Abstract generation method and system applied to semantic mining of newspaper and periodical articles

Abstract

According to the abstract generation method and system applied to semantic mining of the newspaper and periodical articles, the accuracy and efficiency of semantic association mining between the newspaper and periodical articles are remarkably improved through semantic extraction and matching analysis. Specifically, firstly, two articles are subjected to semantic extraction, respective semantic spaces are constructed, and semantic vector units which have great influence on semantic association of the articles are effectively recognized through weight coefficient analysis. Then, through semantic space optimization, a most representative semantic vector unit is further refined, and a feature identification technology is utilized to accurately locate the vector unit with semantic correlation between the two articles; finally, through deep semantic association analysis, an abstract result which accurately reflects the semantic association relationship between the two articles is generated, so that not only are the accuracy and comprehensiveness of abstract generation improved, but also the application efficiency in the fields of information retrieval, content recommendation, public opinion analysis and the like is greatly promoted.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a summary generation method and system for semantic mining of newspaper articles. Background Art

[0002] In the era of information explosion, newspapers, as traditional information dissemination media, still carry a large amount of valuable information. However, with the increasing number of newspaper articles, how to quickly and accurately mine key information from massive articles has become an urgent problem to be solved. Especially when it is necessary to perform semantic association analysis on multiple articles, traditional methods are often inefficient and difficult to accurately capture the semantic connections between articles.

[0003] Most of the existing methods for generating newspaper article summaries are based on keyword extraction or text statistical features. These methods may show certain effects when processing simple texts, but when faced with newspaper articles with complex semantic structures, it is often difficult to accurately understand the core content and semantic associations of the articles. In addition, these methods usually lack in-depth analysis of the semantic relationships between different articles, resulting in the generated summary results often only focusing on the content of a single article, and unable to effectively reflect the semantic connections between multiple articles. Summary of the invention

[0004] In view of the above-mentioned problems, in combination with the first aspect of the present application, an embodiment of the present application provides a summary generation method applied to semantic mining of newspaper articles, the method comprising: Performing semantic extraction on the first newspaper article to generate a first article semantic space, and performing semantic extraction on the second newspaper article to generate a second article semantic space; Performing semantic matching analysis on the first article semantic space to generate weight coefficients of each first semantic vector unit in the first article semantic space, and performing semantic matching analysis on the second article semantic space to generate weight coefficients of each second semantic vector unit in the second article semantic space, wherein the weight coefficients reflect the influence coefficients on the semantic association of newspaper articles; According to the weight coefficients of the first semantic vector units, the first article semantic space is optimized to generate x first semantic vector units, and according to the weight coefficients of the second semantic vector units, the second article semantic space is optimized to generate x second semantic vector units; Performing semantic matching analysis on the x first semantic vector units and the x second semantic vector units to generate first feature identifiers corresponding to the x first semantic vector units and second feature identifiers corresponding to the x second semantic vector units, wherein the first feature identifier is used to locate y first semantic vector units among the x first semantic vector units that have semantic relevance to the x second semantic vector units, and the second feature identifier is used to locate z second semantic vector units among the x second semantic vector units that have semantic relevance to the x first semantic vector units, wherein the first feature identifier and the second feature identifier are used to represent corresponding masks; A semantic association analysis is performed on the y first semantic vector units and the z second semantic vector units to generate a first semantic association summary result between the first newspaper article and the second newspaper article.

[0005] On the other hand, an embodiment of the present application also provides a summary generation system, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0006] Based on the above aspects, the embodiment of the present application significantly improves the accuracy and efficiency of semantic association mining between newspaper articles through semantic extraction and matching analysis. Specifically, semantic extraction is first performed on the two articles respectively, and their respective semantic spaces are constructed. Through weight coefficient analysis, the semantic vector units that have a significant impact on the semantic association of the articles are effectively identified. Subsequently, the most representative semantic vector units are further refined through semantic space optimization, and feature identification technology is used to accurately locate the vector units with semantic relevance between the two articles. Finally, through in-depth semantic relevance analysis, a summary result that accurately reflects the semantic association relationship between the two articles is generated, which not only improves the accuracy and comprehensiveness of summary generation, but also greatly promotes the application efficiency in fields such as information retrieval, content recommendation and public opinion analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 It is a schematic diagram of the execution flow of the summary generation method applied to semantic mining of newspaper articles provided in an embodiment of the present application.

[0008] Figure 2 Schematic diagram of the hardware architecture of the summary generation system provided in the embodiment of the present application. DETAILED DESCRIPTION

[0009] The present application will be described in detail below with reference to the accompanying drawings. Figure 1This is a flow chart of a summary generation method for semantic mining of newspaper articles provided by an embodiment of the present application. The summary generation method for semantic mining of newspaper articles is introduced in detail below.

[0010] Step S110 , performing semantic extraction on the first newspaper article to generate a first article semantic space, and performing semantic extraction on the second newspaper article to generate a second article semantic space.

[0011] Specifically, the first newspaper article / the second newspaper article are text objects to be analyzed, for example, one of them is a film review article about a popular movie (the first newspaper article), and the other is an entertainment news article about the starring actor of the movie (the second newspaper article). The article semantic space maps the semantic information of the newspaper article into a multi-dimensional vector space through semantic extraction, in which each word or phrase corresponds to a vector in the space.

[0012] That is, in this embodiment, for the first newspaper article (film review article), the server will first analyze each word in the first newspaper article. For example, the first newspaper article mentions the plot of the movie, such as "the protagonist encounters a mysterious magical creature during the adventure", and the server will identify the semantics of words such as "protagonist", "adventure", "mystery", and "magic creature", and map these words into a multidimensional semantic space, and each word corresponds to a vector in this space. For words that describe the movie scene, such as "gorgeous visual effects", semantic mapping is also performed. In this way, the semantics of the entire film review article are expressed in the form of a vector in this multidimensional space to form the first article semantic space. This semantic space contains various semantic information of the film review article, including evaluations of the movie plot, picture, performance, etc.

[0013] For the second newspaper article (entertainment news article), if it mentions "the leading actor recently participated in a charity event and shared interesting things about the filming of the movie", the server will perform semantic extraction on words such as "lead actor", "charity event", "filming", and "interesting things". These words and their semantic relationships are represented in another multidimensional space in the form of vectors to generate the second article semantic space. This second article semantic space reflects semantic information such as various activities of the leading actor and events related to the movie in entertainment news.

[0014] Step S120, performing semantic matching analysis on the first article semantic space to generate weight coefficients of each first semantic vector unit in the first article semantic space, and performing semantic matching analysis on the second article semantic space to generate weight coefficients of each second semantic vector unit in the second article semantic space, wherein the weight coefficients reflect the influence coefficients on the semantic association of newspaper articles.

[0015] In this embodiment, after the semantic vector units are optimized in the semantic space, x most representative and important semantic vector units are selected from the article semantic space as objects for subsequent feature identification and association analysis.

[0016] Taking the previous film review article (the first newspaper article) as an example, in the semantic space of the first article, semantic vector units related to "movie plot" may be more important in analyzing the overall semantic association of the film review. If most of the content in the film review revolves around the plot, then the semantic vector unit "movie plot" may be given a higher weight coefficient. For example, for the plot-related semantic vector unit "the protagonist encountered a mysterious magical creature during the adventure", the server analyzes its degree of association with other semantic units in the film review (such as the evaluation of the protagonist's acting skills, the theme expression of the film, etc.). If it is found that many evaluations of acting skills are based on this plot, then this plot-related semantic vector unit will have a greater impact on the semantic association of the entire film review article, and the weight coefficient may be relatively high.

[0017] For entertainment news articles (second newspaper articles), if the semantic vector unit "the leading actor shared interesting stories during the filming of the movie" has a strong correlation with other information about the leading actor in the news (such as the image shaping and public image of the leading actor, etc.), then its weight coefficient in the semantic space of the second article will be relatively high. For example, if this interesting story affects the image of the leading actor in the public's mind and echoes other relevant information, then the server will give it a higher weight coefficient to reflect its importance in the semantic association of this entertainment news.

[0018] Step S130, based on the weight coefficients of the first semantic vector units, the semantic space of the first article is optimized to generate x first semantic vector units, and based on the weight coefficients of the second semantic vector units, the semantic space of the second article is optimized to generate x second semantic vector units.

[0019] Still taking the film review article as an example, assume that the set semantic relevance threshold is based on the experience value obtained by the server in the past processing a large number of film review articles. If x = 5 is determined according to this threshold (the 5 here is just for the convenience of example). The server sorts the semantic vector units in the semantic space of the first article according to the weight coefficient of each first semantic vector unit, in descending order of the weight coefficient (that is, the higher the weight coefficient, the higher the ranking). Then select the first 5 first semantic vector units. For example, in a film review article, the first 5 semantic vector units with the highest weight coefficients may be semantic vector units related to the core plot of the movie, the wonderful performance of the protagonist, the unique theme expression of the movie, the impressive pictures, and the deep emotions conveyed by the movie.

[0020] For entertainment news articles, the same method is used. If x = 5 is determined based on the set semantic relevance threshold and the semantic space dimension of entertainment news articles, the first five second semantic vector units are selected in descending order according to the weight coefficient of each second semantic vector unit. These may be semantic vector units related to the main actor's important activities, the main actor's special contribution in the movie, the main actor's relationship with other actors, the main actor's image highlights, and unique events related to the movie.

[0021] Step S140, performing semantic matching analysis on the x first semantic vector units and the x second semantic vector units, generating a first feature identifier corresponding to the x first semantic vector units and a second feature identifier corresponding to the x second semantic vector units, wherein the first feature identifier is used to locate y first semantic vector units among the x first semantic vector units that have semantic relevance to the x second semantic vector units, and the second feature identifier is used to locate z second semantic vector units among the x second semantic vector units that have semantic relevance to the x first semantic vector units, wherein the first feature identifier and the second feature identifier are used to represent corresponding masks.

[0022] Take the 5 first semantic vector units in the previously selected film review article and the 5 second semantic vector units in the entertainment news article as examples. First, determine the first basic feature identifiers corresponding to the 5 first semantic vector units and the second basic feature identifiers corresponding to the 5 second semantic vector units. For the first basic feature identifier in the film review article, assuming that the film review article has added some vocabulary in the previous dimension conversion (for example, some professional terms in film production are added to better understand the content of the film review), then in the first basic feature identifier, the semantic vectors corresponding to these supplementary vocabulary are marked as non-core content. For example, if the word "montage technique" is added, the semantic vector corresponding to it in the first basic feature identifier is marked as non-core content.

