Hotel matching method and device based on AI, equipment and medium

Abstract

The invention discloses an AI-based hotel matching method, device and equipment and a medium, and relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining to-be-matched first hotel information and second hotel information; based on a predefined prompt word template, performing semantic comparison of names, addresses and calls on the first hotel information and the second hotel information through a preset large model, and generating a comprehensive semantic matching degree score; judging whether the comprehensive semantic matching degree score is higher than a preset confidence threshold or not; and if the comprehensive semantic matching degree score is higher than a preset confidence threshold, judging that the first hotel information is matched with the second hotel information. According to the method, the defects of low efficiency of traditional manual checking and insufficient semantic understanding of a simple rule system are effectively overcome, and when massive, multi-source and non-standard hotel information is handled, the processing efficiency can be remarkably improved synchronously, and the matching accuracy can be reliably guaranteed.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to an AI-based hotel matching method, apparatus, equipment, and medium. Background Technology

[0002] The core business processes of hotel service companies encompass several key aspects, including the onboarding of hotel suppliers, the dynamic maintenance and management of partner hotel information, and the integration of information from multiple data sources. Among these, ensuring that the basic hotel information provided by suppliers matches the corresponding information recorded in the company's internal database is a crucial and fundamental task. The quality of this information matching directly impacts subsequent procurement decisions, the maintenance of partnerships, and the optimization of overall supply chain efficiency, forming the cornerstone of stable business operations.

[0003] Currently, the industry primarily relies on two traditional models when handling such information matching tasks. One model is purely manual verification. While this method can handle some complex semantic changes based on the operator's experience, its processing efficiency falls short when dealing with the batch processing needs of massive amounts of supplier data. The large manpower investment and long processing cycle make this method unsuitable for the high demands of real-time information and business agility in the modern business environment. At the same time, the inherent fatigue and subjectivity of manual operation also become factors that introduce potential errors.

[0004] Another approach involves using automated systems based on keyword matching or pre-defined simple rules. While these systems offer significantly faster processing speeds compared to purely manual methods, their intelligence levels are clearly limited. Specifically, they struggle to effectively understand and handle the semantic ambiguity and diversity of expression prevalent in the real world. For example, for hotel names, the system cannot intelligently identify the identical meaning between "Hilton" and "Hilton Hotels Group"; for address information, it cannot normalize common variations such as "XX Road" and "XX Avenue"; and for contact information, it cannot automatically correct and standardize phone numbers with or without area codes. Due to a lack of deep semantic understanding, these automated systems often maintain a low accuracy rate when processing non-standard, ambiguous, or variant information, with frequent false and missed matches. This severely impacts the reliability of the output results and may pose potential risks for subsequent business decisions. Summary of the Invention

[0005] This invention provides an AI-based hotel matching method, apparatus, device, and medium. The technical problem it aims to solve is: how to design a hotel information matching method that can simultaneously achieve high efficiency and high accuracy, so as to overcome the shortcomings of traditional manual verification and simple rule-based automated systems in terms of low efficiency and insufficient semantic understanding when processing massive, multi-source, and non-standardized hotel information.

[0006] In a first aspect, embodiments of the present invention provide an AI-based hotel matching method, comprising:

[0007] Retrieve information on the first and second hotels to be matched;

[0008] Based on predefined prompt word templates, the semantics of the name, address, and telephone number of the first and second hotel information are compared using a pre-set large model to generate a comprehensive semantic matching score.

[0009] Determine whether the comprehensive semantic matching score is higher than a preset confidence threshold;

[0010] If the comprehensive semantic matching score is higher than the preset confidence threshold, the first hotel information and the second hotel information are determined to match.

[0011] A further technical solution is that the method further includes:

[0012] If the comprehensive semantic matching score is not higher than the preset confidence threshold, and digital identifier features can be extracted from the first hotel information and the second hotel information, consistency verification is performed based on the digital identifier features of the first hotel information and the second hotel information.

[0013] If the consistency verification passes, the first hotel information and the second hotel information are determined to match.

[0014] A further technical solution is that, based on a predefined prompt word template, a semantic comparison of the name, address, and telephone number of the first and second hotel information is performed using a preset large model to generate a comprehensive semantic matching score, including:

[0015] Entity recognition and synonym expansion are performed on the name information in the first hotel information and the second hotel information, and the name similarity is calculated.

[0016] The address information in the first hotel information and the second hotel information is parsed hierarchically and compared layer by layer to calculate the address similarity.

[0017] The format of the telephone information in the first hotel information and the second hotel information is standardized and its validity is verified, and the telephone similarity is calculated.

[0018] The comprehensive semantic matching score is generated by weighting the name similarity, address similarity, and telephone number similarity.

[0019] A further technical solution is that the hierarchical parsing includes:

[0020] The unstructured address string is parsed into hierarchical structured data including province, city, district, street, and house number.

[0021] A further technical solution is that the digital identification feature includes a unique identity identifier (ID), and the consistency verification includes:

[0022] Based on the unique identifier ID of the first hotel information, query the list of associated first matched hotels;

[0023] Obtain at least one target information dimension from which there is a difference between the first hotel information and the second hotel information;

[0024] Determine whether the information of the second hotel in the target information dimension is consistent with the information of any hotel in the first matched hotel list in the corresponding dimension;

[0025] If they match, the consistency verification is deemed successful.

