A system and method for detecting evidence contamination based on reverse mapping localization

Abstract

This invention discloses an evidence contamination detection system and method based on reverse mapping localization, belonging to the field of medical evidence-based technology. The invention retrieves operation records from the evidence storage system's operation logs to trace the source path of the target evidence and generate an evidence flow timeline. By calculating a contamination risk value, it traces the upstream evidence source of the target contamination point, constructing a contamination propagation path tree. Finally, it locates the contamination source based on a contamination source scoring function. This invention can accurately identify evidence contamination problems, trace contamination sources and propagation paths, ensuring the reliability of evidence-based medicine and solving the problem that traditional evidence detection methods cannot accurately identify and trace contaminated evidence.

Description

Technical Field

[0001] This invention relates to the field of medical evidence-based technology, specifically to an evidence contamination detection system and method based on reverse mapping localization. Background Technology

[0002] With the advent of the era of evidence-based medicine, clinical decision-making is increasingly reliant on evidence-based scientific evidence. Currently, the amount of evidence accumulated in evidence-based medicine databases is growing exponentially, and this medical evidence presents a huge challenge to the formulation of clinical strategies.

[0003] Traditional methods of reviewing medical evidence often rely on manual screening or simple literature quality assessment tools, which are no longer sufficient to meet current evidence detection needs. Traditional methods cannot accurately identify evidence contamination issues such as data fabrication and selective reporting, nor can they trace the transmission of contaminated evidence to clinical decisions. This often leads to studies containing erroneous data being included in the evidence system, causing biases in clinical strategies and resulting in patients receiving inappropriate treatment.

[0004] These issues pose significant challenges to the reliability of evidence-based medicine, seriously affecting the scientific validity and safety of clinical diagnosis and treatment, hindering the further development of the evidence-based medicine system, and causing waste of medical resources and health risks to patients. Summary of the Invention

[0005] The purpose of this invention is to provide an evidence contamination detection system and method based on reverse mapping localization to solve the problems raised in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an evidence contamination detection method based on reverse mapping localization, the detection method comprising:

[0007] Step S100: Retrieve all operation records of evidence generation instructions from the operation log of the evidence storage system; extract information from each operation record, and trace the source path of each piece of evidence based on the extracted operation record information;

[0008] Step S200: The source path of each piece of evidence is sorted out and an evidence flow timeline is generated; the pollution propagation probability and evidence matching degree of each evidence point on the evidence flow timeline are calculated; and the pollution risk value of each evidence point is calculated based on the pollution propagation probability and evidence matching degree.

[0009] Step S300: Based on the pollution risk value of each evidence point, determine the target pollution point, trace the upstream evidence source of the target pollution point, and construct a pollution propagation path tree;

[0010] Step S400: Calculate the pollution propagation range of the upper-level associated evidence of the target pollution point in the pollution propagation path tree, and construct a pollution source scoring function by combining the pollution propagation probability and evidence matching degree to locate the pollution source.

[0011] Furthermore, step S100 includes:

[0012] Step S101: Retrieve all operation records of evidence generation instructions from the operation log of the evidence storage system. The operation records include evidence derivation, updating, and referencing.

[0013] Step S102: In the operation record of each evidence generation instruction, capture the generated target evidence; in the operation record of each evidence generation instruction, capture all related evidence called when generating the target evidence, and obtain evidence information for the related evidence, the evidence information including the evidence source of the related evidence; the evidence source of the related evidence includes the evidence chain that generated the related evidence in the evidence storage system;

[0014] Step S103: In the operation record of each evidence generation instruction, capture the operation record Re of the associated evidence when the target evidence is generated. k Where k = 1, 2, or 3; the operation record includes derived operations Re. 1 Update operation Re 2 Reference operation Re 3 Among them, if the operation record of the target evidence is a derived operation Re 1 This indicates that new evidence was obtained through analysis of related evidence; this new evidence is the target evidence. If the operation record of the target evidence is an update operation Re... 2 This indicates that new evidence has been obtained by updating the relevant evidence based on the latest research data where there are discrepancies; if the operation record of the target evidence is a citation operation Re... 3 This indicates that the research findings or data of the related evidence were used by the target evidence;

[0015] Step S104: Capture and mark the operation records of all evidence in the evidence chain of the associated evidence in the evidence storage system to obtain the evidence chain ORG; based on the operation records between the target evidence and the associated evidence, the operation records between the associated evidence and the evidence chain, and the operation records between each piece of evidence in the evidence chain, generate the source path L of the target evidence through reverse mapping. A {A→Re k →B→Re k →ORG};wherein, L A B represents the source path of target evidence A, and B represents related evidence of evidence A;

