Intelligent perfume recommendation and display method based on human-computer interaction and intelligent perfume display cabinet thereof

Abstract

The invention relates to an intelligent perfume recommendation and display method based on man-machine interaction and a perfume display cabinet thereof, a man-machine intelligent interaction module is provided for a user to carry out man-machine interaction through touch, voice or a mobile terminal, and the man-machine interaction comprises the steps that the user inputs preference information, selects a perfume spray bottle and carries out AI intelligent question and answer; an AI intelligent question answering module interacts with the user to perform AI intelligent question answering, analyzes habits, hobbies and life habits of the user based on preset selection questions answered by the user, and generates a personalized perfume recommendation list based on text description information input by the user; according to the selection of the user or the recommendation of the system, controlling the atomization device to spray the fragrance of the corresponding perfume for user experience; a user is supported to select perfume through a mobile terminal; by fusing voice control, AI questions and answers, data analysis and atomization technologies, personalized perfume recommendation and multi-mode perfume experience are realized, and the purchasing efficiency and interestingness of customers are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent perfume display cabinets, and in particular to an intelligent perfume recommendation and display method based on human-computer interaction and an intelligent perfume display cabinet. Background Art

[0002] The main function of a perfume display cabinet is to display perfumes. It usually has multi-layer shelves or display racks that can neatly arrange perfume bottles. It is suitable for various types of perfume sales places, such as shopping mall counters, specialty stores, beauty stores, etc. It is the most common type of perfume display cabinet.

[0003] Existing perfume display cabinets lack intelligent recommendation functions. Faced with dozens of perfumes, the lack of intelligent recommendation guidance makes it difficult for customers to make decisions. Ordinary consumers find it difficult to quickly find products that suit their preferences (such as fragrance, occasion, and season). They rely on the subjective recommendations of store clerks. Customers often need to spend a lot of time to understand and experience them one by one, and it is difficult to quickly find products that meet their preferences and needs. The user experience and shopping guide efficiency are low. Summary of the Invention

[0004] In view of this, the present application provides an intelligent perfume recommendation and display method that can improve user experience and shopping guide efficiency.

[0005] To achieve the above object, the present invention adopts the following technical solutions: An intelligent perfume recommendation and display method based on human-computer interaction includes the following steps: Provides a human-computer intelligent interaction module for users to interact with the computer through touch, voice, or mobile terminals. The human-computer interaction includes user input of preference information, selection of perfume spray bottles, and AI intelligent question and answer. Receive the user's specified perfume model or brand and retrieve the corresponding perfume spray bottle; The AI Q&A module interacts with users to conduct AI Q&A. Based on the preset multiple-choice questions answered by users, the module analyzes user habits, preferences, and lifestyle habits, and generates a personalized perfume recommendation list based on the text description information entered by the user. According to the user's choice or system recommendation, the atomizing device is controlled to spray the fragrance of the corresponding perfume for the user to experience.

[0006] As a preferred solution, the AI intelligent question answering includes the following steps: S100, receiving text description information input by the user through the human-computer intelligent interaction module, including voice questions and answers, answers to multiple-choice questions, or free text feedback; S200: Using an improved TF-IDF algorithm to process the text description information input by the user, including the following steps: Preprocess the text description information entered by the user, including word segmentation, part-of-speech filtering, and synonym merging, to extract keywords related to odor preferences; S300, combining the pre-trained word embedding model to calculate the semantic similarity between the word and the fragrance term, and dynamically adjust the TF-IDF weight; S400, generating a user's odor preference vector based on the keyword's TF-IDF improvement value and sentiment polarity; S500: Perform similarity matching on the odor preference vector and the fragrance feature vector in the perfume database to generate a recommendation list.

[0007] As a preferred solution, the dynamic adjustment of TF-IDF weight includes the following steps: S310, dynamic TF-IDF calculation, weighting the context of local text description information, calculating TF-IDF for each sentence or paragraph in the user description separately, capturing short-range semantic associations; S320, screening key fragrance words: taking the top-N words with the highest TF-IDF improvement values and mapping them to a standardized fragrance vocabulary in the perfume field; As a preferred solution, the method for constructing the odor preference vector includes: S410, fragrance dimension definition method: pre-define an N-dimensional classification system for fragrance, with each dimension corresponding to a fragrance type; S420, vector quantization method: mapping keywords extracted by the user to the classification system, and assigning positive / negative weights according to the TF-IDF improvement value and sentiment polarity; determining the polarity of words through sentiment analysis and adjusting the weight signs; S430, vector normalization method: perform L2 normalization on the vector to eliminate text length differences and obtain a final vector.

