Method and system for searching by combining user input information

Abstract

The invention discloses a method and a system for searching by combining user input information, and particularly relates to the technical field of language paragraption.The method comprises the following steps: firstly, converting a plurality of user inputs into an input semantic structure containing keywords, semantic relationships and context tags; comparing the structures, extracting interactive focuses and generating a semantic consensus score; constructing a fusion path diagram according to the consensus score, the keyword weight and the semantic adaptation degree; semantic fusion and conflict mediation are executed according to the path diagram, and a unified expression structure is generated; and finally, sorting the keywords according to semantic levels and task orientation, and constructing a comprehensive search expression to finish query output. According to the method, through semantic analysis and cross comparison input by multiple users, the consensus content is extracted, the fusion path diagram is constructed, and it is ensured that the fusion sequence is reasonable and semantic consistency is achieved; and finally, a comprehensive semantic expression structure is generated and a search expression is constructed, so that accurate alignment of query contents and user intentions is realized, and representativeness and accuracy of search results are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of language analysis, and more particularly, to a method and system for searching by incorporating user input information. Background Art

[0002] With the rapid development of artificial intelligence, natural language processing, and human-computer interaction technologies, various intelligent systems are increasingly integrating natural language input from multiple users. This is particularly evident in collaborative office systems, group decision-making platforms, intelligent search assistants, and multi-user service terminals. In these practical application scenarios, systems often receive language input from multiple users around the same task objective or query requirement. These inputs vary significantly in terms of surface semantics, pragmatic context, vocabulary selection, and structural expression, exhibiting significant heterogeneity, contextual discreteness, and intent diversity.

[0003] Under current technological conditions, mainstream search systems or semantic understanding frameworks typically perform semantic parsing and query generation based on single-user input models. These processes often focus on methods such as keyword extraction, word frequency weighting, or phrase matching. While recent research has attempted to enhance semantic models by incorporating contextual embeddings or user behavior features, these approaches remain insufficient for processing multiple user inputs simultaneously.

[0004] Specifically, existing technologies generally have the following technical bottlenecks:

[0005] When faced with multiple user inputs, the system often treats them as equivalent or directly concatenates them, failing to build a unified semantic structure to extract consensus. This leads to redundant or conflicting expressions, impacting subsequent retrieval efficiency. Different user inputs carry different contextual tags, such as input time, operation status, and usage path. Traditional parsing methods fail to incorporate these into a unified reference system for comparison and weighing, resulting in a fragmented semantic space. User inputs contain similar but inconsistent keywords or phrases (e.g., "application template" vs. "download contract"), making it difficult for traditional methods to identify and address potential semantic discrepancies. Even if multiple user inputs are semantically consistent, their linguistic expressions can still differ significantly, making existing technologies unable to automatically generate a set of expression structures that both covers the intent of multiple parties and is suitable for structured queries. System-generated query expressions often fail to consider the current task context or the system's preset goals, becoming disconnected from the real business environment and causing search results to deviate from user expectations. Therefore, a method and system for searching by merging user input information is proposed to address the above issues. Summary of the Invention

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for searching by combining user input information comprises the following steps:

[0008] Receive raw input information from multiple users and convert each input into an input semantic structure through semantic parsing, which includes keywords, semantic relationships and context labels;

[0009] Cross-comparison of multiple input semantic structures to identify repeated or semantically similar keywords and phrases, extract interactive focus content, and generate a semantic consensus score based on interactive focus coverage.

[0010] Based on the semantic consensus score, the keyword weights formed in each user's historical usage records and the semantic adaptability in the current scenario are combined to construct an input fusion path graph. This path graph is used to determine the fusion order and fusion method between input semantic structures.

[0011] Fusion actions are performed sequentially along the input fusion path diagram. For each pair of input semantic structures, semantic structure comparison, context consistency comparison, and semantic conflict mediation are performed. When the semantic conflict reaches an irreconcilable level, alternative words are used to unify the expression and generate a comprehensive semantic expression structure.

[0012] Sort the keywords in the comprehensive semantic expression structure according to semantic level and task-oriented importance, construct a comprehensive search expression, execute the query operation and output the search results.

[0013] In a preferred embodiment, when each input is converted into the content of the input semantic structure through semantic parsing, the context tags in the input semantic structure include the following four categories: the timestamp of the input information being received, the current activity state of the user's operating system, a sequence of keywords extracted from the user's three consecutive operation paths before receiving the input information, and the task function type tag in the user's current interface; the above four categories of context tags together constitute the user's current context tag vector in the context of the current task;

[0014] The above four types of context labels together constitute the data set based on the current task context, which is used to constrain the parsing boundaries of keywords and semantic relationships, so that the input semantic structure has a unified background reference dimension during the cross-comparison process; the timestamp corresponds to the task execution cycle node, the operation status corresponds to the functional stage, the operation path represents the behavioral causal chain, and the functional type tag defines the main direction of the current task intention.

[0015] In a preferred embodiment, when generating a semantic consensus score, the semantic consensus score is based on an undirected keyword graph constructed from the interactive focus content, where each node in the graph is a standardized keyword and the edge represents the frequency of the keyword co-occurring in different input semantic structures;

[0016] Assume that the node weight is variable X and the edge frequency is variable Y, then the semantic consensus score S is defined as: S = f(X) × P(Y), where f(X) is the normalized graph embedding vector modulus, and P(Y) is the co-occurrence edge weight defined based on Markov transition probability;

[0017] Specifically, P(Y)=∑p(i→j), where p(i→j) is the conditional probability of continuous co-occurrence between any two keyword nodes in all semantic structures. The consensus score results of all nodes in the graph are used to input the priority path selection in the construction of the fusion path graph.

[0018] In a preferred embodiment, the construction of the input fusion path diagram is based on the set of all input semantic structures that participate in forming the semantic consensus score S. The user input corresponding to each structure in the set is subjected to a fusion evaluation. The fusion evaluation result F is jointly determined by the semantic consensus score S, the keyword weight W of the user input, and the semantic fitness M, where F = s × sigmoid (W + M), s represents all keywords contained in the semantic structure of the user input, and when there is a matching node in the interactive focus graph, the semantic consensus score obtained is averaged; the keyword weight W represents the average value of the product of the frequency of use and click consistency of all keywords input by the user in the historical usage record, and the semantic fitness M represents the average value of the cosine value of the semantic vector angle between all keywords input by the user and the current task context label;

[0019] If the fusion evaluation result F is lower than the set retention threshold T, the corresponding user input information is eliminated from the fusion path construction, and the remaining user input information is sorted from high to low according to the F value and arranged in sequence as a node sequence in the input fusion path graph.

