Matching method for recommending online and offline script playing methods based on AI algorithm

Abstract

The invention belongs to the technical field of commercial entertainment recommendation, and discloses a matching method for recommending online and offline script playing methods based on an AI algorithm, and the method comprises the steps: collecting user immersion experience data, carrying out the multi-dimensional deconstruction of the immersion experience data, and obtaining a user preference pedigree; performing script emotion tension and interaction density analysis on the script content to form a cross-scene experience map; designing a dual-channel sensing network, and performing cross fusion on the user preference pedigree and the cross-scene experience atlas based on the dual-channel sensing network to generate a personalized matching matrix; establishing an immersion degree evaluation parameter set, and performing scene adaptability calculation on the personalized matching matrix to obtain a scene conversion coordination degree index; generating a script recommendation sequence according to the scene conversion coordination degree index, and realizing real-time recommendation adjustment and experience feedback collection through context awareness; the difference between online and offline script experiences is effectively quantified, and a scientific cross-scene matching basis is provided.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of commercial entertainment recommendation, more particularly, to an AI algorithm-based online and offline script play method recommendation method. BACKGROUND

[0002] With the booming development of immersive entertainment industry, script entertainment, as a new cultural consumption form integrating narrative, role-playing and social interaction, has become an important choice for young people's leisure and entertainment in the entertainment environment where digitalization and physical experience are interpenetrating. Precise matching of user preferences and suitable script types is of key significance to improving user experience, enhancing platform stickiness and optimizing industry resource allocation. In recent years, to meet the diversified entertainment needs and improve user selection efficiency, recommendation systems based on user portraits and content tags have been widely used in script entertainment. These systems analyze user historical preferences, social networks and consumption behaviors, and combine script types, difficulty and interactive modes to achieve intelligent matching and personalized recommendation of various play methods such as online text reasoning, offline real scene script and city exploration.

[0003] However, existing script play recommendation technology has significant limitations in dealing with the problem of differentiated adaptation of online and offline script experiences, which is particularly prominent in maintaining the coherence of cross-platform user experience. Specifically, when users experience script entertainment in different online and offline scenarios, due to fundamental differences in physical environment constraints, social interaction modes and immersion perception dimensions, scripts of the same theme or type present completely different experience effects in different carriers. For example, in a suspense and mystery script, online text reasoning relies on logical analysis ability and imagination, while offline real scene interaction emphasizes more on spatial exploration and team collaboration. Existing recommendation systems often simply map users' preferences on online platforms directly to offline recommendations, ignoring the essential differences between the two experience modes. This cross-scenario experience fragmentation not only leads to a decline in recommendation accuracy, but also causes strong cognitive gap and expectation deviation when users migrate from online to offline experience. The current mainstream method to solve this problem is to use a simple type matching and tag classification system to coarsely associate online and offline content through a manually preset rule base. However, this static rule-based solution not only lacks sensitivity to dynamic user preferences, but also requires a large amount of manual experience accumulation and continuous content annotation work, making it difficult to cope with the rapid iteration of script content and the complex changes in user preferences, and unable to meet the growing demand for seamless cross-scenario experience and lack of recommendation diversity, ultimately limiting the innovative development of script entertainment industry in user refinement operation and full-scenario experience design.

[0004] In view of this, the present application proposes an AI algorithm-based online and offline script play method recommendation method to solve the above problems. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art, in order to achieve the above-mentioned purposes, the present application provides the following technical scheme: a method for matching online and offline script play based on an AI algorithm, comprising:

[0006] Step S1: collecting user immersion experience data, multi-dimensionally deconstructing the immersion experience data, and obtaining a user preference spectrum;

[0007] Step S2: analyzing script content for script emotional tension and interaction density, and forming a cross-scene experience graph;

[0008] Step S3: designing a double-channel perception network, cross-fusing the user preference spectrum and the cross-scene experience graph based on the double-channel perception network, and generating a personalized matching matrix;

[0009] Step S4: establishing an immersion degree evaluation parameter set, calculating the scene adaptability of the personalized matching matrix, and obtaining a scene conversion coordination index;

[0010] Step S5: generating a script recommendation sequence according to the scene conversion coordination index, and realizing real-time recommendation adjustment and experience feedback collection through situational perception.

[0011] The technical effects and advantages of the method for matching online and offline script play based on an AI algorithm of the present application are as follows:

[0012] The present application multi-dimensionally deconstructs user immersion experience data to obtain a user preference spectrum containing a cognitive preference index and a social emotional inclination, and comprehensively captures the experience demand characteristics of the user in different scenes; the script content is analyzed for emotional tension and interaction density to form a cross-scene experience graph, overcoming the limitation in the prior art that simple label classification cannot accurately express the online and offline experience differences. Through parallel processing of the cognitive channel and the emotional channel, deep cross-fusion of the user preference spectrum and the cross-scene experience graph is realized to generate a personalized matching matrix, solving the problem that the existing recommendation system simply maps online platform preferences to offline recommendations. Compared with the traditional matching method based on static rules, the present application establishes a multi-dimensional immersion degree evaluation parameter set, calculates the scene adaptability of the personalized matching matrix, and obtains a scene conversion coordination index, effectively quantifying the differences in online and offline script experience, and providing a scientific cross-scene matching basis. Through multi-scale decomposition by wavelet transform, subtle emotional changes of the user are captured, while the traditional method often ignores these microscopic experience differences; in script analysis, through three-dimensional fusion of the narrative complexity index, the emotional tension spectrum, and the scene constraint matrix, accurate expression of the experience effect in different carriers online and offline is realized, solving the experience fragmentation problem caused by differences in physical environment restrictions, social interaction modes, and immersion perception dimensions. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1 The figure is a schematic diagram of the online and offline script play matching method based on the AI algorithm of the present application.

