A fitness management service method and device based on artificial intelligence

Abstract

The present invention relates to the field of fitness management technology, specifically to a fitness management service method and device based on artificial intelligence, including: obtaining the demand text of the target user based on a preset demand template text, obtaining the preset characteristic text of various fitness equipment, and obtaining the personal information text of the target user; and sentence encoding the characteristic text based on a pre-trained sentence feature encoding model to obtain the feature vectors of various fitness equipment. By combining the personal information text, demand text and characteristic text of fitness equipment of the target user, the present invention can tailor-make recommendations for fitness equipment that matches the user's needs and conditions. This personalized recommendation can effectively improve the user's fitness experience, making the user more efficient and comfortable during the fitness process, and through intelligent matching, it can reasonably utilize existing fitness equipment resources to avoid wasting certain equipment due to incompatibility with user needs.

Description

Technical Field

[0001] The present invention relates to the technical field of fitness management, and in particular to a fitness management service method and device based on artificial intelligence. Background Art

[0002] Traditional fitness equipment recommendation methods are often based on standardized selections and lack in-depth consideration of individual user differences. For example, equipment in gyms is usually selected based on general needs rather than customized based on the user's personal health status, fitness goals, and preferences. This can cause some users to be confused when choosing equipment or use equipment that is not suitable for their needs, thereby reducing fitness effects and experience. Traditional methods rely on manual recommendations, and users often need to manually screen or rely on the advice of professionals to choose suitable equipment, which is not only time-consuming but also prone to decision fatigue. Especially for ordinary users without professional knowledge, the process of selecting equipment may be very cumbersome and inefficient. Moreover, due to the large individual differences among users, a single recommendation method is difficult to cover the needs of each user, resulting in a low level of intelligence in the service. Summary of the invention

[0003] The technical problem to be solved by the present invention is to overcome the shortcomings of the above-mentioned prior art and provide a fitness management service method and device based on artificial intelligence.

[0004] The technical solution adopted to solve the above technical problems is: a fitness management service method based on artificial intelligence, including:

[0005] Based on a preset demand template text, obtain the demand text of the target user, obtain the preset characteristic texts of various fitness equipment, and obtain the personal information text of the target user;

[0006] Performing sentence encoding on the characteristic text based on a pre-trained sentence feature encoding model to obtain feature vectors of various fitness equipment;

[0007] Acquire the demand vector of the target user based on the personal information text, the demand text and the feature vectors of the various fitness equipment;

[0008] The demand vector of the target user is matched with the feature vectors of the various fitness equipment to obtain the feature vector that matches the demand vector, that is, to obtain the fitness equipment that matches the target user.

[0009] Preferably, obtaining the demand vector of the target user based on the personal information text, the demand text and the feature vectors of the various fitness equipment includes:

[0010] Obtaining a dense vector of the personal information text based on a preset embedding matrix;

[0011] Performing a word segmentation operation on the demand text to obtain a fitness word sequence corresponding to the demand text;

[0012] Matching the fitness word sequence with a preset fitness map to obtain a keyword set matching the fitness map;

[0013] The word-level feature vector corresponding to the keyword set is acquired based on the feature vectors of the keyword set and the various fitness equipment, that is, the word-level feature vector of the demand text is obtained.

[0014] Preferably, obtaining the demand vector of the target user based on the personal information text, the demand text and the feature vectors of the various fitness equipment further includes:

[0015] Sentence encoding is performed on the requirement text based on a pre-trained sentence feature encoding model to obtain a sentence feature vector of the requirement text;

[0016] Concatenating the word-level feature vector of the demand text and the sentence feature vector of the demand text to obtain a concatenated vector of the demand text;

[0017] The concatenated vector of the demand text and the dense vector of the personal information text are fused to obtain the demand vector of the target user.

[0018] Preferably, the fitness graph is composed of multiple triples, wherein the triples include a head entity, a relationship and a tail entity, wherein the head entity represents a node in the fitness graph and is used to represent a fitness term, the relationship represents the association between the head entity and the tail entity, and the tail entity represents another node in the fitness graph and is used to represent a fitness effect.

[0019] Preferably, the dense vector of the personal information text is expressed as follows:

[0020] ;

[0021] in, A dense vector representing personal information text, Represents personal information text vector, Represents the preset embedding matrix.