[0020] In a preferred embodiment, the fusion action is only performed on user input information that has passed the fusion evaluation function F and has a score not lower than the retention threshold T, that is, the input semantic structure set corresponding to the sorted valid node sequence in the input fusion path graph. In the process of performing the semantic structure comparison, a keyword structure mapping matrix is ​​constructed for each pair of input semantic structures. The rows and columns of the matrix respectively represent the top k keywords in the two input semantic structures. The matrix elements are the structural-level semantic similarity scores between the pair of keywords. The score is calculated based on the product of the semantic dependency path length and the cosine value of the context vector angle. The cosine value of the context vector angle refers to the cosine value of the angle between the context semantic vectors of the two keywords. The context semantic vector is the semantic expression vector of the keyword in the semantic subspace defined by the current context label vector. When the similarity score exceeds the preset matching threshold, the structure merge is performed, otherwise it enters the semantic conflict mediation stage.

[0021] In a preferred embodiment, semantic conflict mediation adopts an alternative expression processing mechanism. For any two keyword pairs that are in the same semantic position in the input semantic structure but whose semantic vector distance exceeds a set distance threshold, a semantic center vector of the keyword pair is constructed, and the term with the smallest distance to the center vector is selected from the semantic neutralization word set as a replacement word for unified expression processing of the keyword pair. The processing process is only executed on the conflicting keyword pairs between the two input semantic structures of the current fusion operation, and the other inputs retain the original structure. The semantic neutralization word set forms a low-density area in the semantic space after cluster analysis, ensuring that the replacement process does not introduce semantic bias.

[0022] In a preferred embodiment, when generating a comprehensive semantic expression structure, the semantic level score and task-oriented importance based on which keywords are sorted are calculated from the following sources: the semantic level score U is the product of the keyword weight W of each keyword in the input fusion path diagram in the calculation of the semantic consensus score S and its number of occurrences in the fusion process;

[0023] The task-oriented importance V is the cosine value of the semantic angle between each keyword and the task-level context label vector; the task-level context label vector is the reference standard extracted from the system task metadata;

[0024] The final ranking weight Q of the keywords is obtained by exponential normalization of the sum of U and V, that is, Q = softmax(U + V), and the softmax function is W(i) = exp(U(i) + V(i)) ÷ ∑exp(U(j) + V(j)), where i and j traverse all keywords involved in the ranking. The ranking weight W determines the output order of the keywords in the comprehensive semantic expression structure, ensuring that the ranking takes into account both the degree of consensus and task orientation.

[0025] In a preferred embodiment, the construction of the comprehensive search expression adopts a segmented splicing generation method, which includes three parts: the main semantic segment, the conflict mediation segment and the context supplement segment;

[0026] The main semantic segment includes the top R keywords ranked by ranking weight Q, representing the main intention content in the fused semantic structure; the conflict mediation segment includes keywords that have participated in the semantic conflict mediation process and have been uniformly processed by alternative expressions, which is used to supplement the description of the semantic consistency of the final expression; the context supplement segment includes a set of keywords with the shortest distance to the task-level context label vector in the semantic space, which is used to bridge the gap between the expression structure and the task environment;

[0027] A comprehensive search expression is formed by splicing the main semantic segment, the context supplement segment, and the conflict mediation segment in order.

[0028] In a preferred embodiment, a system for searching by incorporating user input information includes:

[0029] Semantic parsing module, which is used to receive original input information from multiple users and convert each input into an input semantic structure through semantic parsing, which contains keywords, semantic relations and context labels;

[0030] The focus extraction module is used to cross-compare multiple input semantic structures, identify repeated or semantically similar keywords and phrases, extract interactive focus content, and generate a semantic consensus score based on the interactive focus coverage;

[0031] The path construction module is used to construct an input fusion path diagram based on the semantic consensus score, combining the keyword weight formed in each user's historical usage records and the semantic adaptability in the current scenario, to determine the fusion order and fusion method between input semantic structures;

[0032] The structural fusion module is used to perform fusion actions in sequence along the input fusion path diagram, performing semantic structure comparison, context consistency comparison, and semantic conflict mediation for each pair of input semantic structures. When the semantic conflict reaches an irreconcilable level, it uses alternative words to unify the expression and generate a comprehensive semantic expression structure.

[0033] The expression generation module is used to sort the keywords in the comprehensive semantic expression structure according to the semantic level and task-oriented importance, construct a comprehensive search expression, perform query operations and output search results.

[0034] Technical effects and advantages of the present invention:

[0035] The present invention can achieve efficient fusion of multi-user input information, thereby improving the representativeness and comprehensiveness of query results. By receiving the original input information of multiple users and converting each input into an input semantic structure of a unified structure through semantic parsing, the structure contains keywords, semantic relationships and context labels, so that the input content of different users can be uniformly modeled at the structural level. Subsequently, the system cross-compares these structured semantic inputs, identifies repeated or semantically similar keywords and phrases, and extracts interactive focus content as the basis for consensus, thereby generating a semantic consensus score. This mechanism ensures that in the fusion process, not only the subjectivity of a single user's input is considered, but also based on the semantic overlap between multiple users to form more collectively representative expression content, so that the final query operation is closer to the real needs of most users.

[0036] The present invention provides a scoring-driven input fusion path determination method, which effectively improves the rationality of the fusion order and the consistency of the fusion results. This method constructs an input fusion path diagram for guiding the fusion order and fusion method based on the generated semantic consensus score, combined with the keyword weights formed by each user in the historical usage records and the semantic adaptability in the current scene. The path diagram provides a global sorting and screening basis in the fusion process, so that when performing the fusion action, it can operate in sequence along the preset path, reducing the interference of low-relevance or semantically deviated inputs. In addition, through the guidance of the path diagram, each pair of input semantic structures will complete the comparison, mediation and merging steps according to consistent logic during the fusion process, thereby enhancing the coordination of the fusion results in terms of context understanding and semantic unification.

[0037] The present invention realizes the precise semantic alignment between the query content and the user input by constructing a comprehensive semantic expression structure and generating a comprehensive search expression. After completing the fusion of the input semantic structure, the system will generate a comprehensive semantic expression structure, which is the product of multiple rounds of fusion and conflict mediation, and has the characteristics of semantic integrity and context consistency. On this basis, the system sorts the keywords in the expression structure according to the semantic level and task-oriented importance, and constructs a comprehensive search expression. As the final query instruction, this expression can accurately reflect the common semantics between multiple user inputs and the most representative query intentions, so that the search engine can more effectively capture the key points of information when executing the query operation, thereby outputting more accurate and comprehensive search results. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings;

[0039] Figure 1 This is a schematic diagram of a method for searching by combining user input information in the present invention.

