System for providing artificial intelligence based extended hotel concierge service

Abstract

An AI-based extended hotel concierge service provision system is provided, comprising a user terminal that outputs a pre-set question, inputs a response to the question, and receives travel destination recommendations based on preferences derived from the response, and a concierge service provision server including a transmission unit that sends a question to the user terminal to determine preferences, a reception unit that receives a response to the question, an analysis unit that analyzes preferences by analyzing the response, and a recommendation unit that extracts travel destinations based on preferences and recommends them to the user terminal.

Description

Technology Field

[0001] The present invention relates to an AI-based extended hotel concierge service provision system, and provides a system that provides travel destination recommendations, guide matching, and room service based on AI. Background Technology

[0002] The "Republic of Seoul" phenomenon is intensifying within Korea's tourism industry. This is due not only to the fact that the majority of foreign tourists visiting Korea head to Seoul, but also because government agencies are implementing policies concentrated in the capital region. Seoul accounts for 90% of foreign tourist visits to Korea, a figure more than four times higher than the rate for other regions, and this concentration on Seoul has persisted for 10 years. Furthermore, the return visit rate for Korean tourism is significantly lower compared to China or Japan; while tourists may have been satisfied with their first visit, the inability to find new attractions on subsequent trips plays a major role. Additionally, despite possessing world-class digital infrastructure, issues such as inconvenient applications and language accessibility are cited as factors lowering the return rate. Consequently, Korea's travel balance has recorded a deficit for 120 consecutive months.

[0003] At this time, a method for recommending travel packages or travel courses based on AI has been researched and developed. In this regard, prior art Korean Published Patent No. 2021-0081746 (published July 2, 2021) and Korean Published Patent No. 2025-0014872 (published February 3, 2025) respectively disclose a configuration for extracting travel package products corresponding to gender, age, and travel type entered from a traveler's terminal, which involves analyzing travel tendencies and extracting travel package products by identifying travel tendencies from social media, and a configuration for generating travel courses from travel big data, storing tendency analysis information to map to the travel courses, and generating travel courses by reflecting real-time information.

[0004] However, in the former case, although it is stated that travel preferences are identified through social media, recommendation accuracy may be low because social media posts and travel itineraries do not constitute reliable data representing travel trends. In the latter case, although it is stated that travel courses are generated in response to real-time information, recommendation accuracy is bound to suffer, just like in the former, if preferences are not properly identified. In inbound tourism, foreigners often visit only Seoul and move on immediately to Japan; this is due not only to language barriers but also to the inability to find proper guides and the lack of information or recommendations regarding unique tourist attractions in each region. Therefore, research and development of a system capable of recommending travel destinations tailored to individual preferences is required. The problem to be solved

[0005] One embodiment of the present invention provides an AI-based extended hotel concierge service provision system capable of providing extended concierge services including travel destination recommendation, guide matching, and room service by transmitting a question to a user terminal to recommend a travel destination, receiving a response to the question, recommending a travel destination based on the response and also providing travel destination information, matching a guide to guide the travel destination via O2O (Online to Offline), and providing room service through a digital human. However, the technical problem that this embodiment aims to solve is not limited to the technical problem described above, and other technical problems may exist. means of solving the problem

[0006] As a technical means for achieving the aforementioned technical task, one embodiment of the present invention includes a user terminal that outputs a preset question, inputs a response to the question, and receives a travel destination recommendation based on a preference derived from the response; a transmission unit that transmits a question to the user terminal to determine the preference; a reception unit that receives a response to the question; an analysis unit that analyzes the preference by analyzing the response; and a recommendation unit that extracts a travel destination based on the preference and recommends it to the user terminal. Effects of the invention

[0007] According to any one of the means for solving the problem of the present invention described above, by providing an expanded concierge service including travel destination recommendations, guide matching, and room service, the phenomenon of tourists being concentrated in Seoul is resolved and the development of the local economy is contributed; the return rate of inbound tourism can be increased by discovering and providing content specific to each region; the convenience of tourists can be enhanced by providing real-time language translation and payment services; customer value and customer satisfaction can be increased by providing room service with digital humans; and business owners operating hotels can also find a breakthrough in sales limits and achieve marketing results by providing new services. Brief explanation of the drawing

[0008] FIG. 1 is a drawing for explaining an AI-based extended hotel concierge service provision system according to one embodiment of the present invention. Figure 2 is a block diagram illustrating a concierge service provider server included in the system of Figure 1. FIGS. 3 and 4 are drawings for illustrating an embodiment in which an AI-based extended hotel concierge service according to an embodiment of the present invention is implemented. FIG. 5 is a flowchart illustrating an operation method for providing an AI-based extended hotel concierge service according to an embodiment of the present invention. Specific details for implementing the invention

[0009] Embodiments of the present invention are described below with reference to the attached drawings so that those skilled in the art can easily implement the invention. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. Furthermore, in order to clearly explain the present invention in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification are denoted by similar reference numerals.

