A rating feature classification method for technology company rating models

Abstract

This invention relates to the field of feature analysis, and more particularly to a rating feature classification method for a technology enterprise rating model. The invention sets a search target, obtains several search interfaces associated with the search target, performs visual analysis on the search interfaces, and determines visual clustering regions based on the relative positions of target features. It calculates the correlation parameters between target features and the search target, sets classification labels for each visual clustering region, and performs adaptive selection analysis on target features within the visual clustering regions based on the labels. This includes selecting and removing target features from ambiguous areas of the search target, retaining features with high correlation to the search target, or removing all target features from the visual clustering regions. Subsequently, the retained target features are transmitted to a preset language rating model. This invention can effectively reduce irrelevant features in web pages, improve the efficiency and accuracy of feature extraction from web pages, and ensure the accuracy of feedback from the language rating model.

Description

Technical Field

[0001] This invention relates to the field of feature analysis, and more particularly to a rating feature classification method for a technology company rating model. Background Technology

[0002] With the development of computer science, Natural Language Models (NLP) have emerged. By inputting content into NLP models, they can analyze the content and provide corresponding feedback. Currently, some NLP models are capable of extracting information from web pages and then providing comprehensive feedback based on the extracted information. Based on this, NLP models are widely used in various fields, especially in the field of business consulting. Relevant consulting advice can be given to businesses by NLP models, and then comprehensive feedback can be provided based on the NLP models to support the consultants.

[0003] For example, Chinese Patent Publication No. CN118861222A discloses a consultation processing method, apparatus, storage medium, and computer equipment based on a large language model. The method includes: receiving target consultation text input by a user based on the customer service function during non-working hours of human customer service; constructing target intent analysis prompts based on the target consultation text and a preset intent analysis prompt template, and calling a large language model to generate the consultation intent corresponding to the target consultation text based on the target intent analysis prompts; determining the target processing method for the target consultation text based on the consultation intent, and performing consultation processing according to the target processing method.

[0004] However, the following problems still exist in the existing technology.

[0005] In existing technologies, when a user submits a request for information and the large model responds based on the information on the webpage, the large model's response is slow due to the numerous webpage interfaces and mixed content. Furthermore, irrelevant information may be introduced to interfere with the large model's response, resulting in low response efficiency and unsatisfactory results. Summary of the Invention

[0006] To address this, the present invention provides a rating feature method for a technology enterprise rating model, which overcomes the problems in the prior art where, when a user submits an inquiry and the large model responds based on information from a webpage, the extraction of relevant features from the webpage is slow due to the numerous webpage interfaces and mixed content, and irrelevant information may be introduced to interfere with the feedback of the large model, resulting in low feedback efficiency and poor feedback results.

[0007] To achieve the above objectives, this invention provides a rating feature classification method for a technology enterprise rating model, comprising:

[0008] Set a search target and obtain several search interfaces associated with the search target;

[0009] Visual analysis of the search interface is performed, including identifying several target features on the page, clustering the target features based on the relative positional relationship between the target features, and determining several visual clustering regions.

[0010] Extract some target features from the visual clustering region, and determine the correlation parameters between these features and the retrieval target, including:

[0011] Determine the text segments corresponding to the extracted target features in the visual clustering region;

[0012] The text segments are compared with the search target to determine the semantic relevance of each text segment, and the mean of the semantic relevance is determined as the relevance parameter.

[0013] Classification labels are set for each of the aforementioned visual clustering regions;

[0014] Based on the classification labels, target features in each visual clustering region are analyzed and extracted, including:

[0015] Identify labeled keywords strongly associated with the search target within visual clustering regions, determine the location information of the labeled keywords within the visual clustering regions, and filter labeled keyword groups for retrieving the location of ambiguous regions of the search target, including:

[0016] The distance between each labeled keyword and its nearest neighboring labeled keyword in the reference direction is determined based on the location information of the labeled keywords.

[0017] If the distance between a labeled keyword and its nearest labeled keyword is greater than the predetermined fuzzy region spacing threshold;

[0018] Then the identified keyword and its nearest neighbor identified keyword are determined as the identified keyword group;

[0019] The reference direction is perpendicular to the bottom edge of the search interface;

[0020] Identify the fuzzy regions of the retrieval target within the visual clustering regions, including:

[0021] A reference line parallel to the reference direction is constructed at the center of the visual clustering region;

[0022] Construct a calibration line perpendicular to the reference line on the reference line;

[0023] The area enclosed by each of the calibration lines and the edge of the visual clustering region is defined as the fuzzy region of the retrieval target.