[0040] Figure 2 This is a schematic diagram of a system for searching by combining user input information in the present invention. DETAILED DESCRIPTION

[0041] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0042] Reference Figure 1 - Figure 2 The following examples were obtained:

[0043] Example 1:

[0044] A method for searching by combining user input information comprises the following steps:

[0045] Receive raw input information from multiple users and convert each input into an input semantic structure through semantic parsing. This structure contains keywords, semantic relationships, and context labels. In natural language expression, each user's input has differences in word usage habits, expression order, content granularity, etc. Processing each input through semantic parsing can effectively abstract the original unstructured information into an input semantic structure with a standard expression format. This structure contains three core components: keywords are used to represent the information topic; semantic relationships describe the logical or causal relationship between keywords; and context labels are used to characterize the environmental state when the input occurs, such as task background, operation interface, time nodes, and other information. This structure is not only the basic data unit for subsequent processing, but also ensures that different user inputs are comparable and fusion-capable in the same semantic space.

[0046] Cross-comparison is performed on multiple input semantic structures to identify keywords and phrases that appear repeatedly or have similar semantics, extract interactive focus content, and generate a semantic consensus score based on the interactive focus coverage rate. This step aims to discover the semantic intersection between different user inputs and quantify the degree of consensus of this intersection. By cross-comparing keywords and phrases in multiple input semantic structures, the system can identify content that appears repeatedly or has similar semantics in multiple user expressions, thereby extracting a content set that represents the current interactive focus of multiple user inputs. The interactive focus content reflects the core concerns or expression trends of the user group to a certain extent. The system further generates a semantic consensus score based on the coverage rate of the interactive focus content in all semantic structures to quantify the strength of this consensus at the semantic level. This score serves as the basis for fusion judgment and sorting in subsequent steps and plays a guiding role.

[0047] Based on the semantic consensus score, the weight of each user's keyword in historical usage history and its semantic fit in the current scenario are combined to construct an input fusion path graph. This path graph is used to determine the order and method of fusion between input semantic structures. By jointly evaluating the strength of group semantic consensus and the adaptability of individual inputs to the current task, a fusion scoring mechanism is established for ranking and screening. The semantic consensus score generated in the previous step establishes a node embedding model for each keyword based on an undirected keyword graph. The collective semantic connectivity between keywords is defined by combining node weights and co-occurrence edge frequencies. This score reflects the degree of consensus between the input semantic structures. Building on this, the current step incorporates the weight of each user's keyword in historical usage history to quantify the stability and usage tendency of that user's keyword. Semantic fit is also introduced to represent the degree of semantic fit between the keyword and the current task context label, capturing the applicability of the input in the current scenario. These three factors are combined into a joint evaluation function that serves as the basis for determining whether to participate in the subsequent fusion process and its ranking weight in the fusion path. The fusion path graph consists of all input semantic structures whose evaluation scores exceed the preset retention threshold, and forms a sequence from high to low scores, which serves as a node chain for the next fusion action.

[0048] Fusion actions are performed sequentially along the input fusion path graph. For each pair of input semantic structures, semantic structure comparison, context consistency comparison, and semantic conflict mediation are performed. When semantic conflicts reach irreconcilable levels, substitutions are used to unify the expressions and generate a comprehensive semantic expression structure. After the fusion path graph is finalized, this step begins pairwise fusion between the semantic structures. This process not only merges the content but also prioritizes semantic consistency and contextual coordination. First, semantic structure comparison is performed to identify overlaps or conflicts between the two input structures in terms of keywords and semantic relationships. Second, their contextual labels are evaluated for consistency differences to avoid context mismatches. If semantic conflicts are detected, the system attempts mediation. If mediation fails, meaning the semantic differences exceed the system's set tolerance threshold, substitutions are used to select semantically neutral words to unify the keywords. Finally, a comprehensive semantic expression structure is generated. This unified expression, resulting from the integration of all valid input semantic information, possesses semantic integrity and consistency and serves as the sole source for subsequent search expression construction.

[0049] The keywords in the comprehensive semantic expression structure are sorted according to semantic level and task-oriented importance, and a comprehensive search expression is constructed. The query operation is executed and the search results are output. In order to convert the fused semantic expression content into a search-executable form, two dimensions of keyword sorting are formulated: semantic level and task-oriented importance. The former reflects the node weight and frequency of occurrence of the keyword in the consensus graph, and the latter measures the degree of fit between the keyword and the current task intent. The sorted keywords are used to construct a comprehensive search expression, which includes three parts: the main semantic segment, the conflict mediation segment, and the context supplement segment. They are combined in a fixed splicing order into a standard expression format that can be directly used in the query system. Finally, the system executes the search operation and outputs the retrieval results that are most relevant to the multi-user semantic fusion results, achieving the semantic unified retrieval goal of group intent.

[0050] In the method described in the present invention, when converting each input into an input semantic structure through semantic parsing, it is necessary to extract information from the input context and perform semantic modeling. To this end, the context labels in the input semantic structure include the following four categories:

[0051] The first category is the "timestamp of input receipt," which refers to the specific time when the user input was received by the system. This information is typically expressed as a standard date and time stamp, such as "2025-05-30 14:26:35." This timestamp not only identifies the time the input was received by the system but also determines the stage of the task flow in which the input falls, thus corresponding one-to-one with each node in the task execution cycle. For example, in a process-based task, the input at the initial stage may focus on completely different points than the input at the final stage. The system can use timestamps to identify the behavioral timing characteristics of the input.

[0052] The second category is the "current activity state of the user's operating system," which refers to the system's operational state or interface status at the time the user performs input, such as editing, browsing, searching, submitting, etc. This state information is used to map the functional stages of the task, such as "data entry," "parameter configuration," and "result confirmation," helping the system infer the operational context of the input intent. The operating system activity state is output by the system state management module and is typically represented by a function label, interface identifier, or operation code.

[0053] The third category is "keyword sequences extracted from the user's three-step operation path before receiving input." This refers to the keywords or operation objects involved in the three consecutive operations performed by the system before the user generates the current input. For example, if a user clicks "Find Customer," "View Contract," and "Add Record" in succession, the system will extract the keywords "customer," "contract," and "add" from these three operations, forming a keyword sequence that describes the causal chain of user behavior. This sequence is action-forward and provides semantic context for the current input.