[0026] If there is a discrepancy, then based on the unique identifier ID of the second hotel information, query the associated list of second matched hotels;

[0027] Determine whether the information of the first hotel in the target information dimension is consistent with the information of any hotel in the second matched hotel list in the corresponding dimension;

[0028] If they match, the consistency verification is deemed successful.

[0029] If there is no consistency, the consistency verification is deemed to have failed.

[0030] A further technical solution is that the method further includes:

[0031] If the consistency verification fails, the first hotel information and the second hotel information are marked as a suspicious matching pair and pushed to the preset manual review interface for manual review by the user.

[0032] Receive the results of manual review input by the user.

[0033] A further technical solution is that the method further includes:

[0034] Receive rule optimization information input by the user;

[0035] Based on the rule optimization information, update the thesaurus or matching rules in the prompt word template.

[0036] Secondly, embodiments of the present invention also provide an AI-based hotel matching device, which includes a unit for performing the above-described method.

[0037] Thirdly, embodiments of the present invention also provide a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described method.

[0038] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the above-described method.

[0039] This invention provides an AI-based hotel matching method, apparatus, device, and medium. The method includes: acquiring first and second hotel information to be matched; performing semantic comparison of the name, address, and telephone number of the first and second hotel information using a predefined prompt word template and a pre-set large-scale model to generate a comprehensive semantic matching score; determining whether the comprehensive semantic matching score is higher than a pre-set confidence threshold; and if the comprehensive semantic matching score is higher than the pre-set confidence threshold, determining that the first and second hotel information match. This invention guides a large-scale model to perform deep semantic comparison of hotel names, addresses, and telephone numbers using predefined prompt word templates, generating a quantified comprehensive matching score, and achieving automated decision-making based on a pre-set confidence threshold. This solution effectively overcomes the shortcomings of low efficiency in traditional manual verification and insufficient semantic understanding in simple rule systems. When dealing with massive, multi-source, and non-standardized hotel information, it can simultaneously achieve a significant improvement in processing efficiency and reliable assurance of matching accuracy, providing efficient and accurate technical support for hotel B2B business. Attached Figure Description

[0040] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 A flowchart illustrating an AI-based hotel matching method provided in an embodiment of the present invention;

[0042] Figure 2 This is a schematic block diagram of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0045] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0046] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0047] As used in this specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrases "if determined" or "if [described condition or event] is detected" may be interpreted, depending on the context, as "once determined," "in response to determination," "once [described condition or event] is detected," or "in response to detection of [described condition or event]."

[0048] Please see Figure 1 This invention provides an AI-based hotel matching method, which includes the following steps:

[0049] S1, obtain the information of the first and second hotels to be matched.

[0050] In practice, the first hotel information and the second hotel information to be matched are obtained. The first hotel information and the second hotel information to be matched can be manually entered or retrieved from a preset data source; this invention is not specifically limited to these two types.

[0051] S2, based on a predefined prompt word template, performs a semantic comparison of the name, address, and telephone number of the first hotel information and the second hotel information using a preset large model, and generates a comprehensive semantic matching score.

[0052] In practice, based on predefined prompt word templates, a pre-set large model (such as ChatGPT) is used to perform semantic comparison on the three core elements: name, address, and phone number. This step is crucial for improving the intelligence and accuracy of the matching. The large model refers to a pre-trained language model based on deep learning, possessing natural language understanding and generation capabilities. The predefined prompt word templates ensure the standardization and repeatability of the comparison process, avoiding the instability of results caused by temporary or arbitrary judgments. Using a large model for semantic comparison, rather than traditional keyword string matching, allows the system to understand the actual meaning behind the information. For example, for hotel names, the system can understand the referential relationship between "Hilton" and "Hilton Hotels Group"; for addresses, it can resolve the association between "XX Road" and "XX Avenue" in a specific context, and even handle issues of inconsistent abbreviations, aliases, and formats; for phone numbers, it can identify the same number represented by different country codes and area code formats. This deep semantic understanding capability fundamentally overcomes the inherent deficiency of simple rule systems in terms of insufficient semantic understanding when facing massive, multi-source, and non-standardized information. The comprehensive semantic matching score generated through this step is a quantitative and comprehensive evaluation indicator. It is no longer a simple binary output of "yes" or "no", but a continuous value that reflects the degree of matching probability, providing accurate data basis for subsequent intelligent decision-making.

[0053] In this invention, the prompt word template refers to a predefined structured instruction framework configured to guide and constrain a pre-defined large model to perform standardized information processing tasks. This prompt word template is not simply a collection of keywords, but a complete operational protocol that includes input data formatting specifications, processing logic guidelines, and output structure requirements.

[0054] Specifically, the core functions of this prompt word template are as follows:

[0055] First, the prompt word template defines the structured transformation rules for the input information. It parses and reorganizes the acquired unstructured information about the first and second hotels according to three predefined comparison dimensions: name, address, and phone number, forming a standardized input format that the large model can directly parse. This ensures that information can be systematically and unambiguously fed into the model.

[0056] Furthermore, the prompt word template includes built-in semantic processing guidelines for each comparison dimension. For name information, the template instructs the large model to perform entity recognition and synonym expansion operations; for address information, the prompt word template requires the large model to perform hierarchical parsing and layer-by-layer comparison; for telephone information, the prompt word template specifies the steps for format standardization and validity verification. These guidelines work together to focus the large model's general semantic understanding capabilities and adapt them to the specific domain task of hotel information matching.