[0024] Remove target features from the blurred region of the search target and retain the remaining target features;

[0025] Alternatively, remove all target features from the visual clustering region;

[0026] The retained target features are transferred to the pre-defined language rating model;

[0027] The search target is in text form and is set by the user.

[0028] Furthermore, the process of clustering target features based on their relative positional relationships to determine several visual clustering regions includes:

[0029] Determine the coordinates of several target features on the search interface;

[0030] The relative distances between target features are determined based on the coordinates described above;

[0031] Clustering of each target feature based on relative distance yields several clusters;

[0032] Determine the edge target features in each cluster, and connect the edge target features sequentially to obtain the connected regions;

[0033] The visual clustering region is obtained by expanding the connection region by a predetermined factor.

[0034] In each of the clusters, the relative distance between the target feature and the nearest target feature is less than a predetermined distance threshold.

[0035] Furthermore, the process of extracting partial target features from the visual clustering region includes,

[0036] The text composed of target features in the visual clustering region is divided into several text segments;

[0037] Extract a number of text segments at predetermined extraction intervals.

[0038] Furthermore, setting classification labels for each of the aforementioned visual clustering regions includes,

[0039] If the correlation parameter is greater than or equal to a predetermined correlation standard threshold, then an association label is set for the visual clustering region;

[0040] If the correlation parameter is less than a predetermined correlation standard threshold, then a fuzzy label is set for the visual clustering region.

[0041] Furthermore, based on the classification labels, the target features in each visual clustering region are analyzed and extracted, including...

[0042] If the visual clustering region is an associated label, then identify the labeling keywords that are strongly associated with the search target in the visual clustering region, determine the position information of the labeling keywords in the visual clustering region, filter the labeling keyword group used for the position of the fuzzy region of the search target, determine the fuzzy region of the search target in the visual clustering region, remove the target features in the fuzzy region of the search target, and retain the remaining target features.

[0043] If the visual clustering region is a fuzzy label, then all target features in the visual clustering region are removed.

[0044] Further, several search interfaces associated with the search target are obtained, wherein,

[0045] If the semantic relevance between the title of the search interface and the search target is greater than a predetermined relevance standard threshold, then the search interface is determined to be associated with the search target.

[0046] Furthermore, target features include textual features and symbolic features.

[0047] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention sets a search target, obtains several search interfaces associated with the search target, performs visual analysis on the search interfaces, and determines visual clustering regions based on the relative positions of target features. It calculates the correlation parameters between target features and the search target, sets classification labels for each visual clustering region, and performs adaptive selection analysis on target features within the visual clustering regions based on the labels. This includes selecting and removing target features from ambiguous areas of the search target, retaining features with high correlation to the search target, or removing all target features from the visual clustering regions. Subsequently, the retained target features are transmitted to a preset language rating model. This invention can effectively reduce irrelevant features in web pages, improve the efficiency and accuracy of feature acquisition from web pages, and ensure the accuracy of feedback from the language rating model.

[0048] In particular, this invention performs visual analysis of the search interface to determine visual clustering regions based on the relative positions of target features. In reality, the distribution of different search interfaces varies, and the relevance between the entire search interface's content and the target cannot be determined solely based on the title. Different search interfaces may have various information areas and layouts, and similar content often clusters together. Therefore, performing language analysis on the text of the entire page to determine the relevance to the search content, especially with a large number of interfaces, would waste significant computational resources. Therefore, this invention considers determining visual clustering regions by observing the clustering of target features on the page from a visual perspective. Directly employing semantic association analysis for image-based clustering is relatively simple and efficient. For a single visual clustering region, only a subset of target features are extracted to determine the relevance parameters to the retrieval target. Labels are then assigned to the visual clustering regions, reducing the amount of data analysis. Therefore, this invention determines visual clustering regions and sets classification labels, which can dynamically adapt to different retrieval interface layouts. It can effectively identify and divide visual clustering regions, and thus, when faced with a large number of retrieval interfaces, it can effectively reduce irrelevant features in web pages, improve the efficiency and accuracy of feature extraction from web pages, and ensure the accuracy of the language rating model feedback.