[0010] Throughout the specification, when a part is described as being "connected" to another part, this includes not only cases where they are "directly connected" but also cases where they are "electrically connected" with other elements interposed between them. Furthermore, when a part is described as "including" a component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components, and it should be understood that this does not preclude the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0011] Terms such as “about,” “substantially,” etc., used throughout the specification, are used to mean at or near the stated value when inherent manufacturing and material tolerances are presented in the stated meaning, and are used to prevent unscrupulous infringers from unfairly exploiting the disclosure in which precise or absolute values ​​are mentioned to aid in understanding the invention. Terms such as “step” or “step of” used throughout the specification of the invention do not mean “step for”.

[0012] In this specification, the term "part" includes a unit realized by hardware, a unit realized by software, and a unit realized using both. Additionally, one unit may be realized using two or more pieces of hardware, and two or more units may be realized by one piece of hardware. Meanwhile, "part" is not limited to software or hardware, and "part" may be configured to reside in an addressable storage medium or configured to run on one or more processors. Accordingly, as an example, "part" includes components such as software components, object-oriented software components, class components, and task components, as well as processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. The functions provided within the components and "parts" may be combined into a smaller number of components and "parts" or further separated into additional components and "parts." In addition, the components and '~parts' may be implemented to play one or more CPUs within the device or secure multimedia card.

[0013] Some of the operations or functions described herein as being performed by a terminal, device, or device may instead be performed by a server connected to said terminal, device, or device. Likewise, some of the operations or functions described as being performed by a server may also be performed by a terminal, device, or device connected to said server.

[0014] In this specification, some of the operations or functions described as mapping or matching with a terminal may be interpreted as meaning mapping or matching the terminal's unique number or personal identification information, which is the terminal's identifying data.

[0015] The present invention will be described in detail below with reference to the attached drawings.

[0016] FIG. 1 is a diagram illustrating an AI-based extended hotel concierge service provision system according to an embodiment of the present invention. Referring to FIG. 1, the AI-based extended hotel concierge service provision system (1) may include at least one user terminal (100), a concierge service provision server (300), and at least one advertiser terminal (400). However, since the AI-based extended hotel concierge service provision system (1) of FIG. 1 is merely an embodiment of the present invention, the present invention is not to be interpreted as being limited by FIG. 1.

[0017] At this time, each component of FIG. 1 is generally connected through a network (Network, 200). For example, as shown in FIG. 1, at least one user terminal (100) can be connected to a concierge service provider server (300) through the network (200). And, the concierge service provider server (300) can be connected to at least one user terminal (100) and at least one advertiser terminal (400) through the network (200). Also, at least one advertiser terminal (400) can be connected to the concierge service provider server (300) through the network (200).

[0018] Here, a network refers to a connection structure capable of exchanging information among individual nodes, such as multiple terminals and servers. Examples of such networks include Local Area Networks (LANs), Wide Area Networks (WANs), the World Wide Web (WWW), wired and wireless data networks, telephone networks, and wired and wireless television networks. Examples of wireless data communication networks include, but are not limited to, 3G, 4G, 5G, 3GPP (3rd Generation Partnership Project), 5GPP (5th Generation Partnership Project), 5G NR (New Radio), 6G (6th Generation of Cellular Networks), LTE (Long Term Evolution), WIMAX (World Interoperability for Microwave Access), Wi-Fi, Internet, LAN (Local Area Network), Wireless LAN (Wireless Local Area Network), WAN (Wide Area Network), PAN (Personal Area Network), RF (Radio Frequency), Bluetooth network, NFC (Near-Field Communication) network, satellite broadcasting network, analog broadcasting network, DMB (Digital Multimedia Broadcasting) network, etc.

[0019] In the following, the term "at least one" is defined as a term including both singular and plural forms, and it will be obvious that even if the term "at least one" does not exist, each component may exist in a singular or plural form and may mean singular or plural. Furthermore, whether each component is provided in a singular or plural form may be changed according to the embodiment.

[0020] At least one user terminal (100) may be a terminal of a tourist user who uses a web page, app page, program, or application related to an AI-based extended hotel concierge service to receive recommendations for travel destinations, be matched with a guide, and request room service.

[0021] Here, at least one user terminal (100) may be implemented as a computer capable of connecting to a remote server or terminal via a network. Here, the computer may include, for example, a navigation system, a laptop equipped with a web browser, a desktop, a laptop, etc. At this time, at least one user terminal (100) may be implemented as a terminal capable of connecting to a remote server or terminal via a network. At least one user terminal (100) may include all kinds of handheld-based wireless communication devices, such as navigation, PCS (Personal Communication System), GSM (Global System for Mobile communications), PDC (Personal Digital Cellular), PHS (Personal Handyphone System), PDA (Personal Digital Assistant), IMT (International Mobile Telecommunication)-2000, CDMA (Code Division Multiple Access)-2000, W-CDMA (W-Code Division Multiple Access), Wibro (Wireless Broadband Internet) terminal, smartphone, smartpad, tablet PC, etc.

[0022] The concierge service providing server (300) may be a server that provides an AI-based expanded hotel concierge service web page, app page, program, or application. Additionally, the concierge service providing server (300) may be a server that provides AI-based expanded concierge services, including travel destination recommendations, guide matching, and room service.