[0008] As a preferred solution, generating the recommendation list includes the following steps: S510, converting the fragrance components of each perfume into corresponding fragrance dimension vectors, and performing perfume feature vectorization; S520: The similarity matching is performed by using cosine similarity calculation to obtain the user preference vector and the perfume vector:

[0009] Among them, u is the user vector, v is the perfume vector; S530: Dynamically adjust recommendations. If a user repeatedly expresses "dislike" for a recommended perfume, the weights of other perfumes with high similarity to the perfume are reduced. The TF-IDF semantic weight parameter is optimized based on the user's actual spray experience times, and the recommendation weight is dynamically optimized based on the user's historical feedback data.

[0010] As a preferred solution, the preference information input by the user is received through the human-computer intelligent interaction module, and the preference information includes at least one of usage scenario, smell preference, and emotional needs; Based on the preference information input by the user, a sequence of interactive questions is dynamically generated, and the AI intelligent question-answering module conducts multiple rounds of questions and answers with the user according to the interactive question sequence; the AI intelligent question-answering module includes: a preset perfume preference question bank, and analyzes the user's personality, usage scenario, and scent preference through the user's answers; Based on the user's answers to the interactive question sequence, keywords are extracted and the user's scent preference vector is constructed. The recommendation results are dynamically optimized based on the machine learning algorithm, and the matching accuracy is improved by combining the user's historical selection data; The odor preference vector is matched with the perfume attributes in the perfume database, the matching degree is calculated, and at least three perfumes with the highest matching degree are screened out as recommendation results.

[0011] As a preferred solution, the atomization device includes: multiple independently controlled micro-sprayers, each connected to a variety of perfume storage bottles; according to user selection or system recommendation instructions, the atomized smell of the corresponding perfume is accurately sprayed to avoid odor contamination.

[0012] An intelligent perfume display cabinet for implementing the above method comprises: The cabinet has multiple perfume storage units and atomizing devices built in; Human-computer interaction module, integrating touch screen, voice recognition microphone and AI question-answering speaker; Control module, used to process user input signals, coordinate spray devices and recommendation logic; Wireless communication module, supporting connection with user mobile terminals to achieve remote control; The control module further includes: The semantic processing module is used to run the improved TF-IDF algorithm and word embedding model. The semantic processing module supports offline and cloud-based collaborative computing: in offline mode, a lightweight word embedding model is used; in cloud mode, a BERT deep learning model is used for semantic enhancement. A fragrance vocabulary storage unit, which contains standardized fragrance terms and associated word vectors; Real-time update module, adjust TF-IDF parameters and fragrance vocabulary based on new user feedback data.

[0013] As a preferred solution, the perfume storage unit is of replaceable design and is suitable for perfume spray bottles of different specifications; the atomization device includes independently controlled micro-atomization units, each micro-atomization unit corresponds to a perfume storage unit, and is equipped with a piezoelectric atomization piece to accurately control the amount of fragrance mist released; the cabinet is equipped with a clean airflow system and a perfume inventory management system: the perfume inventory management system monitors the perfume inventory in real time, reminds users to replenish or replace perfumes, and ensures the continuity of perfume recommendations and displays. The clean airflow system is connected to the control module to remove residual odors after each spray to prevent odor mixing and ensure the purity of the odor around the cabinet; it supports users to preview the perfume ingredients, tonality and recommendation reasons through mobile terminals. Users can also rate and comment on recommended perfumes through mobile terminals. The system optimizes the recommendation algorithm based on feedback to improve the accuracy of recommendations; the control module connects to the cloud database through a wireless communication module to regularly update the perfume library and recommendation algorithm; it supports users to scan codes to purchase perfumes after experience, or generate shopping lists through mobile terminals.

[0014] A computer-readable storage medium stores program instructions, wherein when the instructions are executed by a processor, the above-mentioned intelligent perfume recommendation and display method is implemented.