[0014] Figure 2 The figure is a schematic diagram of the online and offline script play matching system based on the AI algorithm of the present application. DETAILED DESCRIPTION

[0015] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0016] Embodiment 1

[0017] Please refer to Figure 1 The online and offline script play matching method based on the AI algorithm of the present embodiment includes the following steps:

[0018] Step S1: Collecting user immersion experience data, multi-dimensionally deconstructing the immersion experience data, and obtaining a user preference pedigree;

[0019] Step S2: Analyzing script emotional tension and interactive density of script content, and forming a cross-scene experience graph;

[0020] Step S3: Designing a double-channel perception network, cross-fusing the user preference pedigree and the cross-scene experience graph based on the double-channel perception network, and generating a personalized matching matrix;

[0021] Step S4: Establishing an immersion degree evaluation parameter set, calculating the scene adaptability of the personalized matching matrix, and obtaining a scene conversion coordination index;

[0022] Step S5: Generating a script recommendation sequence according to the scene conversion coordination index, and realizing real-time recommendation adjustment and experience feedback collection through situational perception.

[0023] Preferably, the process of performing step S1 can specifically include the following steps:

[0024] The user immersion experience data is collected, including plot participation trajectory, role interaction behavior, emotional response signal and decision tendency data; the plot participation trajectory is analyzed in time sequence to obtain narrative preference characteristics; the role interaction behavior is analyzed in social network to obtain interaction mode characteristics; the emotional response signal is calculated in emotion to obtain emotional fluctuation spectrum; the decision tendency data is analyzed in cognitive style to obtain thinking mode characteristics; the narrative preference characteristics and the thinking mode characteristics are interactively coupled to obtain a cognitive preference index; adaptive correlation processing is performed based on the interaction mode characteristics and the emotional fluctuation spectrum to obtain a social emotional tendency; and multi-dimensional cross verification is performed based on the cognitive preference index and the social emotional tendency to construct a user preference spectrum.

[0025] Specifically, first, the user immersion experience data is collected, including plot participation trajectory, role interaction behavior, emotional response signal and decision tendency data. The plot participation trajectory is obtained by recording the reading dwell time, click interaction and plot selection path of the user in the script, the sampling frequency is set to 5Hz, and the subtle changes in the user participation process can be captured; the role interaction behavior is obtained by recording the role dialogue selection of the user on the online platform, the role interaction video analysis and voice interaction in the offline scene, the sampling interval is dialogue level; the emotional response signal is collected by facial expression recognition, voice emotion analysis and text emotion evaluation, the sampling frequency is 1Hz, which is used to monitor the emotional changes of the user; the decision tendency data is collected by recording the selection of the user at key plot nodes, decision time and decision change behavior, the frequency is event triggered.

[0026] The plot participation trajectory is analyzed in time sequence to obtain narrative preference characteristics. Specifically, the user plot participation data is segmented by using a sliding time window (window size is 120 seconds, step is 30 seconds); the participation index in each time window is calculated, including dwell time distribution, interaction frequency and plot jump mode; the attention weight of the user to different narrative elements (such as suspense clues, role conflicts, emotional turning points) is extracted; the participation performance of the user under different narrative rhythms is analyzed to obtain quantitative indicators of preference for fast rhythm or slow rhythm; finally, the narrative preference characteristics including narrative theme preference vector (dimension is 8, corresponding to 8 main narrative types) and narrative structure preference vector (dimension is 5, corresponding to linear, branch, ring, multi-line parallel and open type) are formed.

[0027] Social network analysis was conducted on character interaction behavior to obtain interaction pattern characteristics. First, a user-character interaction network was constructed, with nodes representing users and characters in the script, and edges representing interaction intensity. The frequency, duration, and emotional inclination of user interactions with each character were calculated. Centrality indicators of users in the interaction network were analyzed, including degree centrality (reflecting social activity) and betweenness centrality (reflecting information control ability). User-preferred interaction patterns were extracted, such as dominant (high centrality), collaborative (high connectivity), or observant (low connectivity). Based on the social network analysis results, interaction pattern characteristics were generated, including a social dominance tendency index (values ​​from 0 to 1, with values ​​above 0.7 indicating strong dominance), a character preference vector (dimension equal to the number of characters in the script), and conflict handling style parameters (values ​​across five dimensions: cooperative, competitive, compromising, avoidant, and accommodating).

[0028] Emotional computation is performed on the emotional response signal to obtain the emotional fluctuation spectrum. A suitable wavelet transform function for emotion recognition is selected, such as the Daubechies wavelet (db4), to perform multi-scale decomposition of the emotional response signal, obtaining emotional sub-signals in different frequency bands. Typically, the signal is decomposed to three levels, generating eight sub-bands. The intensity of the emotional sub-signals in different frequency bands is then quantized to generate an emotional intensity distribution map. The intensity calculation formula is E(e)=∑|S_e(t)| 2 Here, E(e) represents the emotional intensity of the e-th sub-band, and S_e(t) represents the signal of the e-th sub-band. Based on the emotional intensity distribution map, the K-means clustering algorithm (K=5, corresponding to five basic emotions: joy, sadness, fear, anger, and surprise) is used to cluster emotional types, selecting key frequency bands containing core emotions. The signals of the key frequency bands are reconstructed to obtain emotional feature signals. The fluctuation frequency (number of emotional changes per minute), duration (single emotion duration), and peak intensity (maximum emotional intensity) of the emotional feature signals are calculated to form an emotional fluctuation spectrum. The emotional fluctuation spectrum can reflect the emotional change patterns of users during the script experience, providing an emotional basis for subsequent personalized recommendations.

[0029] Cognitive style analysis was performed on decision-making tendency data to obtain thinking pattern characteristics. User selection records at key decision points were categorized and statistically analyzed to create a decision type distribution map; the distribution of decision time was analyzed, and the average decision time (ADT) and standard deviation of decision time (DTSD) were calculated; based on decision-changing behavior, a hesitation index was calculated, which is the ratio of the number of changed decisions to the total number of decisions; a decision tree model was used to analyze user decision paths and extract decision rule characteristics; combining the above indicators, principal component analysis was used to construct thinking pattern characteristics, including an intuition-analysis dimension (range -1 to 1, negative values ​​indicate intuitiveness, positive values ​​indicate analysis), a risk preference dimension (range 0 to 10, higher values ​​indicate higher risk preference), and an openness dimension (range 0 to 1, representing the degree of acceptance of new experiences). These thinking pattern characteristics reflect users' cognitive processing methods and decision-making styles in the script experience.