[0057] Finally, the prompt word template specifies the structured paradigm for the output results. It mandates that the output of large models must include a comprehensive semantic matching score calculated based on a weighted average of name similarity, address similarity, and telephone number similarity, and can further include details of differences across each dimension and confidence level indicators. This specification ensures the machine readability, quantifiability, and consistency of the output results, providing a stable and reliable data interface for subsequent threshold judgment and decision-making processes.

[0058] In some preferred embodiments, the above step "based on a predefined prompt word template, performing semantic comparison of the name, address, and telephone number of the first and second hotel information using a preset large model to generate a comprehensive semantic matching score" specifically includes the following steps: performing entity recognition and synonym expansion on the name information in the first and second hotel information to calculate name similarity; performing hierarchical parsing on the address information in the first and second hotel information and comparing it layer by layer to calculate address similarity; performing format standardization and validity verification on the telephone information in the first and second hotel information to calculate telephone similarity; and performing a weighted calculation based on the name similarity, address similarity, and telephone similarity to generate the comprehensive semantic matching score. For example, the weighted calculation is performed based on a pre-configured weight allocation scheme (e.g., name weight 0.5, address weight 0.4, telephone weight 0.1).

[0059] In practice, the semantic comparison process was precisely defined, clarifying the specific processing strategies for the three core dimensions of name, address, and telephone number. This ensures that semantic matching is no longer a black-box operation, but a standardized process that is supported by clear technical means, is interpretable, and highly targeted. Entity recognition and synonym expansion of name information aim to solve problems such as core entity extraction and aliases / abbreviations in hotel naming. For example, accurately identifying whether "Beijing Guomao Hotel" and "China World Hotel" refer to the same entity (if it is a former name or alias), or understanding the equivalence between "Hilton" and "Hilton," thus improving the semantic depth of name comparison.

[0060] Furthermore, hierarchical parsing and layer-by-layer comparison of address information is a radical solution to address non-standard address representations. It decomposes chaotic unstructured address strings (such as "No. 10000 Shennan Avenue, Nanshan District, Shenzhen, Guangdong Province") into structured levels such as province, city, district, street, and house number, thereby systematically eliminating interference caused by differences in format, simplified / traditional characters, and common names, such as "Nanshan District, Shenzhen" and "Shenzhen Nanshan", "Shennan Avenue" and "Shennan Road", and ensuring the essential consistency of geographical location through layer-by-layer comparison.

[0061] Furthermore, standardizing the format and validating the telephone information directly resolves the human error caused by missing area codes, different separators, and different international code formats (such as +86 and 0086) in contact information comparison, ensuring that the compared numbers are essentially the same.

[0062] Furthermore, a weighted calculation is performed based on the similarity across these three dimensions to generate a comprehensive score. This allows the final decision to balance the importance of different information elements (e.g., names and addresses typically have higher weight than phone numbers), resulting in a comprehensive and reasonable overall evaluation. This embodiment significantly improves the accuracy, robustness, and interpretability of semantic comparison through multi-dimensional and structured fine-grained processing.

[0063] In one specific embodiment, firstly, entity recognition and synonym expansion are performed on the name information to calculate name similarity. The name information is input into a preset large model, and the core entity identifier is extracted through a pre-trained named entity recognition module. Simultaneously, the core entity identifier is expanded with synonyms based on a pre-built industry thesaurus to generate a set of name variants. The semantic similarity between the core entity identifier and the set of name variants of the first hotel information and the core entity identifier and the set of name variants of the second hotel information is calculated respectively, and the highest semantic similarity is taken as the name similarity.

[0064] Furthermore, the address information is parsed hierarchically, and the similarity is calculated by comparing each level. An address parsing engine is used to parse the unstructured address strings into hierarchical structured data including province, city, district, street, and house number. An address hierarchy weight mapping table is established, where the weight of higher-level administrative divisions decreases progressively. Similarity calculations are performed on the corresponding level address elements of the first and second hotel information, with lower-level element similarity calculations performed only after higher-level element matching is successful. The address similarity is then obtained by weighted summation based on the similarity scores and corresponding weights of each level.

[0065] Furthermore, the telephone information is standardized in format and validated for validity, and telephone similarity is calculated. Regular expression matching is used to identify the country code, area code, and main number in the telephone number; the telephone number is standardized in format according to preset international telephone coding rules; the standardized telephone number is validated based on a number segment rule base, and invalid numbers are filtered out; the standardized telephone numbers of the first hotel information and the second hotel information are precisely matched. If they are completely identical, the telephone similarity is the highest value; if they are inconsistent, the similarity is the lowest value.

[0066] Furthermore, a weighted calculation is performed based on name similarity, address similarity, and phone number similarity to generate a comprehensive semantic matching score. A pre-configured weighting scheme is read, where the sum of the name weight, address weight, and phone number weight is 1. The similarity scores for each dimension are multiplied by their corresponding weights and then summed to obtain an initial comprehensive score. This initial comprehensive score is then standardized using a preset scoring calibration function, outputting a comprehensive semantic matching score ranging from zero to full marks.

[0067] This embodiment ensures the systematicness and repeatability of semantic comparison by establishing a multi-dimensional and structured information processing flow, providing reliable technical support for subsequent matching decisions.

[0068] In some preferred embodiments, the hierarchical parsing includes parsing an unstructured address string into hierarchical structured data including province, city, district, street, and house number.