[0016] In the above steps, "target evidence" refers to the evidence currently generated in the evidence storage system; "related evidence" is the basis for the generation of "target evidence"; for example, multiple references cited by a paper (target evidence) (related evidence); "evidence chain" records the generation history of "related evidence" itself. Through the evidence chain, the depth of evidence tracing can be ensured, and the source of evidence can be explored more deeply; through reverse mapping, starting from the target evidence, the related data and data chain on which the target evidence depends can be traced in reverse, which can better locate the source of pollution.

[0017] Furthermore, step S200 includes:

[0018] Step S201: Sort all evidence along the source path of the target evidence according to time sequence to generate an evidence flow timeline; the set of evidence points on the evidence flow timeline is E={e1,…,e i ,…,e n}; where each piece of evidence is marked with a timestamp t i e i Let e ​​represent the i-th evidence point. n Indicate the source of evidence; calculate the probability of contamination transmission between two adjacent evidence points. Among them, P i The probability of contamination propagation between the i-th evidence point and the (i+1)-th evidence point on the evidence flow timeline is represented by N, where N represents the total number of times the i-th evidence point is operated on in the operation log of the evidence storage system, and t represents the probability of contamination propagation between the i-th evidence point and the (i+1)-th evidence point on the evidence flow timeline. j t j+1 These represent evidence point e respectively. i e i+1 The timestamp marked, k This represents the operation weight corresponding to the operation record; where k = 1, 2, or 3.

[0019] Step S202: Obtain the complete content of each evidence point in the evidence flow timeline. The complete content of each piece of evidence includes text, chart annotations, and data tables. Semantically vectorize the obtained complete content of each piece of evidence using natural language processing technology. Calculate the content overlap between the semantically vectorized evidence points using cosine similarity and normalize the overlap. Set the credibility of each evidence point based on the total number of operations performed on it in the evidence storage system's operation log; the higher the total number of operations, the higher the credibility. Calculate the evidence matching degree between each evidence point. Among them, M i S represents the degree of evidence matching between the i-th evidence point and the (i+1)-th evidence point. i C represents the degree of overlap between the i-th evidence point and the (i+1)-th evidence point. i Indicates the credibility of the i-th evidence point;

[0020] Step S203: Calculate the pollution risk value for each evidence point based on the probability of pollution transmission and the degree of evidence matching. ; where e→e i This represents all evidence points e that point to evidence point e. i ;

[0021] In the above steps, "e→e i "" indicates all upstream evidence points pointing to the current evidence point e. The comprehensive risk of multi-source pollution is assessed by summing the products of the propagation probability and mismatch degree of each evidence point in the upstream path. For example, if evidence e depends on two related upstream evidences e1 and e2, then the pollution risk value of evidence e is Q = Q (e→e1) +Q (e→e2) .

[0022] Furthermore, step S300 includes:

[0023] Step S301: Preset a pollution risk threshold F, which is set based on the historical pollution data of the evidence storage system; traverse all evidence points on the evidence flow timeline and compare the pollution risk value Q of each evidence point with the pollution risk threshold F; if the pollution risk value Q≥F, then the evidence point is taken as the target pollution point.

[0024] Step S302: Retrieve the source path L for each target contamination point e. e {e→Re k →e i →Re k →ORG};Capture all relevant evidence in the upstream path of the target contamination point e along the source path, forming a set of relevant evidence {u1,u2,…,u}. n}; where each piece of related evidence is marked with an operation record and timestamp related to the target contamination point e; u n This represents the nth piece of evidence linking the target contamination point e;

[0025] Step S303: For each piece of associated evidence in the set of associated evidence, repeat the tracing logic of step S302 to generate a set of associated evidence in the upstream path of each piece of associated evidence; based on the operation records and timestamps marked by each piece of associated evidence, construct a pollution propagation path tree from the target pollution point e to each layer of upstream path, for visualizing the hierarchical relationship of pollution propagation.

[0026] Furthermore, step S400 includes:

[0027] Step S401 includes: starting from the root node target pollution point e of the pollution propagation path tree, traversing the associated evidence at each layer upwards, counting the number of lower-level associated evidence for each associated evidence in the pollution propagation path tree, and calculating the pollution propagation range of each associated evidence. ;in, i Q represents the extent of pollution propagation associated with the i-th piece of evidence in the pollution propagation path tree. i Let m represent the contamination risk value of the i-th related evidence, and m represent the number of lower-level related evidences of the i-th related evidence.

