Information tracing method, system and storage medium

Abstract

The embodiments of the present invention relate to an information tracing method, system, and storage medium. The method comprises: obtaining target subject information of the same type from multiple platforms, and preprocessing the target subject information to obtain a target text; creating a target graph for the target text; performing entity linking processing on the target graph to obtain a publisher set of the target text, wherein the publisher set carries information on the propagation path of the target text; and performing a traceability analysis on the publisher set to determine the target publisher of the target subject information. By tracking and analyzing subject information on multiple platforms across platforms, across time and space, the origin, evolution, and propagation path of the subject information can be determined, and comprehensive information reference and decision support can be provided to users. Thus, the technical effect of multi-platform information tracing can be achieved.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of information identification technology, and in particular to an information tracing method, system, and storage medium. Background Art

[0002] With the continuous development of internet technology, more and more people are using various platforms to publish content and share information, such as social media, blogs, and forums. Thematic information on these platforms (such as discussions about a topic, event, or product) is often widely discussed and forwarded. However, due to the differences in platforms and the dissemination of information, it is difficult to trace the source, evolution, and dissemination path of such thematic information.

[0003] At present, the technology for tracing the source of multi-platform thematic information has made certain progress, but there are still some challenges and shortcomings. The problem of platform differences: There are great differences in the information format, data structure, user behavior, etc. of different platforms, which brings certain difficulties to cross-platform tracing. The problem of data quality: There is a large amount of noise, false information and misinterpretation in the information on the Internet, which affects the accuracy and credibility of the tracing of multi-platform thematic information. The problem of data scale: The amount of information on the Internet is very large. How to effectively process and analyze this data is also a difficulty in tracing the source of multi-platform thematic information. Summary of the Invention

[0004] In view of this, in order to solve the technical problem of tracing the source of the multi-platform theme information mentioned above, an embodiment of the present invention provides an information tracing method, system and storage medium.

[0005] In a first aspect, an embodiment of the present invention provides an information tracing method, comprising:

[0006] Acquire target subject information of the same type from multiple platforms, and pre-process the target subject information to obtain target text;

[0007] Creating a target graph of the target text;

[0008] Performing entity linking processing on the target graph to obtain a publisher set of the target text, wherein the publisher set carries the propagation path information of the target text;

[0009] Perform source tracing analysis on the publisher set to determine the target publisher of the target subject information.

[0010] In one possible implementation, preprocessing the target subject information to obtain the target text includes:

[0011] Extracting data based on cosine similarity of the target subject information to obtain relevant data without missing information;

[0012] The associated data is sorted according to the time parameter to obtain the target text.

[0013] In one possible implementation, creating the target graph of the target text includes:

[0014] Performing entity extraction on the target text to obtain an entity extraction result of the target text;

[0015] Performing attribute extraction on the target text to obtain an attribute extraction result of the target text;

[0016] Based on the entity extraction result, performing entity relationship extraction on the target text to obtain a relationship extraction result of the target text;

[0017] A target graph of the target text is created based on the entity extraction results, the attribute extraction results, and the relationship extraction results.

[0018] In a possible implementation, performing entity linking on the target graph to obtain a publisher set of the target text includes:

[0019] Matching the entity extraction results in the target graph with the standard knowledge base to obtain a quasi-entity set in the standard knowledge base;

[0020] Performing score evaluation on all entities in the quasi-entity set to obtain a quasi-entity score set;

[0021] The entity corresponding to the highest score in the quasi-entity score set is taken as the target entity;

[0022] All target entities corresponding to the target text are taken as a publisher set.

[0023] In a possible implementation, performing source tracing analysis on the publisher set to determine the target publisher of the target topic information includes:

[0024] Obtain the number of comments and forwardings corresponding to each publisher in the publisher set, and obtain the number of propagation paths in the target text;

[0025] Perform weighted averaging on the number of comments, the number of reposts, and the number of propagation paths to obtain a traceability score corresponding to each publisher;

[0026] The traceability scores corresponding to each publisher in the publisher set are compared to determine the target publisher of the target subject information.

[0027] In a possible implementation, comparing the traceability score corresponding to each publisher in the publisher set to determine the target publisher of the target topic information includes:

[0028] Obtaining a first traceability score corresponding to a first publisher in the publisher set, and obtaining a second traceability score corresponding to a second publisher;

[0029] Obtaining a difference between the first traceability score and the second traceability score to obtain a traceability difference;

[0030] Determining whether the traceability difference is greater than a preset first threshold and determining a target publisher, wherein the first threshold represents the degree of traceability similarity between two publishers;

[0031] When the tracing difference is greater than the first threshold, the publisher with the larger tracing score is selected as the target publisher.