[0069] In its implementation, the specific structure of address hierarchical resolution was defined, explicitly specifying it as a five-level structure including province, city, district, street, and house number. This refined structural design provides unprecedented accuracy and fault tolerance for address comparison, establishing a powerful and unified address understanding framework that can systematically resolve the diversity and ambiguity commonly found in Chinese address representations.

[0070] For example, when faced with two addresses, "Beijing Chaoyang District Workers' Stadium North Road" and "Beijing Chaoyang District Gongti North Road," traditional string matching might fail due to differences in character count or abbreviation. However, through the hierarchical parsing defined in this embodiment, the system can accurately parse the former as (Province: Beijing, City: Beijing, District: Chaoyang District, Street: Workers' Stadium North Road, Address: Empty) and the latter as (Province: Beijing, City: Beijing, District: Chaoyang District, Street: Gongti North Road, Address: Empty). During layer-by-layer comparison, the system can identify the equivalence of "Beijing" and "Beijing" at the provincial and municipal levels, and further identify the relationship between "Workers' Stadium North Road" and "Gongti North Road" as the full and abbreviation of the same street through a thesaurus or semantic model. This parsing method allows the system to ignore unimportant differences in expression and focus on key geographical hierarchical information, thereby making accurate consistency judgments at the core levels such as district and street. It not only solves the abbreviation / full name problem, but also effectively handles complex situations such as changes in administrative divisions (such as "county" becoming "district") and the replacement of old and new road names, ensuring the depth, accuracy and reliability of the address comparison process.

[0071] S3, determine whether the comprehensive semantic matching score is higher than the preset confidence threshold.

[0072] In practice, the comprehensive semantic matching score is determined to be higher than a preset confidence threshold. This step introduces a key mechanism for automated decision-making and is one of the core elements for achieving high efficiency. Setting the confidence threshold essentially defines a "high confidence zone" for the system. When the matching score is higher than this threshold, it indicates that the system, based on semantic analysis, has sufficient confidence to determine that the two matches belong to the same hotel, thus allowing for a direct and automatic match. This mechanism enables a large number of clear and explicit matching cases to be processed instantly without any manual intervention, completely freeing human resources from tedious and repetitive simple verification work. The processing speed is increased by orders of magnitude compared to purely manual methods, greatly optimizing the overall process efficiency.

[0073] S4. If the comprehensive semantic matching score is higher than the preset confidence threshold, it is determined that the first hotel information and the second hotel information match.

[0074] In practice, if the comprehensive semantic matching score is higher than the preset confidence threshold, the first hotel information and the second hotel information are determined to match, thus the matching ends, and the first hotel and the second hotel are marked as the same hotel.

[0075] In some preferred embodiments, the method further includes the following steps: if the comprehensive semantic matching score is not higher than a preset confidence threshold, and digital identifier features can be extracted from the first hotel information and the second hotel information, a consistency verification is performed based on the digital identifier features of the first hotel information and the second hotel information; if the consistency verification passes, the first hotel information and the second hotel information are determined to match.

[0076] In specific implementation, an intelligent supplementary verification mechanism for low-confidence situations is added, thereby significantly improving the system's decision-making reliability and overall fault tolerance in complex and ambiguous scenarios. When the comprehensive semantic matching score is low, i.e., when the system lacks confidence in semantic judgment (for example, hotel names are highly similar but address descriptions differ significantly), directly determining a match or non-match carries a high risk of misjudgment. If digital identification features can be extracted from the first and second hotel information, these features (such as a unique internal system code) possess uniqueness and determinism, unaffected by the diversity of natural language expressions. This embodiment, by extracting and verifying the consistency of these digital identification features, can make a final decision when semantic ambiguity arises, using this "digital ID card."

[0077] For example, if a semantic comparison results in a lower score due to the difference in address descriptions between "XX Hotel (Pudong Airport Branch)" and "XX Hotel (Shanghai Pudong International Airport T2 Terminal)," the system can still accurately determine that they are the same hotel if their hotel ID codes are consistent.

[0078] In some preferred embodiments, the digital identification feature includes a unique identifier (ID), and the consistency verification includes: querying a first list of matched hotels based on the unique identifier (ID) of the first hotel information; obtaining at least one target information dimension where there is a difference between the first hotel information and the second hotel information; determining whether the information of the second hotel information in the target information dimension is consistent with the information of any hotel in the first list of matched hotels in the corresponding dimension; if consistent, the consistency verification is deemed to have passed; if inconsistent, querying a second list of matched hotels based on the unique identifier (ID) of the second hotel information; determining whether the information of the first hotel information in the target information dimension is consistent with the information of any hotel in the second list of matched hotels in the corresponding dimension; if consistent, the consistency verification is deemed to have passed; if inconsistent, the consistency verification is deemed to have failed.

[0079] In practical implementation, by introducing a cross-validation mechanism based on a list of matched hotels, an innovative and efficient solution is provided for complex matching scenarios where there are discrepancies in some dimensions of hotel information, when semantic matching confidence is insufficient. This embodiment significantly improves the system's fault tolerance and judgment accuracy in cases of incomplete, inconsistent, or partially erroneous information through an indirect association verification that utilizes historical matching data as a trust bridge.

[0080] Specifically, when the overall semantic matching score is low, it indicates that there are significant textual differences between the first hotel information and the second hotel information in one or more dimensions such as name, address, and telephone number, making it impossible for the system to make a high-confidence judgment based solely on semantic similarity.