[0023] Then, five first semantic vector units are selected according to the first basic feature identifier to generate five first semantic vector units to be optimized. For example, the relevant semantic vector units marked as non-core content are specially processed to obtain the first semantic vector units to be optimized. Similarly, a similar operation is performed on the entertainment news articles to obtain five second semantic vector units to be optimized.

[0024] Next, the server performs semantic relevance analysis on the five first semantic vector units to be tuned through the semantic relevance analysis module, and performs relevance analysis on the five second semantic vector units to be tuned and the semantic vector unit generated by the semantic relevance analysis module through the interactive relevance analysis module to generate five first adjusted semantic vector units. The same operation is performed on the entertainment news articles to generate five second adjusted semantic vector units.

[0025] For example, in a film review article, if a first semantic vector unit to be optimized is about the movie scenes, and after analysis it is found that it has a certain correlation with the interesting stories about the filming of the movie shared by the leading actor in the entertainment news (the content in the second adjusted semantic vector unit), then this first semantic vector unit to be optimized will be adjusted to the first adjusted semantic vector unit.

[0026] After that, semantic matching analysis is performed on the five first adjusted semantic vector units to generate confidence combinations corresponding to each first adjusted semantic vector unit. For example, for the first adjusted semantic vector unit about the plot of a movie, its association with each semantic vector unit in entertainment news is analyzed to obtain the confidence of semantically related content (for example, 0.8, indicating that it is likely to be related) and the confidence of non-semantically related content (for example, 0.2). Similar operations are performed on the five second adjusted semantic vector units.

[0027] Based on these confidence combinations, the five first semantic identification codes corresponding to the five first adjusted semantic vector units and the five second semantic identification codes corresponding to the five second adjusted semantic vector units are determined. For example, for a first adjusted semantic vector unit, if the confidence that it belongs to semantically related content is the highest, then the first semantic identification code may be the semantic content label "plot related".

[0028] Perform logical operations on the five first semantic identification codes to generate five first confidences, and perform logical operations on the five second semantic identification codes to generate five second confidences. Then, filter the five first confidences based on the first basic feature identification to generate a first adjusted feature identification; filter the five second confidences based on the second basic feature identification to generate a second adjusted feature identification. Finally, determine the first feature identification based on the first adjusted feature identification, and determine the second feature identification based on the second adjusted feature identification.

[0029] Step S150 , performing semantic association analysis on the y first semantic vector units and the z second semantic vector units, and generating a first semantic association summary result between the first newspaper article and the second newspaper article.

[0030] Assume that after the previous steps, y = 3 first semantic vector units with semantic relevance to the entertainment news article (second newspaper article) are determined in the film review article (first newspaper article), and z = 2 second semantic vector units with semantic relevance to the film review article are determined in the entertainment news article.

[0031] The server performs semantic relevance analysis on the three first semantic vector units (e.g., about the plot, the performance of the main character, and the theme of the movie) and the two second semantic vector units (e.g., about the contribution of the main character in the movie and the interesting stories shared by the main character during the filming). It analyzes the degree of semantic connection and logical relationship between these semantic vector units.

[0032] For example, a certain plot in a movie is directly related to the filming anecdotes shared by the leading actor, and there is also an intrinsic connection between the leading actor's performance and the leading actor's contribution to the movie. By comprehensively analyzing the relationship between these semantic vector units, the server generates a first semantic association summary result between the first newspaper article (film review article) and the second newspaper article (entertainment news article). This summary result may be a description that points out how the plot, performance, and theme in the film review article are related to the leading actor's contribution and filming anecdotes in the entertainment news article, such as "The wonderful plot mentioned in the film review is a reflection of the leading actor's contribution to the movie, the leading actor's wonderful performance echoes the filming anecdotes shared by the leading actor, and the theme of the movie is related to the leading actor's role positioning in the movie" and other similar semantic association descriptions.

[0033] Based on the above steps, the embodiment of the present application significantly improves the accuracy and efficiency of semantic association mining between newspaper articles through semantic extraction and matching analysis. Specifically, first, semantic extraction is performed on the two articles respectively, and their respective semantic spaces are constructed. Through weight coefficient analysis, the semantic vector units that have a significant impact on the semantic association of the articles are effectively identified. Subsequently, the most representative semantic vector units are further refined through semantic space optimization, and feature identification technology is used to accurately locate the vector units with semantic relevance between the two articles. Finally, through in-depth semantic relevance analysis, a summary result that accurately reflects the semantic association relationship between the two articles is generated, which not only improves the accuracy and comprehensiveness of summary generation, but also greatly promotes the application efficiency in fields such as information retrieval, content recommendation and public opinion analysis.

[0034] In a possible implementation, step S130 includes: Step S131, determining x according to a set semantic relevance threshold.

[0035] Step S132, according to the weight coefficients of the first semantic vector units and based on the order of decreasing the weight coefficients, the first x first semantic vector units in the semantic space of the first article are output as the x first semantic vector units, and according to the weight coefficients of the second semantic vector units and based on the order of decreasing the weight coefficients, the first x second semantic vector units in the semantic space of the second article are output as the x second semantic vector units.

[0036] In a possible implementation, step S131 includes: determining x according to the semantic relevance threshold and the dimension of the first article semantic space, or determining x according to the semantic relevance threshold and the dimension of the second article semantic space.

[0037] In this embodiment, the aforementioned film review article is still taken as the first newspaper article and the entertainment news article is taken as the second newspaper article.

[0038] The server has a set of established standards and procedures for determining x based on the set semantic relevance threshold when processing film review articles (first newspaper articles) and entertainment news articles (second newspaper articles). The semantic relevance threshold is a value pre-set by the server based on a large amount of text processing experience and the requirements for semantic analysis accuracy. This value reflects the lower limit of the degree of relevance that needs to be achieved between semantic vector units when optimizing the semantic space.

[0039] Assume that the semantic relevance threshold is set to a relative value, such as 0.6 (this is just an example value set for ease of understanding). When processing a film review article, the server first needs to determine the dimension of the semantic space of the first article. The dimension of the semantic space of the first article depends on the complexity and diversity of the semantic information in the film review article. For example, the film review article evaluates the plot, performance, picture, music, theme and other aspects in detail. Each aspect contains multiple semantic elements, which together constitute a multidimensional semantic space. Assume that after analysis, the dimension of the semantic space of the first article is 10 (this is just a hypothetical number of dimensions, and different values ​​may be obtained in practice based on specific semantic analysis).

[0040] The server determines x based on the semantic relevance threshold and the dimension of the semantic space of the first article. The specific calculation method may be a complex functional relationship. For example, it may be based on the proportional relationship between the semantic relevance threshold and the dimension, and the number of semantic vector units that meet the threshold requirement is calculated. Assume that according to this calculation method, x = 4 is calculated. This x = 4 means that 4 first semantic vector units will be selected when optimizing the semantic space of the first article.

[0041] The same process is used for entertainment news articles (second newspaper articles). Entertainment news articles may contain semantic information about the personal life of the protagonist, the development of his acting career, movie-related anecdotes, and other aspects. This information constitutes the semantic space of the second article. Assume that the dimension of the semantic space of the second article is 8. Also based on the pre-set semantic relevance threshold of 0.6, the value of x is obtained through a calculation method similar to that of the semantic space of the first article. Here, it is assumed that the calculated value of x = 3 (since the dimension of the semantic space of the second article is different from that of the semantic space of the first article, the calculated value of x may also be different).

[0042] After determining the value of x, it is necessary to output the first x first semantic vector units in the semantic space of the first article as x first semantic vector units based on the descending order of the weight coefficients of each first semantic vector unit. In the film review article, the weight coefficient of each first semantic vector unit reflects the importance of each semantic vector unit in the semantic association of the entire film review article. For example, the semantic vector unit about the plot of the movie may be given a higher weight coefficient because it is the core content of the film review. For example, the plot-related semantic vector unit "the protagonist encountered a mysterious magical creature during the adventure" has an important influence on the explanation of the theme and the shaping of the character image in the entire film review, so the weight coefficient is relatively high. The semantic vector unit about the small details of a supporting role in the movie has a relatively low weight coefficient because it has little influence on the semantic association of the overall film review.

[0043] The server sorts all the first semantic vector units in the semantic space of the first article in descending order according to the weight coefficients. Then the first x (here x = 4) first semantic vector units are selected as the optimized x first semantic vector units. Assume that the first four first semantic vector units are semantic vector units related to the core plot of the movie, the wonderful performance of the protagonist, the unique theme expression of the movie, and the impressive pictures.

[0044] Similarly, for the entertainment news article (the second newspaper article), according to the weight coefficient of each second semantic vector unit, based on the descending order of the weight coefficient, the first x second semantic vector units in the semantic space of the second article are output as x second semantic vector units. In the entertainment news article, for example, the semantic vector unit about the major performing arts activities of the leading actor (such as winning an important award) has a higher weight coefficient because it has a greater impact on the semantic association of the entire entertainment news article; while the semantic vector unit about a small episode of the leading actor at the event (such as the details of the interaction with a small fan) has a relatively low weight coefficient. The server arranges all the second semantic vector units in descending order according to the weight coefficient, and selects the first x (here x = 3) second semantic vector units as the optimized x second semantic vector units. Assume that these three second semantic vector units are semantic vector units related to the important performing arts activities of the leading actor, the special contribution of the leading actor in the movie, and the interesting shooting stories shared by the leading actor.

[0045] In this way, the process of determining x based on the semantic relevance threshold and optimizing the semantic space of the first article and the semantic space of the second article based on the weight coefficient is completed, and x first semantic vector units and x second semantic vector units are generated respectively. This process helps to focus on the part of the article that is most closely related to semantics, laying the foundation for subsequent more accurate semantic analysis and mining of the semantic relationship between the two articles.

[0046] In a possible implementation, step S140 includes: Step S141, determining the first basic feature identifiers corresponding to the x first semantic vector units, and determining the second basic feature identifiers corresponding to the x second semantic vector units.

[0047] Step S142: Filter the x first semantic vector units according to the first basic feature identifier to generate x first semantic vector units to be optimized, and filter the x second semantic vector units according to the second basic feature identifier to generate x second semantic vector units to be optimized.