[0032] In one possible implementation, determining whether the traceability difference is greater than a preset first threshold and determining the target publisher further includes:

[0033] When the tracing difference is less than or equal to the first threshold, obtaining the number of first propagation paths of the first publisher and obtaining the number of second propagation paths of the second publisher;

[0034] Comparing the number of the first propagation paths with the number of the second propagation paths to determine a target publisher;

[0035] When the number of the first propagation paths is greater than or equal to the number of the second propagation paths, determining that the target publisher is the first publisher;

[0036] When the number of the first propagation paths is smaller than the number of the second propagation paths, the target publisher is determined to be the second publisher.

[0037] In one possible implementation, the method further includes:

[0038] The entity information of the target publisher corresponding to the target graph and the propagation path information are displayed through a visualization device.

[0039] In a second aspect, an embodiment of the present invention provides an information tracing system applied to the information tracing method described in the first aspect, comprising:

[0040] A preprocessing module is used to obtain target topic information of the same type from multiple platforms and preprocess the target topic information to obtain target text;

[0041] A graph creation module, used for creating a target graph of the target text;

[0042] An entity linking module is used to perform entity linking processing on the target graph to obtain a publisher set of the target text, wherein the publisher set carries the propagation path information of the target text;

[0043] The tracing module is used to perform tracing analysis on the publisher set to determine the target publisher of the target subject information.

[0044] In a third aspect, an embodiment of the present invention provides a storage medium, which stores one or more programs, and the one or more programs can be executed by one or more processors to implement any information tracing method described in the first aspect.

[0045] The information tracing solution provided by the embodiment of the present invention obtains target subject information of the same type from multiple platforms and pre-processes the target subject information to obtain a target text; creates a target graph for the target text; performs entity linking on the target graph to obtain a publisher set for the target text, the publisher set carrying the propagation path information of the target text; and performs traceability analysis on the publisher set to determine the target publisher of the target subject information. By tracking and analyzing subject information on multiple platforms across platforms, across time and space, the origin, evolution, and propagation path of the subject information is determined, providing users with comprehensive information reference and decision support. This solution can achieve the technical effect of multi-platform information tracing. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0047] Figure 1 A flowchart of an information tracing method provided by an embodiment of the present invention;

[0048] Figure 2 A schematic diagram of a flow chart of another information tracing method provided by an embodiment of the present invention;

[0049] Figure 3 A schematic diagram of a process for determining a target publisher provided by an embodiment of the present invention;

[0050] Figure 4 A schematic diagram of another process for determining a target publisher provided by an embodiment of the present invention;

[0051] Figure 5 A schematic diagram of the structure of an information tracing system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0052] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. 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 shall fall within the scope of protection of the present invention.

[0053] In the embodiments of the present invention, the terms "including" and "having" are intended to convey an open-ended, inclusive meaning and indicate that additional elements / components / etc. may be present in addition to the listed elements / components / etc.; the terms "first" and "second," etc., are used merely as labels and do not limit the quantity of their objects. Furthermore, the various elements and regions in the drawings are shown for schematic purposes only, and thus the present invention is not limited to the sizes or distances shown in the drawings.

[0054] To facilitate understanding of the embodiments of the present invention, specific embodiments will be further explained below with reference to the accompanying drawings. The embodiments do not limit the embodiments of the present invention.

[0055] Figure 1 A flow chart of an information tracing method provided by an embodiment of the present invention. Figure 1 The provided diagram shows that the information tracing method specifically includes:

[0056] S101. Acquire target subject information of the same type from multiple platforms, and pre-process the target subject information to obtain target text.

[0057] The present invention is applied to the field of information processing technology, and particularly to information tracing technology. By collecting information of the same type on multiple platforms, a target text is obtained. A target graph of the target text is then created, and a relationship graph of entities and attributes in the target text is constructed. Based on entity linking, a publisher set is obtained. After comparing and tracing the publisher sets, the target publisher is finally obtained, achieving the technical effect of tracing the source of the topic information on multiple platforms.

[0058] The "multi-platform" mentioned here can be understood as different information publishing platforms, such as social media, blogs, and forums. The target topic information here can be understood as text data that carries thematic characteristics. The "preprocessing" mentioned here can be understood as deduplication and information extraction of the target topic information.

[0059] Furthermore, target topic information of the same emotion category or the same topic stored on multiple platforms is obtained through search software or database, and then the obtained target topic information is deduplicated, de-imaged, de-noised and information extracted to obtain target text about the topic information.

[0060] S102: Create a target map of the target text.

[0061] The target graph mentioned here can be understood as a knowledge graph, which represents the relationship between entities, events, or events and entities in data information.

[0062] Furthermore, after obtaining the target text, relationship extraction is performed based on the entity data and event attributes contained in the organized target text, thereby realizing the association relationship between entities and events in the target text. This association relationship is represented in the form of a graph, and the target graph corresponding to the target text is obtained, preparing for the next step of analyzing the target publisher in the target graph.