[0023] Here, the concierge service providing server (300) may be implemented as a computer capable of connecting to a remote server or terminal via a network. Here, the computer may include, for example, a navigation system, a laptop equipped with a web browser, a desktop, a laptop, etc.

[0024] At least one advertiser terminal (400) may be a terminal of an advertiser who wishes to promote a hotel using a web page, app page, program, or application related to an AI-based extended hotel concierge service. In this case, the advertiser may be a local government in addition to a hotel.

[0025] Here, at least one advertiser terminal (400) may be implemented as a computer capable of connecting to a remote server or terminal via a network. Here, the computer may include, for example, a navigation system, a laptop equipped with a web browser, a desktop, a laptop, etc. At this time, at least one advertiser terminal (400) may be implemented as a terminal capable of connecting to a remote server or terminal via a network. At least one advertiser terminal (400) may include all kinds of handheld-based wireless communication devices, such as navigation, PCS (Personal Communication System), GSM (Global System for Mobile communications), PDC (Personal Digital Cellular), PHS (Personal Handyphone System), PDA (Personal Digital Assistant), IMT (International Mobile Telecommunication)-2000, CDMA (Code Division Multiple Access)-2000, W-CDMA (W-Code Division Multiple Access), Wibro (Wireless Broadband Internet) terminal, smartphone, smartpad, tablet PC, etc.

[0026] FIG. 2 is a block diagram for explaining a concierge service providing server included in the system of FIG. 1, and FIG. 3 and FIG. 4 are drawings for explaining an embodiment in which an AI-based extended hotel concierge service according to an embodiment of the present invention is implemented.

[0027] Referring to FIG. 2, the concierge service providing server (300) may include a transmission unit (310), a reception unit (320), an analysis unit (330), a recommendation unit (340), an information provision unit (350), a guide matching unit (360), a room service unit (370), a robot provision unit (380), and a marketing unit (390).

[0028] When a concierge service providing server (300) or another server (not shown) operating in conjunction with one embodiment of the present invention transmits an AI-based extended hotel concierge service application, program, app page, web page, etc. to at least one user terminal (100) and at least one advertiser terminal (400), the at least one user terminal (100) and at least one advertiser terminal (400) may install or open the AI-based extended hotel concierge service application, program, app page, web page, etc. Additionally, a service program may be operated on at least one user terminal (100) and at least one advertiser terminal (400) using a script executed in a web browser. Here, a web browser refers to a program that enables the use of World Wide Web (WWW) services and receives and displays hypertext described in HTML (Hyper Text Mark-up Language), and includes, for example, Chrome, Microsoft Edge, Safari, Firefox, Whale, UC Browser, etc. In addition, "application" refers to an application on a terminal, and includes, for example, an app running on a mobile terminal (smartphone).

[0029] Referring to FIG. 2, the transmission unit (310) can transmit a question to the user terminal (100) to determine preference. The user terminal (100) can output a pre-set question. At this time, the question may not be a fixed question in advance, but a question that changes according to the user's response. To this end, an initial question set (Set) is provided to determine preference, and a question can be generated and provided based on the response of the user terminal (100) to this question set. This is explained in the receiving unit (320).

[0030] The receiver (320) can receive a response to a question. The user terminal (100) can input a response to a question. First, the question can be divided into two types of questions. One is an initial question set that includes questions to collect basic information such as the user's travel style, budget, and preferred activities, and the other is a subsequent question that is adaptively generated based on the user's response. The former is a fixed question, and the latter is a question generated by an LLM-based generative AI to respond to the response in real time.

[0031] Types of questions detail Type purpose Initial question set Collect basic information such as travel style, budget, and preferred activities static question Getting a rough idea of ​​travel preferences Follow-up questions Questions generated by LLM-based generative AI based on user responses Dynamic questions Identifying detailed travel preferences

[0032] An initial set of questions is configured to identify the user's approximate travel preferences, and subsequent questions are dynamically generated based on the user's previous responses. In this process, LLM-based generative AI utilizes state-of-the-art big language models such as GPT-4 and Claude to identify detailed travel preferences. At this stage, the user's responses are analyzed through LLM-based generative AI, where sentiment and semantic analysis are performed as shown in Fig. 3a (b), and sentiment and semantics are extracted. Based on this, more specific and personalized subsequent questions are generated, and finally, questions that provide highly relevant and useful information are selected. This process is performed iteratively to progressively refine the user's preferences.

[0033] ① Preparation of initial question set ② User response→┌ ③ User Response Analysis (LLM-based Generative AI) ④ Analyzed preferences → ⑤ Generating follow-up questions │ │ ⑧ Selected Question└ ┐┌ ┘⑥ Generated questions ↓ │ ↑↓ Travel style, budget, preferred activities, etc. └→ Sentiment and semantic analysis of responses ⑦ Question Optimization Personalized specific questions ↓ Select highly relevant questions

[0034] The analysis unit (330) can analyze preferences by analyzing the responses. This corresponds to ③ in Table 2. The responses provided by the user undergo preprocessing and vectorization through advanced natural language processing techniques, and are then analyzed in depth through sentiment analysis and semantic analysis. For sentiment analysis, a BERT (Bidirectional Encoder Representations from Transformers) based model can be used to extract positive, negative, and neutral sentiments from the responses. Additionally, for semantic analysis, key keywords and concepts are extracted from the user's responses using the latest embedding technology such as Word2Vec, and the user's preferences can be expressed in a structured form by mapping them to a pre-built travel-related ontology. At this time, an ontology refers to a method in computer science and informatics for systematically defining concepts of a specific domain and the relationships between them. For example, it is a method of identifying the connection relationships between each node using nodes and edges, such as a GNN (Graph Neural Network). The analysis results derived in this way can be used as input to the RAG system described later, which can significantly improve the accuracy of personalized travel destination recommendations.