[0015] The above-mentioned intelligent perfume recommendation and display method based on human-computer interaction and its intelligent perfume display cabinet have the following significant beneficial effects: 1. By providing a human-machine intelligent interaction module that supports multimodal interaction via touch, voice, and mobile terminals, it breaks through the limitations of traditional display cabinets with a single display. Users can freely input their preferences and independently select perfume spray bottles. They can also obtain professional answers through AI intelligent Q&A. This greatly enhances the autonomy and fun of interaction, reduces selection difficulties, shortens average shopping time, and improves user experience and shopping efficiency. 2. The AI intelligent question-answering module, based on a dual analysis mechanism of pre-set multiple-choice questions and text descriptions, can deeply explore user habits, preferences, and lifestyles to generate a list of perfume recommendations that better meet user needs. This improves the matching accuracy of the personalized recommendation algorithm, effectively solving the problem of "one-size-fits-all" recommendations in traditional display cabinets and increasing the success rate of product promotion. 3. The recommended perfume's scent is sprayed in real time through an atomizer, allowing users to directly experience the scent, avoiding the limitations of traditional purchasing based solely on descriptions or packaging; 4. By combining AI Q&A with system recommendations, manual intervention is reduced, making the work of shopping guides easier and enabling an efficient and intelligent perfume recommendation and display process. 5. Support remote selection of perfume and triggering of display cabinet spray via mobile terminals, making it convenient for users to experience fragrance in a contactless scenario and improving flexibility and convenience of use.

[0016] In order to more clearly illustrate the structural features and effects of the present invention, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 Schematic diagram of the framework structure of a smart perfume display cabinet based on human-computer interaction provided by an embodiment of the present invention; Figure 2 This is a flowchart of the AI intelligent question answering provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0019] See also Figure 1 and Figure 2 , showing an intelligent perfume recommendation and display method based on human-computer interaction and an intelligent perfume display cabinet provided by an embodiment of the present invention, including the following steps: A human-computer intelligent interaction module is provided for users to interact with the computer through touch, voice, or mobile terminals. The human-computer interaction includes user input of preference information, selection of perfume spray bottles, and AI intelligent question and answer. A touch screen, buttons, and voice input device are provided. Users can select a perfume recommendation mode through the interface or directly use voice commands to control the perfume recommendation process. The human-computer intelligent interaction module uses natural language processing (NLP) and speech synthesis technology to recognize user voice commands and retrieve corresponding perfume information. Receive the user's designated perfume model or brand and retrieve the corresponding perfume spray bottle; the user specifies the perfume spray bottle through touch or voice input, and the system controls the designated spray bottle to prepare for spraying; The AI Q&A module interacts with users to conduct AI Q&A. Based on the preset multiple-choice questions answered by users, the module analyzes user habits, preferences, and lifestyle habits, and generates a personalized perfume recommendation list based on the text description information entered by the user. Based on the user's choice or system recommendation, the atomizer is controlled to spray the fragrance of the corresponding perfume for the user to experience; the system sprays the recommended perfumes through multiple spray heads for the user to taste, and the spray heads spray on demand to ensure that the user can accurately experience the taste of each perfume.

[0020] Furthermore, the AI intelligent question answering includes the following steps: S100. Receive the text description information input by the user through the human-machine intelligent interaction module, including voice questions and answers, multiple-choice question answers or free text feedback; the multiple-choice question library questions include: gender, main usage scenarios, preferred fragrance notes, fragrance retention duration, spice allergens, skin type (dry, oily, sensitive), usage season, budget range, etc. S200. Process the text description information input by the user using an improved TF-IDF algorithm, including the following steps: Preprocess the text description information input by the user, including word segmentation,词性 filtering, and synonym merging, and extract keywords related to odor preferences; specifically, for word segmentation and词性 filtering: use NLP tools (such as jieba, NLTK, or Spacy) to segment the text, retain nouns and adjectives (such as "fruity fragrance", "fresh", "sweet and greasy"), and剔除 stop words (such as "of", "and"); for synonym merging: merge similar words (such as "fruity fragrance" ≈ "fruit flavor") through a semantic dictionary (such as WordNet) or a word embedding model (Word2Vec). S300. Combine a pre-trained word embedding model to calculate the semantic similarity between words and fragrance note terms, and dynamically adjust the TF-IDF weights. S400. Generate the user's odor preference vector based on the improved TF-IDF value and sentiment polarity of the keywords. S500. Perform a similarity match between the odor preference vector and the fragrance note feature vectors in the perfume database to generate a recommendation list.