[0030] A cognitive preference index is obtained by interactively coupling narrative preference features with thinking pattern features. A narrative-thinking coupling matrix is ​​constructed, where matrix elements represent the matching degree between narrative elements and thinking patterns. Tensor products are performed on the narrative preference feature vector and the thinking pattern feature vector to obtain an initial coupling tensor. Singular value decomposition is applied to reduce the dimensionality of the coupling tensor to extract the main feature patterns. A cognitive preference index vector is calculated, including four dimensions: exploration and discovery tendency (values ​​from 0 to 100), logical reasoning tendency (values ​​from 0 to 100), emotional experience tendency (values ​​from 0 to 100), and social interaction tendency (values ​​from 0 to 100). The cognitive preference index quantifies the degree of user preference for different script types and narrative structures at the cognitive level.

[0031] Social sentiment tendency is obtained through adaptive correlation processing based on interaction pattern features and sentiment fluctuation spectrum. A temporal association rule mining algorithm is employed to analyze the temporal correlation patterns between interactive behavior and sentiment changes; the interaction-sentiment covariance matrix is ​​calculated to quantify the correlation strength between different interaction types and sentiment states; an adaptive weighting algorithm is applied to dynamically adjust the weights of interaction features and sentiment features according to the correlation strength; the weighted features are then fused to generate social sentiment tendency indicators, including a social-driven sentiment index (ranging from -1 to 1, representing the direction and intensity of the influence of social interaction on sentiment) and a sentiment-guided interaction index (ranging from 0 to 2, representing the degree to which sentiment state guides interactive behavior). Social sentiment tendency reflects the correlation pattern between users' social interactions and emotional experiences, providing an important basis for cross-scenario matching.

[0032] A user preference spectrum is constructed using multi-dimensional cross-validation based on cognitive preference index and social sentiment tendency. A multi-dimensional cross-validation matrix is ​​designed to test the consistency and complementarity among different features; low-reliability features and outliers are removed, while high-reliability features are retained; a non-negative matrix factorization algorithm is applied to fuse and reconstruct the multi-dimensional features; finally, a user preference spectrum is constructed, containing feature vectors from cognitive, sentiment, social, and behavioral dimensions, forming a complete user preference feature space. This user preference spectrum comprehensively describes the user's preference characteristics in the script experience, providing a foundation for subsequent personalized recommendations.

[0033] Preferably, the process of performing step S2 may specifically include the following steps:

[0034] Natural language processing is performed on the script content to extract key plot points and emotional markers; based on the plot points, a narrative structure diagram of the script is constructed; complexity analysis is performed on the narrative structure diagram to obtain a narrative complexity index; emotional semantic analysis is performed on the script text to obtain an emotional fluctuation curve; based on the emotional fluctuation curve, an emotional tension value is calculated to construct an emotional tension spectrum; scene elements are extracted from both online and offline scripts to establish a physical interaction mapping table; based on the physical interaction mapping table, interaction density and spatial constraints are analyzed to construct a scene constraint matrix; and the narrative complexity index, emotional tension spectrum, and scene constraint matrix are fused in three dimensions to form a cross-scene experience map.

[0035] Specifically, natural language processing (NLP) was performed on the script content to extract key plot points and sentiment markers. A BERT-based language model was used for semantic analysis of the script text to identify key plot points. Sentiment markers were added to text fragments using a sentiment lexicon and contextual semantic analysis. The sentiment markers employed a three-dimensional VAD (Value-Arousal-Dominance) model, with each dimension ranging from 0 to 1. Decision points and plot turning points in the script were identified through syntactic analysis and dependency parsing. A sequence of plot points and sentiment markers was established as the foundational data for subsequent analysis. During NLP, a finely tuned language model was used to address the script's unique dialogue and narrative structure, achieving an accuracy of 87.5%.

[0036] Based on plot nodes, a narrative structure graph for the script is constructed. Extracted plot nodes are used as nodes in the graph, and causal and temporal relationships between plots are used as edges. The importance weights of nodes are calculated based on centrality indicators and narrative functions. The types and strengths of edges are analyzed to distinguish between necessary and optional paths. Key paths, branching points, and loops in the narrative structure are identified. Finally, a weighted directed graph representation of the narrative structure graph is formed. For large scripts, the narrative structure graph typically contains 30 to 50 nodes and 40 to 70 edges; complex scripts can have over 100 nodes.

[0037] A complexity analysis was performed on the script's narrative structure graph to obtain narrative complexity metrics. Graph theory metrics for the structure graph were calculated, including average degree, clustering coefficient, and path length. The number and depth of narrative branches were analyzed to quantify the complexity of narrative choices. The information entropy between narrative nodes was assessed to reflect the uncertainty of plot development. Combining these metrics, a narrative complexity index was constructed, including linear complexity (values ​​0-10), branch complexity (values ​​0 to 10), and overall complexity (values ​​0 to 100). The narrative complexity index reflects the complexity of the script's structure and is an important reference for assessing user cognitive load.

[0038] Sentimental semantic analysis was performed on the script text to obtain an emotional fluctuation curve. Deep semantic parsing was used to label the script text with emotions, obtaining an initial emotional sequence. A multi-layered emotion classification model was employed for emotion analysis, accurately identifying eight basic emotions (joy, trust, fear, surprise, sadness, disgust, anger, and anticipation) and their intensity. Kalman filtering was applied to the initial emotional sequence for temporal smoothing, with a window size of 5 plot units and a smoothing factor of 0.3, resulting in an emotional flow graph. A peak detection algorithm (with a minimum peak interval set to 5% of the script length and a peak threshold set to 1.5 times the average emotional intensity) was used to identify peaks and troughs in the emotional flow graph, extracting key emotional turning points. Based on these key emotional turning points, an emotional fluctuation curve was constructed, with the script's progression on the horizontal axis and emotional intensity on the vertical axis. The emotional fluctuation curve reflects the dynamic process of emotional development in the script and forms the basis for constructing an emotional tension spectrum.