[0022] Preferably, obtaining the word-level feature vector corresponding to the keyword set based on the keyword set and the feature vectors of the various fitness equipment, that is, obtaining the word-level feature vector of the demand text, includes:

[0023] Linking each keyword in the keyword set to an entity in the fitness graph to obtain a linked entity set corresponding to the keyword set;

[0024] Extracting a neighbor entity of one hop of each link entity in the link entity set in the fitness graph to obtain a neighbor entity set corresponding to the link entity set;

[0025] Matching the feature vectors of the various fitness equipment with the neighbor entity set through an attention mechanism to obtain an attention score corresponding to the neighbor entity set;

[0026] The neighbor entity set is weighted on average based on the attention score to obtain a word-level feature vector corresponding to the keyword set, that is, to obtain a word-level feature vector of the demand text.

[0027] Preferably, the expression of the keyword set is as follows:

[0028] ;

[0029] in, A set of keywords representing the requirement text, Indicates the requirement text, The first keyword in the keyword set of the requirement text Keywords, A collection of entities representing a fitness graph;

[0030] The calculation formula of the attention score is as follows:

[0031] ;

[0032] in, Represents the neighbor entity set The attention scores of neighbor entities, Indicates The feature vector of fitness equipment corresponding to the neighbor entities, Represents the neighbor entity set The feature vectors of neighbor entities, Represents a set of neighbor entities

[0033] The expression of the word-level feature vector is as follows:

[0034] ;

[0035] in, Word-level feature vector representing the requirement text

[0036] Preferably, the training method of the sentence feature encoding model includes:

[0037] The target text is segmented by Jieba word segmentation to obtain multiple word segments, and the multiple word segments are embedded to obtain word segmentation vectors;

[0038] The word segmentation vector is input into a bidirectional long short-term memory network, and the sentence representation of the word segmentation vector is output through the bidirectional long short-term memory network.

[0039] Preferably, the sentence representation of the word segmentation vector is expressed as follows:

[0040] ;

[0041] in, Indicated by Encoded target text No. element vector, Indicated by Encoded target text No. element vector, Indicates the target text No. word vectors, Represents a bidirectional long short-term memory network

[0042] The technical solution adopted to solve the above technical problems is: a fitness management service device based on artificial intelligence, which is applicable to the fitness management service method based on artificial intelligence, including:

[0043] A text collection unit, the text collection unit is used to obtain the target user's demand text based on a preset demand template text, obtain preset characteristic texts of various fitness equipment, and obtain the target user's personal information text;

[0044] A sentence encoding unit, the sentence encoding unit is used to sentence encode the characteristic text based on a pre-trained sentence feature encoding model to obtain feature vectors of various fitness equipment;

[0045] A demand construction unit, the demand construction unit is used to obtain the demand vector of the target user based on the personal information text, the demand text and the feature vectors of the various fitness equipment;

[0046] The equipment matching unit is used to match the demand vector of the target user with the feature vectors of the various fitness equipment to obtain the feature vector that matches the demand vector, that is, to obtain the fitness equipment that matches the target user.

[0047] The beneficial effects of the present invention are as follows: (1) By combining the target user's personal information text, demand text and fitness equipment feature text, the present invention can tailor recommendations for users to match their needs and conditions. This personalized recommendation can effectively improve the user's fitness experience, making the user more efficient and comfortable during the fitness process. By converting text data into feature vectors based on a pre-trained sentence feature encoding model, efficient encoding and comparison of user needs and fitness equipment features are achieved. This precise vector matching can significantly improve the relevance and accuracy of matching results, avoiding the possible mismatching and inefficient recommendation problems in traditional recommendation systems. (2) The present invention automates the selection process of user needs and fitness equipment, processes and analyzes a large amount of information through artificial intelligence algorithms, reduces the need for manual intervention, and users do not need to manually screen and judge the most suitable fitness equipment, and automatically provide The optimal selection improves the intelligence level of the service and can provide customized fitness equipment recommendations based on the individual differences of the target users, reducing the confusion and selection fatigue of users when choosing equipment. In this way, users can more efficiently choose fitness equipment that suits them and improve the effectiveness and fun of fitness. (3) The present invention obtains user demand text and fitness equipment feature text. This method can handle the diverse needs of different users. Whether it is muscle building, fat loss, body shaping or rehabilitation, it can be intelligently matched and recommended according to the specific situation of the user. This flexibility enables it to cover a wider range of user groups and provide solutions for different fitness needs. Through intelligent matching, it can reasonably utilize existing fitness equipment resources to avoid wasting some equipment due to incompatibility with user needs. At the same time, gyms and other fitness facilities can also optimize equipment configuration according to changes in user needs and improve the utilization efficiency of facility resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A schematic flow chart of the steps of an overall method in an embodiment of the present invention;

[0049] Figure 2 A schematic diagram of the device architecture of an overall device in an embodiment of the present invention.