[0028] Step S402: Construct a pollution source scoring function by combining evidence matching degree and pollution transmission probability. The pollution source scoring function is used to score all related evidence on the pollution propagation path tree, the scores of all related evidence are sorted, and the related evidence with the highest score is marked. The marked related evidence is the source of pollution evidence.

[0029] Furthermore, to better implement the above method, an evidence pollution detection system based on reverse mapping location is also provided. This detection system includes: an operation record processing module, a pollution risk calculation module, a pollution path tracing module, and a pollution source location module.

[0030] The operation log processing module is used to retrieve all operation records of evidence generation instructions in the operation log of the evidence storage system, extract information from each operation record, and sort out the source path of each piece of evidence based on the extracted operation record information;

[0031] The pollution risk calculation module analyzes the source path of each piece of evidence, generates an evidence flow timeline, and calculates the probability of pollution transmission, evidence matching degree, and pollution risk value.

[0032] The pollution path tracing module is used to identify target pollution points, trace upstream evidence sources of the target pollution points, and construct a pollution propagation path tree;

[0033] The pollution source location module calculates the pollution propagation range of the upper-level associated evidence of the target pollution point in the pollution propagation path tree, and constructs a pollution source scoring function by combining the pollution propagation probability and evidence matching degree to locate the pollution source.

[0034] Furthermore, the operation record processing module includes: an operation record retrieval unit, an associated evidence capture unit, an operation type marking unit, and a source path generation unit;

[0035] The operation record retrieval unit is used to retrieve all operation records of evidence generation instructions from the operation log of the evidence storage system.

[0036] The associated evidence capture unit captures associated evidence called during the generation of target evidence and obtains the evidence source of the associated evidence;

[0037] Operation type marking unit, used to mark the operation type of the target evidence on the associated evidence;

[0038] The source path generation unit is used to generate the source path of the target evidence through reverse mapping.

[0039] Furthermore, the pollution risk calculation module includes: a timeline generation unit, a pollution probability calculation unit, an evidence matching degree calculation unit, and a risk value calculation unit;

[0040] The timeline generation unit is used to generate a timeline of evidence flow.

[0041] The pollution probability calculation unit is used to calculate the pollution propagation probability of adjacent evidence points on the evidence flow timeline.

[0042] The evidence matching degree calculation unit is used to calculate the evidence matching degree between adjacent evidence points;

[0043] The risk value calculation unit calculates the pollution risk value for each evidence point based on the probability of pollution transmission and the degree of evidence matching.

[0044] Furthermore, the pollution path tracing module includes: a target pollution point identification unit, an upstream evidence tracing unit, and a path tree construction unit;

[0045] The target contamination point identification unit determines the target contamination point based on a preset contamination risk threshold.

[0046] The upstream evidence tracing unit is used to retrieve the source path of the target contamination point and capture upstream related evidence in the source path of the target contamination point to construct a set of related evidence.

[0047] The path tree construction unit constructs a pollution propagation path tree by recursively tracing the upstream paths of each related piece of evidence.

[0048] Furthermore, the pollution source location module includes: a propagation range calculation unit, a scoring function construction unit, and a pollution source location unit;

[0049] The spread range calculation unit counts the number of upper-level related evidences for each related piece of evidence in the pollution spread path tree, and calculates the pollution spread range of each related piece of evidence in combination with the pollution risk value.

[0050] The scoring function construction unit combines evidence matching degree, pollution transmission probability and pollution transmission range to construct a pollution source scoring function;

[0051] The pollution source location unit determines the pollution source of the evidence by judging the pollution source score of the relevant evidence.

[0052] Compared with the prior art, the beneficial effects of the present invention are:

[0053] 1. Precise Tracing of Pollution Sources: By using reverse positioning technology to trace back from the target evidence to its related evidence, this invention can deeply explore the source of evidence and accurately locate the source of pollution.

[0054] 2. Quantitative assessment of pollution risk: The pollution risk value is calculated by calculating indicators such as the probability of pollution transmission and the degree of evidence matching. The assessment also takes into account factors such as time interval, number of operations, and content overlap, making the pollution risk assessment more scientific and accurate, and overcoming the ambiguity of traditional manual assessment.