[0035] The recommendation unit (340) can extract travel destinations based on preferences and recommend them to the user terminal (100). The recommendation unit (340) can generate a preference vector by embedding preferences, input the preference vector as a query into a pre-established travel destination database, such as a RAG (Retrieval-Augmented Generation), search for travel destinations, list the searched results, and recommend them to the user terminal (100). The user terminal (100) can receive recommendations for travel destinations based on preferences derived from the response.

[0036] The RAG system can recommend optimal travel destinations by effectively matching user preferences with vast travel destination data. In this process, the Dense Passage Retrieval (DPR) technique can be utilized to efficiently search for highly relevant travel destination information. DPR is a method designed to ensure high similarity (vector dot product) between related query and search target passages. This concept emerged as language models advanced significantly, enabling the representation of documents or search terms as vectors. Accordingly, a method can be adopted where user preferences are vectorized, and destinations within the database (pre-vectorized data) are compared with these preferences to extract destinations with high similarity. The information retrieved in this way undergoes a re-ranking process to derive the final recommendation results. By applying a powerful pre-trained language model to this process, the accuracy and diversity of recommendations are significantly enhanced.

[0037] At this time, for detailed information on DPR, refer to the paper (Karpukhin, Vladimir, Barlas O?uz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih. "Dense passage retrieval for open-domain question answering." arXiv preprint arXiv:2004.04906 (2020).).

[0038] The information provision unit (340) can provide information about the recommended travel destination by inputting the recommended travel destination into a pre-established Large Language Model (LM)-based Generative Artificial Intelligence (GIA) and generating information about things to see, things to do, and things to eat in the LLM-based Generative Artificial Intelligence, and then providing the travel destination information to the user terminal (100). Once the travel destination is extracted from the recommendation unit (340), information about the travel destination must be provided together so that the user can understand it. If only the travel destination is recommended and no description of the travel destination is provided, the user may hesitate to go to that travel destination. This is because they would have to look up all the information about the travel destination individually. Accordingly, information on how to get to the travel destination, what to do and eat at the travel destination, and what to see must be provided.

[0039] Accordingly, in order to provide detailed and attractive information about recommended travel destinations, one embodiment of the present invention utilizes an LLM-based generative AI to generate descriptions of major attractions, activities, food, etc. This process may consist of four stages: travel destination data collection, data preprocessing, content generation using an LLM-based generative AI, and quality verification. First, travel destination data may be collected from credible travel information sites and review platforms, including the Korea Tourism Organization. After undergoing a cleaning and structuring process, the collected data is input into an LLM-based generative AI, which generates travel destination descriptions categorized into sights, activities, and food. The generated content (travel destination descriptions) may be modified if necessary after undergoing verification of accuracy and appropriateness. In this case, the verification of appropriateness may be a process of cross-verifying whether the LLM-based generative AI has written the document based on evidence without causing hallucinatory symptoms. The verified content may be visualized as images and videos, provided together with text, or text provided as images or videos.

[0040] The guide matching unit (360) receives input of a first language, which is the native language, from the user terminal (100), and receives input of a second language, which is the language of the country of travel, from the user terminal (100), extracts guides who are proficient in the first language and the second language, filters guides based on the location of the user terminal (100), and then transmits the guide list to the user terminal (100). For example, assuming a guest has come to Korea from Russia, one must first find a person who is proficient in both Russian and Korean. In particular, it is best to find a person whose native language is Russian, but even if that is not the case, one must find a person who is fluent in both languages. Then, assuming the user is in Myeongdong, Seoul, one must find a person who has registered Myeongdong as a guideable area. If Guide A lives in Gyeonggi-do but has registered Seoul as a guideable area, Guide A can also be extracted. Once selected in this way, one has selected a person who is fluent in both Russian and Korean and who is able to commute to Myeongdong, Seoul. The results may be randomly listed and sent to the user terminal (100), but the guide list may also be rearranged considering the user's tastes or preferences. For example, if the user is female and prefers female guides, the list may be arranged by placing female guides at the top. Also, if the user considers the guide's experience and attitude, the guides may be rearranged based on the user's guide career or reviews. Alternatively, the guide list may be filtered based on collaborative filtering, and the results may be provided to the user terminal (100).