[0054] The fourth category, "Task Function Type Tags in the User's Current Interface," refers to the labels for the function types that are activated or focused on the system interface during user input, such as "Expense Approval," "Report Generation," and "Parameter Configuration." This tag semantically defines the primary task of the current input intent—that is, the system task goal the input information serves. This tag originates from the system task scheduling module or the interface interaction management module.

[0055] The four types of context labels above collectively constitute the user's current context label vector in the context of the current task. This vector is a high-dimensional semantic vector that represents the semantic environment of the user input. Each label dimension is converted to a vector using a pre-trained semantic embedding model (such as BERT and Word2Vec) and then merged to form a unified input for semantic calculation and comparison.

[0056] Furthermore, these four types of context labels collectively constitute the data set underlying the current task context, constraining the boundaries of keyword and semantic relationship identification during subsequent semantic parsing. Specifically, parsing operations such as keyword extraction, relationship determination, and semantic path generation operate exclusively within the semantic subspace defined by the current context label vector. This ensures that the input semantic structures generated by different user inputs share a unified background reference dimension, preventing divergent parsing structures due to contextual differences.

[0057] For example, if user A, on an e-commerce platform, clicks "Add Product," "Set Price," and "Set Inventory" on the "Product Management" interface at 10:00 AM on May 30th, then enters the "Parameter Configuration" interface and enters the text "Set as Discount Season Default." The contextual tags for this user input are as follows: Timestamp: 2025-05-30 10:00:23; Current Activity Status: Function Module marked "Parameter Configuration in Progress"; Operation Path Keyword Sequence: [Add, Price, Inventory]; Current Function Type: Discount Settings Module. The resulting contextual tag vector focuses on keywords like "product operation" and "discount configuration," enabling the semantic parsing module to more accurately understand that the word "default" refers to the discount template, not other parameter items.

[0058] It should be noted that the current context label vector does not directly participate in the construction of semantic structure by extracting keywords from it, but rather serves as a semantic parsing boundary control factor to limit the scope of the semantic space in which the parsing behavior is performed, playing the role of "limiting the semantic space", "filtering candidates", and "ensuring consistency". The specific functions are as follows:

[0059] Keyword extraction stage: The current context label vector defines the semantic space for parsing candidate words. The system only performs keyword candidate screening among words whose semantic similarity with the vector exceeds a preset value to exclude redundant semantic units that are irrelevant to the current task intent.

[0060] Relationship determination stage: The current context label vector is used to limit the set of candidate semantic relationships. For example, in "approval" tasks, typical semantic relationship templates such as "authorization-rejection" and "initiate-process" will be given priority, thereby preventing the relationship determination process from entering a semantic paradigm irrelevant to the current task.

[0061] During the semantic path generation phase, the search algorithm for semantic path construction (such as those based on graph traversal or dependency tree expansion) only allows semantic jumps between path nodes within the semantic subspace defined by the context labels, ensuring path continuity and context consistency. In summary, the current context label vector is not equivalent to "a set of keywords," but rather a high-dimensional semantic embedding vector. Its role is to limit the spatial dimension and constraint scope of semantic parsing, ensuring that the semantic structures of multiple user inputs maintain a unified parsing standard during the generation phase, fundamentally guaranteeing the feasibility and consistency of cross-comparison and fusion.

[0062] In the method described in the present invention, the semantic consensus score is used to measure the strength of the semantic commonality between different user inputs, and serves as the core basis in the process of constructing the input fusion path map to determine the fusion priority between each input semantic structure. The score is generated before the input fusion stage and is completed based on the undirected keyword graph constructed by the interactive focus content. Specifically, the system first extracts the interactive focus content based on the keywords or semantically similar phrases that appear crosswise in multiple input semantic structures. Interactive focus content refers to expression units that appear repeatedly or have highly similar semantics in two or more input semantic structures, such as an expression unit with a similarity of 0.8 or more, which has group consensus characteristics and represents the core information that users are concerned about in the current scenario.

[0063] Based on these interactive focus points, the system constructs an undirected keyword graph. Each node in this graph represents a standardized keyword. Standardization includes unifying synonyms, eliminating inflection points (such as plural / past tense), and merging domain terms. The edge between any two nodes in the graph represents the frequency of co-occurrence of the keyword pair in different input semantic structures, that is, the number of different user inputs in which the two words appear together.

[0064] On this basis, a node weight X is assigned to each node. The node weight X represents the importance of the keyword's position in the interactive focus graph, which can be obtained by normalizing the modulus after generating a vector using a graph embedding algorithm (such as DeepWa lk, Node2Vec). The normalized vector modulus of the node is represented by f(X), which indicates its structural stability. At the same time, the edge frequency Y represents the co-occurrence strength connecting the keyword with other keywords, that is, the frequency of the word and other words in the graph appearing together in multi-user input, which is used to construct the co-occurrence edge weight. The system uses a Markov transition probability-based approach to weighted modeling of edge frequency, which is defined as: P(Y) = ∑p(i→j), where p(i→j) is the conditional probability between any two keyword nodes i and j that appear in a fixed order in all semantic structures (for example, "first A then B"), and is calculated as follows:

[0065] p(i→j) = count(i followed by j) ÷ count(i appears); that is, the frequency of keyword i followed by keyword j in all user inputs, as a percentage of the total frequency of keyword i. If two words appear multiple times in consecutive positions in the semantic structure, it indicates a high degree of semantic dependency and a high probability of transition. After obtaining f(X) and P(Y), the semantic consensus score S is calculated, defined as: S = f(X) × P(Y). The consensus score S is determined by the product of the structural stability of a keyword node in the graph embedding space and its transition strength in the semantic co-occurrence network. A higher S value indicates a stronger consensus for that keyword across multiple user inputs. Ultimately, the system uses the consensus scores of all nodes in the graph to prioritize paths in the construction of the input fusion path graph. When constructing the path graph, the system considers whether the keywords contained in each input semantic structure match the high-scoring consensus keyword nodes and adjusts the order of nodes in the fusion path accordingly.

[0066] For example, consider the collaborative configuration of a product recommendation system. User A enters "set recommendation logic," User B enters "configure content push," and User C enters "update push rules." After parsing, the system extracts keywords such as "set," "configure," "recommend," "push," and "rules." It finds that "push" appears in all three users or has similar semantics (e.g., "content push," "recommendation push," and "push rules"). Based on this information, the system constructs an interaction focus graph. If "push" and "recommend" co-occur consecutively for users A and B, the system assigns a high edge frequency and calculates p(recommendation → push) to 0.75 and p(push → recommend) to 0.32. For the "push" node, the system obtains a vector modulus of 0.68 from the graph embedding model. The consensus score for the "push" keyword is: S = f(X) × P(Y) = 0.68 × (0.75 + 0.32) = 0.73 (approximately). This score is used as a basis for ranking and fusion decision-making in the subsequent input fusion path construction.