[0055] 3. Visualization of pollution transmission paths: By constructing a pollution transmission path tree, the hierarchical transmission relationship from the pollution source to the target pollution evidence is presented intuitively, which facilitates targeted measures to reduce the scope of pollution. Attached Figure Description

[0056] Figure 1 This is a schematic diagram of the method flow of an evidence contamination detection system and method based on reverse mapping localization according to the present invention;

[0057] Figure 2 This is a schematic diagram of the system structure of an evidence contamination detection system and method based on reverse mapping localization according to the present invention. Detailed Implementation

[0058] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Example 1: As Figure 1 As shown, the present invention provides a technical solution, a method for detecting evidence contamination based on reverse mapping localization, the method comprising:

[0060] Step S100: Retrieve all operation records of evidence generation instructions from the operation log of the evidence storage system; extract information from each operation record, and trace the source path of each piece of evidence based on the extracted operation record information;

[0061] Step S100 includes:

[0062] Step S101: Retrieve all operation records of evidence generation instructions from the operation log of the evidence storage system. The operation records include evidence derivation, updating, and referencing.

[0063] Step S102: In the operation record of each evidence generation instruction, capture the generated target evidence; in the operation record of each evidence generation instruction, capture all related evidence called when generating the target evidence, and obtain evidence information for the related evidence, the evidence information including the evidence source of the related evidence; the evidence source of the related evidence includes the evidence chain that generated the related evidence in the evidence storage system;

[0064] Step S103: In the operation record of each evidence generation instruction, capture the operation record Re of the associated evidence when the target evidence is generated. k Where k = 1, 2, or 3; the operation record includes derived operations Re. 1 Update operation Re 2 Reference operation Re 3 Among them, if the operation record of the target evidence is a derived operation Re 1 This indicates that new evidence was obtained through analysis of related evidence; this new evidence is the target evidence. If the operation record of the target evidence is an update operation Re... 2 This indicates that new evidence has been obtained by updating the relevant evidence based on the latest research data where there are discrepancies; if the operation record of the target evidence is a citation operation Re... 3 This indicates that the research findings or data of the related evidence were used by the target evidence;

[0065] Step S104: Capture and mark the operation records of all evidence in the evidence chain of the associated evidence in the evidence storage system to obtain the evidence chain ORG; based on the operation records between the target evidence and the associated evidence, the operation records between the associated evidence and the evidence chain, and the operation records between each piece of evidence in the evidence chain, generate the source path L of the target evidence through reverse mapping. A {A→Re k →B→Re k →ORG};wherein, L A B represents the source path of target evidence A, and B represents related evidence of evidence A;

[0066] Step S200: The source path of each piece of evidence is sorted out and an evidence flow timeline is generated; the pollution propagation probability and evidence matching degree of each evidence point on the evidence flow timeline are calculated; and the pollution risk value of each evidence point is calculated based on the pollution propagation probability and evidence matching degree.

[0067] Step S200 includes:

[0068] Step S201: Sort all evidence along the source path of the target evidence according to time sequence to generate an evidence flow timeline; the set of evidence points on the evidence flow timeline is E={e1,…,ei ,…,e n}; where each piece of evidence is marked with a timestamp t i e i Let e ​​represent the i-th evidence point. n Indicate the source of evidence; calculate the probability of contamination transmission between two adjacent evidence points. Among them, P i The probability of contamination propagation between the i-th evidence point and the (i+1)-th evidence point on the evidence flow timeline is represented by N, where N represents the total number of times the i-th evidence point is operated on in the operation log of the evidence storage system, and t represents the probability of contamination propagation between the i-th evidence point and the (i+1)-th evidence point on the evidence flow timeline. j t j+1 These represent evidence point e respectively. j e j+1 The timestamp marked, k This represents the operation weight corresponding to the operation record; where k = 1, 2, or 3.

[0069] Step S202: Obtain the complete content of each evidence point in the evidence flow timeline. The complete content of each piece of evidence includes text, chart annotations, and data tables. Semantically vectorize the obtained complete content of each piece of evidence using natural language processing technology. Calculate the content overlap between the semantically vectorized evidence points using cosine similarity and normalize the overlap. Set the credibility of each evidence point based on the total number of operations performed on it in the evidence storage system's operation log; the higher the total number of operations, the higher the credibility. Calculate the evidence matching degree between each evidence point. Among them, M i S represents the degree of evidence matching between the i-th evidence point and the (i+1)-th evidence point. i C represents the degree of overlap between the i-th evidence point and the (i+1)-th evidence point. i Indicates the credibility of the i-th evidence point;