[0041] When a room service request is made from a user terminal (100), the room service department (370) can classify the type of room service and then transmit the request to a terminal (not shown) in charge of the classified room service. At this time, a digital human may be set up to handle the room service, or a screen displaying the digital human may be displayed. At this time, the digital human is a virtual human equipped with a chatbot and digitally implements the human's appearance, voice, movements, etc. It is also a computer-generated 3D version of a human. The chatbot may be based on LLM-based generative AI. At this time, the chatbot recognizes the user's voice utterance, extracts the request from the voice utterance, and classifies the type of request to determine the department to which the request should be assigned. For example, if the type of room service is an order for food or beverage, the task can be assigned to the food or beverage preparation team; if it is a cleaning service, to the cleaning team; if it is a request for equipment repair, to the maintenance team; if it is a request for equipment usage inquiry, to the equipment management team; and if it is a request for a wake-up call, to the hotel lobby.

[0042] The robot providing unit (380) can move the autonomous robot to the location of the room number where room service was requested when the autonomous robot is called from the person in charge terminal and the room number of the room service requested is entered into the autonomous robot, and when it arrives at the location of the room number, it can send an alarm to the user terminal (100) that room service has arrived. For example, if a customer at a hotel requests the purchase of digestive medicine, the person in charge can obtain the digestive medicine, put it in the autonomous robot along with a recovery note and water, and deliver it by pressing the customer's room number. At this time, it is assumed that the autonomous robot possesses all indoor maps, can ride and get off elevators, and includes an interface that can press or select buttons. Here, the autonomous robot is equipped as a follower robot, and if a customer requests the robot to move luggage, the autonomous robot can be kept on standby in the lobby to load the luggage and follow the guest, or the autonomous robot can be made to find the guest's room on its own. Customer satisfaction and the element of fun can be increased because customers do not have to give the tip they would have given to the person while making a request.

[0043] The marketing department (390) can conduct marketing for promotion to inbound and outbound tourists and send a marketing report to an advertiser terminal (400) based on the results of the marketing. When conducting marketing targeting inbound and outbound tourists, influencers may be hired. However, when conducting experience group marketing by hiring influencers, the reliability of the experience group's reviews is often low, making the reviews unreliable. Therefore, to measure the reliability of the experience group's reviews, the gap with general consumer reviews is measured, and for experience groups that differ significantly from consumer reviews, the reliability is set low so that they are not included in the next experience group, thereby improving the reliability of the experience group reviews. To achieve this, star ratings and reviews are extracted, and the reliability of the experience group can be measured not only by comparing star ratings but also by performing sentiment analysis on the reviews (using a BERT-based NLP sentiment analysis model) and calculating the difference between the satisfaction of the experience group and the satisfaction of consumers.

[0044] For example, after preprocessing text review data, a BERT-based NLP sentiment analysis model is used to classify each sentence as positive or negative, and a continuous sentiment score between 0 and 1 is assigned. Subsequently, based on these sentiment scores calculated for each sentence, the satisfaction of the j-th user (including both trial participants and consumers) for the k-th product is calculated. The satisfaction of the j-th user for the k-th product SSkj in the text reviews can be calculated according to Equation 1.

[0045]

[0046] In this case, a higher score in Equation 1 indicates a higher level of satisfaction driven primarily by positive opinions. Specifically, a satisfaction score closer to 1 indicates higher satisfaction, while a score decreases when negative sentiment is dominant. Therefore, a low satisfaction score reflects reviews where negative sentiment is expressed more than positive sentiment. Sentence PSkji represents the positive sentence score of the i-th sentence of the j-th consumer regarding the k-th product, and PNkji represents the negative sentence score of the i-th sentence of the j-th consumer regarding the k-th product.

[0047] As described above, each consumer's text review is classified into positive and negative sentences, and the degree of sentiment is expressed as a sentiment score. For example, Consumer 1 gave a 5-star rating, but expressed more negative sentiment (8 sentences) than positive sentiment (7 sentences) in the text review. As a result, the difference between the sum of the positive and negative scores was negative, indicating that actual satisfaction with the product was low. The example above suggests that it may be difficult to accurately reflect a consumer's positive or negative opinion based solely on star ratings.

[0048] Next, to evaluate the reliability of the experience group reviews, the difference in satisfaction between the two groups (the experience group and consumer reviews) can be statistically verified. Specifically, if a statistically significant difference in satisfaction is found when analyzing the difference between the two groups, it suggests that the reliability of the experience group reviews is low. To verify this, an independent samples T-Test is applied to the text review data to statistically analyze the difference in satisfaction between the two groups. The T-Test is a method that evaluates whether an observed difference is due to chance or is actually a significant difference by comparing the means of two groups, and it is suitable for analyzing the difference in satisfaction between the experience group and consumer reviews in one embodiment of the present invention. Additionally, regarding star ratings, the difference in satisfaction between the two groups can be statistically verified through a non-parametric test using the Mann-Whitney U test. This method is suitable because star rating reviews are considered ordinal data, and the Mann-Whitney U test is suitable for comparing mean differences in ordinal scale data, which is difficult to satisfy the normality assumption for the independent samples T-Test.