[0021] Furthermore, the dynamic adjustment of the TF-IDF weights includes the following steps: S310. Perform dynamic TF-IDF calculation, perform context weighting on the local text description information, and calculate TF-IDF separately for each sentence or paragraph described by the user.

[0022] where α is a smoothing parameter to avoid overflow when DF = 0. Semantic enhancement weight, calculate the similarity between words and perfume fragrance note terms (such as "floral fragrance", "woodsy note") using pre-trained word vectors (such as GloVe) as an additional weight: )) where ∈ the fragrance note word library, SemanticWeight(t) is generated based on word embedding similarity or a topic model, and its value is the maximum cosine similarity between the target word and the words in the fragrance note word library. Calculate the TF value of the word through a local context window to capture short-distance semantic associations, such as when the user emphasizes "hate sweet and greasy". Final weight: ; S320. Filter key fragrance terms: Take the top N terms with the highest TF-IDF improvement values (e.g., N=10) and map them to a standardized fragrance vocabulary in the perfume field (e.g., "fresh → citrus" and "sweet → gourmand"). For example, if a user describes "I like lemon and mint in the summer, but I hate strong musk," the following keywords are extracted: {lemon: 0.8, mint: 0.7, musk: -0.6} (weights include positive and negative preferences). Furthermore, the method for constructing the odor preference vector includes: S410, fragrance dimension definition method: pre-define an N-dimensional classification system for fragrance, with each dimension corresponding to a fragrance type; for example, a 10-dimensional vector: [citrus, floral, woody, gourmand, oriental, marine, green, fruity, musky, spicy]; S420, vector quantization method: Map the keywords extracted by the user to the classification system, and assign positive / negative weights based on the TF-IDF improvement value and sentiment polarity to perform positive / negative preference assignment. For example, if the user clearly likes a word (such as "lemon"), the corresponding dimension (citrus tone) weight is **+TF-IDF improvement value**, and if the user clearly dislikes a word (such as "musk"), the corresponding dimension (musk tone) weight is **-TF-IDF improvement value**. Use sentiment analysis (such as VADER) to determine the polarity of the word and adjust the weight sign to perform neutral description processing; S430, vector normalization method: performing L2 normalization processing on the vector to eliminate text length differences and obtain a final vector; according to: , For example, a user describes a preference vector [0.5, 0.1, -0.3, 0, 0, 0.2, 0.4, 0.6, -0.7, 0], which means a strong preference for fruity / citrus notes and a dislike for musk / woody notes.

[0023] Furthermore, generating the recommendation list includes the following steps: S510. Convert the fragrance components of each perfume (such as product descriptions or perfumer annotations) into corresponding fragrance dimension vectors to perform perfume feature vectorization. For example, the fragrance description of perfume A is "top note: lemon; middle note: jasmine; base note: cedar" → vector [0.7, 0.5, 0.3, 0, 0, 0, 0, 0, 0]; S520: The similarity matching is performed by using cosine similarity calculation to obtain the user preference vector and the perfume vector:

[0024] Among them, u is the user vector, v is the perfume vector; S530: Dynamically adjust recommendations. If a user repeatedly expresses "dislike" for a recommended perfume, the weights of other perfumes with high similarity to the perfume are reduced. The TF-IDF semantic weight parameter is optimized based on the user's actual spray experience times, and the recommendation weight is dynamically optimized based on the user's historical feedback data.

[0025] Practical application examples: User input: "I want a perfume that's perfect for summer, as refreshing as a sea breeze, with a hint of watermelon, and not too sweet." step: S1. Extract keywords: {sea breeze: 0.9, refreshing: 0.8, watermelon: 0.7, sweet: -0.6}; S2, mapped to the fragrance vector: sea breeze → ocean tone, refreshing → citrus tone, watermelon → fruity tone, sweet → gourmet tone; S3. Generate preference vector: [0.8, 0, 0, -0.6, 0, 0.9, 0, 0.7, 0, 0]; S4. Matching database: recommend "marine + fruity" perfumes (such as Jo Malone Blue Bell).