[0048] Step S143, performing semantic association analysis on the x first semantic vector units to be optimized according to the first semantic association analysis module, and generating x first adjusted semantic vector units based on the x second semantic vector units to be optimized and the x semantic vector units generated by the first semantic association analysis module according to the first interactive association analysis module, and performing semantic association analysis on the x second semantic vector units to be optimized according to the second semantic association analysis module, and generating x second adjusted semantic vector units based on the x first semantic vector units to be optimized and the x semantic vector units generated by the second semantic association analysis module according to the second interactive association analysis module.

[0049] Step S144: Adjust the first basic feature identifier according to the x first adjustment semantic vector units to generate a first adjustment feature identifier, and adjust the second basic feature identifier according to the x second adjustment semantic vector units to generate a second adjustment feature identifier.

[0050] Step S145: determining the first feature identifier according to the first adjustment feature identifier, and determining the second feature identifier according to the second adjustment feature identifier.

[0051] In this embodiment, continuing to take the aforementioned film review article (first newspaper article) and entertainment news article (second newspaper article) as examples, the film review article obtains x first semantic vector units after the previous steps, and the entertainment news article also obtains x second semantic vector units.

[0052] First, determine the first basic feature identifiers corresponding to the x first semantic vector units, and determine the second basic feature identifiers corresponding to the x second semantic vector units. For the x first semantic vector units in the film review article, it is assumed that the x first semantic vector units are semantic vector units related to the core plot of the film, the wonderful performance of the protagonist, the unique theme expression of the film, and the impressive picture. When determining the first basic feature identifier, the server will take into account some preprocessing operations that may have been performed on the previous film review article or the structural characteristics of the article itself. For example, if some words about photography technology are added to the semantic vector units related to the film picture in order to better understand the semantic vector units related to the film picture, then in the first basic feature identifier, the part of the semantic vector units related to these supplementary words will be marked as auxiliary content or non-core content. For example, for the semantic vector unit related to the picture, if the word "long lens application" is added, then the semantic vector part corresponding to "long lens application" in the first basic feature identifier will be marked as non-core content to indicate its relatively secondary position in the entire semantic association.

[0053] For the x second semantic vector units in the entertainment news article, it is assumed that these x second semantic vector units are semantic vector units related to the important performing activities of the leading actor, the special contribution of the leading actor in the movie, and the interesting stories shared by the leading actor during the filming. Similar factors are also considered when determining the second basic feature identifier. If some words are added to supplement the performing background information of the leading actor when processing the entertainment news article, such as "experience studying at a drama academy", then in the second basic feature identifier, the semantic vector unit part corresponding to these supplementary words will be marked as non-core content.

[0054] Next, x first semantic vector units are screened according to the first basic feature identifier to generate x first semantic vector units to be optimized, and x second semantic vector units are screened according to the second basic feature identifier to generate x second semantic vector units to be optimized. For film review articles, the server selects the first semantic vector units marked as non-core content or with more auxiliary content according to the mark in the first basic feature identifier for special processing to form x first semantic vector units to be optimized. For example, if there are more non-core content marks corresponding to supplementary vocabulary in the semantic vector unit related to the picture when the first basic feature identifier is determined in front, then this semantic vector unit related to the picture may be selected into the first semantic vector unit to be optimized.

[0055] Entertainment news articles are also screened based on the second basic feature identifier. If there are more non-core content tags corresponding to the supplementary performance background information in the semantic vector unit related to the starring performance activity, then this semantic vector unit may be selected into the x second semantic vector units to be tuned.

[0056] Then, according to the first semantic association analysis module, a semantic association analysis is performed on the x first semantic vector units to be optimized, and according to the first interactive association analysis module, x second semantic vector units to be optimized and x semantic vector units generated by the first semantic association analysis module are used to generate x first adjusted semantic vector units, and according to the second semantic association analysis module, a semantic association analysis is performed on the x second semantic vector units to be optimized, and according to the second interactive association analysis module, x first semantic vector units to be optimized and x semantic vector units generated by the second semantic association analysis module are used to generate x second adjusted semantic vector units.

[0057] On the film review article side, the first semantic association analysis module will deeply analyze the semantic associations between the x first semantic vector units to be adjusted. For example, for the first semantic vector units to be adjusted related to the movie plot and the protagonist's performance, analyze their logical connections, semantic association levels, etc. in the film review article. At the same time, the first interactive association analysis module will consider the interactive relationship between the x second semantic vector units to be adjusted (from the entertainment news article) and the semantic vector units generated by the first semantic association analysis module. Assuming that the first semantic association analysis module finds that the first semantic vector unit to be adjusted related to the movie plot has a potential association with the filming fun shared by the protagonist in the entertainment news article (the content in the second semantic vector unit to be adjusted), then through further analysis and processing by the first interactive association analysis module, the first semantic vector unit to be adjusted related to the movie plot will be adjusted to make it more accurately reflect the semantic association with the entertainment news article, thereby generating one of the x first adjusted semantic vector units.

[0058] For entertainment news articles, the second semantic association analysis module performs semantic association analysis on the x second semantic vector units to be adjusted, such as analyzing the association between the two second semantic vector units to be adjusted, namely, the important performing activities of the leading actor and the special contribution of the leading actor in the movie. The second interactive association analysis module considers the interactive relationship between the x first semantic vector units to be adjusted (from the film review article) and the semantic vector unit generated by the second semantic association analysis module. For example, when analyzing the second semantic vector unit to be adjusted, namely, the special contribution of the leading actor in the movie, it is found that it is associated with the wonderful performance of the protagonist in the film review article (the content in the first semantic vector unit to be adjusted). Through the processing of the second interactive association analysis module, the semantic vector unit of the special contribution of the leading actor in the movie is adjusted, thereby generating one of the x second adjusted semantic vector units.

[0059] Afterwards, the first basic feature identifier is adjusted based on the x first adjusted semantic vector units to generate the first adjusted feature identifier, and the second basic feature identifier is adjusted based on the x second adjusted semantic vector units to generate the second adjusted feature identifier. For film review articles, the server will adjust the first basic feature identifier based on the new semantic relationship and content of the x first adjusted semantic vector units. For example, if a first adjusted semantic vector unit (such as one related to the plot of a movie) has a stronger association with the semantic vector unit in the entertainment news article after the previous adjustment, then in the first adjusted feature identifier, the part originally marked as non-core content may be adjusted according to the new semantic association. If an auxiliary vocabulary previously related to the plot (such as the specific name of a scene) becomes more important under the new semantic association, then the marking of the semantic vector part corresponding to this vocabulary in the first adjusted feature identifier will be adjusted from non-core content to relatively core content.

[0060] For entertainment news articles, the second basic feature identifier is also adjusted based on the x second adjusted semantic vector units. For example, if the second adjusted semantic vector unit of the special contribution of the leading actor in the movie has established a new and closer semantic connection with the semantic vector unit in the film review article after adjustment, then in the second adjusted feature identifier, the marked content related to this semantic vector unit (such as some specific details describing the contribution of the leading actor that were previously marked as non-core content) may be adjusted according to the new semantic connection.

[0061] Finally, the first feature identifier is determined based on the first adjusted feature identifier, and the second feature identifier is determined based on the second adjusted feature identifier. The first feature identifier is further determined based on the first adjusted feature identifier. It integrates all the previous adjustment and analysis results on the first semantic vector unit and can accurately locate the part of the x first semantic vector units that has semantic relevance with the x second semantic vector units. For example, in a film review article, the first feature identifier can clearly point out which first semantic vector units (such as semantic vector units for movie plots, protagonist performances, etc.) have strong semantic relevance with the semantic vector units in entertainment news articles.

[0062] Similarly, the second feature identifier is determined based on the second adjustment feature identifier, and it can accurately locate the part of the x second semantic vector units that has semantic relevance with the x first semantic vector units. For example, in an entertainment news article, the second feature identifier can point out the semantic relevance between semantic vector units such as the leading actor's important performing activities and the leading actor's special contribution in the movie and the semantic vector units in the film review article. Through such a process, the server completes the semantic matching analysis of the x first semantic vector units and the x second semantic vector units, and generates the corresponding first feature identifier and second feature identifier, which helps to further explore the semantic association relationship between the two articles.

[0063] In a possible implementation, before step S110, when the dimension of the first newspaper article and the dimension of the second newspaper article are inconsistent, the first newspaper article and the second newspaper article may be converted to the same dimension.

[0064] Step S141 includes: Step S1411, determining the first basic feature identifier based on the vocabulary supplemented in the dimensional conversion of the first newspaper article, wherein the mark corresponding to the semantic vector of the vocabulary supplemented in the x first semantic vector units in the first basic feature identifier is non-core content.

[0065] Step S1412, determining the second basic feature identifier based on the vocabulary supplemented in the dimensional conversion of the second newspaper article, wherein the mark corresponding to the semantic vector of the vocabulary supplemented in the x second semantic vector units in the second basic feature identifier is non-core content.

[0066] In this embodiment, the aforementioned film review article is still taken as the first newspaper article and the entertainment news article is taken as the second newspaper article.

[0067] Before performing semantic extraction on film review articles to generate the first article semantic space, and before performing semantic extraction on entertainment news articles to generate the second article semantic space, the server needs to first handle the inconsistency between the dimensions of the first newspaper article (film review article) and the second newspaper article (entertainment news article). Film review articles and entertainment news articles may have different semantic dimensions due to the different nature of their content, coverage, and emphasis of expression.

[0068] For example, a film review article may focus on the artistic aspects of the film, constructing semantic information from multiple artistic elements such as plot, performance, picture, music, etc. These elements are interrelated, forming a unique semantic dimension of the film review article. Entertainment news articles, on the other hand, focus more on events related to the artist, such as the activities of the leading actors, image creation, public reaction, etc., which constitute the semantic dimension of entertainment news articles. It is very likely that the semantic dimension of the film review article is composed of multiple sub-dimensions related to film art, while the semantic dimension of the entertainment news article is composed of multiple sub-dimensions related to the activities of the artist. The number and specific content of the sub-dimensions of the two may be different, resulting in inconsistency in the overall dimension.