[0039] Emotional tension values ​​are calculated based on emotional fluctuation curves, and an emotional tension spectrum is constructed. The formula for calculating emotional tension is defined as follows: Where T(i) is the emotional tension value at the i-th time point, I(i) is the emotional intensity, and V(i) is the emotional diversity. The emotional tension rate was calculated; the emotional tension sequence throughout the script was analyzed; statistical characteristics of emotional tension were extracted, including average tension, peak tension, tension fluctuation amplitude, and tension rise / fall rate; based on these characteristics, an emotional tension spectrum was constructed to quantify the emotional fluctuation characteristics of the script. Analysis shows that the emotional tension spectrum of a high-quality script typically has 3 to 5 distinct tension peaks, with the peak interval approximately 20% to 25% of the total script length.

[0040] Scene elements are extracted from both online and offline scripts to establish a physical interaction mapping table. For online scripts, interface interaction elements, multimedia content types, and interaction frequencies are extracted; for offline scripts, physical space layout, prop types, and interpersonal interaction patterns are extracted. The core experience elements and interaction mechanisms of corresponding online and offline scenes are analyzed. A physical interaction mapping table is established to define the correspondence and conversion mechanisms between online and offline elements. The physical interaction mapping table includes multiple dimensions, such as spatial dimension (mapping between virtual and physical space), interaction dimension (mapping between click operations and physical operations), and sensory dimension (mapping between audiovisual experience and multisensory experience).

[0041] Based on the analysis of physical interaction mapping tables, interaction density and spatial constraints are considered to construct a scene constraint matrix. The interaction density index for online and offline scenarios is calculated as: interaction density = number of interactions per unit time × interaction complexity. Spatial constraints in offline scenarios are assessed, including venue area requirements, participant limits, and device dependence. Experience loss and gain during online-offline conversion are analyzed. Finally, a scene constraint matrix is ​​constructed, where matrix elements represent the constraint relationships and compatibility between different scenario characteristics. The scene constraint matrix typically has a 5×5 dimension, corresponding to five main scenario characteristics: spatial requirements, social density, sensory stimulation, interaction complexity, and device dependence.

[0042] A cross-scene experience graph is formed by fusing narrative complexity indicators, emotional tension spectra, and scene constraint matrices in three dimensions. A three-dimensional fusion algorithm is designed to represent the three types of features in a common parameter space; the compatibility index between features is calculated to evaluate the matching degree of features in different dimensions; a cross-dimensional mapping function is constructed to achieve a unified expression of narrative, emotional, and scene characteristics; finally, a cross-scene experience graph is formed, which represents the experience characteristics of the script in different scenarios in the form of a three-dimensional tensor. The cross-scene experience graph provides a comprehensive feature representation for evaluating the adaptability of the script in online and offline environments.

[0043] Preferably, the process of performing step S3 may specifically include the following steps:

[0044] A cognitive channel network and an emotional channel network are constructed. These two networks are then combined to design a dual-channel perception network. Parameters of the dual-channel perception network are adjusted using historical matching cases to obtain an optimized network. An adaptive weight allocation method is used to extract features from the user preference spectrum and cross-scenario experience graph. These extracted features are then input into the optimized dual-channel perception network to obtain scenario suitability prediction results. Based on the scenario suitability prediction results, a preference-experience mapping relationship is constructed. This preference-experience mapping relationship is then collaboratively calculated with preset cross-scenario conversion rules to obtain a personalized matching matrix.

[0045] Specifically, a cognitive channel network is constructed to handle the matching relationship between user cognitive preferences and script narrative complexity. The cognitive channel network employs a multilayer perceptron structure, containing three hidden layers (64, 32, and 16 nodes respectively). The input layer receives user cognitive preference indices and script narrative complexity indicators; the hidden layers use the ReLU activation function to capture complex cognitive matching patterns through nonlinear transformations; the output layer generates a cognitive fit score, reflecting the degree of matching between user cognitive ability and script complexity. The cognitive channel network focuses on handling cognitive-level matching problems such as logical reasoning, puzzle solving, and narrative comprehension within the script.

[0046] An emotional channel network was constructed to assess the coordination between user emotional preferences and the emotional tension of the script. The emotional channel network employs a recurrent neural network structure, specifically a GRU (Gated Recurrent Unit) network, containing two GRU layers (64 units each). The input layer receives temporal data of the user's emotional fluctuation spectrum and the script's emotional tension spectrum. The GRU layers capture the temporal patterns and dynamic characteristics of emotional changes. An attention mechanism layer weights key emotional nodes. The output layer generates an emotional resonance score, representing the degree of matching between the user's emotional preferences and the script's emotional experience. The emotional channel network focuses on addressing the dynamic matching problem of emotional experiences, assessing whether users can obtain the expected emotional stimulation and satisfaction.

[0047] A dual-channel perception network is designed by combining cognitive and emotional channel networks. A channel fusion layer receives the output features from both channels. A gating mechanism dynamically adjusts the weights of the two channels, calculated as: W_c = σ(W·[O_c,O_e] + b), where W_c is the cognitive channel weight, O_c and O_e are the outputs of the cognitive and emotional channels respectively, and σ is the sigmoid function. A cross-channel attention module is designed to achieve mutual enhancement of cognitive and emotional features. Finally, a comprehensive matching score is output through a fully connected layer. This dual-channel perception network integrates matching evaluation from both cognitive and emotional dimensions, enabling a comprehensive assessment of the user's suitability for the script.

