Recursive deep text query method, system and equipment and storage medium

Abstract

The invention discloses a recursive deep text query method, system and device and a storage medium, and the method comprises the steps: firstly creating an initial query and initializing a hierarchical depth, executing a first-layer query, carrying out evolution processing, and then carrying out retrieval; if the information density is insufficient, sub-query is generated to execute second-layer query, and retrieval is performed after second evolution processing; if the second-layer retrieval is contradictory, third-layer query is executed, retrieval is performed after timeline query and information source tracing verification are passed, and a third-layer retrieval result is finally output; through multi-level query and targeted evolution processing, query can be optimized step by step, more information dimensions can be covered, information gaps can be filled up, and retrieval comprehensiveness and accuracy can be improved; and a third-layer verification mechanism can effectively check contradictions by means of timelines and source tracing, so that the accuracy and reliability of results are guaranteed, and better retrieval results are provided for users.

Description

[0001] Technology Neighborhood

[0002] This invention relates to the fields of artificial intelligence and natural language processing, and in particular to a recursive deep text query method, system, device, and storage medium. Background Technology

[0003] In the field of information retrieval, traditional single-query methods have significant shortcomings. On the one hand, their ability to identify and expand entities is limited, making it difficult to comprehensively cover relevant entities and their variations, resulting in missing information in the search results. On the other hand, the lack of effective derivation of relationship chains and generation of semantic variations results in a narrow semantic scope, failing to accurately capture implicit information connections. Furthermore, traditional queries typically lack multi-level evolution and verification mechanisms. When search results are contradictory or lack sufficient information density, they cannot guarantee the accuracy and comprehensiveness of the results through further query optimization and verification, making it difficult to meet users' needs for high-quality, accurate, and comprehensive search results. Summary of the Invention

[0004] The purpose of this invention is to address the technical problems existing in the background art by proposing a recursive deep text query method, system, device, and storage medium.

[0005] To achieve the above-mentioned technical objectives, the technical solution adopted by the present invention is as follows:

[0006] A first implementation of the first aspect of the present invention provides a recursive deep text query method, comprising:

[0007] S101. Create an initial query operation and initialize the query level depth of the query operation;

[0008] S102. Execute the first-level query of the query operation, and perform the first evolutionary processing on the first-level query to execute the first-level retrieval. The evolutionary processing includes entity recognition and expansion, relation chain deduction and semantic variant generation.

[0009] S103. When the information density of the first-level retrieval is insufficient, a set of subqueries is generated and the second-level query is executed. The second-level query is then subjected to a second evolutionary process to execute the second-level retrieval. The second evolutionary process includes refining the information gap and expanding the relevant information dimensions.

[0010] S104. When a contradiction is detected in the second-level search, execute the third-level query to verify the result of the second-level search in order to execute the third-level search, and output the result of the third-level search as the final result. The verification includes timeline query and information source tracing.

[0011] Optionally, in a second implementation of the first aspect of the present invention, the method for evaluating information density is as follows:

[0012] Construct a formula for calculating information density, and a preset threshold;

[0013] The first score of information density is calculated using a formula and compared with a threshold to determine the evaluation result.

[0014] Optionally, in a third implementation of the first aspect of the present invention, the formula for calculating information density is:

[0015] ID = α × Relevance + β × Completeness + γ × Determinism, where ID is the information density;

[0016] Where, relevance = semantic similarity × entity matching degree;

[0017] Completeness = Dimensions of information obtained / Dimensions of information required;

[0018] Certainty = Σ(Source Authority × Consistency Coefficient) / Number of Sources;

[0019] Weighting coefficients: α = 0.4, β = 0.3, γ = 0.3.

[0020] Optionally, in a fourth implementation of the first aspect of the present invention, S103 further includes:

[0021] S1031. After evaluating the information density of the first-level retrieval, determine whether to proceed with query recursion.