[0069] When the server detects this dimensional inconsistency, it needs to convert the film review article and the entertainment news article to the same dimension. This conversion process is not a simple unification of the dimensional values, but a semantic mapping and supplementation method. For a film review article, if its semantic expression in a certain artistic element is relatively in-depth, and the entertainment news article lacks the corresponding semantic dimension in this aspect, the server may supplement some relevant common semantic elements in the semantic construction of the entertainment news article to align its dimension with the film review article to a certain extent. And vice versa.

[0070] Suppose a film review article has a very detailed description of the picture analysis, involving semantic information in multiple dimensions such as color matching and lens language, while an entertainment news article hardly mentions this aspect. In order to convert the two to the same dimension, the server will add some basic semantic elements about the movie picture in the semantic construction of the entertainment news article, such as simply mentioning whether the visual effect of the picture is attractive. Similarly, if the entertainment news article has a detailed description of the fan group of the leading actor, but the film review article does not involve this aspect, the server will add some semantic elements about audience reaction (a broad concept similar to fan group reaction) in the semantic construction of the film review article.

[0071] After completing such a dimensional conversion, the first basic feature identifiers corresponding to the x first semantic vector units are determined, and the second basic feature identifiers corresponding to the x second semantic vector units are determined. For the x first semantic vector units in the film review article, since some words are added during the dimensional conversion process, these added words have special significance in determining the first basic feature identifier. For example, when converting a film review article and an entertainment news article to the same dimension, in order to make the semantics of the film review article more fully aligned with the entertainment news article, some words about the response of the film market are added to the film review article (although the film review itself may seldom directly mention this aspect).

[0072] When determining the first basic feature identifier, for x first semantic vector units, if one of the semantic vector units has a semantic association with the supplementary vocabulary about the film market response, then in the first basic feature identifier, the part corresponding to the semantic vector of the supplementary vocabulary in this semantic vector unit will be marked as non-core content. For example, in a film review article, there is a first semantic vector unit about the expression of the film theme. In the process of dimension conversion, in order to align with the dimension of the film influence of the entertainment news article, some words about the box office performance of the film are supplemented. This box office performance vocabulary is associated with the film theme expression semantic vector unit, but in the semantic system of the entire film review article, compared with the in-depth explanation of the film theme itself, the relationship with the plot and performance, etc., the box office performance is a relatively non-core content, so in the first basic feature identifier, the semantic vector part corresponding to the supplementary vocabulary of box office performance will be marked as non-core content.

[0073] The situation is similar for the x second semantic vector units in the entertainment news article. Assume that when the entertainment news article and the film review article are converted to the same dimension, some basic information words about the film production team are added to the entertainment news article (because the film review article may involve some indirect evaluations of the production team, such as the impact of the director's style on the film, etc.). When determining the second basic feature identifier, for the x second semantic vector units, if a semantic vector unit has a semantic association with the supplemented words about the film production team, then in the second basic feature identifier, the part corresponding to the semantic vector of the supplemented words in this semantic vector unit will be marked as non-core content. For example, in the entertainment news article, there is a second semantic vector unit about the role of the leading actor in the movie. After the film production team-related words are added, a member of the production team (such as the producer) may have a certain influence on the role shaping, but compared with the core content such as the leading actor's own acting skills and the role's positioning in the plot, this part related to the producer will be marked as non-core content in the second basic feature identifier.

[0074] In this way, after processing the inconsistency of the dimensions of the first newspaper article and the second newspaper article and converting them, the first basic feature identifier and the second basic feature identifier are accurately determined according to the vocabulary supplemented in the dimension conversion, laying the foundation for further analysis of the semantic relationship between the two articles. This way of marking non-core content helps to focus more specifically on the core semantic part in the subsequent semantic analysis, and improves the accuracy of semantic matching and association analysis.

[0075] In a possible implementation, step S144 includes: Step S1441, perform semantic matching analysis on the x first adjusted semantic vector units to generate confidence combinations corresponding to each of the x first adjusted semantic vector units, and perform semantic matching analysis on the x second adjusted semantic vector units to generate confidence combinations corresponding to each of the x second adjusted semantic vector units, wherein the confidence combinations reflect the confidences of semantically related content and the confidences of semantically unrelated content. Determine the x first semantic identification codes corresponding to the x first adjusted semantic vector units based on the confidence combinations corresponding to the first adjusted semantic vector units, and determine the x second semantic identification codes corresponding to the x second adjusted semantic vector units based on the confidence combinations corresponding to the second adjusted semantic vector units, wherein the first semantic identification code reflects the semantic content label corresponding to the maximum confidence in the confidence combinations corresponding to the first adjusted semantic vector units, and the second semantic identification code reflects the semantic content label corresponding to the maximum confidence in the confidence combinations corresponding to the second adjusted semantic vector units. Performing logical operations on the x first semantic identification codes to generate x first confidences, and performing logical operations on the x second semantic identification codes to generate x second confidences. Screening the x first confidences based on the first basic feature identification to generate the first adjusted feature identification, and screening the x second confidences based on the second basic feature identification to generate the second adjusted feature identification.

[0076] Let’s continue with the example of the film review article mentioned above as the first newspaper article and the entertainment news article as the second newspaper article.

[0077] In this process, the server first performs a semantic matching analysis on the x first adjustment semantic vector units to generate a confidence combination corresponding to each of the x first adjustment semantic vector units, and simultaneously performs a semantic matching analysis on the x second adjustment semantic vector units to generate a confidence combination corresponding to each of the x second adjustment semantic vector units.

[0078] For the x first adjusted semantic vector units in the film review article, take the first adjusted semantic vector unit about the movie plot as an example. The server will perform a detailed semantic matching analysis on this vector unit and each semantic element in the entertainment news article. For example, a key plot in the movie plot may be associated with the shooting fun shared by the leading actor in the entertainment news article. The server will use a series of semantic analysis algorithms to determine the possibility of this association, thereby generating the confidence of semantically related content and the confidence of non-semantically related content. Assume that after analysis, the confidence of this movie plot being related to the shooting fun of the leading actor is 0.7 (indicating that it is more likely to be related), and the confidence of not belonging to semantically related content is 0.3. For the first adjusted semantic vector unit related to the movie theme, after semantic matching analysis with the content such as the role positioning of the leading actor in the movie in the entertainment news article, the confidence of belonging to semantically related content is 0.6, and the confidence of not belonging to semantically related content is 0.4, etc., so that the corresponding confidence combination is generated for each first adjusted semantic vector unit.

[0079] Similar operations are performed for the x second adjusted semantic vector units in the entertainment news article. For example, for the second adjusted semantic vector unit of the important performing arts activities of the leading actor, the server performs semantic matching analysis with each semantic element in the film review article. If the performing arts activities of the leading actor are matched and analyzed with the content related to the promotional effect of the film in the film review article, the confidence level of the semantically related content is 0.8, and the confidence level of the semantically related content is 0.2. For the second adjusted semantic vector unit of the special contribution of the leading actor in the film, after semantic matching analysis with the content related to the artistic achievements of the film in the film review article, the confidence level of the semantically related content is 0.7, and the confidence level of the semantically related content is 0.3, etc., thereby generating confidence combinations corresponding to each second adjusted semantic vector unit.

[0080] Next, x first semantic identification codes corresponding to x first adjusted semantic vector units are determined based on the confidence combination corresponding to each first adjusted semantic vector unit, and x second semantic identification codes corresponding to x second adjusted semantic vector units are determined based on the confidence combination corresponding to each second adjusted semantic vector unit. For the first adjusted semantic vector unit in the film review article, taking the confidence combination corresponding to the previous movie plot (the confidence of the semantically related content is 0.7, and the confidence of the non-semantically related content is 0.3) as an example, since the confidence of the semantically related content is higher, the corresponding first semantic identification code may be a semantic content label such as "plot-related to entertainment news", indicating that this first adjusted semantic vector unit is mainly identified by the plot and entertainment news-related features in the semantic association with the entertainment news article. For the first adjusted semantic vector unit related to the movie theme, the confidence of the semantically related content in its confidence combination is 0.6, and the confidence of the non-semantically related content is 0.4, and the corresponding first semantic identification code may be "theme-partially related to entertainment news".

[0081] For the second adjusted semantic vector unit in the entertainment news article, the confidence combination corresponding to the important performing activities of the leading actor (the confidence of semantically related content is 0.8, and the confidence of non-semantically related content is 0.2), the corresponding second semantic identification code may be "Performing activities-strongly related to film reviews". The confidence combination corresponding to the special contribution of the leading actor in the movie (the confidence of semantically related content is 0.7, and the confidence of non-semantically related content is 0.3), the corresponding second semantic identification code may be "Special contribution-relatively related to film reviews".

[0082] Then, the x first semantic identification codes are processed by logical operation to generate x first confidences, and the x second semantic identification codes are processed by logical operation to generate x second confidences. Assume that for the x first semantic identification codes in the film review article, the server adopts a weighted summation logical operation method (this is just an example method). If the first semantic identification code "plot-related to entertainment news" is assigned a higher weight, and "theme-partially related to entertainment news" is assigned a lower weight, a value is obtained after weighted summation calculation as one of the x first confidences. For the x second semantic identification codes in the entertainment news article, a similar logical operation method is also used. For example, the second semantic identification code "performing activities-strongly related to film reviews" corresponding to the important performing activities of the leading actor has a higher weight, and the second semantic identification code "special contribution-relatively related to film reviews" corresponding to the special contribution of the leading actor in the movie has a lower weight, and a value is obtained after weighted summation calculation as one of the x second confidences.

[0083] Afterwards, x first confidences are screened based on the first basic feature identifier to generate a first adjustment feature identifier, and x second confidences are screened based on the second basic feature identifier to generate a second adjustment feature identifier. For film review articles, the first basic feature identifier marks which are core contents and which are non-core contents. Assume that in the first basic feature identifier, the part related to the plot of the movie is marked as the core content, and the part related to the market response of the movie (the content supplemented in the previous dimension conversion) is marked as the non-core content. When screening x first confidences, if the first semantic identification code corresponding to a first confidence is related to the plot and has a higher weight in the logical operation, then this first confidence will be given priority consideration when generating the first adjustment feature identifier; and if the first semantic identification code corresponding to a first confidence is related to the market response of the movie and has a lower weight in the logical operation, then when screening according to the first basic feature identifier, it will have less impact on the first adjustment feature identifier.