[0063] S103. Perform entity linking processing on the target graph to obtain the publisher set of the target text. The publisher set carries the propagation path information of the target text.

[0064] The entity linking process mentioned here can be understood as the operation of linking the entity objects extracted from the target text to the corresponding correct entity objects in the knowledge base. Given the entity reference items, a set of candidate entity objects that meet the requirements are selected from the knowledge base, and then the reference items are linked to the correct entity objects through similarity calculation. The publisher set mentioned here can be understood as the set of entity linking results in the target graph, representing the correct entity obtained by each entity link in the target graph as publisher information.

[0065] Furthermore, entity linking is performed on all entities in the target graph, and entities with high similarity values ​​corresponding to each entity are found from the knowledge base as publisher information, thereby obtaining the publisher information of all entities. All the publisher information obtained is used as a publisher set, which contains a set of text propagation path information corresponding to each entity, in preparation for the next step of determining the target publisher.

[0066] S104: Perform source tracing analysis on the publisher set to determine the target publisher of the target subject information.

[0067] The traceability analysis mentioned here can be understood as the process of determining the source of information by using information such as the time of publication and forwarding volume of each publisher's text information. The target publisher here can be understood as the publisher source of the text forwarding information within the entire set of publishers included in the target graph.

[0068] Furthermore, based on the publisher set corresponding to all entities in the target graph, we determine whether each publisher is the source publisher. We analyze the source of the text publishing based on the publishing time and forwarding volume information of each publisher's corresponding text information, find the original publisher, and use it as the target publisher to achieve the purpose of tracing the source.

[0069] The information tracing solution provided by the embodiment of the present invention obtains target subject information of the same type from multiple platforms and pre-processes the target subject information to obtain the target text; creates a target graph for the target text; performs entity linking on the target graph to obtain a set of publishers of the target text, which carries information about the target text's dissemination path; and performs traceability analysis on the publisher set to determine the target publisher of the target subject information. By tracking and analyzing subject information on multiple platforms across platforms, across time and space, the origin, evolution, and dissemination path of the subject information is determined, providing users with comprehensive information reference and decision support. This solution can achieve the technical effect of multi-platform information tracing.

[0070] Figure 2 A flowchart of another information tracing method provided by an embodiment of the present invention. Figure 2 This is introduced based on the previous embodiment. Figure 2 The diagram provided, the information tracing method specifically includes:

[0071] S201. Obtain target topic information of the same type from multiple platforms.

[0072] S202: Extract data based on cosine similarity of target topic information to obtain relevant data without missing information.

[0073] S203: Sort the associated data according to the time parameter to obtain the target text.

[0074] The present invention is applied to the field of information processing technology, and particularly to information tracing technology. By collecting information of the same type on multiple platforms, a target text is obtained. A target graph of the target text is then created, and a relationship graph of entities and attributes in the target text is constructed. Based on entity linking, a publisher set is obtained. After comparing and tracing the publisher sets, the target publisher is finally obtained, achieving the technical effect of tracing the source of the topic information on multiple platforms.

[0075] The "multi-platform" mentioned here can be understood as different information publishing outlets, such as social media, blogs, and forums. The target topic information here can be understood as text data with a specific theme. Preprocessing here can be understood as deduplication and information extraction performed on the target topic information. Data extraction here can be understood as the process of feature extraction, which involves analyzing the content of the target topic information to extract topics. Linked data here can be understood as the data information obtained after data preprocessing and missing value processing, representing information such as the publisher, publication time, publication location, and publication content of the target topic information.

[0076] Furthermore, by searching software or databases to obtain target topic information of the same sentiment category or theme stored on multiple platforms, cosine similarity calculations are performed on the obtained target topic information, and preliminary data extraction is performed on the target topic information to eliminate missing information. The extracted data is then sorted by platform release time and topic extraction is performed to obtain the target text information.

[0077] In a possible example scenario, target topic information is collected from multiple platforms; crawlers are used to centrally collect the release status of a certain type of topic information content on major Internet platforms.

[0078] The collected content is shown in Table 1. At the same time, some labels are manually added to the content of the target subject information, usually recorded as Paper_Num, to clarify the content of the target subject information and provide support for subsequent filtering of the collected content.

[0079] Table 1

[0080]

[0081] After setting the target topic information collection format, we perform text similarity calculations on the target topic information. This involves calculating the cosine similarity between the collected target topic information and the pre-labeled target topic information. We also set a threshold parameter, A, with a range of [0, 1] and a default value of 0.8. We then filter the collected target topic information to identify those with low relevance.

[0082] The collected target subject information is recorded as Get_Num, i represents a target subject information in the target subject information. j represents a target subject information in Paper_Num, and the cosine similarity is calculated using Formula 1 to obtain cos(i,j):

[0083]

[0084] Calculate the cosine similarity between i and each of the marked Paper_Num and sum them up. As shown in formula 2:

[0085]

[0086] if If the value is greater than the threshold parameter A, the target subject information is retained to achieve the purpose of information filtering.