[0081] First, the system intelligently identifies the specific discrepancies between the information of the first and second hotels, such as address or name. Then, the system performs a core cross-validation: it checks whether the problematic address (or name) of the second hotel appears in the trusted association list of the first hotel. If so, a strong chain of indirect evidence is generated: although the addresses of the two hotels do not directly match, the address of the second hotel is the same as a confirmed associated address of the first hotel, strongly suggesting that the second and first hotels essentially refer to the same entity. This design aims to resolve direct information conflicts caused by hotel relocation, name changes, or branch information confusion. For example, Hotel A (ID A1) currently has the address "2231 Pujiang Avenue," while Hotel B (ID B1) currently has the address "Tiangong Building, Pujiang Avenue," showing a direct address discrepancy. However, by querying A1's matched list, a historical match, Hotel C, is found, whose address happens to be "Tiangong Building, Pujiang Avenue." This discovery immediately establishes an association between B1 and A1, allowing the system to determine if they match.

[0082] To further ensure the completeness of the verification, this embodiment also incorporates a two-way verification process. If the query in the first direction fails, the system automatically switches directions, querying the matched list of the second hotel to check if the discrepancy information of the first hotel exists in that list. This two-way checking mechanism ensures that regardless of which side the information discrepancy occurs in, the system has the opportunity to discover potential correlations, thereby maximizing the uncovering of true matching relationships hidden by superficial differences.

[0083] Therefore, this embodiment greatly enhances the robustness of the system in handling "dirty data" and complex business scenarios. It expands the basis for matching decisions from isolated, potentially inaccurate data points to the entire historical matching relationship network. This adds a layer of more flexible and intelligent verification based on the relationship graph on top of semantic matching, ultimately significantly reducing the misjudgment rate caused by local information errors or changes, and improving the accuracy and reliability of the overall matching system.

[0084] In some preferred embodiments, the method further includes the following steps: if the consistency verification fails, the first hotel information and the second hotel information are marked as a suspicious matching pair and pushed to a preset manual review interface for manual review by the user; and the manual review result input by the user is received.

[0085] In practice, a manual review process was introduced as a final guarantee and supplement to the automated process. This constructed an efficient human-machine collaborative working mode, achieving an optimal balance between matching accuracy and the ability to handle complex problems. When the semantic matching confidence is low and the digital identifier feature verification also fails, it means that the current case exceeds the boundary of the system's automated processing capabilities and belongs to a complex, ambiguous, or potentially erroneous case. For example, two hotels may have significantly changed all information that can be automatically compared due to business restructuring, brand changes, etc., but they are still related according to business logic. In this case, if the system forcibly makes an automatic judgment, it is very easy to produce errors.

[0086] This embodiment cleverly leverages the strengths of human experts in understanding complex contexts, applying domain knowledge, and making flexible inferences by marking such suspicious matches and pushing them to a human review interface. Experts can comprehensively utilize external information, historical records, and business experience for in-depth analysis, thereby resolving "special cases" that machines cannot handle. This design creates a virtuous cycle: machines efficiently handle over 80% of routine and clear cases, freeing up human resources; humans can then focus on tackling the remaining 20% ​​of highly complex cases, ensuring the overall quality of the final output. This avoids the drawback of purely automated systems sacrificing accuracy in complex scenarios when pursuing efficiency, and also overcomes the inefficiency of purely manual processing. The ultimate result is that the system achieves near-100% accuracy while maintaining high efficiency, providing the highest reliability guarantee for business decisions.

[0087] In some preferred embodiments, the method further includes the following steps: receiving rule optimization information input by a user; and updating the thesaurus or matching rules in the prompt word template based on the rule optimization information.

[0088] In its implementation, a rule optimization mechanism based on human feedback was added. This transformed the entire system from a static tool into a dynamic intelligent system with continuous learning and self-evolution capabilities. It can continuously accumulate and solidify the knowledge and experience of human experts, thereby gradually expanding the scope of automated processing and continuously improving its accuracy. When humans process suspicious matching pairs, the rule optimization information input by users (i.e., domain experts), such as marking "Economic and Technological Development Zone" and "New Area" as synonyms in a specific context, or indicating a specific address abbreviation rule, is extremely valuable domain knowledge. By receiving this information and updating the thesaurus or matching rules in the prompt word template, the system essentially transforms the expert's one-time judgment into a general capability for sustainable system application. For example, the first time encountering a match between "XX Software Park" and "XX Science and Technology Park" might require human intervention, but after the expert confirms their equivalence, the system updates the thesaurus, and subsequent similar cases can be automatically and accurately matched. This closed-loop feedback mechanism creates a positive cycle of "practice-learning-optimization." As the system processes more data and receives more human feedback, its built-in semantic understanding rules and comparison logic will become increasingly sophisticated and closer to real-world business needs. Ultimately, this will lead to a continuous increase in the system's automation rate (i.e., the proportion of data that can be processed correctly without human intervention), a corresponding reduction in the workload of manual review, and an optimization of overall operating costs. At the same time, the system's intelligence and business adaptability will also continuously improve, ultimately forming a virtuous cycle of becoming smarter and more efficient with continued use.