[0084] For entertainment news articles, the core and non-core parts of the content related to the leading actor are marked in the second basic feature identifier. For example, the content related to the leading actor's acting skills is marked as the core content, and the gossip about the leading actor (the related content supplemented in the previous dimension conversion) is marked as the non-core content. When screening x second confidences, if the second semantic identification code corresponding to a second confidence is related to acting skills and has a higher weight in the logical operation, then this second confidence will be given priority consideration when generating the second adjustment feature identifier; and if the second semantic identification code corresponding to a second confidence is related to gossip and has a lower weight in the logical operation, then when screening according to the second basic feature identifier, it will have less impact on the second adjustment feature identifier.

[0085] Alternatively, step S1442, according to the x first adjustment semantic vector units, the x second adjustment semantic vector units, the first adjustment feature identifier and the second adjustment feature identifier, the x first adjustment semantic vector units are adjusted to generate x third adjustment semantic vector units, and according to the x first adjustment semantic vector units, the x second adjustment semantic vector units, the first adjustment feature identifier and the second adjustment feature identifier, the x second adjustment semantic vector units are adjusted to generate x fourth adjustment semantic vector units. According to the x third adjustment semantic vector units, the first adjustment feature identifier is adjusted to generate a third adjustment feature identifier, and according to the x fourth adjustment semantic vector units, the second adjustment feature identifier is adjusted to generate a fourth adjustment feature identifier. The first feature identifier is determined according to the third adjustment feature identifier, and the second feature identifier is determined according to the fourth adjustment feature identifier, the first feature identifier is the adjustment feature identifier obtained when the number of adjustment cycles of the first basic feature identifier is not less than the set order, and the second feature identifier is the adjustment feature identifier obtained when the number of adjustment cycles of the second basic feature identifier is not less than the set order.

[0086] In another case, based on the x first adjusted semantic vector units, the x second adjusted semantic vector units, the first adjusted feature identifiers, and the second adjusted feature identifiers, the x first adjusted semantic vector units are adjusted to generate x third adjusted semantic vector units, and the x second adjusted semantic vector units are adjusted to generate x fourth adjusted semantic vector units. For the x first adjusted semantic vector units in the film review article, assuming that the first adjusted semantic vector unit related to the movie plot, based on the shooting anecdotes it shares with the leading actors in the entertainment news article (from the x second adjusted semantic vector units) and the current first adjusted feature identifier (for example, marking the importance of the plot-related part), if it is found that a certain detail in the plot has a deeper semantic association with the shooting anecdotes but has not been fully reflected before, then the first adjusted semantic vector unit related to the movie plot is adjusted to generate one of the x third adjusted semantic vector units.

[0087] For the x second adjusted semantic vector units in the entertainment news article, for example, the second adjusted semantic vector unit of the special contribution of the leading actor in the movie, according to its relationship with the movie theme in the film review article (from the x first adjusted semantic vector units) and the current second adjusted feature identifier (for example, marking the position of the special contribution in the overall semantics), if it is found that the special contribution has a new semantic connection with the movie theme that has not been mined, this semantic vector unit is adjusted to generate one of the x fourth adjusted semantic vector units.

[0088] Finally, the first adjustment feature identifier is adjusted based on the x third adjustment semantic vector units to generate the third adjustment feature identifier, and the second adjustment feature identifier is adjusted based on the x fourth adjustment semantic vector units to generate the fourth adjustment feature identifier. For the film review article, based on the new semantic relationship and content in the x third adjustment semantic vector units, if a third adjustment semantic vector unit (such as movie plot-related) has new important information in the semantic association with the entertainment news article, then in the first adjustment feature identifier, the original mark of the plot-related part (such as the degree of importance, etc.) will be adjusted according to this new situation to generate the third adjustment feature identifier.

[0089] For entertainment news articles, based on the new semantic relationships and contents in the x fourth adjusted semantic vector units, if a fourth adjusted semantic vector unit (such as one related to the special contribution of the leading actor in the movie) has a new situation in the semantic association with the film review article, then in the second adjusted feature identifier, the original mark of the special contribution related part will be adjusted according to the new situation to generate the fourth adjusted feature identifier. The first feature identifier is determined based on the third adjusted feature identifier, and the second feature identifier is determined based on the fourth adjusted feature identifier. Here, the first feature identifier is an adjusted feature identifier obtained after the first basic feature identifier has been adjusted multiple times (the number of adjustment cycles is not less than the set order), and the second feature identifier is an adjusted feature identifier obtained after the second basic feature identifier has been adjusted multiple times (the number of adjustment cycles is not less than the set order). They can more accurately locate the semantic correlation between the x first semantic vector units and the x second semantic vector units.

[0090] In a possible implementation, step S110 includes: performing semantic extraction on the first newspaper article to generate the first article semantic space and the third article semantic space, and performing semantic extraction on the second newspaper article to generate the second article semantic space and the fourth article semantic space, the dimension of the first article semantic space is different from the dimension of the third article semantic space, and the dimension of the second article semantic space is different from the dimension of the fourth article semantic space.

[0091] Wherein, after step S150, the method further includes: Step S160, based on the first semantic association summary result, the dimension of the first article semantic space, the dimension of the third article semantic space and the third article semantic space, perform semantic space optimization on the third article semantic space to generate y newspaper article fragments corresponding to the y first semantic vector units; and based on the first semantic association summary result, the dimension of the second article semantic space, the dimension of the fourth article semantic space and the fourth article semantic space, perform semantic space optimization on the fourth article semantic space to generate z newspaper article fragments corresponding to the z second semantic vector units.

[0092] Step S170, semantically enhancing the y newspaper article fragments based on the y newspaper article fragments and the z newspaper article fragments to generate y enhanced fragments, and semantically enhancing the z newspaper article fragments based on the y newspaper article fragments and the z newspaper article fragments to generate z enhanced fragments.

[0093] Step S180, extracting the core semantics of the y enhanced segments to generate y core semantics, and performing semantic compression or expansion processing on the z enhanced segments based on the semantic depth error between the third article semantic space and the fourth article semantic space to generate z adjusted segments.

[0094] Step S190, performing semantic positioning processing on the z adjustment segments, generating z positioning semantics corresponding to the z adjustment segments, performing semantic association analysis on the y core semantics and the z positioning semantics, and generating a second semantic association summary result between the first newspaper article and the second newspaper article.

[0095] Wherein, step S190 includes: Step S191: Obtain the semantic importance coefficient of each semantic unit in the first adjustment segment of the z adjustment segments.

[0096] Step S192: weightedly fuse the semantic importance coefficient and semantic content of each semantic unit in the first adjusted segment to generate the positioning semantics of the first adjusted segment.

[0097] Let’s still take the aforementioned film review article as the first newspaper article and the entertainment news article as the second newspaper article as an example.

[0098] When the server performs semantic extraction on the first newspaper article (film review article), it will generate the first article semantic space and the third article semantic space. Film review articles contain rich and diverse semantic information, and due to different semantic analysis angles and requirements, semantic spaces of different dimensions will be formed.

[0099] For the generation of the semantic space of the first article, the server will focus on the main semantic elements in the film review article, such as evaluations of the plot, performance, picture, music, etc. It will analyze the semantic relationship between each word and sentence, and map these elements into a multidimensional space. For example, the semantics of the plot may form a subspace in this space, which contains semantic vectors such as the story's beginning, development, turning point, and ending, the rationality of the plot, and the relevance of the plot to the theme; the performance aspect includes semantic vectors such as the actor's interpretation of the role, the level of acting skills, and cooperation with other actors. These semantic vectors together constitute the semantic space of the first article, which focuses on the semantic expression of the core content of the film review article.

[0100] The generation of the semantic space of the third article may start from a different perspective. It is assumed that it will pay more attention to the background information in the film review article, the relevant cultural elements cited, or some supplementary content that assists in understanding the film review. For example, the semantic relationship construction of the relevant plot of the original novel from which the film was adapted (although not the core of the film review, it serves as supplementary information), the cultural background involved in the film (such as the influence of cultural customs in a specific region on the plot of the film), etc. In this way, the dimension of the semantic space of the third article is different from that of the semantic space of the first article, because they cover different semantic focuses and angles.

[0101] Similarly, for the second newspaper article (entertainment news article), the server generates the second article semantic space and the fourth article semantic space when performing semantic extraction. The second article semantic space focuses on the core content in the entertainment news article, such as the semantic information of the starring actor's performing activities, public image, and relationship with other artists. For example, the starring actor's performing activities may include recent movies and variety shows, etc. These contents form their own semantic vectors in the semantic space, such as the type of movies, role positioning in variety shows, etc., which together constitute the second article semantic space.

[0102] The fourth article semantic space is constructed from a different semantic perspective, and may focus on some peripheral information in the entertainment news article, such as the background of the source of the report, the time of the news release and the relationship between the overall environment of the entertainment industry at that time, etc. This makes the fourth article semantic space different from the second article semantic space in dimension, and it contains some auxiliary or background semantic elements.

[0103] After the first semantic association summary result between the first newspaper article and the second newspaper article is generated, a series of operations are performed next.

[0104] First, based on the first semantic association summary result, the dimension of the first article semantic space, the dimension of the third article semantic space, and the third article semantic space, the semantic space of the third article is optimized to generate y newspaper article fragments corresponding to y first semantic vector units. The first semantic association summary result contains the main semantic association information between the film review article and the entertainment news article. For example, the first semantic association summary result shows that the movie plot in the film review article is strongly associated with the performance of the leading actor in the movie in the entertainment news article. Based on this result, when optimizing the semantic space of the third article, the server will focus on the background information or supplementary content related to the movie plot. If there are cultural background elements related to the movie plot in the third article semantic space (for example, cultural elements that the movie plot takes place in a certain historical period), and these elements have a certain potential connection with the entertainment news article in the first semantic association summary result, the server will select these parts and generate y newspaper article fragments corresponding to y first semantic vector units.

[0105] At the same time, based on the first semantic association summary result, the dimension of the second article semantic space, the dimension of the fourth article semantic space, and the fourth article semantic space, the semantic space of the fourth article is optimized to generate z newspaper article fragments corresponding to the z second semantic vector units. For example, the first semantic association summary result shows that the image of the leading actor in the entertainment news article has some correlation with the theme of the movie in the film review article. Then when optimizing the fourth article semantic space, the server will search for background information related to the image of the leading actor (such as the relationship between the image strategy of the leading actor at the time of the news release and the overall atmosphere of the entertainment industry at that time), select the relevant parts, and generate z newspaper article fragments corresponding to the z second semantic vector units.