[0049] In particular, this invention adaptively employs different analysis and extraction methods for visual clustering regions of different classification tags in the search interface. For associated tags, it considers the effectiveness of the text in the entire visual clustering region, determines the position of the labeled keywords, and then determines the labeled keyword group to further determine the fuzzy region of the search target. In reality, the content corresponding to the text in the visual clustering region is not necessarily related to the search target. Usually, if the text is the main text describing the search target, the labeled keywords that are strongly related to the search target will be distributed relatively evenly in the text. If there is content unrelated to the search target in the visual clustering region, there may not be labeled keywords that are strongly related to the search target in this part of the content. Therefore, for the special case where the labeled keywords are widely spaced, the fuzzy region of the search target is determined, and the target features in it are removed. Thus, when facing a large number of search interfaces, it can effectively reduce irrelevant features in the webpage, improve the efficiency and accuracy of feature acquisition from the webpage, and ensure the accuracy of the feedback from the language rating model.

[0050] In particular, for visual clustering regions with fuzzy labels, all target features are directly removed. Finally, the retained target features are transmitted to a pre-defined language rating model, which effectively filters out irrelevant information and improves the efficiency and accuracy of information acquisition. Attached Figure Description

[0051] Figure 1 This is a schematic diagram illustrating the steps of a rating feature classification method for a technology company rating model according to an embodiment of the invention.

[0052] Figure 2 A logical decision diagram for identifying the label category of the current model region in an embodiment of the invention;

[0053] Figure 3 This is a logic block diagram illustrating the analysis of the visual clustering region based on the region label in an embodiment of the invention.

[0054] Figure 4 A logical decision diagram for filtering calibrated keyword groups that reference the spacing of fuzzy regions, as shown in the embodiments of the invention.

[0055] Figure 5 This is a logical block diagram illustrating the association between the search interface and the search target in an embodiment of the invention. Detailed Implementation

[0056] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0057] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0058] Please see Figure 1 The diagram illustrates the steps of a rating feature classification method for a technology enterprise rating model according to an embodiment of the present invention. The rating feature classification method for a technology enterprise rating model according to an embodiment of the present invention includes:

[0059] Step S1: Set the search target and obtain several search interfaces associated with the search target;

[0060] Step S2 involves performing visual analysis of the search interface, including identifying several target features on the page, clustering the target features based on their relative positional relationships, and determining several visual clustering regions.

[0061] Step S3: Extract some target features from the visual clustering region and determine the correlation parameters between these features and the retrieval target, including...

[0062] Determine the text segments corresponding to the extracted target features in the visual clustering region;

[0063] The text segments are compared with the search target to determine the semantic relevance of each text segment, and the mean of the semantic relevance is determined as the relevance parameter.

[0064] Classification labels are set for each of the aforementioned visual clustering regions;

[0065] Step S4: Analyze and extract target features from each visual clustering region based on the classification labels, including...

[0066] Identify labeled keywords strongly associated with the search target within visual clustering regions, determine the location information of the labeled keywords within the visual clustering regions, and filter labeled keyword groups for retrieving the location of ambiguous regions of the search target, including:

[0067] The distance between each labeled keyword and its nearest neighboring labeled keyword in the reference direction is determined based on the location information of the labeled keywords.

[0068] If the distance between a labeled keyword and its nearest labeled keyword is greater than the predetermined fuzzy region spacing threshold;

[0069] Then the identified keyword and its nearest neighbor identified keyword are determined as the identified keyword group;

[0070] The reference direction is perpendicular to the bottom edge of the search interface;

[0071] Identify the fuzzy regions of the retrieval target within the visual clustering regions, including:

[0072] A reference line parallel to the reference direction is constructed at the center of the visual clustering region;

[0073] Construct a calibration line perpendicular to the reference line on the reference line;

[0074] The area enclosed by each of the calibration lines and the edge of the visual clustering region is defined as the fuzzy region of the retrieval target.

[0075] Remove target features from the blurred region of the search target and retain the remaining target features;

[0076] Alternatively, remove all target features from the visual clustering region;

[0077] Step S5: Transfer the retained target features to the preset language rating model;

[0078] The search target is in text form and is set by the user.

[0079] Specifically, users can pre-set the required search targets. For example, if they need to obtain evaluation information of a certain company, the search target can be set as "how the company is in certain aspects". Those skilled in the art can set search targets according to their needs, which will not be elaborated here.