[0087] Statistics are collected for each target topic category. Missing values ​​in a category are recorded as Missing_Value, which ranges from [0, 1]. The threshold parameter B ranges from [0, 1]. If Missing_Value is greater than 0.8, the category is deleted. Missing values ​​are filled in by calculating the mean, mode, or [minimum, maximum] interval of the column values, among other methods.

[0088] The target subject information is sorted according to the publishing time and the collection time. When the publishing time is the same, the information is sorted in chronological order according to the collection time. This completes the content sorting of the target subject information, implements the preprocessing process of the target subject information, and obtains the target text.

[0089] S204: Perform entity extraction on the target text to obtain the entity extraction result of the target text.

[0090] S205: Extract attributes from the target text to obtain the attribute extraction result of the target text.

[0091] S206: Based on the entity extraction results, entity relationship extraction is performed on the target text to obtain a relationship extraction result of the target text.

[0092] S207: Create a target graph of the target text based on the entity extraction results, attribute extraction results, and relationship extraction results.

[0093] Entity extraction can be understood as the process of extracting information from the target text according to a set entity paradigm. Attribute extraction can be understood as the process of extracting information from the target text according to a set attribute paradigm. Relationship extraction can be understood as the process of extracting information from the target text according to a set relationship paradigm.

[0094] Furthermore, by performing entity extraction, attribute extraction and relationship extraction on the target text, the corresponding entity extraction results, attribute extraction results and relationship extraction results are obtained. The three are integrated to obtain the target graph, which represents the upstream and downstream relationship graph between the entity relationships and attribute relationships in the target text.

[0095] Optionally, there are multiple ways to extract entities. Entities, attributes, and relationships are extracted from the sorted target text using triples. Usually, a paradigm is set that can be expressed as the following triple:

[0096] (subject, predicate, object)

[0097] Among them, subject is the subject (also called the subject), whose value is usually an entity or event; predicate is the predicate (also called the predicate), whose value is usually a relationship or attribute; object is the object (also called the object), whose value can be an entity, event, concept, or ordinary value (such as a number, string, etc.). According to the set paradigm, data is extracted from the target text, and then the triples representing the target text are obtained.

[0098] The goal of attribute extraction is to collect attribute information about a specific entity from various information sources. For example, for a public figure, one can obtain information such as their nickname, birthday, nationality, and educational background from publicly available online information. There are various methods for calculating attribute similarity, including commonly used methods such as edit distance, set similarity, and vector-based similarity.

[0099] Edit distance: Levenstein, Wagner and Fisher, Edit Distance with Afine Gaps;

[0100] Set similarity calculation: Jaccard coefficient, Dice;

[0101] Vector-based similarity calculation: Cosine similarity, TFIDF similarity.

[0102] Pattern matching-based extraction methods, also known as rule-based extraction methods, are methods that extract entities and attributes from text based on a series of pre-constructed rules. This method first defines relevant extraction rules, such as defining relevant standard tags or writing regular expressions, and then matches these rules with the target text. The matching results are used to obtain the extracted entities and their attributes.

[0103] Entity-attribute extraction methods based on pattern matching can be divided into three types according to the method of defining patterns: extraction based on manual definition, extraction based on supervised learning, and extraction based on iteration. The manual definition method is based on the manual definition of a series of patterns manually defined by personnel with professional knowledge in related fields. The learning-based method first collects relevant corpus to form a large-scale corpus, and then automatically obtains patterns through standard unstructured example training to build a knowledge base with a large number of entity-attributes. The iteration-based method is to first define template tuples, and then iterate these template tuples to automatically generate patterns, thereby extracting entity-attributes.

[0104] The entity-attribute extraction method based on relationship classification transforms the attribute extraction problem into a relationship classification problem. First, the two extracted entities are considered as a sample, and the direct relationship between the entities is considered as a label. Then, by constructing sample features, the samples are classified based on these features, and the classification results are used as the relationship attributes between the entities. Relationship-based extraction methods are usually carried out with the help of machine learning methods, such as support vector machines (SVM) and neural networks. Classification models are learned by training on large corpora, thereby extracting entity-attributes. Relationship-based classification methods can be divided into remote supervision methods and fully supervised methods according to the construction method of their corpus. Remote supervision methods basically use machines to build the corpus, while fully supervised methods use humans to build the corpus.

[0105] Clustering-based entity-attribute extraction methods transform the attribute extraction problem into a clustering problem. First, entity feature vectors are constructed. These feature vectors are then clustered using relevant methods. The resulting clusters represent the entity's attributes. For example, weakly supervised clustering methods can be used for category attributes, while unsupervised clustering methods can be used for product attributes. The specific implementation methods are not detailed here.