[0106] Next, semantic enhancement is performed on the y newspaper article segments according to the y newspaper article segments and the z newspaper article segments to generate y enhanced segments, and semantic enhancement is performed on the z newspaper article segments to generate z enhanced segments. For the y newspaper article segments (from the third article semantic space related to the film review article), assuming that one of the newspaper article segments is about the cultural background information of the movie plot, the server will use the information in the z newspaper article segments (from the fourth article semantic space related to the entertainment news article) to perform semantic enhancement. For example, if the leading actor in the entertainment news article mentioned his understanding of the cultural background of the movie plot in an interview, the server will integrate this information into the newspaper article segment about the cultural background information of the movie plot to generate an enhanced segment. Similarly, for the z newspaper article segments, for example, one of them is a segment about the relationship between the leading actor's image strategy and the atmosphere of the entertainment industry, semantic enhancement is performed using the information related to the movie theme in the y newspaper article segments (if the movie theme involves the reflection of social phenomena, and this social phenomenon has a certain relationship with the atmosphere of the entertainment industry at that time) to generate one of the z enhanced segments.

[0107] Then, the core semantics of the y enhanced segments are extracted to generate y core semantics. For each of the y enhanced segments, the server will deeply analyze its semantic structure to find the core part that best represents the semantic essence of the segment. For example, for an enhanced segment about the cultural background information of the movie plot after semantic enhancement, its core semantics may be the influence of the specific cultural background on the motivation of the character's behavior in the movie plot. At the same time, according to the semantic depth error between the semantic space of the third article and the semantic space of the fourth article, the z enhanced segments are semantically compressed or expanded to generate z adjusted segments. Assuming that the semantic depth of the semantic space of the third article related to the movie plot is deeper than the semantic depth of the semantic space of the fourth article related to the image of the leading actor (that is, there is a semantic depth error), for a z enhanced segment about the relationship between the image strategy of the leading actor and the atmosphere of the entertainment industry, if the semantic information in this segment is relatively shallow, the server may perform expansion processing based on the semantic depth error, add more information related to the image of the leading actor and related to the semantics of the film review article, and generate one of the z adjusted segments; on the contrary, if the semantic information in this segment is too complicated and needs to be simplified based on the semantic depth error, semantic compression processing is performed.

[0108] After that, the z adjustment segments are processed for semantic positioning to generate z positioning semantics corresponding to the z adjustment segments. Taking the first adjustment segment among the z adjustment segments as an example, the server first obtains the semantic importance coefficients of each semantic unit in the first adjustment segment. For example, in a first adjustment segment about the adjustment of the image strategy of the leading actor, the semantic unit of the image positioning change of the leading actor may have a higher semantic importance coefficient because it has an important impact on the semantic core of the entire segment; while the semantic unit of the news source reporting the image positioning change may have a lower semantic importance coefficient. Then, the server performs weighted fusion on the semantic importance coefficients and semantic contents of each semantic unit in the first adjustment segment to generate the positioning semantics of the first adjustment segment. For example, the semantic content of the semantic unit of the image positioning change of the leading actor is multiplied by its higher semantic importance coefficient, and the semantic content of the semantic unit of the news source is multiplied by its lower semantic importance coefficient. The result obtained after such weighted fusion is the positioning semantics of the first adjustment segment.

[0109] Finally, semantic association analysis is performed on the y core semantics and z positioning semantics to generate a second semantic association summary result between the first newspaper article and the second newspaper article. For example, there may be a certain logical connection between the influence of the film plot cultural background on the character's behavioral motivation in the y core semantics and the change in the image positioning of the leading actor in the z positioning semantics. It may be that the cultural background in the film plot affects the audience's expectations of the leading actor's image, thereby prompting the leading actor to adjust his image positioning. By performing such a comprehensive semantic association analysis on all y core semantics and z positioning semantics, the server generates a deeper and more comprehensive second semantic association summary result, which further reveals the deep semantic connection between the film review article and the entertainment news article. It is not just a superficial connection, but also includes the semantic relationship hidden behind the article after operations such as semantic space optimization, semantic enhancement, and semantic adjustment.

[0110] In a possible implementation, the method further includes: Step A110, performing semantic extraction on the first template newspaper article to generate a first template article semantic space, and performing semantic extraction on the second template newspaper article to generate a second template article semantic space.

[0111] Step A120, based on the deep learning network, perform semantic matching analysis on the semantic space of the first template article to generate weight coefficients of each first semantic vector unit in the semantic space of the first template article, and based on the deep learning network, perform semantic matching analysis on the semantic space of the second template article to generate weight coefficients of each second semantic vector unit in the semantic space of the second template article, wherein the weight coefficients reflect the influence coefficients on the semantic association of the template newspaper article.

[0112] Step A130, based on the weight coefficients of the first semantic vector units, the semantic space of the first template article is optimized to generate x first semantic vector units, and based on the weight coefficients of the second semantic vector units, the semantic space of the second template article is optimized to generate x second semantic vector units.

[0113] Step A140, based on the semantic matching analysis network, perform semantic matching analysis on the x first semantic vector units and the x second semantic vector units, generate first feature identifiers corresponding to the x first semantic vector units and second feature identifiers corresponding to the x second semantic vector units, the first feature identifier is used to locate y first semantic vector units among the x first semantic vector units that have semantic relevance to the x second semantic vector units, and the second feature identifier is used to locate z second semantic vector units among the x second semantic vector units that have semantic relevance to the x first semantic vector units.

[0114] Step A150, optimizing the deep learning network according to the weight coefficients of the first semantic vector units and the weight coefficients of the second semantic vector units, and optimizing the semantic matching analysis network according to the first feature identifier and the second feature identifier.

[0115] In a possible implementation, step A150 includes: Step A151, determining a first training cost based on the error between a feature identifier reflecting whether a first semantic vector unit in a first template article semantic space has a topic feature vector and a weight coefficient of the first semantic vector unit in the first template article semantic space. And determining a second training cost based on the error between a feature identifier reflecting whether a second semantic vector unit in a second template article semantic space has a topic feature vector and a weight coefficient of the second semantic vector unit in the second template article semantic space.

[0116] Step A152: optimizing the deep learning network according to the first training cost and the second training cost.

[0117] Step A153, determining a third training cost based on the error between the first feature identifier and the semantically relevant feature identifier, and determining a fourth training cost based on the error between the second feature identifier and the semantically relevant feature identifier. The semantically relevant feature identifier is a feature identifier that reflects whether the semantically relevant part of the x first semantic vector units and the x second semantic vector units has a topic feature vector.

[0118] Step A154: optimizing the semantic matching analysis network according to the third training cost and the fourth training cost.

[0119] In this embodiment, the first template newspaper article and the second template newspaper article are compared with the previous example of the film review article and the entertainment news article. Assume that the first template newspaper article is a retrospective film review template article about a classic movie series, and the second template newspaper article is a retrospective entertainment news template article about the acting career of the leading actor of the movie series.

[0120] First, the server performs semantic extraction on the first template newspaper article to generate the semantic space of the first template article, and performs semantic extraction on the second template newspaper article to generate the semantic space of the second template article.

[0121] For the first template newspaper article (classic film series retrospective film review template article), the server will perform semantic analysis on various elements in the article when performing semantic extraction. For example, for each movie in the film series, its respective plot characteristics, the overall theme evolution of the series, the inheritance and changes of performance styles in different movies, the development of picture styles, the continuity and innovation of music styles, etc. will be analyzed. These elements are intertwined at the semantic level to construct a complex semantic network, and finally form the semantic space of the first template article. This space covers all possible semantic categories involved in the film review template article when evaluating the film series, from micro-plot details to macro-cultural influence of the series of movies.

[0122] For the second template newspaper article (entertainment news template article that reviews the actor's acting career), the server also performs comprehensive semantic extraction. It analyzes the actor's representative works at each stage of his acting career, role changes, image shaping process, status changes in the entertainment industry, and the evolution of cooperative relationships with other actors and directors. These elements construct the semantic space of the second template article, which comprehensively presents the semantic information architecture of the entertainment news template article when reporting on the actor's acting career.

[0123] Next, based on the deep learning network, a semantic matching analysis is performed on the semantic space of the first template article to generate the weight coefficient of each first semantic vector unit in the semantic space of the first template article, and a semantic matching analysis is performed on the semantic space of the second template article to generate the weight coefficient of each second semantic vector unit in the semantic space of the second template article.

[0124] In the deep learning network, a large number of semantic models and algorithms related to film reviews and entertainment news have been pre-trained. For each first semantic vector unit in the semantic space of the first template article, such as the semantic vector unit about the plot features of the movie in the retrospective film review template article of the movie series, the deep learning network will analyze how closely it is associated with the overall semantics of the film review. If the plot feature is one of the core attractions of the series of movies and is in a key exposition part in the entire film review template article, then the weight coefficient of this semantic vector unit will be higher. This means that it has a greater impact on the semantic association of the template newspaper article. For example, the plot reversal in a certain movie is a classic element of the series, which has been emphasized many times in the film review and is associated with the success of the movie, then the semantic vector unit related to this plot reversal will be given a higher weight coefficient.

[0125] The same is true for each second semantic vector unit in the semantic space of the second template article. In the entertainment news template article that reviews the starring actor's acting career, the semantic vector unit related to the starring actor's representative work at a certain important stage of his acting career will have a higher weight coefficient if this work has a milestone significance for the starring actor's acting career and is the focus of description in the entertainment news template article. For example, if the starring actor wins an international award for a certain movie, the semantic vector unit related to this movie will have a great influence coefficient in the semantic association of the entire entertainment news template article, and will be given a higher weight coefficient.

[0126] Then, based on the weight coefficient of each first semantic vector unit, the semantic space of the first template article is optimized to generate x first semantic vector units, and based on the weight coefficient of each second semantic vector unit, the semantic space of the second template article is optimized to generate x second semantic vector units.

[0127] Assume that the value of x is determined by calculation or according to preset rules (the way x is determined here is similar to the way film review articles and entertainment news articles are processed before, which may be related to factors such as the semantic relevance threshold and the dimension of the template article semantic space). For the first template article semantic space, the weight coefficients of each first semantic vector unit are arranged in descending order, and the first x first semantic vector units are selected. For example, in a retrospective film review template article of a movie series, the first x semantic vector units with the highest weight coefficients may be semantic vector units related to the plots of the core works of the series, the most representative performances, the most unique picture style, and the elements that have a decisive influence on the theme of the series.