[0050] Figure numerals: 1. text collection unit; 2. sentence encoding unit; 3. demand building unit; 4. equipment matching unit. DETAILED DESCRIPTION

[0051] Embodiment 1, as Figure 1 As shown, the present invention proposes a fitness management service method based on artificial intelligence, comprising:

[0052] S1. Obtaining the target user's demand text based on a preset demand template text, obtaining the preset characteristic texts of various fitness equipment, and obtaining the target user's personal information text;

[0053] S2. Sentence encoding is performed on the characteristic text based on a pre-trained sentence feature encoding model to obtain feature vectors of various fitness equipment;

[0054] S3, obtaining the target user's demand vector based on the personal information text, demand text and feature vectors of various fitness equipment;

[0055] S4. Matching the demand vector of the target user with the feature vectors of various fitness equipment to obtain a feature vector that matches the demand vector, that is, obtaining fitness equipment that matches the target user.

[0056] In the present invention, the demand template text is a pre-set text format or template that can help identify user needs. This template may include various common types of user fitness needs, such as wanting to gain muscle, lose fat, improve physical fitness, etc. Through this template, the user's needs can be obtained in a more systematic manner; the target user's demand text refers to the specific fitness demand information collected from the user, which may be obtained through questionnaires, dialog boxes or other interactive forms. These texts describe the user's fitness goals, interests and requirements, such as whether equipment assistance is needed, the frequency of exercise, etc.; the fitness equipment characteristic text contains various information and characteristics about different fitness equipment, such as their functions, usage methods, applicable target muscle groups, difficulty, etc. For example, the characteristic text of a treadmill may include "suitable for aerobic exercise, can improve cardiopulmonary function, adjust exercise intensity, etc."; the sentence feature encoding model is a machine learning model based on natural language processing (NLP) technology, which is used to convert input text (such as fitness equipment characteristic text) into a feature vector. Through this model, the text can be converted into a numerical form (vector), and these numerical values ​​can capture the semantic information contained in the text; the feature vector is a numerical representation extracted from the text, which can be regarded as a mathematical representation of the text content. After encoding, each text (such as "treadmill" or "muscle building training") will become a vector with a specific dimension. This vector represents various semantic features of the text and can be used for subsequent matching and comparison; personal information text is personal data about the target user, including their age, gender, weight, health status, fitness history, training experience, etc. This information can help the system recommend the most suitable fitness plan or equipment for the user; the user's demand vector is compared with the feature vector of each equipment, and the most matching equipment is usually found by calculating their similarity. Common similarity calculation methods include but are not limited to cosine similarity, Euclidean distance and Manhattan distance.

[0057] Embodiment 2, a fitness management service method based on artificial intelligence proposed by the present invention, compared with embodiment 1, this embodiment further includes: obtaining a demand vector of a target user based on personal information text, demand text and feature vectors of various fitness equipment, including:

[0058] A1. Obtain the dense vector of personal information text based on the preset embedding matrix;

[0059] A2. Perform word segmentation on the demand text to obtain a fitness word sequence corresponding to the demand text;

[0060] A3, matching the fitness word sequence with the preset fitness map to obtain a keyword set matching the fitness map;

[0061] A4. Based on the keyword set and the feature vectors of various fitness equipment, the word-level feature vector corresponding to the keyword set is obtained, that is, the word-level feature vector of the demand text is obtained.