[0106] After entity-attribute extraction of the target text, a series of discrete named entities are obtained. To obtain semantic topic information, it is also necessary to extract the association relationships between entities from the relevant text information, and connect the entities through relationships to form a network knowledge structure. Specific relationship extraction can be achieved through supervised entity relationship extraction, semi-supervised entity relationship extraction, unsupervised entity relationship extraction, and open entity relationship extraction.

[0107] Supervised learning methods train machine learning models based on labeled training data and then identify relationship types in test data. Supervised learning methods include rule-based methods, feature-based methods, and kernel function-based methods.

[0108] The rule-based method needs to summarize the corresponding rules or templates through manual or machine learning methods according to the different fields involved in the corpus to be processed, and use the template matching method to extract entity relations. The feature vector-based method extracts useful information (including lexical information and grammatical information) from the context of the relational sentence instance as features, constructs feature vectors, and trains the entity relationship extraction model by calculating the similarity of the feature vectors. Kernel-based entity relationship extraction methods, including word sequence kernel function methods, dependency tree kernel function methods, shortest path dependency tree kernel function methods, convolution tree kernel function methods, and their combined kernel function methods, can complement feature-based entity relationship extraction methods.

[0109] Bootstrapping-based semi-supervised entity relationship extraction methods summarize entity relationship sequence patterns from the context containing relationship seeds, and use relationship sequence patterns to discover more relationship seed instances to form a new relationship seed set. Co-learning-based methods use two conditionally independent feature sets to provide different and complementary information, thereby reducing labeling errors.

[0110] Supervised entity relationship extraction methods do not rely on annotated entity relationship corpus, including relationship instance clustering and relationship type word selection. Based on the context in which the entity pairs appear, entity pairs with high similarity are clustered into one category, and representative words are selected to label this relationship.

[0111] Open entity relationship extraction methods automatically complete the relationship type discovery and relationship extraction tasks. They use external domain-independent entity knowledge bases (such as DBPedia, YAGO, OpenCyc, FreeBase, or other domain knowledge bases) to map high-quality entity relationship instances to large-scale text. Training data is obtained from this text alignment method, and then supervised learning methods are used to solve the relationship extraction problem.

[0112] The extraction results of the three are obtained through a variety of entity extraction methods, attribute extraction methods and relationship extraction methods. A relationship graph is constructed based on the entity extraction results, attribute extraction results and relationship extraction results to obtain the target graph of the target text, preparing for the next step of analyzing the traceability of entities in the graph.

[0113] S208: Match the entity extraction results in the target graph with the standard knowledge base to obtain the quasi-entity set in the standard knowledge base.

[0114] S209 , performing score evaluation on all entities in the quasi-entity set to obtain a quasi-entity score set.

[0115] The standard knowledge base here can be understood as the correct entity relationship database stored by the system. The quasi-entity set here can be understood as the set of entities initially identified through matching. The score evaluation here can be understood as the process of calculating the score of each quasi-entity through a given operation and evaluating the correct entity based on the score.

[0116] After obtaining the target graph, the entity extraction results in the target graph are matched with the correct entities stored in the standard knowledge base. Those within the threshold range are determined to be quasi-entities, and the quasi-entity set corresponding to each entity is obtained. All entities in the target text are then evaluated and calculated to obtain a quasi-entity score set, preparing for the next step of evaluating the target entity.

[0117] S210: The entity corresponding to the highest score in the quasi-entity score set is taken as the target entity.

[0118] S211. Take all target entities corresponding to the target text as a publisher set.

[0119] Among them, the publisher set carries the propagation path information of the target text.

[0120] Furthermore, based on the quasi-entity score set, the entity with the highest score is determined and used as the final correct entity for the entity in the target graph. This correct entity is then used as the target entity. This method is used to judge all entities in the target graph, obtaining the set corresponding to the target entity as the target entity set. This target entity set is used as the publisher set, and the publisher set information in the target graph is obtained, preparing for the next step of determining the target publisher.

[0121] In one possible example scenario, consider the entity "Zhang San" in the target graph, with 1,000 forwarded text messages and a publication date of 2022-09-10. By comparing the entity "Zhang San," "Zhang Sanqiang," "Zhang Sanguo," "Zhang Sanli," and "Li San" stored in the standard knowledge base, a threshold of 70% is set. If the similarity between Zhang San and Li San is 50%, which is lower than the threshold, Li San is filtered out. If the similarity between Zhang Sanqiang and Zhang San reaches 80%, which is higher than the 70% threshold, Zhang Sanqiang is designated as a quasi-entity. Similarly, all quasi-entities in the standard knowledge base are obtained. Further matching is performed by comparing publication date and text message volume, confirming that the correct entity is Zhang San. A relationship graph is constructed for the text message published on 2022-09-10 with 1,000 forwarded text messages, thereby obtaining the target entity. The entire target graph is then matched to obtain the publisher set corresponding to all correct entities.