[0089] This invention proposes an AI-based hotel matching method, comprising: acquiring first hotel information and second hotel information to be matched; based on a predefined prompt word template, performing semantic comparison of the name, address, and telephone number of the first and second hotel information using a preset large-scale model to generate a comprehensive semantic matching score; determining whether the comprehensive semantic matching score is higher than a preset confidence threshold; if the comprehensive semantic matching score is higher than the preset confidence threshold, determining that the first hotel information and the second hotel information match. This invention guides a large-scale model to perform deep semantic comparison of hotel names, addresses, and telephone numbers using a predefined prompt word template, generating a quantified comprehensive matching score, and achieving automated decision-making based on a preset confidence threshold. This solution effectively overcomes the shortcomings of low efficiency in traditional manual verification and insufficient semantic understanding in simple rule systems. When dealing with massive, multi-source, and non-standardized hotel information, it can simultaneously achieve a significant improvement in processing efficiency and a reliable guarantee of matching accuracy, providing efficient and accurate technical support for hotel B2B business.

[0090] Corresponding to the above-described AI-based hotel matching method, the present invention also provides an AI-based hotel matching device. This AI-based hotel matching device includes a unit for executing the aforementioned AI-based hotel matching method, and can be configured in a desktop computer, tablet computer, laptop computer, or other terminal. Specifically, the AI-based hotel matching device includes:

[0091] The acquisition unit is used to acquire information about the first hotel and the second hotel to be matched.

[0092] The comparison unit is used to perform semantic comparison of the name, address and telephone number of the first hotel information and the second hotel information based on the predefined prompt word template and a preset large model, and generate a comprehensive semantic matching score.

[0093] The judgment unit is used to determine whether the comprehensive semantic matching score is higher than a preset confidence threshold.

[0094] The determination unit is used to determine that the first hotel information and the second hotel information match if the comprehensive semantic matching score is higher than a preset confidence threshold.

[0095] In some preferred embodiments, the AI-based hotel matching device further includes:

[0096] The verification unit is used to perform consistency verification based on the digital identifier features of the first hotel information and the second hotel information if the comprehensive semantic matching score is not higher than the preset confidence threshold and the digital identifier features can be extracted from the first hotel information and the second hotel information.

[0097] The determination unit is used to determine that the first hotel information and the second hotel information match if the consistency verification passes.

[0098] In some preferred embodiments, the step of using a predefined prompt word template to perform semantic comparison of the name, address, and telephone number of the first and second hotel information through a preset large model to generate a comprehensive semantic matching score includes:

[0099] Entity recognition and synonym expansion are performed on the name information in the first hotel information and the second hotel information, and the name similarity is calculated.

[0100] The address information in the first hotel information and the second hotel information is parsed hierarchically and compared layer by layer to calculate the address similarity.

[0101] The format of the telephone information in the first hotel information and the second hotel information is standardized and its validity is verified, and the telephone similarity is calculated.

[0102] The comprehensive semantic matching score is generated by weighting the name similarity, address similarity, and telephone number similarity.

[0103] In some preferred embodiments, the hierarchical parsing includes:

[0104] The unstructured address string is parsed into hierarchical structured data including province, city, district, street, and house number.

[0105] In some preferred embodiments, the digital identification feature includes a unique identity identifier (ID), and the consistency verification includes:

[0106] Based on the unique identifier ID of the first hotel information, query the list of associated first matched hotels;

[0107] Obtain at least one target information dimension from which there is a difference between the first hotel information and the second hotel information;

[0108] Determine whether the information of the second hotel in the target information dimension is consistent with the information of any hotel in the first matched hotel list in the corresponding dimension;

[0109] If they match, the consistency verification is deemed successful.

[0110] If there is a discrepancy, then based on the unique identifier ID of the second hotel information, query the associated list of second matched hotels;

[0111] Determine whether the information of the first hotel in the target information dimension is consistent with the information of any hotel in the second matched hotel list in the corresponding dimension;

[0112] If they match, the consistency verification is deemed successful.

[0113] If there is no consistency, the consistency verification is deemed to have failed.

[0114] In some preferred embodiments, the AI-based hotel matching device further includes:

[0115] The marking unit is used to mark the first hotel information and the second hotel information as a suspicious matching pair if the consistency verification fails, and push them to the preset manual review interface for manual review by the user.

[0116] The first receiving unit is used to receive the manual review results input by the user.

[0117] In some preferred embodiments, the AI-based hotel matching device further includes:

[0118] The second receiving unit is used to receive rule optimization information input by the user;

[0119] An optimization unit is used to update the thesaurus or matching rules in the prompt word template based on the rule optimization information.

[0120] It should be noted that those skilled in the art can clearly understand that the specific implementation process of the above-mentioned AI-based hotel matching device and its various units can be referred to the corresponding descriptions in the foregoing method embodiments. For the sake of convenience and brevity, these details will not be repeated here.

[0121] The aforementioned AI-based hotel matching device can be implemented as a computer program, which can, for example... Figure 2 It runs on the computer device shown.

[0122] Please see Figure 2 , Figure 2 This is a schematic block diagram of a computer device provided in an embodiment of this application. The computer device 500 can be a terminal or a server. The terminal can be an electronic device with communication functions, such as a smartphone, tablet, laptop, desktop computer, personal digital assistant, or wearable device. The server can be a standalone server or a server cluster composed of multiple servers.

[0123] The computer device 500 includes a processor 502, a memory, and a network interface 505 connected via a system bus 501. The memory may include a non-volatile storage medium 503 and internal memory 504.

[0124] The non-volatile storage medium 503 may store an operating system 5031 and a computer program 5032. When the computer program 5032 is executed, it causes the processor 502 to execute an AI-based hotel matching method.