[0049] After a statistically significant difference between two groups is confirmed, a procedure is required to statistically analyze how large that difference actually is. This process plays an important role in evaluating the actual impact of the difference, even if a difference exists between the two groups. To this end, in one embodiment of the present invention, the effect size is analyzed, and specifically, the magnitude of the difference in satisfaction between the two groups is quantitatively evaluated using Cohen's d method. Cohen's d is a statistical indicator that measures the effect size between two groups and is suitable for the reliability evaluation of the present invention because it evaluates the actual magnitude of the difference in satisfaction. After measuring reliability in this way, the reliability of each marketing agency is identified, and when selecting a marketing agency or influencer, advertising can be conducted by selecting a marketing agency or influencer for whom the difference between the two groups is not significant.

[0050] Hotel Recommendations

[0051] In addition, in one embodiment of the present invention, when a user selects a hotel before or after arriving in Korea, a method of recommending a high-rated or good hotel by predicting the user's rating of the hotel may be further utilized. To this end, reviews left by the user at previous hotels may be referenced. Although reviews left by the user contain rich information regarding the user's preferences and interests, they rely only on sentiment and cannot reflect the interaction between the user and the hotel or the user's specific personal characteristics. Accordingly, in one embodiment of the present invention, recommendations are to be made by focusing on the following three items.

[0052] First, the User-Hotel Interaction Extractor embeds User IDs and Hotel IDs to extract interactions between the user and the hotel, and generates interaction vectors through Self-Attention. Second, the Review Feature Extractor aims to identify features inherent in user reviews; it applies BERT embeddings to each review to extract embedding vectors that consider the context of the review. Finally, Rating Prediction combines the vectors output by the User-Hotel Interaction Extractor and the Review Feature Extractor to predict user preferences.

[0053] Predicted rating Actual rating ┌Rating Prediction└ (hat)yui ←… yui ↑ MLP layer x … ↑ MLP Layer 1 ↑ ┌→ + ←┐ ┌User-Hotel Interaction Extractor└ ■□■□ ▤▥▤▥ ┐Review Feature Extractor┘ ↑ ↑ Self-Attention BERT ↑ ▲▲▲▲ Concatenation Vector Segment embedding ↑ ↑ + User embedding vector Item embedding vector Location embedding ↑ ↑ + ⓞ①…ⓞ ①ⓞ…ⓞ Word embeddings ▲ ▲ ▲ User ID Hotel ID Review text

[0054] The specific details of each part are as follows.

[0055] User-Hotel Interaction Extractor

[0056] The user-hotel interaction extractor aims to learn complex interactions between a user and a hotel by applying self-attention. First, to effectively learn user-hotel interactions, the user ID and hotel ID are converted into dense vectors. This process is implemented by applying embedding techniques to the user ID and hotel ID information, and the resulting embedding vectors are used as latent vectors for the user and the hotel. This process can be expressed as shown in Equation 2.

[0057]

[0058] v in mathematical equation 2 U u and v I i represent the user sparse vector and the hotel sparse vector, respectively, and P T and Q T represents the weights applied during the embedding process. Additionally, pu and qi represent the user latent factor vector and the hotel latent factor vector, respectively, calculated through the operation of sparse vectors and weights. The two latent vectors generated as a result of the aforementioned process can be defined as follows to identify the complex interactions and key aspects between the user and the hotel.

[0059]

[0060]

[0061]

[0062] z in Equation 3 represents the concatenation of the latent factor vectors of the user and hotel, performed to minimize information loss. The concatenated latent factor vector is then used in the Self-Attention Layer, from which a vector reflecting the user's interest is extracted. In Equations 4 and 5, Wh and Wo represent the weights for each layer, while bh and bo represent the biases. The Attention Output hh represents the user's interest expressed as a probability after passing through the Softmax activation function. The identified user interest in the hotel is represented as Vui through the ReLU (Rectified Linear Unit) activation function. Each activation function is expressed as shown in Equations 6 and 7.

[0063]

[0064]

[0065] The user-hotel interaction vector, which reflects user interest, is then combined with review text features extracted by the review feature extractor and used to predict the final rating.

[0066] Review Feature Extractor

[0067] The Review Feature Extractor aims to generate review feature vectors—which represent user preference characteristics within the text—by utilizing BERT on user reviews. Various techniques, such as RNNs and CNNs, exist for extracting semantic features from text sequences. However, because CNNs focus on local information, they often ignore contextual information, which can lead to information loss. While RNN-based language models can consider contextual information, they fail to capture core features and can introduce noise by considering all information and sequences. Although research is underway to complement these single models, existing studies suffer from the disadvantage of being unable to learn deep features because they learn context unidirectionally and simply combine data. BERT overcomes the shortcomings of existing methodologies and can extract accurate features by grasping the deep contextual information of the text.

[0068] First, the input text sequence is tokenized using BERT's proprietary tokenizer. The BERT tokenizer utilizes a WordPiece model that separates words into smaller units called subwords. During tokenization, special tokens [CLS] are added to the beginning of sentences and [SEP] tokens to the end to aid in distinguishing between sentences. These generated tokens are then truncated according to a pre-specified sequence length or padded with [PAD] tokens. Subsequently, embedding is performed using a BERT model pre-trained on the large-scale datasets of Wikipedia and BookCorpus. Deep bidirectional representations are learned through a Masked Language Model approach, which randomly masks parts of the input tokens with [MASK] tokens and predicts the original form of those tokens.