[0122] Optionally, after obtaining the target graph, the target graph is analyzed and mined. Through entity linking, a set of candidate entity objects is selected from the knowledge base based on the given entity referents. The referents are then linked to the correct entity objects through similarity calculations, and the entity with the highest referent is scored as the target entity. Entity linking is the process of linking entity objects extracted from unstructured data (such as text) or semi-structured data (such as tables) to the corresponding correct entity objects in the knowledge base.

[0123] S212. Obtain the number of comments and forwardings corresponding to each publisher in the publisher set, and obtain the number of propagation paths in the target text.

[0124] S213. Perform weighted averaging on the number of comments, the number of reposts, and the number of propagation paths to obtain a corresponding traceability score for each publisher.

[0125] The weighted average processing mentioned here can be understood as a weight calculation process. The traceability score mentioned here can be understood as the probability score of each publisher being the target publisher. The higher the traceability score, the greater the spread of the text information, and the more representative it is of the traceability publisher.

[0126] Furthermore, based on the constructed target graph, combined with the publisher's attributes such as the number of comments and the number of reposts, and at the same time analyzing the number of propagation paths of the content published in the publisher set, a traceability possibility calculation algorithm is constructed to comprehensively calculate the source of the publication of such target topic information, thereby providing support for the traceability of topic information.

[0127] The number of comments is recorded as Number_of_comments, the number of forwardings is recorded as Number_of_forwarding, and the number of propagation paths is recorded as Number_of_propagation_paths.

[0128] The traceability score is obtained as a comprehensive score as shown in Formula 3:

[0129] Composite score (composite_score) =

[0130] αNumber_of_comments+βNumber_of_forwarding+

[0131] γNumber_of_propagation_paths Formula 3

[0132] Where α, β, and γ are the proportional coefficients for the normalization of the target subject information, satisfying the condition: α + β + γ = 1.

[0133] S214: Compare the traceability scores corresponding to each publisher in the publisher set to determine the target publisher of the target subject information.

[0134] Furthermore, by comparing the traceability scores corresponding to each publisher in the publisher set, the highest score is found, and the publisher with the highest traceability score is used as the target publisher, thereby realizing the traceability processing of the target subject information and obtaining the traceability publisher information.

[0135] S215. Display the entity information and propagation path information of the target publisher corresponding to the target graph through a visualization device.

[0136] The visualization device mentioned here can be understood as a display interface or message reminder interface of a designated area.

[0137] Furthermore, after analyzing the target publisher of the target topic information, the entity information, attribute information and entity-attribute relationship information of the target graph are obtained according to the entity-attribute relationship corresponding to the target publisher, and displayed through a visualization device to provide users with clear traceability processing results, and provide a reference basis for subsequent real-time tracking and judgment of the forwarding direction of the target topic information.

[0138] The steps of determining the target publisher in step S214 specifically include:

[0139] S301: Obtain a first tracing score corresponding to a first publisher in a publisher set, and obtain a second tracing score corresponding to a second publisher.

[0140] The first publisher mentioned here can be understood as a publisher number in the publisher set. The first traceability score mentioned here can be understood as the traceability score calculated by the first publisher.

[0141] S302: Obtain the difference between the first traceability score and the second traceability score to obtain the traceability difference.

[0142] S303: Determine whether the traceability difference is greater than a preset first threshold, and determine the target publisher. The first threshold represents the degree of traceability similarity between two publishers.

[0143] The provenance difference mentioned here can be understood as the interpolation between the provenance scores corresponding to any two publishers in the publisher set. The first threshold mentioned here can be understood as the degree of provenance similarity between the two publishers.

[0144] Furthermore, two publishers are randomly found from the publisher set, their traceability scores are calculated respectively, and the interpolation of the traceability scores of the two is used as the basis for determining whether the two publishers are the target publishers.

[0145] S304: When the traceability difference is greater than a first threshold, the publisher with a larger traceability score is selected as the target publisher.

[0146] Furthermore, when the interpolation of the traceability scores corresponding to any two publishers in the publisher set is greater than the first threshold, it indicates that there is a great difference between the two publishers. The larger the traceability score, the earlier the corresponding publisher published and the more text information forwarded. The publisher with a large traceability score is selected as the target publisher.

[0147] In a possible example scenario, publisher 1 and publisher 2 are obtained from the publisher set. Through tracing calculation, the comprehensive score corresponding to the traceability score of publisher 1 is 1000 points, and the comprehensive score corresponding to the traceability score of publisher 2 is 3000 points. The first threshold is set to 500. Because 3000-1000=2000, the traceability difference of 2000 is much larger than the threshold of 500. It can be clearly judged that the number of forwardings, forwarding time or forwarding paths corresponding to publisher 2 is greater, and it is more likely to be the source of the target topic information. Publisher 2 is temporarily used as the target publisher. Through the same judgment method, all publishers in the publisher set are compared, and finally the publisher with the highest traceability score is used as the target publisher.