[0026] Furthermore, the preference information input by the user is received through the human-computer intelligent interaction module, wherein the preference information includes at least one of usage scenario, smell preference, and emotional need; Based on the preference information input by the user, a sequence of interactive questions is dynamically generated, and the AI intelligent question-answering module conducts multiple rounds of questions and answers with the user according to the interactive question sequence. The AI intelligent question-answering module includes: a preset perfume preference selection question bank, which is constructed based on a fragrance classification system and includes main categories such as floral, oriental, and fresh. The user's answers are used to analyze their personality, usage scenarios, and scent preferences. Based on the user's answers to the interactive question sequence, keywords are extracted and the user's scent preference vector is constructed. The recommendation results are dynamically optimized based on the machine learning algorithm, and the matching accuracy is improved by combining the user's historical selection data; The odor preference vector is matched with the perfume attributes in the perfume database, the matching degree is calculated, and at least three perfumes with the highest matching degree are screened out as recommendation results.

[0027] Furthermore, the atomization device includes: a plurality of independently controlled micro-sprayers, each connected to a plurality of perfume storage bottles; according to user selection or system recommendation instructions, the atomized smell of the corresponding perfume is accurately sprayed to avoid odor contamination.

[0028] An intelligent perfume display cabinet for implementing the above method comprises: The cabinet 1 has multiple perfume storage units 2 and multiple atomizing devices 3 built in; the perfume storage units 2 are correspondingly connected to the atomizing devices 3; Human-computer interaction module 4, integrating touch screen 41, voice recognition microphone 42 and AI question-answering speaker 43; The control module 5 is used to process user input signals, coordinate the spray device and the recommendation logic; the human-computer interaction module 4 is connected to the control module 5; The wireless communication module 6 supports connection with the user's mobile terminal 7 to achieve remote control; the wireless communication module 6 is connected to the control module 5; The control module 5 further includes: Semantic processing module 51, used to run the improved TF-IDF algorithm and word embedding model; the semantic processing module 51 supports offline and cloud-based collaborative computing: in offline mode, a lightweight word embedding model (such as Word2Vec) is used; in cloud-based mode, deep models such as BERT are called for semantic enhancement; A fragrance vocabulary storage unit 52, containing standardized fragrance terms and associated word vectors; The real-time updating module 53 adjusts the TF-IDF parameters and the fragrance vocabulary according to the newly added feedback data from users.

[0029] Furthermore, the perfume storage unit 2 is of replaceable design and is suitable for perfume spray bottles of different specifications; the atomization device 3 includes independently controlled micro-atomization units, each micro-atomization unit corresponds to a perfume storage unit 2, and is equipped with a piezoelectric atomization piece to accurately control the amount of fragrance mist released; the cabinet 1 is equipped with a clean airflow system 9 and a perfume inventory management system 10: the perfume inventory management system 10 and the clean airflow system 9 are respectively connected to the control module 5, the perfume inventory management system 10 monitors the perfume inventory in real time, reminds users to replenish or replace perfumes, ensures the continuity of perfume recommendations and displays, removes residual odors after each spray, prevents odor mixing, and ensures the purity of the odor around the cabinet 1; supports users to preview perfume ingredients, tonality and recommendation reasons through the mobile terminal 7, and users can also rate and comment on the recommended perfumes through the mobile terminal 7. The system optimizes the recommendation algorithm based on feedback to improve the accuracy of recommendations; the control module 5 is connected to the cloud database 8 to regularly update the perfume library and recommendation algorithm; supports users to scan codes to purchase perfumes after experience, or generate shopping lists through the mobile terminal 7.

[0030] A computer-readable storage medium stores program instructions, wherein when the instructions are executed by a processor, the above-mentioned intelligent perfume recommendation and display method is implemented.