[0062] In this embodiment, a dense vector refers to a vector containing many non-zero elements. In text processing (such as NLP), it is often used to represent the semantics of a word or sentence. Unlike a sparse vector, each element of a dense vector usually contains a meaningful numerical value, which can better capture the contextual information of the word or sentence. An embedding matrix is ​​a matrix in which each row represents the vector of a word or concept. Common embedding methods include Word2Vec, GloVe, etc. These embedding matrices capture the semantic features of words by mapping them to a low-dimensional vector space. In the text processing process, pre-trained embedding matrices are usually used to convert words into dense vectors. Word segmentation is a basic step in natural language processing, which refers to the process of dividing a continuous text into words or subwords. For example, after word segmentation, "muscle building training equipment" may be divided into "muscle building" and "training equipment". The sequence obtained after word segmentation will help the system understand each important component in the text. A fitness graph is a knowledge graph that contains concepts related to the fitness field and their relationships. For example, the graph may include The relationship between "muscle building" and "dumbbells", or the connection between "aerobics" and "treadmills", provides structured association information for different fitness activities, equipment, and goals; the keyword set is fitness-related keywords extracted from the demand text. Through word segmentation and matching with the fitness map, the extracted keyword set will contain the core fitness goals or preferences in user needs; the word-level feature vector refers to a numerical vector associated with a specific word, which represents the semantic information of the word. By matching the keywords with the equipment features, the feature vector of each word can be obtained, thereby reflecting the fitness goals or equipment features in the demand text.

[0063] In an optional embodiment, obtaining the target user's demand vector based on the personal information text, the demand text and the feature vectors of various fitness equipment also includes:

[0064] A5. Sentence encoding is performed on the requirement text based on the pre-trained sentence feature encoding model to obtain a sentence feature vector of the requirement text;

[0065] A6. Concatenate the word-level feature vector of the requirement text and the sentence feature vector of the requirement text to obtain a concatenated vector of the requirement text;

[0066] A7. Fuse the concatenated vector of the demand text and the dense vector of the personal information text to obtain the demand vector of the target user.

[0067] It should be noted that the sentence feature encoding model is a model used to convert a sentence or text into a vector representation of a fixed length. Common sentence encoding methods include BERT, GPT, Sentence-BERT, etc. These models can convert the meaning of the entire sentence into a "feature vector" by understanding the grammar and semantics of the sentence for subsequent analysis or matching tasks; concatenating vectors refers to connecting multiple vectors in sequence into a new vector. In this example, the word-level feature vector and sentence feature vector of the requirement text are concatenated together to form a new vector representation. This concatenation method is usually used to combine information at different levels (such as word level and sentence level) to obtain a richer representation.

[0068] In an optional embodiment, the fitness graph is composed of multiple triples, wherein the triple includes a head entity, a relationship, and a tail entity, wherein the head entity represents a node in the fitness graph and is used to represent fitness terms, the relationship represents the association between the head entity and the tail entity, and the tail entity represents another node in the fitness graph and is used to represent fitness effects.

[0069] It should be noted that the fitness graph is a structured knowledge representation system used to store and organize various information about fitness. It is similar to a graph, consisting of nodes (Entities) and edges (Relations). Nodes represent entities, and edges represent the relationships between these entities. In the field of fitness, a fitness graph can contain information about fitness movements, equipment, exercise effects, fitness goals, etc. Nodes (entities) may be fitness exercises, equipment, skills, etc., and edges represent the relationships between these elements, such as how a certain exercise is related to fitness effects, or what type of training a certain equipment is suitable for; a triple is a basic data structure of a knowledge graph. The structure is used to represent the relationship between entities. Each triplet contains a head entity, a relationship, and a tail entity. The first element in the tuple usually represents a node or concept. In the fitness graph, this can be a specific fitness word or action (such as "squat", "push-up"); the second element in the triplet represents the connection or association between the head entity and the tail entity. The relationship describes how entities interact or connect with each other. For example, "improve", "enhance", "consume", etc. represent the relationship between an action and an effect; the third element in the triplet usually represents another node related to the head entity. In the fitness graph, the tail entity usually represents a fitness effect or result (for example, "muscle gain", "fat burning").

[0070] In an optional embodiment, the expression of the dense vector of the personal information text is as follows:

[0071] ;

[0072] in, A dense vector representing personal information text, Represents personal information text vector, Represents the preset embedding matrix.

[0073] In an optional embodiment, based on the keyword set and the feature vectors of various fitness equipment, the word-level feature vector corresponding to the keyword set is obtained, that is, the word-level feature vector of the demand text is obtained, including:

[0074] B1. Link each keyword in the keyword set to an entity in the fitness graph to obtain a linked entity set corresponding to the keyword set;

[0075] B2. extracting the neighbor entity of one hop of each link entity in the link entity set in the fitness graph to obtain the neighbor entity set corresponding to the link entity set;

[0076] B3. Match the feature vectors of various fitness equipment with the neighbor entity set through the attention mechanism to obtain the attention score corresponding to the neighbor entity set;

[0077] B4. Based on the attention score, the neighbor entity set is weighted evenly to obtain the word-level feature vector corresponding to the keyword set, that is, the word-level feature vector of the demand text is obtained.