[0067] In the method described herein, the input fusion path graph is constructed based on the semantic consensus score generated in the previous stage. It further integrates the historical behavioral characteristics of each user input and the current task semantic context to determine and rank the fusion value of the user input information. The construction process begins with the set of all input semantic structures that contribute to the semantic consensus score S. This set includes all user input information that the system has received and semantically parsed. Each user input is converted into an input semantic structure through semantic parsing, consisting of three core elements: keywords, semantic relationships, and contextual tags. These structures are uniformly embedded in the same semantic vector space to ensure a consistent foundation for subsequent comparison and fusion operations. For each input semantic structure in this set, the system performs a fusion evaluation to quantify its contribution to the construction of the final comprehensive expression structure. The fusion evaluation result F is the quantitative output of this evaluation and is defined as follows: F = s × sigmoid(W + M); where the variable s refers to all keywords contained in the user input semantic structure, and the semantic consensus score corresponding to the subset of matching nodes in the interaction focus graph constructed in the previous stage is averaged. This value reflects the degree of fit between the user input and the semantic focus of group interaction and is global consensus-oriented.

[0068] Keyword weight W is defined as the average of the product of the frequency of use and click consistency across all keywords entered by the user in their historical usage history. Frequency represents how often a keyword appears in past user interactions, while click consistency indicates the degree of consistency between the behavior (e.g., click, confirm, browse) triggered by the keyword and the system's expected task path. The product of these two measures the "behavioral value" of the keyword.

[0069] Semantic fit, M, is defined as the average cosine of the angles between all keywords entered by the user and the current task context labels in the semantic vector space. This value indicates whether the input content fits the current task context. If the direction of a keyword's semantic vector is similar to that of the context label vector, its contribution is high.

[0070] In actual calculations, the system first extracts the vector expression of each keyword through a semantic parsing process, then extracts its corresponding historical records based on the user behavior database, and finally calculates the W and M values ​​respectively, and performs nonlinear normalization on them through the sigmoid function to ensure that the output score range is controlled and has discrimination. Once the fusion evaluation value F is calculated, the system compares the value with the retention threshold T preset by the system: if F is lower than the threshold T, it means that the input is not comprehensively evaluated in terms of consensus, historical validity and task adaptability, and the system will remove the input information from the input fusion path graph construction; if F is higher than or equal to the threshold T, the input is retained, and all retained inputs are sorted from high to low according to their F values, and arranged in order as a node sequence in the input fusion path graph. The node sequence in the path graph not only represents the order in which each user input structure participates in the fusion, but also determines the pairing logic and mediation process of the fusion action in the subsequent structural fusion stage. Through the above-mentioned fusion evaluation mechanism, the present invention effectively filters out redundant and scene-deviating user inputs, retains only high semantic quality and representative input data, and enhances the stability and accuracy of the comprehensive semantic expression structure.

[0071] Take an actual scenario as an example: User A inputs "set recommendation rules", User B inputs "optimize push configuration", and User C inputs "update marketing logic". The system performs semantic analysis on the three inputs, extracts keywords such as "set", "recommend", "push", "configuration", "marketing", etc., and forms their respective input semantic structures. The interactive focus graph shows that "recommend", "push", and "configuration" co-occur frequently in multiple inputs, so the corresponding keywords obtain a higher consensus score S in the graph. Assuming that in user B's input, "push" and "configuration" both match the interactive focus graph nodes, the obtained s is 0.85; at the same time, the user has a high historical frequency of use of the "push" keyword, and it is often used in the push function module, with a high W value of 0.73; the keywords in their input have a high semantic similarity with the current context label (such as "content operation interface"), with an M value of 0.68; then:

[0072] F = 0.85 × sigmoid(0.73 + 0.68) ≈ 0.85 × sigmoid(1.41) ≈ 0.85 × 0.80 ≈ 0.68. If the system sets the retention threshold T to 0.65, user B's input is retained and placed higher in the input fusion path graph. A similar evaluation method is applied to all user inputs, forming a fusion path graph with stable ranking, clear source, and semantic coherence.

[0073] In the method described in this invention, fusion actions are performed only on user input information that has passed the fusion evaluation function F and has a score no lower than the retention threshold T. As previously mentioned, the fusion evaluation function F reflects the comprehensive semantic value of an input semantic structure. Only when its F score exceeds the preset threshold T can it enter the subsequent fusion processing stage. All input semantic structures that meet this condition are sorted from high to low according to their F score, forming a valid node sequence in the input fusion path graph.

[0074] The set of input semantic structures corresponding to the valid node sequences in the input fusion path graph is the target set for subsequent structural fusion operations. Based on the order of the structures in the sequence, the system extracts adjacent pairs of structures and performs semantic structure comparisons on each pair to determine whether they have sufficient semantic structural similarity, thereby deciding whether to merge them or enter the conflict mediation process.

[0075] During the semantic structure comparison process, the system constructs a keyword structure mapping matrix for each pair of input semantic structures. This matrix is ​​a two-dimensional mapping table, with its rows and columns representing the top k (e.g., 10) keywords in the two input semantic structures. Each element in the matrix represents the structural-level semantic similarity score between the pair of keywords. This score quantifies the degree of semantic logical structure match between the keywords in the two different structures. Specifically, the structural-level semantic similarity score is calculated as follows: Similarity score = semantic dependency path length × context vector angle cosine value; semantic dependency path length refers to the length of the semantic dependency path between the keyword to be compared and the other keywords in each semantic structure. The shorter the path, the stronger the semantic core of the word in the structure. The cosine of the context vector angle refers to the cosine of the angle between the context semantic vectors of the two keywords (cosinesimilarity), which is used to measure the similarity of their semantic environments. A value closer to 1 indicates a more consistent semantic environment.

[0076] By multiplying these two metrics, the resulting score combines semantic status and contextual consistency, assessing whether a keyword pair exhibits structural equivalence across two semantic structures. For all scoring items in the matrix, if any score exceeds a preset matching threshold, the system deems the feasibility of structural fusion between the input semantic structures.

[0077] Once the above criteria are met, the system performs a structural merge operation, unifying the identical or similar keyword nodes in the two semantic structures and merging, simplifying, and uniformly encoding their associated semantic relationships and contextual tags. The purpose of the structural merge operation is to consolidate duplicate information in the semantic expression space and improve the simplicity and consistency of the overall structure.