[0031] The above-mentioned intelligent perfume recommendation and display method based on human-computer interaction and its intelligent perfume display cabinet have the following significant beneficial effects: 1. By providing a human-machine intelligent interaction module that supports multimodal interaction via touch, voice, and mobile terminals, it breaks through the limitations of traditional display cabinets with a single display. Users can freely input their preferences and independently select perfume spray bottles. They can also obtain professional answers through AI intelligent Q&A. This greatly enhances the autonomy and fun of interaction, reduces selection difficulties, shortens average shopping time, and improves user experience and shopping efficiency. 2. The AI intelligent question-answering module, based on a dual analysis mechanism of pre-set multiple-choice questions and text descriptions, can deeply explore user habits, preferences, and lifestyles to generate a list of perfume recommendations that better meet user needs. This improves the matching accuracy of the personalized recommendation algorithm, effectively solving the problem of "one-size-fits-all" recommendations in traditional display cabinets and increasing the success rate of product promotion. 3. The recommended perfume's scent is sprayed in real time through an atomizer, allowing users to directly experience the scent, avoiding the limitations of traditional purchasing based solely on descriptions or packaging; 4. By combining AI Q&A with system recommendations, manual intervention is reduced, making the work of shopping guides easier and enabling an efficient and intelligent perfume recommendation and display process. 5. Support remote selection of perfume and triggering of display cabinet spray via mobile terminals, making it convenient for users to experience fragrance in a contactless scenario and improving flexibility and convenience of use.

[0032] It should be noted that the present invention is not limited to the above-mentioned embodiments. Based on the creative spirit of the present invention, those skilled in the art can also make other changes. These changes made based on the creative spirit of the present invention should be included in the scope of protection required by the present invention.

Claims

1. An intelligent perfume recommendation and display method based on human-computer interaction, characterized in that: The following steps are involved: Provides a human-computer intelligent interaction module for users to interact with the computer through touch, voice, or mobile terminals. The human-computer interaction includes user input of preference information, selection of perfume spray bottles, and AI intelligent question and answer. Receive the user's specified perfume model or brand and retrieve the corresponding perfume spray bottle; The AI Q&A module interacts with users to conduct AI Q&A. Based on the preset multiple-choice questions answered by users, the module analyzes user habits, preferences, and lifestyle habits, and generates a personalized perfume recommendation list based on the text description information entered by the user. According to the user's choice or system recommendation, the atomizing device is controlled to spray the fragrance of the corresponding perfume for the user to experience.

2. The intelligent perfume recommendation and display method based on human-computer interaction according to claim 1, characterized in that: The AI intelligent question answering includes the following steps: S100, receiving text description information input by the user through the human-computer intelligent interaction module, including voice questions and answers, answers to multiple-choice questions, or free text feedback; S200: Using an improved TF-IDF algorithm to process the text description information input by the user, including the following steps: Preprocess the text description information entered by the user, including word segmentation, part-of-speech filtering, and synonym merging, to extract keywords related to odor preferences; S300, combining the pre-trained word embedding model to calculate the semantic similarity between the word and the fragrance term, and dynamically adjust the TF-IDF weight; S400, generating a user's odor preference vector based on the keyword's TF-IDF improvement value and sentiment polarity; S500: Perform similarity matching on the odor preference vector and the fragrance feature vector in the perfume database to generate a recommendation list.

3. The intelligent perfume recommendation and display method based on human-computer interaction according to claim 2, characterized in that: The dynamic adjustment of TF-IDF weight includes the following steps: S310, dynamic TF-IDF calculation, weighting the context of local text description information, calculating TF-IDF for each sentence or paragraph in the user description separately, capturing short-range semantic associations; S320 , screening key fragrance words: taking the top-N words with the highest TF-IDF improvement value and mapping them to the standardized fragrance vocabulary in the perfume field.

4. The intelligent perfume recommendation and display method based on human-computer interaction according to claim 3, characterized in that: The method for constructing the odor preference vector includes: S410, fragrance dimension definition method: pre-define an N-dimensional classification system for fragrance, with each dimension corresponding to a fragrance type; S420, vector quantization method: mapping keywords extracted by the user to the classification system, and assigning positive / negative weights according to the TF-IDF improvement value and sentiment polarity; determining the polarity of words through sentiment analysis and adjusting the weight signs; S430, vector normalization method: perform L2 normalization processing on the vector to eliminate the difference in text length and obtain the final vector.