[0069] <Rating Prediction>

[0070] Finally, in the Rating Prediction process, a rating representing user preference is predicted based on the vectors extracted from the user-hotel interaction extractor and the review feature extractor. To utilize the previously extracted user-hotel interaction vector and review feature vector for prediction, a combination of vectors is performed using a process similar to Equation 8. Here, Vui represents the vector learned from user-hotel interactions, and s represents the vector learned from the features of review text.

[0071]

[0072] The vector generated through Equation 8 is input into a Multi-Layer Perceptron (MLP) for rating prediction, and is repeated as shown in Equation 9 for final rating prediction. In this process, ReLU can be used as the activation function to pass the operation result to the next layer. Additionally, to eliminate subjective deviations in user preference ratings and predict user preference as a value within the range of 1 to 5, the user preference rating can be adjusted to the range of 0 to 1 through linear transformation. Therefore, Sigmoid can be used as the final output function.

[0073]

[0074] Here, W T i , W T L is the weight for each layer, b i , b L represents bias. The final predicted value (hat)yui extracted through Equation 9 can be evaluated for recommendation performance after being converted to the original range of 1 to 5 by applying the linear transformation of Equation 10 so that it can be located within the range of the existing rating.

[0075]

[0076] Of course, it goes without saying that hotels can be recommended in various ways in addition to the methods mentioned above.

[0077] Hereinafter, the operation process according to the configuration of the concierge service providing server of FIG. 2 described above will be explained in detail with reference to FIG. 3 and FIG. 4. However, it is obvious that the embodiment is merely one of the various embodiments of the present invention and is not limited thereto.

[0078] Referring to FIG. 3a, (a) a concierge service provider server (300) can identify the user's travel preferences as shown in (b) through a process in which the server provides an initial set of questions to the user terminal (100) and receives a response thereto, analyzes the response to generate the next question, and provides it to the user terminal (100). Additionally, the concierge service provider server (300) can search for and recommend travel destinations based on preferences as shown in (c), and can generate and provide information about the recommended travel destinations as shown in (d). Furthermore, as shown in (a) of FIG. 3b, a guide can be matched, and as shown in (b), AI-based room service can be provided. Additionally, as shown in (c), an autonomous driving robot can be used in conjunction with room service, and (d) influencer marketing can be conducted and the results provided to the advertiser terminal (400).

[0079] To solve the problems of domestic travel as shown in Fig. 4a as shown in Fig. 4b, content is generated as shown in Figs. 4c and 4d. An AI concierge service is provided as shown in Fig. 4f, and travel plans are proposed by recommending accommodations, restaurants, and tourist information based on social media as shown in Fig. 4g. Guide matching is performed as shown in Fig. 4h, and guides are matched as shown in Fig. 4i. Simultaneous interpretation can be provided as shown in Figs. 4i and 4j. An integrated service (tentatively named Journeye) as shown in Fig. 4k can be provided using the platform of the present invention (tentatively named Journeye) that combines all these services. A business model as shown in Fig. 4l can be established. Since the OTA market has currently reached its limits as shown in Figs. 4m to 4o, travel curation similar to that of an OTA is provided to enable a rapid response to customer consumption patterns. The service process may be as shown in Fig. 4p, but is not limited thereto.

[0080] A service roadmap according to one embodiment of the present invention is as shown in FIG. 4q, and global expansion is possible as shown in FIG. 4t by using a strategy as shown in FIG. 4s with a core function as shown in FIG. 4r. A platform according to one embodiment of the present invention can be promoted with marketing strategies as shown in FIG. 4u to FIG. 4w. However, it is not limited thereto.

[0081] As for the details regarding the AI-based extended hotel concierge service provision method of FIGS. 2 to 4 that are not described, they are identical to or can be easily inferred from the details described above regarding the AI-based extended hotel concierge service provision method through FIG. 1, so further explanation will be omitted.

[0082] FIG. 5 is a diagram illustrating the process of transmitting and receiving data between each component included in the AI-based extended hotel concierge service provision system of FIG. 1 according to an embodiment of the present invention. Hereinafter, an example of the process of transmitting and receiving data between each component will be described through FIG. 5, but the present invention is not to be interpreted as being limited to such an embodiment, and it is obvious to those skilled in the art that the process of transmitting and receiving data illustrated in FIG. 5 may be changed according to various embodiments described above.

[0083] Referring to FIG. 5, the concierge service provider server sends a question to the user terminal to determine preferences (S5100).

[0084] Then, the concierge service provider server receives a response to the question (S5200) and analyzes the response to analyze the preference (S5300).

[0085] In addition, the concierge service provider server extracts travel destinations based on preferences and recommends them to the user terminal (S5400).

[0086] The order of the steps described above (S5100~S5400) is merely an example and is not limited thereto. That is, the order of the steps described above (S5100~S5400) may vary, and some of these steps may be executed simultaneously or deleted.