[0070] Step S203: Calculate the pollution risk value for each evidence point based on the probability of pollution transmission and the degree of evidence matching. ; where e→e i This represents all evidence points e that point to evidence point e. i ;

[0071] Step S300: Based on the pollution risk value of each evidence point, determine the target pollution point, trace the upstream evidence source of the target pollution point, and construct a pollution propagation path tree;

[0072] Step S300 includes:

[0073] Step S301: Preset a pollution risk threshold F, which is set based on the historical pollution data of the evidence storage system; traverse all evidence points on the evidence flow timeline and compare the pollution risk value Q of each evidence point with the pollution risk threshold F; if the pollution risk value Q≥F, then the evidence point is taken as the target pollution point.

[0074] Step S302: Retrieve the source path L for each target contamination point e. e {e→Re k →e i →Re k →ORG};Capture all relevant evidence in the upstream path of the target contamination point e along the source path, forming a set of relevant evidence {u1,u2,…,u}. n}; where each piece of related evidence is marked with an operation record and timestamp related to the target contamination point e; u n This represents the nth piece of evidence linking the target contamination point e;

[0075] Step S303: For each piece of related evidence in the set of related evidence, repeat the tracing logic of step S302 to generate a set of related evidence in the upstream path of each piece of related evidence; based on the operation records and timestamps marked by each piece of related evidence, construct a pollution propagation path tree from the target pollution point e to each layer of upstream path for visualizing the hierarchical relationship of pollution propagation;

[0076] Step S400: Calculate the pollution propagation range of the upper-level associated evidence of the target pollution point in the pollution propagation path tree, and construct a pollution source scoring function by combining the pollution propagation probability and evidence matching degree to locate the pollution source;

[0077] Step S400 includes:

[0078] Step S401 includes: starting from the root node target pollution point e of the pollution propagation path tree, traversing the associated evidence at each layer upwards, counting the number of lower-level associated evidence for each associated evidence in the pollution propagation path tree, and calculating the pollution propagation range of each associated evidence. ;in, i Q represents the extent of pollution propagation associated with the i-th piece of evidence in the pollution propagation path tree. i Let m represent the contamination risk value of the i-th related evidence, and m represent the number of lower-level related evidences of the i-th related evidence.

[0079] Step S402: Construct a pollution source scoring function by combining evidence matching degree and pollution transmission probability. The pollution source scoring function is used to score all related evidence on the pollution propagation path tree, the scores of all related evidence are sorted, and the related evidence with the highest score is marked. The marked related evidence is the pollution source.

[0080] Example 2: Figure 2 As shown, in order to better implement the above method, an evidence pollution detection system based on reverse mapping location is also provided. The detection system includes: an operation record processing module, a pollution risk calculation module, a pollution path tracing module, and a pollution source location module.

[0081] The operation log processing module is used to retrieve all operation records of evidence generation instructions in the operation log of the evidence storage system, extract information from each operation record, and sort out the source path of each piece of evidence based on the extracted operation record information;

[0082] The pollution risk calculation module analyzes the source path of each piece of evidence, generates an evidence flow timeline, and calculates the probability of pollution transmission, evidence matching degree, and pollution risk value.

[0083] The pollution path tracing module is used to identify target pollution points, trace upstream evidence sources of the target pollution points, and construct a pollution propagation path tree;

[0084] The pollution source location module calculates the pollution propagation range of the upper-level associated evidence of the target pollution point in the pollution propagation path tree, and constructs a pollution source scoring function by combining the pollution propagation probability and evidence matching degree to locate the pollution source.

[0085] The operation record processing module includes: an operation record retrieval unit, an associated evidence capture unit, an operation type marking unit, and a source path generation unit;

[0086] The operation record retrieval unit is used to retrieve all operation records of evidence generation instructions from the operation log of the evidence storage system.

[0087] The associated evidence capture unit captures associated evidence called during the generation of target evidence and obtains the evidence source of the associated evidence;

[0088] Operation type marking unit, used to mark the operation type of the target evidence on the associated evidence;

[0089] The source path generation unit is used to generate the source path of the target evidence through reverse mapping.

[0090] The pollution risk calculation module includes: a time axis generation unit, a pollution probability calculation unit, an evidence matching degree calculation unit, and a risk value calculation unit.

[0091] The timeline generation unit is used to generate a timeline of evidence flow.