[0078] If no keyword pair within a pair of input semantic structures has a score exceeding the matching threshold, this indicates significant differences in their semantic structures. In this case, the system will not force structural fusion, but instead direct the pair to the semantic conflict mediation stage. This conflict mediation mechanism will be implemented later in the process, such as using alternative expressions or generating semantically neutral words.

[0079] For example, let's take two adjacent structures ranked in the input fusion path graph. Suppose user A's input is "Optimize product recommendation logic" and user B's input is "Configure recommendation content parameters." After semantic parsing, the top k keywords for each are: Structure A: optimization, product, recommendation, logic; Structure B: configuration, recommendation, content, parameters. The system constructs a keyword-structure mapping matrix, with rows and columns corresponding to the top four keywords in the two structures. After pairwise comparison, the system finds that the "recommendation-recommendation" keyword pair has a contextual semantic vector cosine value of 0.94, a semantic dependency path of 1, and a product score of 0.94, exceeding the preset matching threshold of 0.80. Therefore, this keyword pair is considered structurally fusible, and its corresponding semantic relationships, such as "subject-predicate" and "object-verb," ​​are merged. The remaining keywords, such as "optimization-configuration" and "logic-parameter," lack sufficient similarity and do not meet the fusion criteria, requiring further mediation.

[0080] It should be noted that the context semantic vector is a context constraint vector representation derived by modeling with the current context label vector and combining it with the keyword semantic vector. It can be regarded as the semantic expression vector of the keyword in the semantic subspace defined by the current context label vector. The current context label vector is composed of four types of context labels: timestamp, operating system status, user operation path keyword sequence, and task function type label. It represents the semantic boundary of the current task background. This vector participates in multiple links such as semantic parsing and semantic fitness calculation. The context semantic vector is used to measure the specific semantic position of a keyword in the current semantic environment. It is essentially derived by jointly modeling the basic semantic vector of the keyword and the current context label vector, and represents the expression result of the keyword in the context space. The contextual semantic vector refers to the semantic embedding representation generated by combining the keyword in the semantic structure to which it belongs with the current context label vector. It is used to reflect the actual semantic projection result of the keyword in the current task scenario. The current task scenario refers to the semantic environment context directly related to the input behavior recorded and extracted by the system when receiving each user input information. Specifically, it includes characteristic factors such as the user's functional interface, task stage, behavior path and interaction time, which determine the goal or task intention to be achieved by the input.

[0081] In the method of the present invention, in order to ensure that the comprehensive semantic expression structure generated by different user input information during the fusion process has consistency and stability, a semantic conflict mediation mechanism is specially designed. This mechanism is applicable to the situation where the two input semantic structures cannot be directly merged due to differences in semantic expression during the structural fusion process. The mediation mechanism adopts an alternative expression processing mechanism, which aims to reconcile the different words in the two semantic structures by finding suitable neutral expressions in the semantic space, thereby achieving semantic unification and avoiding direct discarding or forced merging to cause expression deviation. The judgment criteria for semantic conflict are as follows: when the system performs the fusion action, it analyzes the keywords inside the two adjacent input semantic structures in the current input fusion path diagram one by one. If it is detected that the distance between the semantic vectors of two keywords in the same semantic position in the high-dimensional semantic space exceeds the set distance threshold, then the keyword pair is judged to have a semantic conflict.

[0082] Keyword pair: refers to a group of keywords that have a corresponding relationship in terms of grammatical position, semantic role, etc. in two input semantic structures, such as "customer" and "user" in the subject position; Same semantic position: refers to two keywords that have similar dependency grammatical roles or semantic labels in their respective structures, such as subject to subject, verb to verb, etc.; Semantic vector distance: defined by the Euclidean distance or cosine angle between vectors generated by embedding models (such as Word2Vec, BERT, etc.); Set distance threshold: the maximum acceptable semantic deviation value predefined by the system, used to determine whether to trigger the mediation mechanism. Once the semantic conflict between keyword pairs is identified, the system will perform the following mediation steps:

[0083] Constructing a semantic center vector: The semantic vectors of the conflicting keyword pairs are averaged or weighted (e.g., weighted by the semantic level scores of the keywords) to obtain a semantic center vector. This vector lies between the two words and represents a potential neutral semantic direction.

[0084] Searching for candidate items within a semantically neutral word set: The system pre-generates a semantically neutral word set through cluster analysis in the semantic space. This set consists of words with relatively ambiguous semantic expressions, low semantic density, and broad applicability across multiple tasks. This set often includes general terms such as "object," "participant," and "target item."

[0085] Select the term with the minimum distance as the replacement word: From the semantic neutral word set, the system calculates the distance between the semantic vector of each term and the semantic center vector in step 1, and selects the term with the minimum distance as the replacement word for the keyword pair, thereby achieving semantically unified expression without obviously favoring any original expression.

[0086] Replace conflicting keywords in the original structure: The system only replaces the keyword pair in the two input semantic structures involved in the current fusion operation. Other input semantic structures not involved in the current fusion operation will retain the original structure unchanged, ensuring that the fusion behavior is localized and does not spread to non-conflicting areas.

[0087] Continue to perform structural fusion: After completing the unification of alternative words, rebuild the keyword mapping matrix of the structure pair. If there is no conflict at this time, complete the structural merger, otherwise continue to the next round of mediation.

[0088] In the proposed search method for incorporating user input, keyword ranking plays a crucial role in generating a comprehensive semantic representation structure. To ensure a balanced ranking between expressive consistency and task orientation, the system weights keywords based on two core scoring metrics: semantic rank score (U) and task-orientation importance (V).

[0089] Composition and calculation of semantic level score U: The semantic level score U of a keyword reflects the degree of semantic consensus reflected in the user group input, which is derived from its overall contribution to the semantic consensus score calculation. Specifically: U = accumulated keyword weight W × number of occurrences; where: the accumulated keyword weight W refers to the direct summation of the keyword weight W values ​​assigned to each input when a specific keyword appears in all user inputs participating in the construction of the semantic consensus score S; the keyword weight W itself refers to the average product of the frequency of use and click consistency of all keywords in the user's input in the current use record; since multiple user inputs may contain the same keyword at the same time, the keyword will receive multiple W values, which are accumulated as the first item in the semantic level score U; the number of occurrences refers to the cumulative frequency of the keyword's appearance when it is structurally matched, extracted, involved in conflict mediation, or structurally sorted during the fusion process. For example, the keyword "configuration" is identified as a high-weight keyword in the input of 5 users, and the corresponding keyword weight W values ​​are 0.81, 0.77, 0.92, 0.68, and 0.88, respectively. The cumulative value of W for the keyword "configuration" is: W_total = 0.81 + 0.77 + 0.92 + 0.68 + 0.88 = 4.06; if the keyword appears 7 times in total in the structural fusion stage, its semantic level score is: U = 4.06 × 7 = 28.42; this score will be used as the first dimension score in the subsequent sorting.