[0022] Optionally, in the fifth implementation of the first aspect of the present invention, determining whether to enter the query recursion includes:

[0023] S10311. Determine whether the similarity of the preset query vector is greater than the first preset value;

[0024] S10312. If it is less than the second preset value, then execute query redirection, recreate a new query operation, calculate the information gain of the new query operation, and when the information gain is greater than the second preset value, perform semantic similarity check, calculate and update the query vector similarity. When the information gain is less than the second preset value, stop the query recursive operation.

[0025] S10313. If the value is greater than the given value, then if all computing resources are found to be compatible, execute the query recursive operation; if all computing resources are found to be incompatible, stop the query recursive operation.

[0026] Optionally, in a sixth implementation of the first aspect of the present invention, the contradiction detection in S104 includes:

[0027] Perform contradiction detection on the second-level search;

[0028] If a contradiction exists, proceed to step S104;

[0029] If there is no contradiction, the information density in the second-level retrieval is evaluated to obtain the second score of the information density in the second-level retrieval. The second score is then compared with the threshold to determine whether the score is sufficient.

[0030] If sufficient, the results of the second-level search will be output as the final result.

[0031] Optionally, in the seventh implementation of the first aspect of the present invention, before outputting the final result, it further includes:

[0032] Based on a pre-set multi-layer result fusion algorithm, the information in the final result is weighted, conflict resolution rules are configured, and deduplication is performed.

[0033] A first implementation of the second aspect of the present invention provides a recursive deep text query system, comprising:

[0034] The initial module is used to create the initial query operation and initialize the query hierarchy depth of the query operation;

[0035] The first-level query module is used to execute the first-level query operation. It performs the first evolutionary processing on the first-level query to execute the first-level retrieval. The evolutionary processing includes entity recognition and expansion, relation chain inference and semantic variant generation.

[0036] The second-level query module is used to generate a set of subqueries and execute the second-level query when the information density of the first-level retrieval is insufficient. The second-level query is then subjected to a second evolutionary process to execute the second-level retrieval. The second evolutionary process includes refining the information gap and expanding the relevant information dimensions.

[0037] The three-layer query module is used to execute a third-layer query when a contradiction is detected in the second-layer retrieval. This verifies the results of the second-layer retrieval before executing the third-layer retrieval, and outputs the results of the third-layer retrieval as the final result. The verification includes timeline query and information source tracing.

[0038] A first implementation of the third aspect of the present invention provides a recursive deep text query device, the recursive deep text query device comprising: a memory and at least one processor, the memory storing instructions, and the memory and the at least one processor being interconnected via a line;

[0039] The at least one processor invokes the instructions in the memory to cause the recursive deep text query device to perform the recursive deep text query method as described in any one of the first aspects of the present invention.

[0040] A first implementation of the fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the recursive deep text query method as described in any one of the first aspects of the present invention.

[0041] Compared with the prior art, the present invention has the following beneficial technical effects:

[0042] By first creating an initial query and initializing the hierarchical depth, the first-level query is executed and refined before retrieval. If the information density is insufficient, a subquery is generated and a second-level query is executed, followed by a second refinement before retrieval. If there are contradictions in the second-level retrieval, a third-level query is executed, and the retrieval is verified through timeline querying and information source tracing, ultimately outputting the third-level retrieval results. Through multi-level queries and targeted refinement, the query can be gradually optimized to cover more information dimensions, fill information gaps, and improve the comprehensiveness and accuracy of retrieval. The third-level verification mechanism, with the help of timeline and source tracing, can effectively identify contradictions, ensure the accuracy and reliability of the results, and provide users with higher-quality retrieval results. Attached Figure Description

[0043] Figure 1 This is a schematic diagram of the first embodiment of the recursive deep text query method in this invention;

[0044] Figure 2 This is a schematic diagram of the fourth embodiment of the recursive deep text query method in this invention;

[0045] Figure 3 This is a schematic diagram of the fifth embodiment of the recursive deep text query method in this invention;