[0087] As for the details regarding the AI-based extended hotel concierge service provision method of Fig. 5 that are not described, they are identical to or can be easily inferred from the details described above regarding the AI-based extended hotel concierge service provision method through Figs. 1 to 4, so further explanation will be omitted.

[0088] The method for providing an AI-based extended hotel concierge service according to one embodiment described through FIG. 5 may also be implemented in the form of a recording medium containing computer-executable instructions, such as an application or program module executed by a computer. A computer-readable medium may be any available medium accessible by a computer and includes both volatile and non-volatile media, as well as removable and inseparable media. Additionally, a computer-readable medium may include all computer storage media. Computer storage media include both volatile and non-volatile, removable and inseparable media implemented by any method or technique for storing information, such as computer-readable instructions, data structures, program modules, or other data.

[0089] The method for providing an AI-based extended hotel concierge service according to one embodiment of the present invention described above may be executed by an application basically installed on a terminal (which may include a program included in a platform or operating system, etc., basically installed on the terminal), or by an application (i.e., a program) directly installed by a user on a master terminal through an application providing server, such as an application store server, an application, or a web server related to the service. In this sense, the method for providing an AI-based extended hotel concierge service according to one embodiment of the present invention described above may be implemented as an application (i.e., a program) that is basically installed on a terminal or directly installed by a user, and may be recorded on a computer-readable recording medium such as a terminal.

[0090] The foregoing description of the present invention is for illustrative purposes only, and those skilled in the art will understand that other specific forms can be easily modified without altering the technical spirit or essential features of the present invention. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single unit may be implemented in a distributed manner, and components described as distributed may likewise be implemented in a combined form.

[0091] The scope of the present invention is defined by the claims set forth below rather than by the detailed description above, and all modifications or variations derived from the meaning and scope of the claims and equivalent concepts thereof should be interpreted as being included within the scope of the present invention.

Claims

Claim 1 A user terminal that outputs a pre-set question, inputs a response to the said question, and receives recommendations for travel destinations based on preferences derived from the said response; A transmission unit that transmits a question to the user terminal to determine preferences; a reception unit that receives a response to the question; an analysis unit that analyzes preferences by analyzing the response; a recommendation unit that extracts travel destinations based on preferences and recommends them to the user terminal; an information provision unit that inputs the recommended travel destinations into a pre-established LLM (Large Language Model)-based Generative AI (Generative Artificial Intelligence) to provide information on the recommended travel destinations, induces the LLM-based Generative AI to generate information on sights, activities, and food, and then provides this information as travel destination information to the user terminal; a guide matching unit that receives a first language, which is the native language, from the user terminal and receives a second language, which is the language of the country being traveled to, extracts guides capable of speaking the first and second languages, filters the guides based on the location of the user terminal, and then transmits the guide list to the user terminal; and when room service is requested from the user terminal, classifies the type of room service and transmits the request to the terminal of the person in charge of the classified room service. A concierge service providing server comprising: a room service unit; a robot providing unit that, when the autonomous driving robot is called from the person in charge terminal and the room number of the room requesting the room service is entered into the autonomous driving robot, moves the autonomous driving robot to the location of the room number and transmits an alarm to the user terminal that the room service has arrived when it arrives at the location of the room number; and a marketing unit that conducts marketing for promotion to tourists of inbound and outbound tourism and transmits a marketing report to an advertiser terminal based on the results of the marketing; and the receiving unitThe questions transmitted to the user terminal are divided into two types of questions: a first question and a second question. The first question is configured as a fixed initial question to collect basic information regarding the user's travel style, budget, and preferred activities for the purpose of identifying the user's approximate travel preferences. The second question is configured as a dynamic follow-up question that is dynamically generated in real-time by the LLM-based generative AI based on the user's response to the first question for the purpose of identifying the user's detailed travel preferences. When analyzing the user's response through the LLM-based generative AI, sentiment analysis using a BERT (Bidirectional Encoder Representations from Transformers)-based model and semantic analysis using Word2Vec are performed to extract the sentiment and meaning of the user's response, and a personalized dynamic follow-up question is generated based on the extracted sentiment and meaning. The recommendation unit generates a preference vector by embedding the preferences and queries the preference vector into a pre-established travel destination database, RAG (Retrieval-Augmented Generation) After inputting, search for a travel destination, list the search results, and recommend them to the user terminal. When the guide matching unit transmits the guide list to the user terminal, it transmits a guide list containing randomly listed filtered guides, transmits a guide list reordered considering the user's tastes or preferences, or provides the results of filtering the guide list based on collaborative filtering to the user terminal. When reordering, if the user is female and prefers female guides, it reorders by placing female guides at the top; if the user considers the guide's experience and attitude, it reorders the filtered guides based on the guide's career or reviews. The robot providing unit ensures that the autonomous driving robot possesses an indoor map.An AI-based extended hotel concierge service provision system that enables an autonomous robot to autonomously find the room where room service was requested by including an interface that allows it to enter and exit the elevator and press elevator buttons, and enables the autonomous robot to carry the customer's luggage and follow the customer when the customer requests luggage carrying. Claim 2 delete Claim 3 delete Claim 4 delete Claim 5 delete Claim 6 delete Claim 7 delete

Images