[0090] Composition and calculation of task-oriented importance V: The task-oriented importance V of a keyword is used to measure the degree of fit between its semantic orientation and the current task goal of the system. The calculation method is as follows: V = the cosine value of the semantic angle between the keyword semantic vector and the task-level context label vector, where: the task-level context label vector is a semantic embedding vector generated by a set of label contents (such as task goals, functional module categories, system business nodes, etc.) extracted from the system task metadata; the angle calculated with it is the embedding vector of each keyword to be sorted in the semantic space; the closer the cosine similarity value is to 1, the more the keyword can reflect the current task goal and the higher the sorting priority.

[0091] It should be noted that the two-layer structure of context tags: In order to take into account both the user input context and the system task background, the present invention introduces a two-level structure of context tags:

[0092] User-level context labels: Labels attached to each user input; including input timestamp, operation path keyword sequence, current system state, and task function type tag; used to constrain the boundaries and scope of the user input during the semantic parsing phase;

[0093] Task-level context tags: extracted from the system's global task metadata; used to unify the sorting direction of each keyword in the comprehensive semantic expression structure; participate in the calculation of the angle between the keyword and the system task semantics alignment.

[0094] Final keyword ranking weight Q: The system sums the two aforementioned scoring dimensions and processes them through the exponential normalization function softmax to obtain the final ranking weight Q: Q = softmax(U + V). The calculation traverses all keywords participating in the ranking. The softmax function ensures that the keyword ranking weights are normalized, so that each keyword has a clear relative ranking in the output sequence of the comprehensive semantic expression structure. This ranking mechanism not only preserves the consistent recognition of keywords in user input (U), but also strengthens their alignment with the system's task objectives (V). This allows the generated search expression to more accurately guide downstream query behavior, avoiding semantic dispersion and task deviation.

[0095] In the proposed method for searching by integrating user input, a comprehensive search expression is constructed using a segmented concatenation approach, ensuring the integrity, consistency, and task-orientation of the fused semantic information. This expression consists of three parts: a primary semantic segment, a conflict mediation segment, and a contextual supplement segment. These segments are concatenated in this order to create an expression with precise query capabilities.

[0096] The main semantic segment represents the most consensus-based and task-specific set of keywords within the fused input semantic structure and is the core component of the comprehensive search expression. Source: Extract the top R keywords ranked by weight Q from the fused semantic structure. Keywords in this segment generally represent the primary entity, action, or constraint of the search intent. For example, if the user input focuses on "contract template application," "authorization signature," and "auto-submit," the top R keywords might be "authorization," "template," and "auto-submit."

[0097] The conflict mediation segment is used to provide supplementary explanations for semantic conflicts that arose during the initial fusion process but have been uniformly addressed, improving semantic consistency and system interpretability. Source: All keywords involved in the semantic conflict mediation process. Keyword screening criteria: Term terms identified as conflicting keyword pairs during the fusion operation and uniformly expressed through the substitution mechanism. The substitution process involves constructing a semantic center vector for the conflicting keyword pair and selecting the term with the smallest distance to the center vector from the semantically neutralized word set as the final replacement expression. This segment is not used for core search control but contributes to the semantic integrity of the search expression. Example: If multiple users use "contract template download" and "template application form" respectively, and "application form" and "download" are replaced with the neutralizing word "acquire" during structural fusion, "acquire" will enter this segment.

[0098] The contextual supplement segment is used to enhance the semantic alignment between the search expression and the task environment, ensuring that the query results are consistent with the current task objectives. Source: The set of keywords with the shortest distance to the task-level context label vector in semantic space. The task-level context label vector is generated by extracting system task metadata and serves as the global reference background for the current query intent. Keywords in this segment do not necessarily appear explicitly in user input, but are generated from the system task structure or representative corpus through contextual reasoning. For example, if the current task is "contract archiving," but the user does not explicitly mention the keyword "archiving," the system will supplement the context label vector with terms such as "archiving," "process," and "classification" to increase the task coverage of the search intent.

[0099] The final constructed comprehensive search expression follows the following concatenation order: main semantic segment → contextual supplementation segment → conflict mediation segment. This sequence is designed to prioritize task-centricity in the expression structure; fill in gaps in the contextual understanding; and finally integrate the objections expressed in the unified expression to ensure semantic consistency. Furthermore, each segment is automatically generated based on structured semantic elements, and the expression supports standard query interfaces, such as structured search requests based on Boolean expressions or natural language conversion. This mechanism enhances the accuracy of search results and the alignment with user intent, while also improving the system's responsiveness to multi-user input fusion queries.

[0100] Embodiment 2: A system for searching by combining user input information, comprising:

[0101] Semantic parsing module, which is used to receive original input information from multiple users and convert each input into an input semantic structure through semantic parsing, which contains keywords, semantic relations and context labels;

[0102] The focus extraction module is used to cross-compare multiple input semantic structures, identify repeated or semantically similar keywords and phrases, extract interactive focus content, and generate a semantic consensus score based on the interactive focus coverage;

[0103] The path construction module is used to construct an input fusion path diagram based on the semantic consensus score, combining the keyword weight formed in each user's historical usage records and the semantic adaptability in the current scenario, to determine the fusion order and fusion method between input semantic structures;

[0104] The structural fusion module is used to perform fusion actions in sequence along the input fusion path diagram, performing semantic structure comparison, context consistency comparison, and semantic conflict mediation for each pair of input semantic structures. When the semantic conflict reaches an irreconcilable level, it uses alternative words to unify the expression and generate a comprehensive semantic expression structure.

[0105] The expression generation module is used to sort the keywords in the comprehensive semantic expression structure according to the semantic level and task-oriented importance, construct a comprehensive search expression, perform query operations and output search results.