[0125] The processor 502 provides computing and control capabilities to support the operation of the entire computer device 500.

[0126] The internal memory 504 provides an environment for the execution of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute an AI-based hotel matching method.

[0127] The network interface 505 is used for network communication with other devices. Those skilled in the art will understand that the above structure is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device 500 to which the present application is applied. A specific computer device 500 may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements.

[0128] The processor 502 is used to run a computer program 5032 stored in the memory to perform the following steps:

[0129] Retrieve information on the first and second hotels to be matched;

[0130] Based on predefined prompt word templates, the semantics of the name, address, and telephone number of the first and second hotel information are compared using a pre-set large model to generate a comprehensive semantic matching score.

[0131] Determine whether the comprehensive semantic matching score is higher than a preset confidence threshold;

[0132] If the comprehensive semantic matching score is higher than the preset confidence threshold, the first hotel information and the second hotel information are determined to match.

[0133] In some preferred embodiments, the method further includes:

[0134] If the comprehensive semantic matching score is not higher than the preset confidence threshold, and digital identifier features can be extracted from the first hotel information and the second hotel information, consistency verification is performed based on the digital identifier features of the first hotel information and the second hotel information.

[0135] If the consistency verification passes, the first hotel information and the second hotel information are determined to match.

[0136] In some preferred embodiments, the step of using a predefined prompt word template to perform semantic comparison of the name, address, and telephone number of the first and second hotel information through a preset large model to generate a comprehensive semantic matching score includes:

[0137] Entity recognition and synonym expansion are performed on the name information in the first hotel information and the second hotel information, and the name similarity is calculated.

[0138] The address information in the first hotel information and the second hotel information is parsed hierarchically and compared layer by layer to calculate the address similarity.

[0139] The format of the telephone information in the first hotel information and the second hotel information is standardized and its validity is verified, and the telephone similarity is calculated.

[0140] The comprehensive semantic matching score is generated by weighting the name similarity, address similarity, and telephone number similarity.

[0141] In some preferred embodiments, the hierarchical parsing includes:

[0142] The unstructured address string is parsed into hierarchical structured data including province, city, district, street, and house number.

[0143] In some preferred embodiments, the digital identification feature includes a unique identity identifier (ID), and the consistency verification includes:

[0144] Based on the unique identifier ID of the first hotel information, query the list of associated first matched hotels;

[0145] Obtain at least one target information dimension from which there is a difference between the first hotel information and the second hotel information;

[0146] Determine whether the information of the second hotel in the target information dimension is consistent with the information of any hotel in the first matched hotel list in the corresponding dimension;

[0147] If they match, the consistency verification is deemed successful.

[0148] If there is a discrepancy, then based on the unique identifier ID of the second hotel information, query the associated list of second matched hotels;

[0149] Determine whether the information of the first hotel in the target information dimension is consistent with the information of any hotel in the second matched hotel list in the corresponding dimension;

[0150] If they match, the consistency verification is deemed successful.

[0151] If there is no consistency, the consistency verification is deemed to have failed.

[0152] In some preferred embodiments, the method further includes:

[0153] If the consistency verification fails, the first hotel information and the second hotel information are marked as a suspicious matching pair and pushed to the preset manual review interface for manual review by the user.

[0154] Receive the results of manual review input by the user.

[0155] In some preferred embodiments, the method further includes:

[0156] Receive rule optimization information input by the user;

[0157] Based on the rule optimization information, update the thesaurus or matching rules in the prompt word template.

[0158] It should be understood that, in the embodiments of this application, the processor 502 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0159] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program may be stored in a storage medium, which is a computer-readable storage medium. The computer program is executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.

[0160] Therefore, the present invention also provides a storage medium. This storage medium can be a computer-readable storage medium. The storage medium stores a computer program. When executed by a processor, the computer program causes the processor to perform the following steps:

[0161] Retrieve information on the first and second hotels to be matched;

[0162] Based on predefined prompt word templates, the semantics of the name, address, and telephone number of the first and second hotel information are compared using a pre-set large model to generate a comprehensive semantic matching score.

[0163] Determine whether the comprehensive semantic matching score is higher than a preset confidence threshold;

[0164] If the comprehensive semantic matching score is higher than the preset confidence threshold, the first hotel information and the second hotel information are determined to match.

[0165] In some preferred embodiments, the method further includes:

[0166] If the comprehensive semantic matching score is not higher than the preset confidence threshold, and digital identifier features can be extracted from the first hotel information and the second hotel information, consistency verification is performed based on the digital identifier features of the first hotel information and the second hotel information.

[0167] If the consistency verification passes, the first hotel information and the second hotel information are determined to match.

[0168] In some preferred embodiments, the step of using a predefined prompt word template to perform semantic comparison of the name, address, and telephone number of the first and second hotel information through a preset large model to generate a comprehensive semantic matching score includes:

[0169] Entity recognition and synonym expansion are performed on the name information in the first hotel information and the second hotel information, and the name similarity is calculated.

[0170] The address information in the first hotel information and the second hotel information is parsed hierarchically and compared layer by layer to calculate the address similarity.

[0171] The format of the telephone information in the first hotel information and the second hotel information is standardized and its validity is verified, and the telephone similarity is calculated.

[0172] The comprehensive semantic matching score is generated by weighting the name similarity, address similarity, and telephone number similarity.