[0080] Specifically, there are no restrictions on the specific form of the language rating model. Those skilled in the art can use existing open-source natural language models with authorization, as long as they can provide feedback based on the consultation content input by the user and the information extracted from the search interface. This will not be elaborated further.

[0081] Specifically, there are no restrictions on the method of identifying target features. In practice, target features are text features and symbol features. For example, an OCR model can be used to identify text features. After the text features are calibrated, their coordinates can be located. Of course, other methods can also be used, which will not be elaborated here.

[0082] Specifically, the process of clustering target features based on their relative positional relationships to determine several visual clustering regions includes:

[0083] Determine the coordinates of several target features on the search interface;

[0084] The relative distances between target features are determined based on the coordinates described above;

[0085] Clustering of each target feature based on relative distance yields several clusters;

[0086] Determine the edge target features in each cluster, and connect the edge target features sequentially to obtain the connected regions;

[0087] It is understandable that the edge target features in the search interface are the target features located at the edge of the region where the cluster is located, and must meet at least one edge condition. The edge conditions include minimum vertical coordinate, maximum vertical coordinate, maximum horizontal coordinate, and minimum horizontal coordinate.

[0088] The visual clustering region is obtained by expanding the connection region by a predetermined factor.

[0089] In each of the clusters, the relative distance between the target feature and the nearest target feature is less than a predetermined distance threshold.

[0090] Specifically, the purpose of increasing the predetermined factor is to preserve the target features as much as possible without interfering with other visual clustering regions. Preferably, the predetermined factor is set to 1.1 times.

[0091] Specifically, the predetermined distance threshold is a pre-set value, wherein,

[0092] Several search interfaces are obtained in advance as samples. The average distance between each target feature and the nearest target feature in the search interface is determined. 1.15 times the average distance is set as a predetermined distance threshold.

[0093] Specifically, the process of extracting partial target features from visual clustering regions includes,

[0094] The text composed of target features in the visual clustering region is divided into several text segments;

[0095] Extract a number of text segments at predetermined extraction intervals.

[0096] In practice, when dividing the text composed of target features in the visual clustering region into several text segments, there is no limit to the length of the text segments. Each text segment can be a sentence, a paragraph, or a text unit divided according to specific rules.

[0097] In practice, the predetermined extraction interval is selected within the interval [2, 5], that is, one text segment is extracted every predetermined extraction interval.

[0098] This invention sets a search target, obtains several search interfaces associated with the search target, performs visual analysis on the search interfaces, and determines visual clustering regions based on the relative positions of target features. It calculates the correlation parameters between target features and the search target, assigns classification labels to each visual clustering region, and performs adaptive selection analysis on target features within the visual clustering regions based on the labels. This includes selecting and removing target features from ambiguous areas of the search target, retaining features with high correlation to the search target, or removing all target features from the visual clustering regions. The retained target features are then transmitted to a pre-set language rating model. This invention effectively reduces irrelevant features in web pages, improves the efficiency and accuracy of feature extraction from web pages, and ensures the accuracy of the language rating model's feedback.

[0099] In practice, this invention does not limit the method for calculating semantic relevance. The semantic relevance can be measured by calculating the mean cosine similarity between the word vectors of the text segment and the retrieval target. Of course, other methods can also be used, as long as the semantic relevance of each text segment corresponding to some target features can be calculated by comparing it with the retrieval target. This will not be elaborated further.

[0100] Please see Figure 2 As shown, it is a logical decision diagram for identifying the label category of the current model region in an embodiment of the invention. Specifically, setting classification labels for each of the visual clustering regions includes,

[0101] If the correlation parameter is greater than or equal to a predetermined correlation standard threshold, then an association label is set for the visual clustering region;

[0102] If the correlation parameter is less than a predetermined correlation standard threshold, then a fuzzy label is set for the visual clustering region.

[0103] In practice, the predetermined correlation standard threshold of this invention is pre-determined, wherein,

[0104] Those skilled in the art can collect a large number of corpus samples related to the retrieval target, split the corpus samples into text segments, calculate the average semantic relevance between each text segment in the corpus sample and the retrieval target, and set the product of the average semantic relevance and the error coefficient as a predetermined relevance standard threshold, with the error coefficient in the range [0.75, 0, 85].