[0148] The step S303 for obtaining the target publisher specifically includes:

[0149] S401: When the traceability difference is less than or equal to a first threshold, obtain the number of first propagation paths of the first publisher and obtain the number of second propagation paths of the second publisher.

[0150] S402: Compare the number of the first propagation paths with the number of the second propagation paths to determine the target publisher.

[0151] S403: When the number of the first propagation paths is greater than or equal to the number of the second propagation paths, determine that the target publisher is the first publisher.

[0152] S404: When the number of the first propagation paths is less than the number of the second propagation paths, determine that the target publisher is the second publisher.

[0153] The number of propagation paths mentioned here can be understood as the total number of paths through which the target topic information is forwarded by multiple platforms and multiple publishers.

[0154] Furthermore, when the difference in the two tracing scores between two publishers in the publisher set is less than a first threshold, the difference in the publication time and the number of forwarding paths representing the text information of the two publishers is very small, and the target publisher cannot be determined based on the tracing score. By setting the priority of the number of propagation paths as a reference for further determining the target publisher, based on the comparison of the two propagation path numbers corresponding to the two publishers, the publisher with the larger number of propagation paths is selected as the target publisher, thus achieving tracing analysis of the target topic information.

[0155] In a possible example scenario, the number of propagation paths obtained from the publisher set for publisher A is 100, and the number of propagation paths obtained for publisher B is 40. The traceability score A corresponding to publisher A is calculated to be 1000, and the traceability score B corresponding to publisher B is calculated to be 1050. The first threshold is set to 200. Since 1050-1000=50, the traceability difference of 50 is less than the first threshold of 200. Due to the balance of the proportional coefficient, the true traceability results of publishers A and B cannot be determined by the comprehensive scores of the two. Here, the priority of the number of propagation paths is used as the basis for judging the target publisher. Since the number of propagation paths of publisher A is greater than the number of propagation paths of publisher B, it is judged that publisher A is closer to the target publisher than publisher B. Publisher A is taken as the target publisher, and the source publisher of the target topic information is obtained as publisher A, thereby achieving the technical effect of information traceability under multiple platforms.

[0156] Another information tracing method provided by an embodiment of the present invention collects target subject information on multiple platforms, and then obtains the target text through data extraction and time sorting; obtains the target graph corresponding to the target text through entity-attribute extraction of the target text, and then obtains the target entity in the target graph based on entity linking processing, and then obtains the publisher set; then calculates the traceability score of each publisher and compares it, and sets the number of propagation paths as priority to obtain the target publisher of the target subject information, and conducts cross-platform, cross-temporal and cross-space tracking and analysis of the subject information on multiple platforms to determine the origin, evolution and propagation path of the subject information, and provide users with comprehensive information reference and decision support, thereby achieving the technical effect of multi-platform information tracing.

[0157] Figure 5 A schematic diagram of the structure of an information tracing system provided by an embodiment of the present invention. Figure 5 According to the diagram provided, the information traceability system specifically includes:

[0158] A preprocessing module 51 is used to obtain target subject information of the same type from multiple platforms and preprocess the target subject information to obtain a target text;

[0159] A graph creation module 52 is used to create a target graph of the target text;

[0160] The entity linking module 53 is used to perform entity linking processing on the target graph to obtain the publisher set of the target text, which carries the propagation path information of the target text;

[0161] The source tracing module 54 is used to perform source tracing analysis on the publisher set to determine the target publisher of the target subject information.

[0162] The information tracing system provided in this embodiment can be as follows Figure 5 The information traceability system shown in can be executed as follows Figure 1-2 All steps of the information traceability method in the Figure 1-2 For details on the technical effects of the information tracing method shown, please refer to Figure 1-2 For the sake of brevity, the relevant description will not be repeated here.

[0163] An embodiment of the present invention further provides a storage medium (computer-readable storage medium). The storage medium stores one or more programs. The storage medium may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as read-only memory, flash memory, hard disk, or solid-state drive; and the memory may also include a combination of the aforementioned types of memory.

[0164] When one or more programs in the storage medium can be executed by one or more processors, the above-mentioned information tracing method executed on the information tracing device side can be implemented.

[0165] The processor is used to execute the information tracing program stored in the memory to implement the following steps of the information tracing method performed on the information tracing device side:

[0166] The same type of target topic information is obtained from multiple platforms, and the target topic information is preprocessed to obtain the target text; a target graph of the target text is created; entity linking is performed on the target graph to obtain a publisher set of the target text, which carries the propagation path information of the target text; and traceability analysis is performed on the publisher set to determine the target publisher of the target topic information.