[0048] The parameters of the dual-channel perceptual network were tuned using historical matching cases to obtain an optimized dual-channel perceptual network. A user experience feedback database containing 4,500 user experience records was collected, each including user characteristics, script characteristics, scene type, and satisfaction rating. The dataset was divided into training, validation, and test sets in an 8:1:1 ratio. The loss function was defined as the weighted sum of mean squared error and cross-entropy, with a weight ratio of 7:3. The Adam optimizer was used for parameter optimization, with an initial learning rate of 0.001 and a learning rate decay strategy, decreasing by 10% every 50 epochs. An early stopping mechanism was implemented, stopping training when the validation set loss no longer decreased for 15 consecutive epochs. The final matching accuracy on the test set reached 83.7%. The parameter tuning process enabled the dual-channel perceptual network to accurately predict user satisfaction with script experiences in different scenarios.

[0049] An adaptive weighting method is employed to extract features from user preference profiles and cross-scenario experience graphs. These extracted features are then input into an optimized dual-channel perceptual network to obtain scenario suitability prediction results. An adaptive weighting algorithm is designed to dynamically adjust feature weights based on user historical behavior and current scenario characteristics. Key preference features are extracted from the user preference profile, including cognitive dimensions (exploration and discovery tendency, logical reasoning tendency) and emotional dimensions (emotional experience tendency, social interaction tendency). Script features are extracted from the cross-scenario experience graph, including narrative complexity, emotional tension spectrum, and scenario constraint features. The extracted features are input into the dual-channel perceptual network according to cognitive and emotional channels respectively. The network outputs scenario suitability prediction results, including online suitability and offline suitability scores. The scenario suitability prediction results reflect the degree of matching between users and specific scripts in online and offline environments.

[0050] Based on the scenario suitability prediction results, a preference-experience mapping relationship is constructed. The correspondence between user preference features and script experience features is analyzed; a mapping function set is constructed, defining the mapping rules from the user preference space to the script experience space; dedicated mapping models are established for different scenario types (online text, online multimedia, offline real-world scenarios, etc.); and a preference-experience mapping relationship matrix is ​​formed, where matrix elements represent the sensitivity of specific preference features to specific experience features. The preference-experience mapping relationship provides a theoretical basis for personalized recommendations, achieving precise matching between user needs and script experiences.

[0051] The preference-experience mapping relationship is collaboratively calculated with preset cross-scenario conversion rules to obtain a personalized matching matrix. The preset cross-scenario conversion rules are formulated based on expert experience and user research, and include scenario conversion gain rules and scenario conversion loss rules. Gain rules define experience elements that may be enhanced during the online-to-offline conversion process (e.g., a social interaction intensity enhancement coefficient of 1.5 to 2.0, and a sensory stimulation richness enhancement coefficient of 2.0 to 3.0). Loss rules define experience elements that may be weakened (e.g., a narrative rhythm control weakening coefficient of 0.6 to 0.8, and an information accessibility weakening coefficient of 0.5 to 0.7). Tensor operations are performed on the preference-experience mapping relationship and the cross-scenario conversion rules to obtain a conversion adaptation tensor. Based on the conversion adaptation tensor, a personalized matching matrix is ​​calculated. The matrix dimension is M×N, where M is the number of users and N is the number of scripts. Matrix elements represent the degree of matching between a specific user and a specific script in different scenarios. The personalized matching matrix is ​​the core data structure for subsequent recommendation calculations.

[0052] Preferably, the process of performing step S4 may specifically include the following steps:

[0053] A multi-dimensional immersion assessment parameter set is established, including cognitive engagement index, emotional resonance index, social interaction index, and physical participation index. Based on the multi-dimensional immersion assessment parameter set, the personalized matching matrix is ​​transformed to obtain scene adaptability parameters. The various scene adaptability parameters are balanced to obtain standardized adaptability indicators. The standardized adaptability indicators are fused using multiple thresholds to obtain a comprehensive scene coordination degree. Based on the comprehensive scene coordination degree, scene conversion levels are classified. The stability of the standardized adaptability indicators is assessed to obtain a scene conversion risk index. Based on the scene conversion level and the scene conversion risk index, a scene conversion coordination degree index is generated.

[0054] Specifically, a multi-dimensional immersion assessment parameter set is established to comprehensively evaluate users' immersive experiences in different scenarios. The cognitive engagement index includes attention concentration, puzzle-solving participation, and narrative comprehension depth, reflecting the user's level of cognitive engagement; the emotional resonance index includes emotional response intensity, emotional variation amplitude, and emotional satisfaction, reflecting the quality of the emotional experience evoked by the script; the social interaction index includes interaction frequency, character relationship development, and group collaboration intensity, reflecting the richness of the social experience; and the physical participation index includes physical activity level, environmental exploration level, and sensory stimulation diversity, reflecting the level of participation in offline physical experiences. This multi-dimensional immersion assessment parameter set comprehensively describes the characteristics of users' immersive experiences from four dimensions: cognitive, emotional, social, and physical.

[0055] Based on a multi-dimensional immersion assessment parameter set, the personalized matching matrix is ​​transformed to obtain scene adaptability parameters. A scene adaptability transformation function is designed to map the matching degree values ​​in the personalized matching matrix to the immersion parameter space. For online scenarios, the cognitive investment index and emotional resonance index are the main evaluation metrics, with conversion weights of 0.4 and 0.35, respectively. For offline scenarios, the social interaction index and physical participation index are the main evaluation metrics, with conversion weights of 0.3 and 0.35, respectively. The expected immersion values ​​of users under different scene conversion paths are calculated, resulting in a scene adaptability parameter set, including cognitive adaptability parameters, emotional adaptability parameters, social adaptability parameters, and physical adaptability parameters. The scene adaptability parameters reflect the changing characteristics of users' immersion experience across various dimensions during scene conversion.