[0128] For the second template article semantic space, the weight coefficients of the second semantic vector units are also sorted in descending order, and the first x second semantic vector units are selected. For example, in the entertainment news template article that reviews the actor's acting career, the semantic vector units may be related to the actor's most famous representative works, roles that have a significant impact on his image, key cooperative relationships, etc.

[0129] Afterwards, semantic matching analysis is performed on the x first semantic vector units and the x second semantic vector units according to the semantic matching analysis network to generate first feature identifiers corresponding to the x first semantic vector units and second feature identifiers corresponding to the x second semantic vector units.

[0130] The semantic matching analysis network will perform detailed semantic comparison and association analysis on the x first semantic vector units and the x second semantic vector units. Take the x first semantic vector units in the movie series review template article and the x second semantic vector units in the actor's acting career review entertainment news template article as examples. For a semantic vector unit in the x first semantic vector units that is related to the classic plot of a movie in the series, the semantic matching analysis network will search for the part with semantic relevance in the x second semantic vector units. If in the actor's acting career review entertainment news template article, the actor's performance experience in the movie is an important turning point in his acting career, then in the semantic matching analysis process, this second semantic vector unit related to the actor's performance experience has semantic relevance to the first semantic vector unit related to the movie plot.

[0131] Through such semantic matching analysis, a first feature identifier is generated for the x first semantic vector units, and this first feature identifier can locate the y first semantic vector units among the x first semantic vector units that have semantic relevance to the x second semantic vector units. For example, the first feature identifier may mark which plot, performance, picture style and other semantic vector units in the film series review template article have semantic connections with the relevant semantic vector units in the actor's acting career review template article. Similarly, a second feature identifier is generated for the x second semantic vector units to locate the z second semantic vector units among the x second semantic vector units that have semantic relevance to the x first semantic vector units.

[0132] Next, the deep learning network is optimized according to the weight coefficients of the first semantic vector units and the weight coefficients of the second semantic vector units, and the semantic matching analysis network is optimized according to the first feature identifier and the second feature identifier.

[0133] In terms of optimizing the deep learning network based on the weight coefficients of each first semantic vector unit and the weight coefficients of each second semantic vector unit. First, the first training cost is determined based on the error between the feature identifier reflecting whether the first semantic vector unit in the semantic space of the first template article has a theme feature vector and the weight coefficient of the first semantic vector unit in the semantic space of the first template article. For example, in a retrospective film review template article of a movie series, if there is a first semantic vector unit that deeply explores the theme of the series of movies, its weight coefficient represents its importance in the semantic association of the film review. The feature identifier reflecting whether it has a theme feature vector may be based on some predefined rules or model judgments. If there is a large error between the weight coefficient of this semantic vector unit and the expected weight coefficient based on its theme feature vector, for example, the actual weight coefficient is lower than the expected weight coefficient, this means that the deep learning network may have a deviation in judging the importance of this semantic vector unit, which will result in a certain training cost.

[0134] Similarly, the second training cost is determined based on the error between the feature identifier reflecting whether the second semantic vector unit in the semantic space of the second template article has the theme feature vector and the weight coefficient of the second semantic vector unit in the semantic space of the second template article. For example, in the entertainment news template article of the retrospective of the starring actor's acting career, if the weight coefficient of the second semantic vector unit related to the core element of the starring actor's image creation is inconsistent with the expected weight coefficient determined based on its theme feature vector, the second training cost will be generated.

[0135] Then, based on the first training cost and the second training cost, the deep learning network is optimized. This optimization process may involve adjusting various parameters in the deep learning network, such as the weights and biases of the neural network. If the first training cost is large, it means that there are many problems in the judgment of the weight coefficients of the semantic vector units in the semantic space of the first template article, and the network will adjust the parameters in the direction of reducing this training cost. For example, if the first training cost is large due to inaccurate judgment of the weight coefficients of the semantic vector units related to the plot features of the movie, the network may adjust the parameters related to the semantic analysis of the plot features in order to more accurately determine the weight coefficients.

[0136] In terms of optimizing the semantic matching analysis network based on the first feature identifier and the second feature identifier. The third training cost is determined based on the error between the first feature identifier and the semantically relevant feature identifier, and the fourth training cost is determined based on the error between the second feature identifier and the semantically relevant feature identifier. The semantically relevant feature identifier is a feature identifier that reflects whether the semantic relevance part of the x first semantic vector units and the x second semantic vector units exists in the subject feature vector. For example, if the semantic relevance of the semantic vector units related to the plot of the movie and the performance of the leading actor marked by the first feature identifier deviates from the semantic relevance expected based on the semantically relevant feature identifier, a third training cost will be generated.

[0137] Similarly, for the second feature identifier, if the semantic relevance of the marked semantic vector unit related to the leading actor's representative work and the film style is inconsistent with the expected semantic related feature identifier, a fourth training cost will be generated. Finally, based on the third training cost and the fourth training cost, the semantic matching analysis network is optimized. This optimization process may involve adjusting parameters such as the matching algorithm and the semantic distance calculation method in the semantic matching analysis network to improve the accuracy of the semantic matching analysis, so that the first feature identifier and the second feature identifier can more accurately reflect the semantic relevance between the x first semantic vector units and the x second semantic vector units. By continuously optimizing the deep learning network and the semantic matching analysis network in this way, the accuracy and efficiency of the semantic analysis of template newspaper articles can be improved, thereby providing a better model and algorithm basis for processing the actual semantic analysis of newspaper articles.

[0138] Figure 2 The hardware structure of the summary generation system 100 provided in the embodiment of the present application for implementing the summary generation method for semantic mining of newspaper articles is shown as follows: Figure 2 As shown, the summary generation system 100 may include a processor 110 , a machine-readable storage medium 120 , a bus 130 , and a communication unit 140 .

[0139] In one possible design, the summary generation system 100 may be a single server or a server group. The server group may be centralized or distributed (for example, the summary generation system 100 may be a distributed system). In some embodiments, the summary generation system 100 may be local or remote. For example, the summary generation system 100 may access information and / or data stored in the machine-readable storage medium 120 via a network. For another example, the summary generation system 100 may be directly connected to the machine-readable storage medium 120 to access the stored information and / or data. In some embodiments, the summary generation system 100 may be implemented on a summary generation system. By way of example only, the summary generation system may include a private cloud, a semantically relevant cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-layer cloud, etc., or any aggregation thereof.

[0140] The machine-readable storage medium 120 may store data and / or instructions. In some embodiments, the machine-readable storage medium 120 may store data obtained from an external terminal. In some embodiments, the machine-readable storage medium 120 may store data and / or instructions that the summary generation system 100 uses to execute or use to complete the exemplary methods described in this application.

[0141] In the specific implementation process, one or more processors 110 execute computer executable instructions stored in the machine-readable storage medium 120, so that the processor 110 can execute the summary generation method applied to semantic mining of newspaper articles in the above method embodiment. The processor 110, the machine-readable storage medium 120 and the communication unit 140 are connected through the bus 130, and the processor 110 can be used to control the sending and receiving actions of the communication unit 140.

[0142] The specific implementation process of the processor 110 can refer to the various method embodiments executed by the summary generation system 100 above. The implementation principles and technical effects are similar and will not be described in detail in this embodiment.

[0143] In addition, an embodiment of the present application further provides a readable storage medium, in which computer executable instructions are set. When a processor executes the computer executable instructions, the summary generation method applied to semantic mining of newspaper articles as above is implemented.

[0144] It should be noted that in order to simplify the description disclosed in this application and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present application, multiple features are sometimes combined into one embodiment, drawings, or descriptions thereof. Similarly, it should be noted that in order to simplify the description disclosed in this application and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present application, multiple features are sometimes combined into one embodiment, drawings, or descriptions thereof.

Claims

1. A summary generation method applied to semantic mining of newspaper articles, characterized in that: The method comprises: Performing semantic extraction on the first newspaper article to generate a first article semantic space, and performing semantic extraction on the second newspaper article to generate a second article semantic space; Performing semantic matching analysis on the first article semantic space to generate weight coefficients of each first semantic vector unit in the first article semantic space, and performing semantic matching analysis on the second article semantic space to generate weight coefficients of each second semantic vector unit in the second article semantic space, wherein the weight coefficients reflect the influence coefficients on the semantic association of newspaper articles; According to the weight coefficients of the first semantic vector units, the first article semantic space is optimized to generate x first semantic vector units, and according to the weight coefficients of the second semantic vector units, the second article semantic space is optimized to generate x second semantic vector units; A semantic matching analysis is performed on the x first semantic vector units and the x second semantic vector units to generate a first semantic association summary result between the first newspaper article and the second newspaper article.

2. The method for generating abstracts for semantic mining of newspaper articles according to claim 1, characterized in that: The step of performing semantic matching analysis on the x first semantic vector units and the x second semantic vector units to generate a first semantic association summary result between the first newspaper article and the second newspaper article includes: Performing semantic matching analysis on the x first semantic vector units and the x second semantic vector units to generate first feature identifiers corresponding to the x first semantic vector units and second feature identifiers corresponding to the x second semantic vector units, wherein the first feature identifier is used to locate y first semantic vector units among the x first semantic vector units that have semantic relevance to the x second semantic vector units, and the second feature identifier is used to locate z second semantic vector units among the x second semantic vector units that have semantic relevance to the x first semantic vector units, wherein the first feature identifier and the second feature identifier are used to represent corresponding masks; A semantic association analysis is performed on the y first semantic vector units and the z second semantic vector units to generate a first semantic association summary result between the first newspaper article and the second newspaper article.

3. The method for generating abstracts for semantic mining of newspaper articles according to claim 1, characterized in that: The method of performing semantic space optimization on the first article semantic space according to the weight coefficients of the first semantic vector units to generate x first semantic vector units, and performing semantic space optimization on the second article semantic space according to the weight coefficients of the second semantic vector units to generate x second semantic vector units includes: Determine x according to the semantic relevance threshold and the dimension of the first article semantic space, or determine x according to the semantic relevance threshold and the dimension of the second article semantic space; According to the weight coefficients of the first semantic vector units, based on the order of decreasing the weight coefficients, the first x first semantic vector units in the semantic space of the first article are output as the x first semantic vector units; and according to the weight coefficients of the second semantic vector units, based on the order of decreasing the weight coefficients, the first x second semantic vector units in the semantic space of the second article are output as the x second semantic vector units.