[0092] The pollution probability calculation unit is used to calculate the pollution propagation probability of adjacent evidence points on the evidence flow timeline.

[0093] The evidence matching degree calculation unit is used to calculate the evidence matching degree between adjacent evidence points;

[0094] The risk value calculation unit calculates the pollution risk value for each evidence point based on the probability of pollution transmission and the degree of evidence matching.

[0095] The pollution path tracing module includes: a target pollution point identification unit, an upstream evidence tracing unit, and a path tree construction unit;

[0096] The target contamination point identification unit determines the target contamination point based on a preset contamination risk threshold.

[0097] The upstream evidence tracing unit is used to retrieve the source path of the target contamination point and capture upstream related evidence in the source path of the target contamination point to construct a set of related evidence.

[0098] The path tree construction unit recursively traces the upstream paths of each piece of related evidence to construct a pollution propagation path tree;

[0099] The pollution source location module includes: a propagation range calculation unit, a scoring function construction unit, and a pollution source location unit.

[0100] The spread range calculation unit counts the number of upper-level related evidences for each related piece of evidence in the pollution spread path tree, and calculates the pollution spread range of each related piece of evidence in combination with the pollution risk value.

[0101] The scoring function construction unit combines evidence matching degree, pollution transmission probability and pollution transmission range to construct a pollution source scoring function;

[0102] The pollution source location unit determines the pollution source of the evidence by judging the pollution source score of the relevant evidence;

[0103] In an embodiment of the present invention, the generation record of evidence A is retrieved from the operation log of the evidence storage system. The generation record records that evidence A is generated through derivative operation analysis of related evidence B. Evidence related to B references basic research data C in the evidence chain ORG through a reference operation. This basic research evidence C is a direct related evidence of B in the evidence chain. The source path of evidence A is generated through reverse mapping: A→Re1→B→Re3→ORG. The timestamps on the evidence chain are C(2022.1), B(2023.5), and A(2024.1), generating the evidence flow timeline: A(2024.1)→B(2023.5)→C(2022.1). According to the evidence flow timeline, A→B: the time interval is 8 months, and B is operated on N=50 times. 1 =3 Propagation probability P B→A =(3×8) / 50=0.48; B→C: time interval 16 months, C is operated on N=100 times, 3 =1 Propagation probability P B→C =(1×16) / 100=0.16; The content overlap between B and A is S i =0.7, the credibility of B is C i =0.8, M B =(0.7×0.8×3) / 8=0.21; The content overlap between C and B is S i =0.9, the confidence level of C is C i =0.9, M C =(0.9×0.9×1) / 16=0.05; therefore, the pollution risk value Q of evidence A is... A =0.48×(1-0.21)+0.16×(1-0.05)=0.53; The system's preset pollution risk threshold F=0.3, Q A =0.53>0.3, so A is marked as the target contamination point; tracing the upstream evidence chain, we capture related evidence B and C, and construct a contamination propagation path tree with target contamination point A as the root and related evidence B and C as branches; in the contamination propagation path tree, the number of related evidence below C is m=2, so the contamination propagation range of C is... C =0.16×2=0.32; In the contamination path tree, the number of lower-level related evidence for B is m=1, so the contamination propagation range of B is... B =0.48×1=0.48; G is calculated according to the pollution source scoring function. C =(1-0.05)×0.16×0.32=0.05, G B =(1-0.21)×0.48×0.48=0.18>0.05, therefore B is determined to be the source of evidence contamination.