[0167] Professionals should also be further aware that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0168] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be implemented using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0169] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An information tracing method, characterized in that: include: Acquire target subject information of the same type from multiple platforms, and pre-process the target subject information to obtain target text; Creating a target graph of the target text; Performing entity linking processing on the target graph to obtain a publisher set of the target text, wherein the publisher set carries the propagation path information of the target text; Performing a traceability analysis on the publisher set to determine the target publisher of the target topic information, including: obtaining the number of comments and forwardings corresponding to each publisher in the publisher set, and obtaining the number of propagation paths in the target text; performing a weighted average processing on the number of comments, the number of forwardings, and the number of propagation paths to obtain a traceability score corresponding to each publisher; and comparing the traceability scores corresponding to each publisher in the publisher set to determine the target publisher of the target topic information; Comparing the traceability scores corresponding to each publisher in the publisher set to determine the target publisher of the target topic information includes: obtaining a first traceability score corresponding to a first publisher in the publisher set, and obtaining a second traceability score corresponding to a second publisher; obtaining a difference between the first traceability score and the second traceability score to obtain a traceability difference; determining whether the traceability difference is greater than a pre-set first threshold to determine the target publisher, the first threshold representing the degree of traceability similarity between two publishers; when the traceability difference is greater than the first threshold, selecting the publisher with the larger traceability score as the target publisher; The determining whether the traceability difference is greater than a preset first threshold and determining the target publisher includes: When the traceability difference is less than or equal to the first threshold, obtain the first propagation path number of the first publisher and the second propagation path number of the second publisher; compare the first propagation path number with the second propagation path number to determine the target publisher; when the first propagation path number is greater than or equal to the second propagation path number, determine that the target publisher is the first publisher; when the first propagation path number is less than the second propagation path number, determine that the target publisher is the second publisher.

2. The method according to claim 1, characterized in that The preprocessing of the target subject information to obtain the target text includes: Extracting data based on cosine similarity of the target subject information to obtain relevant data without missing information; The associated data is sorted according to the time parameter to obtain the target text.

3. The method according to claim 1, characterized in that The step of creating a target graph of the target text includes: Performing entity extraction on the target text to obtain an entity extraction result of the target text; Performing attribute extraction on the target text to obtain an attribute extraction result of the target text; Based on the entity extraction result, performing entity relationship extraction on the target text to obtain a relationship extraction result of the target text; A target graph of the target text is created based on the entity extraction results, the attribute extraction results, and the relationship extraction results.

4. The method according to claim 3, characterized in that The performing entity linking processing on the target graph to obtain a publisher set of the target text includes: Matching the entity extraction results in the target graph with the standard knowledge base to obtain a quasi-entity set in the standard knowledge base; Performing score evaluation on all entities in the quasi-entity set to obtain a quasi-entity score set; The entity corresponding to the highest score in the quasi-entity score set is taken as the target entity; All target entities corresponding to the target text are taken as a publisher set.

5. The method according to claim 1, wherein The method further comprises: The entity information of the target publisher corresponding to the target graph and the propagation path information are displayed through a visualization device.

6. An information tracing system applied to the information tracing method according to claim 1, characterized in that: include: A preprocessing module is used to obtain target topic information of the same type from multiple platforms and preprocess the target topic information to obtain target text; A graph creation module, used for creating a target graph of the target text; An entity linking module is used to perform entity linking processing on the target graph to obtain a publisher set of the target text, wherein the publisher set carries the propagation path information of the target text; A traceability module is used to perform traceability analysis on the publisher set to determine the target publisher of the target topic information, including: obtaining the number of comments and forwardings corresponding to each publisher in the publisher set, and obtaining the number of propagation paths in the target text; performing weighted average processing on the number of comments, forwardings, and propagation paths to obtain a traceability score corresponding to each publisher; and comparing the traceability scores corresponding to each publisher in the publisher set to determine the target publisher of the target topic information; The traceability module is specifically configured to obtain a first traceability score corresponding to a first publisher in the publisher set, and a second traceability score corresponding to a second publisher; obtain the difference between the first traceability score and the second traceability score to obtain a traceability difference; determine whether the traceability difference is greater than a pre-set first threshold, and determine a target publisher, where the first threshold represents the degree of traceability similarity between the two publishers; and when the traceability difference is greater than the first threshold, select the publisher with the larger traceability score as the target publisher. The tracing module is specifically used to obtain the first propagation path number of the first publisher and the second propagation path number of the second publisher when the tracing difference is less than or equal to the first threshold; compare the first propagation path number with the second propagation path number to determine the target publisher; when the first propagation path number is greater than or equal to the second propagation path number, determine that the target publisher is the first publisher; when the first propagation path number is less than the second propagation path number, determine that the target publisher is the second publisher.

7. A storage medium, characterized in that: The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the information tracing method according to any one of claims 1 to 5.

Images