[0046] Figure 4 This is a schematic diagram of an embodiment of the recursive deep text query system of the present invention;

[0047] Figure 5 This is a schematic diagram of an embodiment of the recursive deep text query device in this invention;

[0048] Figure 6 This is a schematic diagram of the execution flow of the recursive deep text query method in an embodiment of the present invention;

[0049] Figure 7 This is a schematic diagram of the execution flow for determining whether to enter the query recursion method in an embodiment of the present invention;

[0050] Figure 8 This is a schematic diagram illustrating the execution flow of the collaborative mechanism in practical application according to an embodiment of the present invention. Detailed Implementation

[0051] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0052] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figures 1-3 as well as Figure 6 The recursive deep text query method in this embodiment of the invention includes:

[0053] S101. Create an initial query operation and initialize the query level depth of the query operation;

[0054] In this embodiment, the initial query operation is created to clarify the starting state of the entire query process, providing an initial query object and a basic hierarchical identifier for subsequent query operations. This serves as the initial query, and the query operation also includes a built-in record of the current query operation's hierarchy, represented in this embodiment as `depth`, which is initially set to 0, indicating the very beginning hierarchy. In subsequent processes, when performing operations such as second-level recursive queries, `depth` increases accordingly to distinguish different query hierarchies, facilitating control and management of query depth. For example, it can be used to determine whether to continue generating subqueries by checking if conditions such as "depth < 5" are met.

[0055] S102. Execute the first-level query of the query operation, and perform the first evolutionary processing on the first-level query to execute the first-level retrieval. The evolutionary processing includes entity recognition and expansion, relation chain deduction and semantic variant generation.

[0056] In this embodiment, query evolution follows a three-stage pattern of "expansion-focusing-verification":

[0057] Phase 1 (L1-L2): Expanded Exploration

[0058] Entity expansion: Identify the core entities in the query and expand their aliases and abbreviations;

[0059] Relationship expansion: Possible relationship types based on common sense reasoning;

[0060] Example: "Relationship between Company A and Company B" → ["Company A invests in Company B", "Company A cooperates with Company B", "Company A competes with Company B"]

[0061] Phase Two (L3-L4): Focus on Deepening

[0062] Targeted queries: Generate precise queries based on clues discovered through L1-L2 analysis;

[0063] Additional details: Inquire about specific time, amount, terms, and other details;

[0064] Example: "Company A invests in Company B" → ["Amount of investment by Company A in Company B", "Percentage of shares held by Company A in Company B", "Date of signing of the investment agreement"]

[0065] The third stage (L5): Cross-verification

[0066] Multi-source verification: Verify key information from different sources;

[0067] Temporal verification: Verify the time logic of events;

[0068] Examples: "Official announcement of Company A and Company B", "Third-party report on the relationship between A and B".

[0069] S103. When the information density evaluated in the first-level retrieval is insufficient, generate a sub-query set and execute the second-level query, and perform a second evolutionary process on the second-level query to execute the second-level retrieval, where the second evolutionary process includes refinement of the information gap and expansion of relevant information dimensions;

[0070] In this embodiment, the method for evaluating information density is as follows:

[0071] Construct a calculation formula for information density and a preset threshold;

[0072] Calculate the first score value of the information density through the calculation formula and compare it with the threshold to determine the evaluation result.

[0073] The calculation formula for information density is:

[0074] ID = α × relevance + β × integrity + γ × certainty, where ID is the information density;

[0075] Among them, relevance = semantic similarity × entity matching degree;

[0076] Integrity = acquired information dimensions / required information dimensions;

[0077] Certainty = Σ(source authority × consistency coefficient) / number of sources;

[0078] Weight coefficients: α = 0.4, β = 0.3, γ = 0.3.

[0079] In this embodiment, the setting of the threshold is as follows:

[0080] ID > 0.8: Sufficient information, stop recursion;

[0081] 0.5 < ID ≤ 0.8: Optionally go deeper;

[0082] ID ≤ 0.5: Must continue recursion.