[0106] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0107] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0108] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0109] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0110] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for searching by combining user input information, characterized in that: The following steps are involved: Receive raw input information from multiple users and convert each input into an input semantic structure through semantic parsing, which includes keywords, semantic relationships and context labels; Cross-comparison of multiple input semantic structures to identify repeated or semantically similar keywords and phrases, extract interactive focus content, and generate a semantic consensus score based on interactive focus coverage. Based on the semantic consensus score, the keyword weights formed in each user's historical usage records and the semantic adaptability in the current scenario are combined to construct an input fusion path graph. This path graph is used to determine the fusion order and fusion method between input semantic structures. Fusion actions are performed sequentially along the input fusion path diagram. For each pair of input semantic structures, semantic structure comparison, context consistency comparison, and semantic conflict mediation are performed. When the semantic conflict reaches an irreconcilable level, alternative words are used to unify the expression and generate a comprehensive semantic expression structure. Sort the keywords in the comprehensive semantic expression structure according to semantic level and task-oriented importance, construct a comprehensive search expression, execute the query operation and output the search results.

2. The method for searching by combining user input information according to claim 1, characterized in that: When each input is converted into the content of the input semantic structure through semantic parsing, the context tags in the input semantic structure include the following four categories: the timestamp of the input information being received, the current activity status of the user's operating system, the keyword sequence extracted from the user's three consecutive operation paths before receiving the input information, and the task function type tag in the user's current interface; the above four types of context tags together constitute the user's current context tag vector in the current task context.

3. The method for searching by combining user input information according to claim 2, characterized in that: When generating a semantic consensus score, the semantic consensus score is based on an undirected keyword graph constructed from the interactive focus content, where each node is a standardized keyword and the edge represents the frequency of the keyword co-occurring in different input semantic structures; Assume that the node weight is variable X and the edge frequency is variable Y, then the semantic consensus score S is defined as: S = f(X) × P(Y), where f(X) is the normalized graph embedding vector modulus and P(Y) is the co-occurrence edge weight defined based on Markov transition probability.

4. The method for searching by combining user input information according to claim 3, characterized in that: The input fusion path graph is constructed based on the set of all input semantic structures that contribute to the semantic consensus score S. The user input corresponding to each structure in the set is fused and evaluated. The fusion evaluation result F is determined by the semantic consensus score S, the keyword weight W of the user input, and the semantic fitness M, where F = s × sigmoid(W + M), s represents all keywords included in the semantic structure of the user input. When there is a matching node in the interaction focus graph, the semantic consensus score obtained is averaged. The keyword weight W represents the average product of the frequency of use and click consistency of all keywords input by the user in the historical usage record. The semantic fitness M represents the average cosine value of the semantic vector angle between all keywords input by the user and the current task context label. If the fusion evaluation result F is lower than the set retention threshold T, the corresponding user input information is eliminated from the fusion path construction, and the remaining user input information is sorted from high to low according to the F value and arranged in sequence as a node sequence in the input fusion path graph.

5. The method for searching by combining user input information according to claim 4, characterized in that: The fusion action is only performed on user input information that has passed the fusion evaluation function F and has a score not lower than the retention threshold T, that is, the set of input semantic structures corresponding to the sorted valid node sequence in the input fusion path graph. In the process of executing the semantic structure comparison, a keyword structure mapping matrix is ​​constructed for each pair of input semantic structures. The rows and columns of the matrix represent the top k keywords in the two input semantic structures respectively. The matrix elements are the structural-level semantic similarity scores between the pair of keywords. The score is calculated based on the product of the length of the semantic dependency path and the cosine value of the context vector angle. The cosine value of the context vector angle refers to the cosine value of the angle between the context semantic vectors of the two keywords. The context semantic vector is the semantic expression vector of the keyword in the semantic subspace defined by the current context label vector. When the similarity score exceeds the preset matching threshold, the structure merge is performed, otherwise it enters the semantic conflict mediation stage.

6. The method for searching by combining user input information according to claim 5, characterized in that: Semantic conflict mediation adopts an alternative expression processing mechanism. For any two keyword pairs that are in the same semantic position in the input semantic structure but whose semantic vector distance exceeds the set distance threshold, the semantic center vector of the keyword pair is constructed, and the term with the smallest distance to the center vector is selected from the semantic word set as the alternative word for unified expression processing of the keyword pair. The processing process is only executed on the conflicting keyword pairs between the two input semantic structures of the current fusion operation, and the other inputs retain the original structure.

7. The method for searching by combining user input information according to claim 6, characterized in that: When generating the comprehensive semantic expression structure, the semantic level score and task-oriented importance based on which keywords are ranked are calculated from the following sources: the semantic level score U is the cumulative value of the keyword weight W of each keyword in the input fusion path diagram in the calculation of the semantic consensus score S, and then the product of this cumulative value and the number of times it appears in the fusion process; The task-oriented importance V is the cosine value of the semantic angle between each keyword and the task-level context label vector; The task-level context label vector is a reference standard extracted from the system task metadata; The final ranking weight Q of the keyword is obtained by exponential normalization of the sum of U and V, that is, Q = softmax(U+V).

8. The method for searching by combining user input information according to claim 7, characterized in that: The construction of comprehensive search expression adopts segmented splicing generation method, which consists of three parts: main semantic segment, conflict mediation segment and context supplement segment. The main semantic segment includes the top R keywords ranked by ranking weight Q, representing the main intention content in the fused semantic structure; the conflict mediation segment includes keywords that have participated in the semantic conflict mediation process and have been uniformly processed by alternative expressions, which is used to supplement the description of the semantic consistency of the final expression; the context supplement segment includes a set of keywords with the shortest distance to the task-level context label vector in the semantic space, which is used to bridge the gap between the expression structure and the task environment; A comprehensive search expression is formed by splicing the main semantic segment, the context supplement segment, and the conflict mediation segment in order.

9. A system for searching by combining user input information, for implementing the method for searching by combining user input information according to any one of claims 1 to 8, characterized in that: include: Semantic parsing module, which is used to receive original input information from multiple users and convert each input into an input semantic structure through semantic parsing, which contains keywords, semantic relations and context labels; The focus extraction module is used to cross-compare multiple input semantic structures, identify repeated or semantically similar keywords and phrases, extract interactive focus content, and generate a semantic consensus score based on the interactive focus coverage; The path construction module is used to construct an input fusion path diagram based on the semantic consensus score, combining the keyword weight formed in each user's historical usage records and the semantic adaptability in the current scenario, to determine the fusion order and fusion method between input semantic structures; The structural fusion module is used to perform fusion actions in sequence along the input fusion path diagram, performing semantic structure comparison, context consistency comparison, and semantic conflict mediation for each pair of input semantic structures. When the semantic conflict reaches an irreconcilable level, it uses alternative words to unify the expression and generate a comprehensive semantic expression structure. The expression generation module is used to sort the keywords in the comprehensive semantic expression structure according to the semantic level and task-oriented importance, construct a comprehensive search expression, perform query operations and output search results.

Images