5. The intelligent perfume recommendation and display method based on human-computer interaction according to claim 4, characterized in that: Generating the recommendation list comprises the following steps: S510, converting the fragrance components of each perfume into corresponding fragrance dimension vectors, and performing perfume feature vectorization; S520: The similarity matching is performed by using cosine similarity calculation to obtain the user preference vector and the perfume vector: Among them, u is the user vector, v is the perfume vector; S530: Dynamically adjust recommendations. If a user repeatedly expresses "dislike" for a recommended perfume, the weights of other perfumes with high similarity to that perfume are reduced. The TF-IDF semantic weight parameter is optimized based on the user's actual number of spray experiences, and the recommendation weight is dynamically optimized based on historical user feedback data.

6. The intelligent perfume recommendation and display method based on human-computer interaction according to claim 2, characterized in that: Receiving preference information input by the user through the human-computer intelligent interaction module, wherein the preference information includes at least one of usage scenario, smell preference, and emotional needs; Based on the preference information input by the user, a sequence of interactive questions is dynamically generated, and the AI intelligent question-answering module conducts multiple rounds of questions and answers with the user according to the sequence of interactive questions; The AI intelligent question-answering module includes: a preset perfume preference question bank, which analyzes the user's personality, usage scenarios and scent preferences through the user's answers; Based on the user's answers to the interactive question sequence, keywords are extracted and a user's scent preference vector is constructed, and the recommendation results are dynamically optimized based on a machine learning algorithm; The odor preference vector is matched with the perfume attributes in the perfume database, the matching degree is calculated, and at least three perfumes with the highest matching degree are screened out as recommendation results.

7. The intelligent perfume recommendation and display method based on human-computer interaction according to claim 1, characterized in that: The atomization device includes: multiple independently controlled micro-sprayers, which are respectively connected to multiple perfume storage bottles; according to user selection or system recommendation instructions, the atomized smell of the corresponding perfume is accurately sprayed to avoid odor contamination.

8. An intelligent perfume display cabinet implementing the method according to any one of claims 1 to 7, characterized in that: include: A cabinet (1) having a plurality of fragrance storage units (2) and an atomizing device (3) built therein; Human-computer interaction module (4), integrated with touch screen (41), voice recognition microphone (42) and AI question-answering speaker (43); A control module (5) for processing user input signals, coordinating the spray device and the recommendation logic; A wireless communication module (6) supports connection with a user mobile terminal (7) to achieve remote control; The control module (5) further includes: A semantic processing module (51) is used to run an improved TF-IDF algorithm and a word embedding model; the semantic processing module (51) supports offline and cloud-based collaborative computing: a lightweight word embedding model is used in offline mode; and a BERT deep model is called in cloud-based mode for semantic enhancement; A fragrance vocabulary storage unit (52), comprising standardized fragrance terms and associated word vectors; The real-time update module (53) adjusts the TF-IDF parameters and the fragrance vocabulary according to the user's newly added feedback data.

9. The intelligent perfume display cabinet according to claim 8, characterized in that: The perfume storage unit (2) is of replaceable design and is suitable for perfume spray bottles of different specifications; the atomization device (3) includes independently controlled micro-atomization units, each micro-atomization unit corresponds to a perfume storage unit (2), and is equipped with a piezoelectric atomization plate to accurately control the amount of fragrance mist released; the cabinet (1) is equipped with a clean airflow system (9) and a perfume inventory management system (10), the perfume inventory management system (10) monitors the perfume inventory in real time and reminds users to replenish or replace perfumes, the clean airflow system (9) is connected to the control module (5), and removes residual odors after each spray to prevent odor mixing; supports users to preview perfume ingredients, tonality and recommendation reasons through mobile terminals (7), and users can also rate and comment on recommended perfumes through mobile terminals (7). The system optimizes the recommendation algorithm based on feedback to improve the accuracy of recommendations; the control module (5) is connected to the cloud database (8) through the wireless communication module (6) to regularly update the perfume library and recommendation algorithm; supports users to scan codes to purchase perfumes after experience, or generate shopping lists through the mobile terminal (7).

10. A computer-readable storage medium storing program instructions, characterized in that: When the instructions are executed by the processor, the smart perfume recommendation and display method according to any one of claims 1 to 7 is implemented.

Images