[0083] Furthermore, step S103 can also be performed as follows:

[0084] S1031. After evaluating the information density of the first-level retrieval, determine whether to proceed with query recursion.

[0085] Furthermore, determining whether to proceed with the query recursion includes:

[0086] S10311. Determine whether the similarity of the preset query vector is greater than the first preset value;

[0087] S10312. If it is less than the second preset value, then execute query redirection, recreate a new query operation, calculate the information gain of the new query operation, and when the information gain is greater than the second preset value, perform semantic similarity check, calculate and update the query vector similarity. When the information gain is less than the second preset value, stop the query recursive operation.

[0088] S10313. If the value is greater than the given value, then if all computing resources are found to be compatible, execute the query recursive operation; if all computing resources are found to be incompatible, stop the query recursive operation.

[0089] See appendix Figure 7 In this embodiment, the query vector similarity is cos(Vi,V0), and the first preset value is 0.6. The formula for calculating the information gain of a new query operation is ΔI = I(Ri) - I(Ri-1), where I(Ri) represents the information content of the current query result, and I(Ri-1) represents the information content of the previous query result. It should be noted that when ΔI > 0.1, it can ensure that each layer of query brings valuable new information. The design of cosine similarity > 0.6 is to prevent the query from deviating from the topic. The computing resources include the number of query depth layers, the query time of each layer, and the total memory resources. The maximum number of query depth layers is 5, the query time of each layer needs to be controlled within 3 seconds, and the total memory resources can be adjusted according to the actual business situation. Here, it is set to not exceed 1GB. Figure 7 It can be seen that when testing computing resources, it is necessary to test the number of query depths, the query time of each layer, and the total memory resources. Only when all these tests are passed can the query recursion operation be executed to enter the next layer of query; otherwise, the query recursion operation is stopped.

[0090] S104. When a contradiction is detected in the second-level search, execute the third-level query to verify the result of the second-level search in order to execute the third-level search, and output the result of the third-level search as the final result. The verification includes timeline query and information source tracing.

[0091] In this embodiment, when a contradiction is detected in the second-level retrieval, a third-level query is required to verify the results of the second level in order to obtain accurate and reliable results. The verification methods include timeline query and information source tracing. Timeline query is to query relevant content along the time sequence of information generation or development to check whether there are conflicts or mutual corroboration between information at different time points, thereby judging the rationality of the contradictions in the second-level results. Information source tracing is to trace the original source of the information, check the reliability and authority of the source, and whether the information provided by the source contradicts the second-level results, thereby verifying the accuracy of the second-level results.

[0092] As for performing the third-level search, the results obtained from the third-level search are used as the final output results to ensure that the final output results have been fully verified and have high credibility.

[0093] Furthermore, the "contradiction detection" in step S104 includes:

[0094] Perform contradiction detection on the second-level search;

[0095] If a contradiction exists, proceed to step S104;

[0096] If there is no contradiction, the information density in the second-level retrieval is evaluated to obtain the second score of the information density in the second-level retrieval. The second score is then compared with the threshold to determine whether the score is sufficient.

[0097] If sufficient, the results of the second-level search will be output as the final result.

[0098] Furthermore, before outputting the final result, it also includes:

[0099] Based on a pre-set multi-layer result fusion algorithm, the information in the final result is weighted, conflict resolution rules are configured, and deduplication is performed.

[0100] In this embodiment, the fusion strategy adopts a weighted voting mechanism:

[0101] 1. Hierarchical weight allocation:

[0102] Base layer: Weight 0.6, provides basic facts.

[0103] Deeper Layer: Weight 0.3, provides detailed information

[0104] Validation layer: weight 0.1, provides validation information.

[0105] 2. Conflict resolution rules:

[0106] Time priority: Latest information has the highest priority.