[0104] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for detecting evidence contamination based on reverse mapping localization, characterized in that: The detection method includes: Step S100: Retrieve all operation records of evidence generation instructions from the operation log of the evidence storage system; extract information from each operation record, and trace the source path of each piece of evidence based on the extracted operation record information; Step S200: The source path of each piece of evidence is sorted out and an evidence flow timeline is generated; the pollution propagation probability and evidence matching degree of each evidence point on the evidence flow timeline are calculated; and the pollution risk value of each evidence point is calculated based on the pollution propagation probability and evidence matching degree. Step S300: Based on the pollution risk value of each evidence node, determine the target pollution point, trace the upstream evidence source of the target pollution node, and construct a pollution propagation path tree; Step S400: Calculate the pollution propagation range of the upper-level associated evidence of the target pollution point in the pollution propagation path tree, and construct a pollution source scoring function by combining the pollution propagation probability and evidence matching degree to locate the pollution source; Step S100 includes: Step S101: Retrieve all operation records of evidence generation instructions from the operation log of the evidence storage system. The operation records include evidence derivation, updating, and referencing. Step S102: In the operation record of each evidence generation instruction, capture the generated target evidence; in the operation record of each evidence generation instruction, capture all related evidence called when generating the target evidence, and obtain evidence information for the related evidence, the evidence information including the evidence source of the related evidence; the evidence source of the related evidence includes the evidence chain that generated the related evidence in the evidence storage system; Step S103: In the operation record of each evidence generation instruction, capture the operation record Re of the associated evidence when the target evidence is generated. k Where k = 1, 2, or 3; the operation record includes derived operations Re. 1 Update operation Re 2 Reference operation Re 3 Among them, if the operation record of the target evidence is a derived operation Re 1 This indicates that new evidence was obtained through analysis of related evidence; this new evidence is the target evidence. If the operation record of the target evidence is an update operation Re... 2 This indicates that new evidence has been obtained by updating the relevant evidence based on the latest research data where there are discrepancies; if the operation record of the target evidence is a citation operation Re... 3 This indicates that the research findings or data of the related evidence were used by the target evidence; Step S104: Capture and mark the operation records of all evidence in the evidence chain of the associated evidence in the evidence storage system to obtain the evidence chain ORG; based on the operation records between the target evidence and the associated evidence, the operation records between the associated evidence and the evidence chain, and the operation records between each piece of evidence in the evidence chain, generate the source path L of the target evidence through reverse mapping. A {A→Re k →B→Re k →ORG};wherein, L A B indicates the source path of target evidence A, and B indicates related evidence of target evidence A.

2. The evidence contamination detection method based on reverse mapping localization according to claim 1, characterized in that: Step S200 includes: Step S201: Sort all evidence along the source path of the target evidence according to time sequence to generate an evidence flow timeline; the set of evidence points on the evidence flow timeline is E={e1,…,e i ,…,e n }; where each piece of evidence is marked with a timestamp t i e i Let e ​​represent the i-th evidence point. n Indicate the source of evidence; calculate the probability of contamination transmission between two adjacent evidence points. Among them, P i The probability of contamination propagation between the i-th evidence point and the (i+1)-th evidence point on the evidence flow timeline is represented by N, where N represents the total number of times the i-th evidence point is operated on in the operation log of the evidence storage system, and t represents the probability of contamination propagation between the i-th evidence point and the (i+1)-th evidence point on the evidence flow timeline. j t j+1 These represent evidence point e respectively. j e j+1 The timestamp of the tag, w k This represents the operation weight corresponding to the operation record; where k = 1, 2, or 3. Step S202: Obtain the complete content of each evidence point in the evidence flow timeline. The complete content of each piece of evidence includes text, chart annotations, and data tables. Semantically vectorize the obtained complete content of each piece of evidence using natural language processing technology. Calculate the content overlap between the semantically vectorized evidence points using cosine similarity and normalize the overlap. Set the credibility of each evidence point based on the total number of operations performed on it in the evidence storage system's operation log; the higher the total number of operations, the higher the credibility. Calculate the evidence matching degree between each evidence point. Among them, M i S represents the degree of evidence matching between the i-th evidence point and the (i+1)-th evidence point. i C represents the degree of overlap between the i-th evidence point and the (i+1)-th evidence point. i Indicates the credibility of the i-th evidence point; Step S203: Calculate the pollution risk value for each evidence point based on the probability of pollution transmission and the degree of evidence matching. ; where e→e i This represents all evidence points e that point to evidence point e. i .

3. The evidence contamination detection method based on reverse mapping localization according to claim 1, characterized in that: Step S300 includes: Step S301: Preset a pollution risk threshold F, which is set based on the historical pollution data of the evidence storage system; traverse all evidence points on the evidence flow timeline and compare the pollution risk value Q of each evidence point with the pollution risk threshold F; if the pollution risk value Q≥F, then the evidence point is taken as the target pollution point. Step S302: Retrieve the source path L for each target contamination point e. e {e→Re k →e i →Re k →ORG}; Capture all relevant evidence in the upstream path of the target contamination point e along the source path, forming a set of relevant evidence {u1,u2,…,u}. n }; where each piece of related evidence is marked with an operation record and timestamp related to the target contamination point e; u n This represents the nth piece of evidence linking the target contamination point e; Step S303: For each piece of associated evidence in the set of associated evidence, repeat the tracing logic of step S302 to generate a set of associated evidence in the upstream path of each piece of associated evidence; based on the operation records and timestamps marked by each piece of associated evidence, construct a pollution propagation path tree from the target pollution point e to each layer of upstream path, for visualizing the hierarchical relationship of pollution propagation.