4. The method for generating abstracts for semantic mining of newspaper articles according to any one of claims 1 to 3, characterized in that: The performing semantic matching analysis on the x first semantic vector units and the x second semantic vector units to generate first feature identifiers corresponding to the x first semantic vector units and second feature identifiers corresponding to the x second semantic vector units includes: Determine the first basic feature identifiers corresponding to the x first semantic vector units, and determine the second basic feature identifiers corresponding to the x second semantic vector units; Filtering the x first semantic vector units according to the first basic feature identifier to generate x first semantic vector units to be optimized, and filtering the x second semantic vector units according to the second basic feature identifier to generate x second semantic vector units to be optimized; Performing semantic relevance analysis on the x first semantic vector units to be optimized according to the first semantic relevance analysis module, and generating x first adjusted semantic vector units based on the x second semantic vector units to be optimized and the x semantic vector units generated by the first semantic relevance analysis module according to the first interactive relevance analysis module, and performing semantic relevance analysis on the x second semantic vector units to be optimized according to the second semantic relevance analysis module, and generating x second adjusted semantic vector units based on the x first semantic vector units to be optimized and the x semantic vector units generated by the second semantic relevance analysis module; According to the x first adjustment semantic vector units, adjusting the first basic feature identifier to generate a first adjustment feature identifier, and according to the x second adjustment semantic vector units, adjusting the second basic feature identifier to generate a second adjustment feature identifier; The first feature identifier is determined according to the first adjustment feature identifier, and the second feature identifier is determined according to the second adjustment feature identifier.

5. The method for generating abstracts for semantic mining of newspaper articles according to claim 4, characterized in that: Before performing semantic extraction on the first newspaper article to generate the first article semantic space, and performing semantic extraction on the second newspaper article to generate the second article semantic space, the method further includes: When the dimension of the first newspaper article and the dimension of the second newspaper article are inconsistent, converting the first newspaper article and the second newspaper article to the same dimension; The determining of the first basic feature identifiers corresponding to the x first semantic vector units, and the determining of the second basic feature identifiers corresponding to the x second semantic vector units, include: Determine the first basic feature identifier according to the vocabulary supplemented in the dimensional conversion of the first newspaper article, wherein the mark corresponding to the semantic vector of the vocabulary supplemented in the x first semantic vector units in the first basic feature identifier is non-core content; The second basic feature identifier is determined based on the vocabulary supplemented in the dimensional conversion of the second newspaper article, and the mark corresponding to the semantic vector of the vocabulary supplemented in the x second semantic vector units in the second basic feature identifier is non-core content.

6. The method for generating abstracts for semantic mining of newspaper articles according to claim 4, characterized in that: The adjusting the first basic feature identifier according to the x first adjustment semantic vector units to generate a first adjustment feature identifier, and adjusting the second basic feature identifier according to the x second adjustment semantic vector units to generate a second adjustment feature identifier, include: Performing semantic matching analysis on the x first adjusted semantic vector units to generate a confidence combination corresponding to each of the x first adjusted semantic vector units, and performing semantic matching analysis on the x second adjusted semantic vector units to generate a confidence combination corresponding to each of the x second adjusted semantic vector units, wherein the confidence combination reflects the confidence of the semantically related content and the confidence of the semantically non-related content; Determine x first semantic identification codes corresponding to the x first adjusted semantic vector units based on the confidence combinations corresponding to the first adjusted semantic vector units, and determine x second semantic identification codes corresponding to the x second adjusted semantic vector units based on the confidence combinations corresponding to the second adjusted semantic vector units, wherein the first semantic identification code reflects the semantic content label corresponding to the maximum confidence in the confidence combinations corresponding to the first adjusted semantic vector units, and the second semantic identification code reflects the semantic content label corresponding to the maximum confidence in the confidence combinations corresponding to the second adjusted semantic vector units; Performing logic operation processing on the x first semantic identification codes to generate x first confidences, and performing logic operation processing on the x second semantic identification codes to generate x second confidences; Filtering the x first confidences according to the first basic feature identifier to generate the first adjusted feature identifier, and filtering the x second confidences according to the second basic feature identifier to generate the second adjusted feature identifier; Alternatively, the x first adjusted semantic vector units are adjusted according to the x second adjusted semantic vector units, the first adjusted feature identifier and the second adjusted feature identifier to generate x third adjusted semantic vector units, and the x second adjusted semantic vector units are adjusted according to the x first adjusted semantic vector units, the x second adjusted semantic vector units, the first adjusted feature identifier and the second adjusted feature identifier to generate x fourth adjusted semantic vector units; According to the x third adjustment semantic vector units, adjusting the first adjustment feature identifier to generate a third adjustment feature identifier, and according to the x fourth adjustment semantic vector units, adjusting the second adjustment feature identifier to generate a fourth adjustment feature identifier; The first feature identifier is determined based on the third adjustment feature identifier, and the second feature identifier is determined based on the fourth adjustment feature identifier, the first feature identifier being an adjustment feature identifier obtained when the number of adjustment cycles of the first basic feature identifier is not less than a set order, and the second feature identifier is an adjustment feature identifier obtained when the number of adjustment cycles of the second basic feature identifier is not less than the set order.

7. The method for generating abstracts for semantic mining of newspaper articles according to any one of claims 1 to 3, characterized in that: The step of performing semantic extraction on the first newspaper article to generate the first article semantic space, and performing semantic extraction on the second newspaper article to generate the second article semantic space, comprises: Performing semantic extraction on the first newspaper article to generate the first article semantic space and the third article semantic space, and performing semantic extraction on the second newspaper article to generate the second article semantic space and the fourth article semantic space, wherein the dimension of the first article semantic space is different from the dimension of the third article semantic space, and the dimension of the second article semantic space is different from the dimension of the fourth article semantic space; After generating a first semantic association summary result between the first newspaper article and the second newspaper article, the method further includes: According to the first semantic association summary result, the dimension of the first article semantic space, the dimension of the third article semantic space and the third article semantic space, the third article semantic space is optimized to generate y newspaper article segments corresponding to the y first semantic vector units; and according to the first semantic association summary result, the dimension of the second article semantic space, the dimension of the fourth article semantic space and the fourth article semantic space, the fourth article semantic space is optimized to generate z newspaper article segments corresponding to the z second semantic vector units; According to the y newspaper article segments and the z newspaper article segments, the y newspaper article segments are semantically enhanced to generate y enhanced segments, and according to the y newspaper article segments and the z newspaper article segments, the z newspaper article segments are semantically enhanced to generate z enhanced segments; Extracting the core semantics of the y enhanced segments to generate y core semantics, and performing semantic compression or expansion processing on the z enhanced segments according to the semantic depth error between the third article semantic space and the fourth article semantic space to generate z adjusted segments; Performing semantic positioning processing on the z adjustment segments to generate z positioning semantics corresponding to the z adjustment segments; Performing semantic association analysis on the y core semantics and the z positioning semantics to generate a second semantic association summary result between the first newspaper article and the second newspaper article; The performing semantic positioning processing on the z adjustment segments to generate z positioning semantics corresponding to the z adjustment segments includes: Obtaining the semantic importance coefficient of each semantic unit in the first adjustment segment of the z adjustment segments; The semantic importance coefficient and the semantic content of each semantic unit in the first adjustment segment are weightedly fused to generate the positioning semantics of the first adjustment segment.

8. The method for generating abstracts for semantic mining of newspaper articles according to any one of claims 1 to 3, characterized in that: The method further comprises: Performing semantic extraction on the first template newspaper article to generate the semantic space of the first template article, and performing semantic extraction on the second template newspaper article to generate the semantic space of the second template article; According to the deep learning network, a semantic matching analysis is performed on the semantic space of the first template article to generate a weight coefficient of each first semantic vector unit in the semantic space of the first template article; and according to the deep learning network, a semantic matching analysis is performed on the semantic space of the second template article to generate a weight coefficient of each second semantic vector unit in the semantic space of the second template article, wherein the weight coefficient reflects an influence coefficient on the semantic association of the template newspaper article; According to the weight coefficients of the first semantic vector units, the semantic space of the first template article is optimized to generate x first semantic vector units, and according to the weight coefficients of the second semantic vector units, the semantic space of the second template article is optimized to generate x second semantic vector units; According to the semantic matching analysis network, a semantic matching analysis is performed on the x first semantic vector units and the x second semantic vector units to generate a first feature identifier corresponding to the x first semantic vector units and a second feature identifier corresponding to the x second semantic vector units, wherein the first feature identifier is used to locate y first semantic vector units having semantic relevance with the x second semantic vector units among the x first semantic vector units, and the second feature identifier is used to locate z second semantic vector units having semantic relevance with the x first semantic vector units among the x second semantic vector units; The deep learning network is optimized according to the weight coefficients of the first semantic vector units and the weight coefficients of the second semantic vector units, and the semantic matching analysis network is optimized according to the first feature identifier and the second feature identifier.

9. The method for generating abstracts for semantic mining of newspaper articles according to claim 8, characterized in that: The step of optimizing the deep learning network according to the weight coefficients of the first semantic vector units and the weight coefficients of the second semantic vector units includes: Determine a first training cost based on an error between a feature identifier reflecting whether a first semantic vector unit in a semantic space of a first template article has a topic feature vector and a weight coefficient of the first semantic vector unit in the semantic space of the first template article; and determine a second training cost based on an error between a feature identifier reflecting whether a second semantic vector unit in a semantic space of a second template article has a topic feature vector and a weight coefficient of the second semantic vector unit in the semantic space of the second template article; Optimizing the deep learning network according to the first training cost and the second training cost; The step of optimizing the semantic matching analysis network according to the first feature identifier and the second feature identifier includes: Determine a third training cost based on an error between the first feature identifier and the semantically relevant feature identifier, and determine a fourth training cost based on an error between the second feature identifier and the semantically relevant feature identifier; wherein the semantically relevant feature identifier is a feature identifier that reflects whether a topic feature vector exists in the semantically relevant part of the x first semantic vector units and the x second semantic vector units; The semantic matching analysis network is optimized according to the third training cost and the fourth training cost.

10. A summary generation system, characterized in that: The summary generation system includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the summary generation method applied to semantic mining of newspaper articles as described in any one of claims 1 to 9 above.

Images