[0056] The adaptive parameters for each scenario are balanced to obtain standardized adaptive indices. Since the dimensions and ranges of adaptive parameters differ across dimensions, standardization is necessary. The Z-score standardization method is used to convert the parameters into a standard normal distribution with a mean of 0 and a standard deviation of 1. Considering individual user differences, personalized correction coefficients are introduced. These coefficients are calculated based on historical user experience data and reflect the user's sensitivity to each dimension of experience. The standardized parameters are then adjusted using these correction coefficients to obtain personalized standardized adaptive indices. Standardization makes the adaptive indices across different dimensions comparable, laying the foundation for subsequent integration.

[0057] Standardized adaptability indicators are fused using multiple thresholds to obtain a comprehensive scene coordination degree. A multi-level fuzzy inference function is constructed to map each standardized adaptability indicator to a scene fit degree. The mapping process is based on fuzzy set theory, using trapezoidal membership functions to define three levels: low, medium, and high. The consistency coefficient of each scene fit degree is calculated to evaluate the degree of coordination between different dimensions of indicators. The consistency coefficient is calculated using weighted Kappa coefficients. Based on the consistency coefficients, the weights of each standardized adaptability indicator are adjusted, with dimensions with high consistency receiving higher weights. The Delphi aggregation method is used to fuse the adjusted scene fit degrees to calculate the comprehensive scene coordination degree, with a value ranging from 0 to 100. The Delphi aggregation method references the evaluation criteria of 25 script experience experts to ensure the professionalism and reliability of the fusion results.

[0058] Based on comprehensive scenario coordination, scenario conversion levels are categorized. According to the distribution characteristics of comprehensive scenario coordination and user experience research data, thresholds for scenario conversion level categorization are set; scenario conversion levels are divided into five levels: Excellent (coordination 90-100), Good (coordination 75-90), Moderate (coordination 60-75), Barely acceptable (coordination 45-60), and Unsuitable (coordination 0-45); different levels correspond to different recommendation strategies and conversion support measures; the level categorization thresholds are dynamically adjusted to adapt to changes in user groups and scenario characteristics. Scenario conversion levels provide a clear decision-making basis for the recommendation system, facilitating the implementation of tiered recommendation strategies.

[0059] A stability assessment of standardized adaptability indicators is conducted to obtain a scenario conversion risk index. Based on the volatility and uncertainty of the standardized adaptability indicators, a scenario conversion risk model is established. The Monte Carlo method is used to simulate random fluctuations in user experience, and the stability probability is calculated through 10,000 iterations. The risk levels of different scenario conversion paths are evaluated, including experience discontinuity risk, expectation gap risk, and maladaptation risk. By integrating various risk factors, the scenario conversion risk index is calculated, with a value ranging from 0 to 100, where a higher value indicates greater risk. The scenario conversion risk index provides a basis for risk control in recommendation decisions and helps avoid negative user experiences.

[0060] Based on the scenario conversion level and scenario conversion risk index, a scenario conversion coordination index is generated. A comprehensive evaluation function is designed to integrate the scenario conversion level (main indicator) and the scenario conversion risk index (correction indicator) into a single evaluation index; the scenario conversion coordination index is calculated using the weighted geometric mean method, with the formula: STCI = TE α ×(100-RI) β The STCI (Scene Conversion Coordination Index) is the scenario conversion level score, TE (Scene Conversion Grade Score) is the scenario conversion risk index, and α and β are weighting coefficients (empirical values ​​of 0.7 and 0.3, respectively). The resulting scenario conversion coordination index ranges from 0 to 100, directly reflecting the script's suitability for online and offline scenario conversion. The scenario conversion coordination index is the core basis for generating script recommendation sequences, comprehensively considering both matching degree and risk factors.

[0061] Preferably, the process of performing step S5 may specifically include the following steps:

[0062] Based on the scenario conversion coordination index, the optimal scenario conversion path is identified, candidate scripts in the script library are matched, and the best script sequence is derived. Recommendation timing is determined based on script type and user experience stage assessment, generating a recommendation priority queue. Based on the recommendation priority queue, tiered experience path suggestions are formulated. Experience expectation guidance is generated based on script type. A script recommendation sequence is constructed based on the tiered experience path suggestions and experience expectation guidance. User experience parameters are monitored in real time, and a preset experience deviation threshold is established. When abnormal user experience parameters are detected, a recommendation reconstruction mechanism is triggered, and adjusted script gameplay matching results are pushed out.

[0063] Specifically, based on the scenario conversion coordination index, the optimal scenario conversion path is identified, candidate scripts in the script library are matched, and the optimal script sequence is derived. The user's scenario conversion coordination index matrix is ​​analyzed to identify the scenario conversion path with the highest coordination, such as a progressive path from online text to online multimedia to offline real-world experiences, or an alternating path from lightweight offline experiences to in-depth online experiences to immersive offline experiences. Based on the optimal conversion path, candidate scripts with high matching degrees are selected from the script library (containing over 10,000 online scripts and over 3,000 offline scripts). A sequence optimization algorithm is applied, considering the thematic coherence, difficulty progression, and experience diversity between scripts, to generate the optimal script sequence. The optimal script sequence typically contains 3 to 7 scripts, forming a complete experience path that meets the user's needs throughout the entire process from beginner to in-depth participation.

[0064] The timing of recommendations is determined based on script type and user experience stage assessment, generating a recommendation priority queue. The process involves analyzing the user's current experience stage, including four phases: initial exploration, familiarization, immersion, and mastery; evaluating the suitability of different script types for each stage and establishing a stage-type adaptation matrix; predicting the optimal recommendation timing based on the user's script consumption cycle and pace preferences; and generating the recommendation priority queue by comprehensively ranking scripts according to urgency and relevance. Scripts in the recommendation priority queue are then ordered by recommendation priority to ensure that the most suitable content is pushed when the user most needs it and is most receptive to suggestions.