[0078] It should be noted that linked entities refer to the process of finding and mapping or linking each keyword in the keyword set with the corresponding entity in the fitness graph. Each keyword is linked to one or more entities in the graph. For example, "squat" may be linked to the "squat" node in the graph; neighbor entities refer to other entities that are directly connected to an entity in a graph through one hop (one edge). For example, in the graph, if "squat" is directly connected to "increase thigh strength", then "increase thigh strength" can be regarded as a neighbor entity of "squat"; the attention mechanism is a technology that calculates a weighted average by assigning different importance weights (i.e., attention scores) to the input data. It is often used in neural networks, especially in natural language processing tasks, to improve the performance of the model by learning which parts are more important; the word-level feature vector is a numerical representation of a keyword or vocabulary in the demand text. By linking entities, extracting neighbor entities, calculating attention scores and weighted averaging, the feature vector corresponding to each keyword is finally obtained.

[0079] In an optional embodiment, the expression of the keyword set is as follows:

[0080] ;

[0081] A set of keywords representing the requirement text, Indicates the requirement text, The first keyword in the keyword set of the requirement text Keywords, A collection of entities representing a fitness graph;

[0082] The calculation formula of attention score is as follows:

[0083] ;

[0084] in, Represents the neighbor entity set The attention scores of neighbor entities, Indicates The feature vector of fitness equipment corresponding to the neighbor entities, Represents the neighbor entity set The feature vectors of neighbor entities, Represents a set of neighbor entities

[0085] The expression of word-level feature vector is as follows:

[0086] ;

[0087] in, Word-level feature vector representing the requirement text

[0088] In an optional embodiment, a method for training a sentence feature encoding model includes:

[0089] C1. Segment the target text through Jieba word segmentation to obtain multiple word segments, and embed the multiple word segments to obtain word segmentation vectors;

[0090] C2. Input the word segmentation vector into the bidirectional long short-term memory network, and output the sentence representation of the word segmentation vector through the bidirectional long short-term memory network.

[0091] In an optional embodiment, the sentence representation of the word segmentation vector is expressed as follows:

[0092] ;

[0093] in, Indicated by Encoded target text No. element vector, Indicated by Encoded target text No. element vector, Indicates the target text No. word vectors, Represents a bidirectional long short-term memory network.

[0094] It should be noted that the bidirectional long short-term memory network is a variant of the long short-term memory network (LSTM). It enhances the expressiveness of the model by simultaneously considering the forward and backward information of the sequence. It is particularly suitable for scenarios that need to consider contextual information, such as processing time series data and natural language processing (NLP) tasks.

[0095] Embodiment three, as Figure 2 As shown, the present invention proposes an artificial intelligence-based fitness management service device, which is applicable to the artificial intelligence-based fitness management service method, including:

[0096] The text collection unit 1 is used to obtain the target user's demand text based on the preset demand template text, obtain the preset characteristic text of various fitness equipment, and obtain the target user's personal information text;

[0097] A sentence encoding unit 2, which is used to perform sentence encoding on the characteristic text based on a pre-trained sentence feature encoding model to obtain feature vectors of various fitness equipment;

[0098] Demand construction unit 3, which is used to obtain the demand vector of the target user based on the personal information text, the demand text and the feature vectors of various fitness equipment;

[0099] The equipment matching unit 4 is used to match the demand vector of the target user with the feature vectors of various fitness equipment to obtain the feature vector matching the demand vector, that is, to obtain the fitness equipment matching the target user.

[0100] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto, and various changes can be made within the knowledge scope of technicians in the relevant technical field without departing from the purpose of the present invention.