4. The evidence contamination detection method based on reverse mapping localization according to claim 1, characterized in that: Step 400 includes: Step S401 includes: starting from the root node target pollution point e of the pollution propagation path tree, traversing the associated evidence at each layer upwards, counting the number of lower-level associated evidence for each associated evidence in the pollution propagation path tree, and calculating the pollution propagation range of each associated evidence. Among them, g i Q represents the extent of pollution propagation associated with the i-th piece of evidence in the pollution propagation path tree. i Let m represent the contamination risk value of the i-th related evidence, and m represent the number of lower-level related evidences of the i-th related evidence. Step S402: Construct a pollution source scoring function by combining evidence matching degree and pollution transmission probability. The pollution source scoring function is used to score all related evidence on the pollution propagation path tree, the scores of all related evidence are sorted, and the related evidence with the highest score is marked. The marked related evidence is the source of pollution evidence.

5. A system for detecting evidence contamination based on reverse mapping localization, used to execute the method for detecting evidence contamination based on reverse mapping localization as described in any one of claims 1-4, characterized in that: The detection system includes: an operation record processing module, a pollution risk calculation module, a pollution path tracing module, and a pollution source location module; The operation record processing module is used to retrieve all operation records of evidence generation instructions in the operation log of the evidence storage system, extract information from each operation record, and sort out the source path of each piece of evidence based on the extracted operation record information; The pollution risk calculation module sorts out the source path of each piece of evidence, generates an evidence flow timeline, and calculates the probability of pollution transmission, evidence matching degree, and pollution risk value. The pollution path tracing module is used to identify target pollution points, trace upstream evidence sources of target pollution points, and construct a pollution propagation path tree; The pollution source localization module calculates the pollution propagation range of the upper-level associated evidence of the target pollution point in the pollution propagation path tree, and constructs a pollution source scoring function by combining the pollution propagation probability and the evidence matching degree to locate the pollution source.

6. The evidence contamination detection system based on reverse mapping localization according to claim 5, characterized in that: The operation record processing module includes: an operation record retrieval unit, an associated evidence capture unit, an operation type marking unit, and a source path generation unit; The operation record retrieval unit is used to retrieve all operation records of evidence generation instructions from the operation log of the evidence storage system. The associated evidence capturing unit captures associated evidence called during the generation of target evidence and obtains the evidence source of the associated evidence; The operation type marking unit is used to mark the operation type of the target evidence on the associated evidence; The source path generation unit is used to generate the source path of the target evidence through reverse mapping.

7. The evidence contamination detection system based on reverse mapping localization according to claim 5, characterized in that: The pollution risk calculation module includes: a time axis generation unit, a pollution probability calculation unit, an evidence matching degree calculation unit, and a risk value calculation unit; The timeline generation unit is used to generate an evidence flow timeline; The pollution probability calculation unit is used to calculate the pollution propagation probability of adjacent evidence points on the evidence flow timeline; The evidence matching degree calculation unit is used to calculate the evidence matching degree between adjacent evidence points; The risk value calculation unit calculates the pollution risk value for each evidence point based on the probability of pollution transmission and the degree of evidence matching.

8. The evidence contamination detection system based on reverse mapping localization according to claim 5, characterized in that: The pollution path tracing module includes: a target pollution point identification unit, an upstream evidence tracing unit, and a path tree construction unit; The target contamination point identification unit determines the target contamination point based on a preset contamination risk threshold. The upstream evidence tracing unit is used to retrieve the source path of the target contamination point and capture the upstream related evidence in the source path of the target contamination point to construct a set of related evidence. The path tree construction unit constructs a pollution propagation path tree by recursively tracing the upstream paths of each piece of related evidence.

9. The evidence contamination detection system based on reverse mapping localization according to claim 5, characterized in that: The pollution source location module includes: a propagation range calculation unit, a scoring function construction unit, and a pollution source location unit; The propagation range calculation unit counts the number of upper-level related evidences for each related evidence in the pollution propagation path tree, and calculates the pollution propagation range of each related evidence in combination with the pollution risk value. The scoring function construction unit combines evidence matching degree, pollution transmission probability and pollution transmission range to construct a pollution source scoring function; The pollution source location unit determines the pollution source of the evidence by judging the pollution source score of the associated evidence.

Images