[0065] Based on a recommendation priority queue, a tiered experience path is proposed. An experience ladder model is designed, dividing the user's online-to-offline experience into multiple progressive stages. For each stage, the most suitable scenario type is selected from the recommendation priority queue. The smooth transition in difficulty gradient and experience intensity is considered. Combined with the user's time availability and geographical location, feasible experience path recommendations are formulated. This tiered experience path considers both the gradual development of user capabilities and the smooth transition between scenarios, avoiding discomfort caused by gaps in cross-scenario experiences.

[0066] Based on script type, experience expectation guidance is generated. Typical experience characteristics and participation patterns of different script types are analyzed; personalized experience expectation guidance is generated for each recommended script, including experience preparation suggestions, participation method tips, and key experience reminders; special attention is paid to online-offline scenario conversion points, providing adaptive suggestions and techniques; a comparison and feedback mechanism between experience expectations and actual experiences is designed to help users form accurate experience expectations. Experience expectation guidance effectively reduces the risk of user disappointment due to expectation discrepancies, especially during the online-offline scenario conversion process.

[0067] A script recommendation sequence is constructed based on tiered experience path suggestions and experience expectation guidance. The experience path and expectation guidance are integrated to form a complete recommendation solution; a recommendation display template is designed, including basic script information, matching reasons, experience expectations, and the optimal experience path; the recommendation presentation format is optimized for different platforms (mobile applications, web pages, offline spaces); a complete script recommendation sequence is built, including both current recommendations and potential future recommendation directions. This complete recommendation sequence provides users with a clear script experience path, enhancing long-term user engagement and platform stickiness.

[0068] The system monitors user experience parameters in real time and presets experience deviation thresholds. When abnormal user experience parameters are detected, a recommendation reconstruction mechanism is triggered, and adjusted script gameplay matching results are pushed to the user. A real-time experience monitoring system is deployed to collect user behavior data and feedback signals during the script experience. Preset experience deviation thresholds include a participation decline threshold (continuous interaction intervals exceeding twice the normal value), an emotional reaction abnormality threshold (emotional fluctuations exceeding the expected range by 30%), and a progress abnormality threshold (completion speed deviating from the expected range by 40%). When monitored parameters exceed preset thresholds, the recommendation reconstruction mechanism is triggered. The user preference model is updated based on real-time data, and the matching degree and recommendation sequence are recalculated. Adjusted script gameplay matching results are pushed to users through multiple channels (in-app push, SMS, email, etc.). The real-time monitoring and dynamic adjustment mechanism ensures the responsiveness and adaptability of the recommendation system, enabling timely responses to changes and problems in user experience.

[0069] This embodiment deconstructs user immersion experience data from multiple dimensions to obtain a user preference spectrum including cognitive preference index and social emotional tendency, comprehensively capturing the characteristics of user experience needs in different scenarios. It analyzes the emotional tension and interaction density of script content to form a cross-scenario experience map, overcoming the limitations of existing technologies that rely solely on simple label classification, which cannot accurately express the differences between online and offline experiences. Through parallel processing of cognitive and emotional channels, it achieves deep cross-fusion of the user preference spectrum and the cross-scenario experience map, generating a personalized matching matrix. This solves the problem of existing recommendation systems simply mapping online platform preferences directly to offline recommendations. Compared with traditional matching methods based on static rules, this invention establishes a multi-dimensional immersion evaluation parameter set, performs scenario adaptability calculations on the personalized matching matrix, and derives a scenario conversion coordination index, effectively quantifying the differences between online and offline script experiences and providing a scientific basis for cross-scenario matching. By using wavelet transform for multi-scale decomposition, subtle emotional changes in users can be captured, whereas traditional methods often overlook these micro-experience differences. In script analysis, the three-dimensional fusion of narrative complexity index, emotional tension spectrum and scene constraint matrix enables accurate expression of the experience effect in different online and offline carriers, solving the problem of experience fragmentation caused by differences in physical environment, social interaction mode and immersive perception dimension.

[0070] Example 2

[0071] Please see Figure 2 As shown, for parts not described in detail in this embodiment, please refer to the description in Embodiment 1. An AI algorithm-based system for recommending online and offline script gameplay matching is provided, including:

[0072] Data acquisition module: Collects user immersive experience data, deconstructs the immersive experience data from multiple dimensions, and obtains the user preference spectrum;

[0073] Script Analysis Module: Analyzes the emotional tension and interaction density of the script content to form a cross-scene experience map;

[0074] Data fusion module: Design a dual-channel perception network, and cross-fusion user preference spectrum and cross-scene experience map based on the dual-channel perception network to generate a personalized matching matrix;

[0075] Adaptation assessment module: Establish an immersion assessment parameter set, perform scene adaptability calculation on the personalized matching matrix, and obtain the scene conversion coordination index;

[0076] Recommendation generation module: Generates a script recommendation sequence based on the scene transformation coordination index, and realizes real-time recommendation adjustment and experience feedback collection through context awareness.

[0077] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0078] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0079] In the description of this invention, it should be understood that the terms "first," "second," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance.

[0080] In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0081] In the description of this invention, "several" means one or more, and "a large number" means two or more.

[0082] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0083] All formulas in this manual are dimensionless and calculated numerically. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0084] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for matching online and offline script gameplay based on AI algorithm recommendation, characterized in that, include: Step S1: Collect user immersion experience data, deconstruct the immersion experience data from multiple dimensions, and obtain the user preference spectrum; Step S2: Analyze the emotional tension and interaction density of the script content to form a cross-scene experience map; Step S3: Design a dual-channel perception network, and cross-fuse the user preference spectrum and cross-scene experience map based on the dual-channel perception network to generate a personalized matching matrix; Step S4: Establish an immersion evaluation parameter set, perform scene adaptability calculation on the personalized matching matrix, and obtain the scene conversion coordination index; Step S5: Generate a script recommendation sequence based on the scene transformation coordination index, and realize real-time recommendation adjustment and experience feedback collection through context awareness.

