform client, travel products belonging to different product sets are displayed from a group of travel products corresponding to the triggered summary descriptive terms. The group of travel products is divided into multiple product sets according to at least one of the following dimensions: product type, product price, and geographical region of product delivery location.

[0084] Optionally, the product recommendation information also includes a description of the features of the travel product, and the display unit 902 is specifically used for: For each travel product corresponding to the triggered summary descriptive term, the feature description information of that travel product will be displayed in the display area corresponding to each travel product. Among them, the characteristic description information of a set of travel products corresponding to any summative descriptive term includes dynamic information, which matches the core focus dimension of the summative descriptive term.

[0085] Optionally, a recommendation display unit 903 is also included, used for: When the summary descriptor is displayed in the first display area of ​​the platform client, for any triggered summary descriptor, at least one recommended descriptor associated with the summary descriptor is displayed in the second display area of ​​the platform client. When any recommended descriptor is triggered, that recommended descriptor is added to the first display area.

[0086] Optionally, the acquisition unit 901 is specifically used for: The first input window is displayed on the platform client; In response to the user's information input operation in the first input window, determine the travel-related information input by the user; or, In response to an input triggering operation performed by the user in the first input window, the first input window is updated to a second input window, and in response to an information input operation performed by the user in the second input window, the travel-related information input by the user is determined; wherein the size of the editable area of ​​the second input window is larger than the size of the editable area of ​​the first input window.

[0087] Based on the same concept as the methods described above, this specification also provides an electronic device, including: a processor; a memory for storing processor-executable instructions; wherein the processor performs the steps of the method as described in any of the above embodiments by executing the executable instructions.

[0088] Based on the same concept as the methods described above, this specification also provides a computer-readable storage medium having computer instructions stored thereon that, when executed by a processor, implement the steps of the methods as described in any of the above embodiments.

[0089] Based on the same concept as the methods described above, this specification also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the methods as described in any of the above embodiments.

Claims

1. A method for displaying information about travel products, characterized in that, Applied to a travel service platform, the method includes: Determine the travel-related information input by the user, and obtain product recommendation information generated by the product recommendation model based on the travel-related information. The product recommendation information is used to indicate at least one set of travel products and summary descriptive words for each set of travel products. The summary descriptive terms are displayed on the platform client, and corresponding travel products are displayed for the triggered summary descriptive terms.

2. The method according to claim 1, characterized in that, The travel-related information is used to characterize the user's generalized travel intention, and the product recommendation information is generated by the product recommendation model based on the generalized travel intention; The generalized travel intent is used to characterize the user's travel intent from at least one of the following dimensions: travel destination, travel time period, travel budget, core travel activities, travel style orientation, travelers, and mode of transportation.

3. The method according to claim 2, characterized in that, The product recommendation model generates product recommendation information corresponding to the travel-related information, including: Identify various travel products related to the generalized travel intention, and construct prompt words based on the corresponding product description information; input the prompt words into the product recommendation model to obtain product recommendation information for the generalized travel intention output by the model; or, From the product recommendation information cached by the travel service platform for various travel intentions, product recommendation information for the generalized travel intention is queried; wherein, the product recommendation information for each cached travel intention is generated by the product recommendation model in response to the prompt words for that travel intention, and the prompt words are constructed based on the product description information of various travel products related to that travel intention.

4. The method according to claim 3, characterized in that, When the generalized travel intent is used to represent the user's travel intent from the perspective of travel destination, the prompts constructed for the generalized travel intent include product description information of travel products related to the travel destination; and / or, The prompts constructed for the generalized travel intent include product descriptions of travel products related to the generalized travel intent and non-product-related contextual information related to the generalized travel intent.

5. The method according to claim 1, characterized in that, The travel products to be displayed are determined based on the triggered summary descriptive terms, including: When the product recommendation information includes product identifiers for at least one set of travel products and summary descriptions for each set of travel products, the travel product represented by the product identifier corresponding to the triggered summary description is determined as the travel product to be displayed; or, When the product recommendation information includes at least one set of filtering rules for travel products and summary descriptive terms for each set of travel products, the travel products to be displayed are selected from the travel products maintained by the travel service platform according to the filtering rules corresponding to the triggered summary descriptive terms.

6. The method according to claim 1, characterized in that, The display of corresponding travel products based on the triggered summary descriptive terms includes: In the same display area of ​​the platform client, travel products corresponding to the triggered summary descriptive terms are displayed in a centralized manner; or, In different display areas of the platform client, travel products belonging to different product sets are displayed from a group of travel products corresponding to the triggered summary descriptive terms. The group of travel products is divided into multiple product sets according to at least one of the following dimensions: product type, product price, and geographical region of product delivery location.

7. The method according to claim 1, characterized in that, The product recommendation information also includes feature descriptions of the travel products, and the travel products corresponding to the triggered summary descriptive terms include: For each travel product corresponding to the triggered summary descriptive term, the feature description information of that travel product will be displayed in the display area corresponding to each travel product. Among them, the characteristic description information of a set of travel products corresponding to any summative descriptive term includes dynamic information, which matches the core focus dimension of the summative descriptive term.

8. The method according to claim 1, characterized in that, Also includes: When the summary descriptor is displayed in the first display area of ​​the platform client, for any triggered summary descriptor, at least one recommended descriptor associated with the summary descriptor is displayed in the second display area of ​​the platform client. When any recommended descriptor is triggered, that recommended descriptor is added to the first display area.

9. The method according to claim 1, characterized in that, The process of determining the travel-related information input by the user includes: The first input window is displayed on the platform client; In response to the user's information input operation in the first input window, determine the travel-related information input by the user; or, In response to an input triggering operation performed by the user in the first input window, the first input window is updated to a second input window, and in response to an information input operation performed by the user in the second input window, the travel-related information input by the user is determined; wherein the size of the editable area of ​​the second input window is larger than the size of the editable area of ​​the first input window.

10. A computer program product, characterized in that, Includes a computer program / instructions that, when executed by a processor, implement the steps of the method as described in any one of claims 1-9.

Images