[0107] Source priority: Official > Authoritative media > General sources

[0108] Majority principle: Information verified by multiple independent sources is more credible.

[0109] 3. Deduplication strategy: Fast deduplication based on semantic hashing while retaining the most complete version of information.

[0110] In this embodiment, an initial query is created and the hierarchical depth is initialized. The first-level query is executed and then refined before retrieval. If the information density is insufficient, a subquery is generated and a second-level query is executed, followed by a second refinement before retrieval. If there are contradictions in the second-level retrieval, a third-level query is executed, and the retrieval is verified through timeline querying and information source tracing. Finally, the third-level retrieval results are output. Through multi-level queries and targeted refinement, the query can be gradually optimized to cover more information dimensions, fill information gaps, and improve the comprehensiveness and accuracy of the retrieval. The verification mechanism of the third level, with the help of timeline and source tracing, can effectively eliminate contradictions, ensure the accuracy and reliability of the results, and provide users with better retrieval results.

[0111] As attached Figure 8 As shown, the collaborative mechanism of this recursive deep text query method in practical applications is specifically implemented as follows:

[0112] The system obtains the user's query information and selects a strategy, which has three scenarios: insufficient information, unclear relationship, and both.

[0113] If the information is insufficient, a recursive query is executed. During the recursive query, a contradiction detection is performed. If a contradiction exists, the exclusion mechanism in the fact exclusion determination mechanism is invoked to collect exclusion evidence. If there is no contradiction, the recursive query continues and the final evidence is output.

[0114] If the relationship is unclear, the fact exclusion determination mechanism is activated. This involves determining if the required information is insufficient. If insufficient, a recursive query is invoked. If sufficient, exclusion evidence is collected.

[0115] If both conditions are met, the execution flow for both scenarios will be executed simultaneously.

[0116] Finally, the collected exclusionary evidence and final evidence are fused together using an evidence fusion algorithm to construct an enhanced context. An answer is generated using a pre-set LLM model and then verified. If the verification passes, the answer is output directly. If the verification fails, the deep proof algorithm is called and the process returns to re-execute the strategy selection.

[0117] The recursive deep text query method in the embodiments of the present invention has been described above. The recursive deep text query system in the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 4Recursive deep text query systems include:

[0118] Initial module 201 is used to create the initial query operation and initialize the query hierarchy depth of the query operation;

[0119] The first-level query module 202 is used to perform the first-level query operation and to perform the first-level retrieval on the first-level query. The evolution process includes entity recognition and expansion, relation chain deduction and semantic variant generation.

[0120] The second-level query module 203 is used to generate a set of subqueries and execute the second-level query when the information density of the first-level retrieval is insufficient, and to perform a second evolutionary process on the second-level query in order to execute the second-level retrieval. The second evolutionary process includes refining the information gap and expanding the relevant information dimensions.

[0121] The three-layer query module 204 is used to execute a third-layer query when a contradiction is detected in the second-layer retrieval, to verify the results of the second-layer retrieval, and to output the results of the third-layer retrieval as the final result. The verification includes timeline query and information source tracing.

[0122] The evaluation method for "information density" in the second-level query module 203 is as follows:

[0123] Construct a formula for calculating information density, and a preset threshold;

[0124] The first score of information density is calculated using a formula and compared with a threshold to determine the evaluation result.

[0125] The formula for calculating information density is as follows:

[0126] ID = α × Relevance + β × Completeness + γ × Determinism, where ID is the information density;

[0127] Where, relevance = semantic similarity × entity matching degree;

[0128] Completeness = Dimensions of information obtained / Dimensions of information required;

[0129] Certainty = Σ(Source Authority × Consistency Coefficient) / Number of Sources;

[0130] Weighting coefficients: α = 0.4, β = 0.3, γ = 0.3.

[0131] The second-level query module 203 can also execute:

[0132] After evaluating the information density of the first-level retrieval, a decision is made on whether to proceed with the query recursion.