Claims

1. A fitness management service method based on artificial intelligence, characterized in that: include: Based on the preset demand template text, the target user's demand text is obtained, the preset characteristic texts of various fitness equipment are obtained, and the personal information text of the target user is obtained; Performing sentence encoding on the characteristic text based on a pre-trained sentence feature encoding model to obtain feature vectors of various fitness equipment; Acquire the demand vector of the target user based on the personal information text, the demand text and the feature vectors of the various fitness equipment; Matching the demand vector of the target user with the feature vectors of the various fitness equipment to obtain the feature vector that matches the demand vector, that is, to obtain the fitness equipment that matches the target user; Acquiring the demand vector of the target user based on the personal information text, the demand text and the feature vectors of the various fitness equipment includes: Obtaining a dense vector of the personal information text based on a preset embedding matrix; Performing a word segmentation operation on the demand text to obtain a fitness word sequence corresponding to the demand text; Matching the fitness word sequence with a preset fitness map to obtain a keyword set matching the fitness map; Based on the keyword set and the feature vectors of the various fitness equipment, a word-level feature vector corresponding to the keyword set is obtained, that is, a word-level feature vector of the demand text is obtained; The method further comprises: obtaining the demand vector of the target user based on the personal information text, the demand text and the feature vectors of the various fitness equipments; Sentence encoding is performed on the requirement text based on a pre-trained sentence feature encoding model to obtain a sentence feature vector of the requirement text; Concatenating the word-level feature vector of the demand text and the sentence feature vector of the demand text to obtain a concatenated vector of the demand text; Fusing the concatenated vector of the demand text and the dense vector of the personal information text to obtain the demand vector of the target user; The fitness graph is composed of a plurality of triples, wherein the triples include a head entity, a relationship, and a tail entity, wherein the head entity represents a node in the fitness graph and is used to represent a fitness term, the relationship represents an association between the head entity and the tail entity, and the tail entity represents another node in the fitness graph and is used to represent a fitness effect; Acquiring the word-level feature vector corresponding to the keyword set based on the feature vectors of the keyword set and the various fitness equipment, that is, obtaining the word-level feature vector of the demand text, includes: Linking each keyword in the keyword set to an entity in the fitness graph to obtain a linked entity set corresponding to the keyword set; Extracting a neighbor entity of one hop of each link entity in the link entity set in the fitness graph to obtain a neighbor entity set corresponding to the link entity set; Matching the feature vectors of the various fitness equipment with the neighbor entity set through an attention mechanism to obtain an attention score corresponding to the neighbor entity set; The neighbor entity set is weighted on average based on the attention score to obtain a word-level feature vector corresponding to the keyword set, that is, to obtain a word-level feature vector of the demand text.

2. The fitness management service method based on artificial intelligence according to claim 1, characterized in that: The expression of the dense vector of the personal information text is as follows: ； in, A dense vector representing personal information text, Represents personal information text vector, Represents the preset embedding matrix.

3. The fitness management service method based on artificial intelligence according to claim 2, characterized in that: The expression of the keyword set is as follows: ； in, A set of keywords representing the requirement text, Indicates the requirement text, The first keyword in the keyword set of the requirement text Keywords, A collection of entities representing fitness graphs; The calculation formula of the attention score is as follows: ； in, Represents the neighbor entity set The attention scores of neighbor entities, Indicates The neighbor entities correspond to , Represents the neighbor entity set represents a set of neighbor entities; The expression of the word-level feature vector is as follows: ； in, Word-level feature vector representing the requirement text.

4. The fitness management service method based on artificial intelligence according to claim 3, characterized in that: The training method of the sentence feature encoding model comprises: The target text is segmented by Jieba word segmentation to obtain multiple word segments, and the multiple word segments are embedded to obtain word segmentation vectors; The word segmentation vector is input into a bidirectional long short-term memory network, and the sentence representation of the word segmentation vector is output through the bidirectional long short-term memory network.

5. The fitness management service method based on artificial intelligence according to claim 4, characterized in that: The expression of the sentence representation of the word segmentation vector is as follows: ； in, Indicated by Encoded target text No. Indicated by Encoded target text No. element vector, Indicates the target text No. Represents a bidirectional long short-term memory network.

6. An artificial intelligence-based fitness management service device, which is applicable to an artificial intelligence-based fitness management service method according to any one of claims 1 to 5, characterized in that: include: A text collection unit (1), the text collection unit (1) is used to obtain the target user's demand text based on a preset demand template text, obtain preset characteristic texts of various fitness equipment, and obtain the target user's personal information text; A sentence encoding unit (2), the sentence encoding unit (2) is used to perform sentence encoding on the characteristic text based on a pre-trained sentence feature encoding model to obtain feature vectors of various fitness equipment; A demand construction unit (3), the demand construction unit (3) being used to obtain a demand vector of the target user based on the personal information text, the demand text and the feature vectors of the various fitness equipment; An equipment matching unit (4), the equipment matching unit (4) is used to match the demand vector of the target user with the feature vectors of the various fitness equipment to obtain the feature vector that matches the demand vector, that is, to obtain the fitness equipment that matches the target user.

Images