[0173] In some preferred embodiments, the hierarchical parsing includes:

[0174] The unstructured address string is parsed into hierarchical structured data including province, city, district, street, and house number.

[0175] In some preferred embodiments, the digital identification feature includes a unique identity identifier (ID), and the consistency verification includes:

[0176] Based on the unique identifier ID of the first hotel information, query the list of associated first matched hotels;

[0177] Obtain at least one target information dimension from which there is a difference between the first hotel information and the second hotel information;

[0178] Determine whether the information of the second hotel in the target information dimension is consistent with the information of any hotel in the first matched hotel list in the corresponding dimension;

[0179] If they match, the consistency verification is deemed successful.

[0180] If there is a discrepancy, then based on the unique identifier ID of the second hotel information, query the associated list of second matched hotels;

[0181] Determine whether the information of the first hotel in the target information dimension is consistent with the information of any hotel in the second matched hotel list in the corresponding dimension;

[0182] If they match, the consistency verification is deemed successful.

[0183] If there is no consistency, the consistency verification is deemed to have failed.

[0184] In some preferred embodiments, the method further includes:

[0185] If the consistency verification fails, the first hotel information and the second hotel information are marked as a suspicious matching pair and pushed to the preset manual review interface for manual review by the user.

[0186] Receive the results of manual review input by the user.

[0187] In some preferred embodiments, the method further includes:

[0188] Receive rule optimization information input by the user;

[0189] Based on the rule optimization information, update the thesaurus or matching rules in the prompt word template.

[0190] The storage medium is a physical, non-transient storage medium, such as a USB flash drive, external hard drive, read-only memory (ROM), magnetic disk, or optical disk, or any other physical storage medium capable of storing program code. The computer-readable storage medium can be non-volatile or volatile.

[0191] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0192] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the division of each unit is merely a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0193] The steps in the method of this invention can be adjusted, merged, or reduced in order according to actual needs. The units in the device of this invention can be merged, divided, or reduced according to actual needs. Furthermore, the functional units in the various embodiments of this 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.

[0194] 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 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, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0195] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0196] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Since these modifications and variations fall within the scope of the claims and their equivalents, this invention also intends to include these modifications and variations.

[0197] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An AI-based hotel matching method, characterized in that, include: Retrieve information on the first and second hotels to be matched; Based on predefined prompt word templates, the semantics of the name, address, and telephone number of the first and second hotel information are compared using a pre-set large model to generate a comprehensive semantic matching score. Determine whether the comprehensive semantic matching score is higher than a preset confidence threshold; If the comprehensive semantic matching score is higher than the preset confidence threshold, the first hotel information and the second hotel information are determined to match.

2. The AI-based hotel matching method according to claim 1, characterized in that, The method further includes: If the comprehensive semantic matching score is not higher than the preset confidence threshold, and digital identifier features can be extracted from the first hotel information and the second hotel information, consistency verification is performed based on the digital identifier features of the first hotel information and the second hotel information. If the consistency verification passes, the first hotel information and the second hotel information are determined to match.

3. The AI-based hotel matching method according to claim 1, characterized in that, The method, based on predefined prompt word templates, performs semantic comparison of the names, addresses, and phone numbers of the first and second hotel information using a pre-defined large model, generating a comprehensive semantic matching score, including: Entity recognition and synonym expansion are performed on the name information in the first hotel information and the second hotel information, and the name similarity is calculated. The address information in the first hotel information and the second hotel information is parsed hierarchically and compared layer by layer to calculate the address similarity. The format of the telephone information in the first hotel information and the second hotel information is standardized and its validity is verified, and the telephone similarity is calculated. The comprehensive semantic matching score is generated by weighting the name similarity, address similarity, and telephone number similarity.

4. The AI-based hotel matching method according to claim 3, characterized in that, The hierarchical parsing includes: The unstructured address string is parsed into hierarchical structured data including province, city, district, street, and house number.

5. The AI-based hotel matching method according to claim 2, characterized in that, The digital identification feature includes a unique identity identifier (ID), and the consistency verification includes: Based on the unique identifier ID of the first hotel information, query the list of associated first matched hotels; Obtain at least one target information dimension from which there is a difference between the first hotel information and the second hotel information; Determine whether the information of the second hotel in the target information dimension is consistent with the information of any hotel in the first matched hotel list in the corresponding dimension; If they match, the consistency verification is deemed successful. If there is a discrepancy, then based on the unique identifier ID of the second hotel information, query the associated list of second matched hotels; Determine whether the information of the first hotel in the target information dimension is consistent with the information of any hotel in the second matched hotel list in the corresponding dimension; If they match, the consistency verification is deemed successful. If there is no consistency, the consistency verification is deemed to have failed.

6. The AI-based hotel matching method according to claim 2, characterized in that, The method further includes: If the consistency verification fails, the first hotel information and the second hotel information are marked as a suspicious matching pair and pushed to the preset manual review interface for manual review by the user. Receive the results of manual review input by the user.

7. The AI-based hotel matching method according to claim 6, characterized in that, The method further includes: Receive rule optimization information input by the user; Based on the rule optimization information, update the thesaurus or matching rules in the prompt word template.

8. An AI-based hotel matching device, characterized in that, Includes a unit for performing the method as described in any one of claims 1-7.

9. A computer device, characterized in that, The computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program that, when executed by a processor, can implement the method as described in any one of claims 1-7.

Images