[0133] Among them, "determining whether to enter the query recursion" includes:

[0134] Determine whether the similarity of the preset query vector is greater than the first preset value;

[0135] If the information gain is less than the second preset value, then query redirection is performed, a new query operation is created, the information gain of the new query operation is calculated, and when the information gain is greater than the second preset value, semantic similarity check is performed, and query vector similarity is calculated and updated. When the information gain is less than the second preset value, the query recursion operation is stopped.

[0136] If the value is greater than the given value, the query recursive operation is executed when all computing resources are found to be compatible; the query recursive operation is stopped when all computing resources are found to be incompatible.

[0137] The "contradiction detection" in the three-layer query module 204 includes:

[0138] Perform contradiction detection on the second-level search;

[0139] If a contradiction exists, proceed to step S104;

[0140] If there is no contradiction, the information density in the second-level retrieval is evaluated to obtain the second score of the information density in the second-level retrieval. The second score is then compared with the threshold to determine whether the score is sufficient.

[0141] If sufficient, the results of the second-level search will be output as the final result.

[0142] Before outputting the final result, the following is also included:

[0143] Based on a pre-set multi-layer result fusion algorithm, the information in the final result is weighted, conflict resolution rules are configured, and deduplication is performed.

[0144] In this embodiment, an initial query is created and the hierarchical depth is initialized. The first-level query is executed and then refined before retrieval. If the information density is insufficient, a subquery is generated and a second-level query is executed, followed by a second refinement before retrieval. If there are contradictions in the second-level retrieval, a third-level query is executed, and the retrieval is verified through timeline querying and information source tracing. Finally, the third-level retrieval results are output. Through multi-level queries and targeted refinement, the query can be gradually optimized to cover more information dimensions, fill information gaps, and improve the comprehensiveness and accuracy of the retrieval. The verification mechanism of the third level, with the help of timeline and source tracing, can effectively eliminate contradictions, ensure the accuracy and reliability of the results, and provide users with better retrieval results.

[0145] The above is attached Figure 4The recursive deep text query method in this embodiment of the invention is described in detail from the perspective of unitized functional entities. The recursive deep text query device in this embodiment of the invention is described in detail from the perspective of hardware processing.

[0146] Figure 5 This is a schematic diagram of the structure of a recursive deep text query device 300 provided in an embodiment of the present invention. The recursive deep text query device 300 can vary significantly due to different configurations or performance. It may include one or more central processing units (CPUs) 310 (e.g., one or more processors) and a memory 320, and one or more storage media 330 (e.g., one or more mass storage devices) for storing application programs 333 or data 332. The memory 320 and storage media 330 can be temporary or persistent storage. The program stored in the storage media 330 may include one or more units (not shown in the diagram), each unit may include a series of instruction operations on the recursive deep text query device 300. Furthermore, the processor 310 may be configured to communicate with the storage media 330 and execute the series of instruction operations in the storage media 330 on the recursive deep text query device 300.

[0147] The recursive deep text query device 300 may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input / output interfaces 360, and / or one or more operating systems 331, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 5 The recursive deep text query device structure shown does not constitute a limitation on communication protocol devices based on LAN projection. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0148] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the recursive deep text query method.

[0149] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0150] The above describes a recursive deep text query method or various implementations in conjunction with specific content, and does not imply that the specific implementation of this invention is limited to these descriptions. Any methods or structures that are similar to or identical to those of this invention, or any technical deductions or substitutions made based on the concept of this invention, should be considered within the scope of protection of this invention.