2. The method for matching online and offline script gameplay based on AI algorithm recommendation according to claim 1, characterized in that, Step S1 includes: This study collects user immersion experience data, including plot participation trajectory, character interaction behavior, emotional response signals, and decision-making tendency data. It performs time-series analysis on the plot participation trajectory to obtain narrative preference characteristics; conducts social network analysis on character interaction behavior to obtain interaction pattern characteristics; performs sentiment calculation on emotional response signals to obtain an emotional fluctuation spectrum; performs cognitive style analysis on decision-making tendency data to obtain thinking pattern characteristics; and performs interactive coupling between narrative preference characteristics and thinking pattern characteristics to obtain a cognitive preference index. Based on the interaction pattern characteristics and the emotional fluctuation spectrum, it performs adaptive correlation processing to obtain social emotional tendency. Finally, it conducts multi-dimensional cross-validation based on the cognitive preference index and social emotional tendency to construct a user preference spectrum.

3. The method for matching online and offline script gameplay based on AI algorithm recommendation according to claim 2, characterized in that, The process of performing emotion calculations on the emotion response signals to obtain the emotion fluctuation spectrum includes: A suitable wavelet transform function for emotion recognition is selected, and the emotion response signal is decomposed into multiple scales to obtain emotion sub-signals in different frequency bands. The intensity of the emotion sub-signals in different frequency bands is quantized to obtain an emotion intensity distribution map. Based on the emotion intensity distribution map, emotion type clustering is performed to select key frequency bands containing core emotions. The key frequency band signals are reconstructed to obtain emotion feature signals. The fluctuation frequency, duration, and peak intensity of the emotion feature signals are calculated to form an emotion fluctuation spectrum.

4. The method for matching online and offline script gameplay based on AI algorithm recommendation according to claim 1, characterized in that, Step S2 includes: Natural language processing is performed on the script content to extract key plot points and emotional markers; based on the plot points, a narrative structure diagram of the script is constructed; complexity analysis is performed on the narrative structure diagram to obtain a narrative complexity index; emotional semantic analysis is performed on the script text to obtain an emotional fluctuation curve; based on the emotional fluctuation curve, an emotional tension value is calculated to construct an emotional tension spectrum; scene elements are extracted from both online and offline scripts to establish a physical interaction mapping table; based on the physical interaction mapping table, interaction density and spatial constraints are analyzed to construct a scene constraint matrix; and the narrative complexity index, emotional tension spectrum, and scene constraint matrix are fused in three dimensions to form a cross-scene experience map.

5. The method for matching online and offline script gameplay based on AI algorithm recommendation according to claim 4, characterized in that, The process of performing emotional semantic analysis on the script text to obtain an emotional fluctuation curve includes: Deep semantic parsing is used to perform sentiment tagging on the script text to obtain an initial sentiment sequence; the initial sentiment sequence is then subjected to temporal smoothing to obtain a sentiment flow graph; peak and trough identification is performed on the sentiment flow graph to extract key sentiment turning points; and sentiment fluctuation curves are constructed based on the key sentiment turning points.

6. The method for matching online and offline script gameplay based on AI algorithm recommendation according to claim 1, characterized in that, Step S3 includes: A cognitive channel network and an emotional channel network are constructed. These two networks are then combined to design a dual-channel perception network. Parameters of the dual-channel perception network are adjusted using historical matching cases to obtain an optimized network. An adaptive weight allocation method is used to extract features from the user preference spectrum and cross-scenario experience graph. These extracted features are then input into the optimized dual-channel perception network to obtain scenario suitability prediction results. Based on the scenario suitability prediction results, a preference-experience mapping relationship is constructed. This preference-experience mapping relationship is then collaboratively calculated with preset cross-scenario conversion rules to obtain a personalized matching matrix.

7. The method for matching online and offline script gameplay based on AI algorithm recommendation according to claim 1, characterized in that, Step S4 includes: A multi-dimensional immersion assessment parameter set is established, including cognitive engagement index, emotional resonance index, social interaction index, and physical participation index. Based on the multi-dimensional immersion assessment parameter set, the personalized matching matrix is ​​transformed to obtain scene adaptability parameters. The various scene adaptability parameters are balanced to obtain standardized adaptability indicators. The standardized adaptability indicators are fused using multiple thresholds to obtain a comprehensive scene coordination degree. Based on the comprehensive scene coordination degree, scene conversion levels are classified. The stability of the standardized adaptability indicators is assessed to obtain a scene conversion risk index. Based on the scene conversion level and the scene conversion risk index, a scene conversion coordination degree index is generated.

8. The method for matching online and offline script gameplay based on AI algorithm recommendation according to claim 7, characterized in that, The process of fusing standardized adaptability indicators across multiple thresholds to obtain a comprehensive scenario coordination degree includes: A multi-level fuzzy inference function is constructed to map each standardized adaptability index to a scene fit. The consistency coefficient of each scene fit is calculated, and the weights of each standardized adaptability index are adjusted based on the consistency coefficient to obtain the adjusted scene fit. The Delphi aggregation method is used to fuse the adjusted scene fit to obtain the comprehensive scene coordination degree.

9. The method for matching online and offline script gameplay based on AI algorithm recommendation according to claim 1, characterized in that, The step of generating a script recommendation sequence based on the scene transformation coordination index includes: Based on the scenario conversion coordination index, the optimal scenario conversion path is identified, candidate scripts in the script library are matched, and the best script sequence is derived. The timing of recommendation is determined according to the script type and user experience stage assessment, and a recommendation priority queue is generated. Based on the recommendation priority queue, a tiered experience path suggestion is formulated. Based on the script type, experience expectation guidance is generated. The script recommendation sequence is constructed based on the tiered experience path suggestion and experience expectation guidance.

10. The method for matching online and offline script gameplay based on AI algorithm recommendation according to claim 1, characterized in that, The method of achieving real-time recommendation adjustment and experience feedback collection through context awareness includes: Real-time monitoring of user experience parameters, preset thresholds for deviation from user experience, and triggering a recommendation reconstruction mechanism when abnormal user experience parameters are detected, pushing the adjusted script gameplay matching results.

Images