[0105] Please see Figure 3 The diagram shown is a logical block diagram illustrating the analysis of visual clustering regions based on the region labels according to an embodiment of the invention. Specifically, the analysis and extraction of target features in each visual clustering region based on the classification labels includes...

[0106] If the visual clustering region is an associated label, then identify the labeling keywords that are strongly associated with the search target in the visual clustering region, determine the position information of the labeling keywords in the visual clustering region, filter the labeling keyword group used for the position of the fuzzy region of the search target, determine the fuzzy region of the search target in the visual clustering region, remove the target features in the fuzzy region of the search target, and retain the remaining target features.

[0107] If the visual clustering region is a fuzzy label, then all target features in the visual clustering region are removed.

[0108] This invention performs visual analysis of the search interface and determines visual clustering regions based on the relative positions of target features. In reality, the distribution of different search interfaces varies, and the relevance of the entire search interface content to the target cannot be determined solely based on the title. Different search interfaces may have various information areas and layouts, and similar content often clusters together. Therefore, performing language analysis on the text of the entire page to determine the relevance to the search content would be computationally wasteful when dealing with a large number of interfaces. Therefore, this invention considers determining visual clustering regions by observing the clustering of target features on the page from a visual perspective, rather than directly using semantic association analysis. Clustering analysis from an image perspective is simpler and improves analysis efficiency. For a single visual clustering region, only a portion of the target features are extracted to determine the relevance parameters to the search target, and labels are set for the visual clustering regions, reducing the amount of data analysis. Therefore, this invention determines visual clustering regions and sets classification labels, which can dynamically adapt to different search interface layouts, effectively identify and divide visual clustering regions, and thus effectively reduce irrelevant features on web pages when dealing with a large number of search interfaces, improving the efficiency and accuracy of feature extraction from web pages and ensuring the accuracy of the language rating model feedback.

[0109] Please see Figure 4 As shown, it is a logic decision diagram for filtering the labeled keyword groups used as a reference for the spacing of fuzzy regions in an embodiment of the invention.

[0110] In practice, the fuzzy region spacing threshold of the present invention is predetermined. Several long texts associated with the search target are obtained in advance. The maximum spacing between the labeled keyword and the nearest labeled keyword in each long text is calculated. The mean of the maximum spacing is solved. The product of the mean of the maximum spacing and the amplification factor is used as the fuzzy region spacing threshold. The amplification factor is selected in the interval [1.2, 1.5].

[0111] When determining the labeling keywords that are strongly associated with the search target, the semantic relevance of each keyword to the search target is calculated. If the semantic relevance is greater than a predetermined strong semantic relevance threshold, the keyword is determined to be a labeling keyword.

[0112] In practice, the strong semantic relevance threshold is determined based on the standard relevance threshold, and is set to 1.5 times the standard relevance threshold.

[0113] In practice, the cosine similarity method is also used to determine the relevance between keywords and the search target. The keywords and several keywords in the search target are vectorized, and the mean cosine similarity between the keywords and the search target is calculated. The mean cosine similarity is then used to determine the relevance between the keywords and the search target.

[0114] It is understandable that the constructed fuzzy area of ​​the search target is a rectangular area.

[0115] Specifically, please refer to Figure 5 As shown, Figure 5 This is a logical block diagram illustrating the association between a search interface and a search target in an embodiment of the invention. Several search interfaces associated with the search target are obtained, wherein...

[0116] If the semantic relevance between the title of the search interface and the search target is greater than a predetermined relevance standard threshold, then the search interface is determined to be associated with the search target.

[0117] Specifically, target features include textual features and symbolic features.

[0118] This invention adaptively employs different analysis and extraction methods for visual clustering regions of different category tags in the search interface. For associated tags, it considers the effectiveness of the text in the entire visual clustering region to determine the position of the labeled keywords, and then determines the labeled keyword group to further determine the fuzzy region of the search target. In reality, the content corresponding to the text in the visual clustering region is not necessarily related to the search target. Usually, if the text is the main text describing the search target, the labeled keywords that are strongly related to the search target will be relatively evenly distributed in the text. If there is content unrelated to the search target in the visual clustering region, there may not be labeled keywords that are strongly related to the search target in this part of the content. Therefore, for the special case where the labeled keywords are widely spaced, the fuzzy region of the search target is determined, and the target features in it are removed. Thus, when facing a large number of search interfaces, it can effectively reduce irrelevant features in web pages, improve the efficiency and accuracy of feature extraction from web pages, and ensure the accuracy of the feedback from the language rating model.