Claims

1. A recursive deep text query method, characterized in that, include: S101. Create an initial query operation and initialize the query level depth of the query operation; S102. Execute the first-level query of the query operation, and perform a first evolutionary process on the first-level query to execute the first-level retrieval, wherein the evolutionary process includes entity recognition and expansion, relation chain deduction and semantic variant generation; S103. When the information density of the first-level retrieval is insufficient, a sub-query set is generated and a second-level query is executed. The second-level query is then subjected to a second evolutionary process to execute the second-level retrieval. The second evolutionary process includes refining the information gap and expanding the relevant information dimensions. S104. When a contradiction is detected in the second-level retrieval, a third-level query is executed to verify the result of the second-level retrieval in order to execute the third-level retrieval, and the result of the third-level retrieval is output as the final result. The verification includes timeline query and information source tracing.

2. The recursive deep text query method according to claim 1, characterized in that, The method for evaluating the information density is as follows: Construct the calculation formula for the information density, and a preset threshold; The first score of the information density is calculated using the formula and compared with the threshold to determine the evaluation result.

3. The recursive deep text query method according to claim 2, characterized in that, The formula for calculating the information density is: ID = α × Relevance + β × Completeness + γ × Determinism, where ID is the information density; Where, relevance = semantic similarity × entity matching degree; Completeness = Dimensions of information obtained / Dimensions of information required; Certainty = Σ(Source Authority × Consistency Coefficient) / Number of Sources; Weighting coefficients: α = 0.4, β = 0.3, γ = 0.

3.

4. The recursive deep text query method according to claim 3, characterized in that, S103 further includes: After evaluating the information density of the first-level retrieval, a decision is made as to whether to proceed with the query recursion.

5. The recursive deep text query method according to claim 4, characterized in that, The determination of whether to enter the query recursion includes: S10311. Determine whether the similarity of the preset query vector is greater than the first preset value; S10312. If the information gain is less than the second preset value, then perform query redirection, recreate a new query operation, calculate the information gain of the new query operation, perform semantic similarity check when the information gain is greater than the second preset value, calculate and update the query vector similarity, and stop the query recursive operation when the information gain is less than the second preset value. S10313. If the value is greater than the given value, then if all computing resources are found to be compatible, execute the query recursive operation; if all computing resources are found to be incompatible, stop the query recursive operation.

6. The recursive deep text query method according to claim 5, characterized in that, The contradiction detection in S104 includes: Perform contradiction detection on the second-level search; If a contradiction exists, continue with step S104. If there is no contradiction, the information density in the second-level retrieval is evaluated to obtain a second score of the information density in the second-level retrieval, and the second score is compared with the threshold to determine whether the score is sufficient. If sufficient, the results of the second-level search will be output as the final result.

7. The recursive deep text query method according to claim 6, characterized in that, Before outputting the final result, it also includes: Based on a pre-set multi-layer result fusion algorithm, the information in the final result is weighted, conflict resolution rules are configured, and deduplication is performed.

8. A recursive deep text query system, characterized in that, include: The initial module is used to create an initial query operation and initialize the query hierarchy depth of the query operation; A first-level query module is used to execute the first-level query of the query operation, and to perform a first-level evolutionary process on the first-level query to execute the first-level retrieval. The evolutionary process includes entity recognition and expansion, relationship chain deduction and semantic variant generation. The second-level query module is used to generate a set of subqueries and execute the second-level query when the information density of the first-level retrieval is insufficient. The second-level query is then subjected to a second evolutionary process to execute the second-level retrieval. The second evolutionary process includes refining the information gap and expanding the relevant information dimensions. The three-layer query module is used to execute a third-layer query when a contradiction is detected in the second-layer retrieval. This verifies the result of the second-layer retrieval, executes the third-layer retrieval, and outputs the result of the third-layer retrieval as the final result. The verification includes timeline query and information source tracing.

9. A recursive deep text query device, characterized in that, The recursive deep text query device includes: a memory and at least one processor, wherein the memory stores instructions, and the memory and the at least one processor are interconnected via a line; The at least one processor invokes the instructions in the memory to cause the recursive deep text query device to perform the recursive deep text query method as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program thereon, characterized in that, When the computer program is executed by the processor, it implements the recursive deep text query method as described in any one of claims 1-7.

Images