[0119] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A rating feature classification method for a technology enterprise rating model, characterized in that, include: Set a search target and obtain several search interfaces associated with the search target; Visual analysis of the search interface is performed, including identifying several target features on the page, clustering the target features based on the relative positional relationship between the target features, and determining several visual clustering regions. Extract some target features from the visual clustering region, and determine the correlation parameters between these features and the retrieval target, including: Determine the text segments corresponding to the extracted target features in the visual clustering region; The text segments are compared with the search target to determine the semantic relevance of each text segment, and the mean of the semantic relevance is determined as the relevance parameter. Classification labels are set for each of the aforementioned visual clustering regions; Based on the classification labels, target features in each visual clustering region are analyzed and extracted, including: Identify labeled keywords strongly associated with the search target within visual clustering regions, determine the location information of the labeled keywords within the visual clustering regions, and filter labeled keyword groups for retrieving the location of ambiguous regions of the search target, including: The distance between each labeled keyword and its nearest neighboring labeled keyword in the reference direction is determined based on the location information of the labeled keywords. If the distance between a labeled keyword and its nearest labeled keyword is greater than the predetermined fuzzy region spacing threshold; Then the identified keyword and its nearest neighbor identified keyword are determined as the identified keyword group; The reference direction is perpendicular to the bottom edge of the search interface; Identify the fuzzy regions of the retrieval target within the visual clustering regions, including: A reference line parallel to the reference direction is constructed at the center of the visual clustering region; Construct a calibration line perpendicular to the reference line on the reference line; The area enclosed by each of the calibration lines and the edge of the visual clustering region is defined as the fuzzy region of the retrieval target. Remove target features from the blurred region of the search target and retain the remaining target features; Alternatively, remove all target features from the visual clustering region; The retained target features are transferred to the pre-defined language rating model; The search target is in text form and is set by the user.

2. The rating feature classification method for a technology enterprise rating model according to claim 1, characterized in that, The process of clustering target features based on their relative positional relationships to determine several visual clustering regions includes: Determine the coordinates of several target features on the search interface; The relative distances between target features are determined based on the coordinates described above; Clustering of each target feature based on relative distance yields several clusters; Determine the edge target features in each cluster, and connect the edge target features sequentially to obtain the connected regions; The visual clustering region is obtained by expanding the connection region by a predetermined factor. In each of the clusters, the relative distance between the target feature and the nearest target feature is less than a predetermined distance threshold.

3. The rating feature classification method for a technology enterprise rating model according to claim 1, characterized in that, The process of extracting partial target features from visual clustering regions includes, The text composed of target features in the visual clustering region is divided into several text segments; Extract a number of text segments at predetermined extraction intervals.

4. The rating feature classification method for a technology enterprise rating model according to claim 1, characterized in that, Setting classification labels for each of the aforementioned visual clustering regions includes, If the correlation parameter is greater than or equal to a predetermined correlation standard threshold, then an association label is set for the visual clustering region; If the correlation parameter is less than a predetermined correlation standard threshold, then a fuzzy label is set for the visual clustering region.

5. The rating feature classification method for a technology enterprise rating model according to claim 1, characterized in that, Based on the classification labels, the target features in each visual clustering region are analyzed and extracted, including... If the visual clustering region is an associated label, then identify the labeling keywords that are strongly associated with the search target in the visual clustering region, determine the position information of the labeling keywords in the visual clustering region, filter the labeling keyword group used for the position of the fuzzy region of the search target, determine the fuzzy region of the search target in the visual clustering region, remove the target features in the fuzzy region of the search target, and retain the remaining target features. If the visual clustering region is a fuzzy label, then all target features in the visual clustering region are removed.

6. The rating feature classification method for a technology enterprise rating model according to claim 1, characterized in that, Obtain several search interfaces associated with the search target, wherein, If the semantic relevance between the title of the search interface and the search target is greater than a predetermined relevance standard threshold, then the search interface is determined to be associated with the search target.

7. The rating feature classification method for a technology enterprise rating model according to claim 1, characterized in that, Target features include textual features and symbolic features.

Images