Enterprise Recommendation Method, Recommendation Device, and Electronic Device Based on Knowledge Graph

Abstract

The present invention discloses an enterprise recommendation method, a recommendation device and an electronic device based on a knowledge graph. Among them, the recommendation method includes: collecting at least one piece of news information associated with an enterprise object, where each piece of news information includes at least: a news lead and comment information; analyzing an emotional label corresponding to the news information based on the comment information; constructing a public opinion knowledge graph corresponding to the target event indicated by the news information based on the news lead; establishing an association relationship between the public opinion knowledge graph of the target event and the enterprise knowledge graph of the target enterprise object to obtain a graph association relationship; and if the emotional label is a positive label, outputting an enterprise recommendation list based on the graph association relationship. The present invention solves the technical problem in the related art that the impact on other associated enterprises is not considered according to public opinion, resulting in the recommended enterprise list not meeting the investment promotion requirements.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular, to an enterprise recommendation method, a recommendation device, and an electronic device based on a knowledge graph. Background Art

[0002] With the rapid development of Internet technology, online media has become the main platform for information dissemination and people's mutual communication, and has also become the main carrier for the formation and spread of online public opinion. Taking microblog public opinion as an example, the number of microblog users is huge, and microblog media has characteristics such as virtualization, rapidity, diversification, openness, anonymity, and interactivity. The explosion of some enterprise public opinions is extremely likely to cause a social response and also affect the government's investment promotion work.

[0003] In related technologies, existing public opinion analysis systems generally start from the perspective of enterprises, providing enterprises with real-time query of public opinion, real-time tracking and analysis of the development trend of public opinion, and multi-form automatic early warning, etc., so that enterprises can timely handle public opinion. However, the existing public opinion analysis systems have the following defects: (1) They do not track and analyze the public opinions of each enterprise from a more fine-grained knowledge level to determine whether there is an impact on other enterprises; (2) They do not consider from the perspective of the government whether the public opinion information of these enterprises will bring investment promotion clues to the staff when the government is engaged in investment promotion work.

[0004] The existing investment promotion models include: industrial chain investment promotion, on-site investment promotion, investment promotion by business, investment promotion by friendship, investment promotion by events and entrusted investment promotion. Among them, industrial chain investment promotion adopts the chain leader system, with major officials of local governments serving as "chain leaders", focusing on characteristic industries, and focusing on attracting a number of benchmark enterprises in the industrial chain and targeted high-quality enterprises to form industrial clusters, so as to attract better-quality enterprises; on-site investment promotion is to visit and talk with well-known enterprises, communicate and exchange with various chambers of commerce and associations, strengthen contacts with successful people abroad, and improve regional network maps; investment promotion by business is to fully mobilize the enthusiasm of the investment promotion work of the settled enterprises, and introduce partners and upstream and downstream supporting enterprises; investment promotion by friendship is investment promotion through events such as returning home during the New Year and other holidays; event investment promotion is to hold some investment promotion activities; entrusted investment promotion is to formulate incentive and reward mechanisms for intermediary investment promotion and entrusted investment promotion, to carry out commercialization, and to hire a group of intermediary agencies as investment promotion representatives, investment promotion ambassadors, investment promotion consultants, etc. However, the current investment promotion method has the following defects in the face of hundreds of millions of enterprises: (1) Since the attributes of the enterprises under consideration are relatively single, the investment promotion effect will be poor; (2) The government's publicity costs are high, but the results are minimal, and it is impossible to determine the target customers to be attracted; (3) Due to information overload, it is difficult for the government to fully understand the interests and hobbies of an enterprise, as well as the enterprise's characteristic industries, and merchants cannot fully obtain relevant policy information from the government; (4) The impact on other related enterprises is not considered based on public opinion, and it is impossible to provide a recommended list based on public opinion, resulting in the recommended list of enterprises not meeting the investment promotion needs.

[0005] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0006] The embodiments of the present invention provide a method and device for recommending enterprises based on knowledge graphs, and an electronic device, so as to at least solve the technical problem in the related art that the impact on other related enterprises is not considered according to public opinion, resulting in the recommended enterprise list not meeting the investment promotion needs.

[0007] According to one aspect of an embodiment of the present invention, there is provided an enterprise recommendation method based on a knowledge graph, including: collecting at least one piece of news information associated with an enterprise object, where each piece of news information includes at least: a news lead and comment information; analyzing an emotion tag corresponding to the news information based on the comment information; constructing a public opinion knowledge graph corresponding to a target event indicated by the news information based on the news lead; establishing an association relationship between the public opinion knowledge graph of the target event and the enterprise knowledge graph of the target enterprise object to obtain a graph association relationship; if the emotion tag is a positive tag, outputting an enterprise recommendation list based on the graph association relationship, where the recommendation list includes associated enterprise objects related to the target enterprise object in the public opinion knowledge graph and associated enterprise objects related to the target enterprise object in the enterprise knowledge graph.

[0008] Optionally, the step of collecting at least one piece of news information associated with an enterprise object includes: collecting a plurality of news titles with a sorting sequence greater than a preset sequence threshold to obtain a title set; performing word segmentation processing and part-of-speech tagging processing on each news title in the title set to obtain a word set; traversing the noun words in the word set, and if there is a noun word of the enterprise object in the word set, collecting the news lead and comment information in the target news corresponding to the target news title to obtain the news information.

[0009] Optionally, the step of analyzing an emotion tag corresponding to the news information based on the comment information includes: extracting keywords in the comment information; performing emotion analysis on each piece of comment information to obtain an emotion distribution statistical chart corresponding to the news information; adding an emotion tag to the news information based on the keywords and the emotion distribution statistical chart.

[0010] Optionally, the step of constructing a public opinion knowledge graph corresponding to a target event indicated by the news information based on the news lead includes: performing information extraction processing on the news lead to obtain object information of the target enterprise object that appears in the news information, where the information extraction processing method includes at least one of the following: entity extraction, attribute extraction, relationship extraction; constructing a public opinion knowledge graph corresponding to the target event indicated by the news information based on the object information of the target enterprise object.

[0011] Optionally, after constructing a public opinion knowledge graph corresponding to a target event indicated by the news information based on the news lead, the enterprise recommendation method further includes: associating the public opinion knowledge graph with a public opinion knowledge base to obtain an updated public opinion knowledge graph, where the public opinion knowledge base pre-stores the public opinion knowledge graph obtained in the historical process; storing the updated public opinion knowledge graph in a graph database.

[0012] Optionally, before establishing the association relationship between the public opinion knowledge graph of the target event and the enterprise knowledge graph of the target enterprise object, the enterprise recommendation method further includes: obtaining the attribute information of the target enterprise object and enterprise relationship data; constructing an enterprise knowledge graph based on the attribute information and the enterprise relationship data.

[0013] Optionally, after constructing the enterprise knowledge graph, the enterprise recommendation method further includes: obtaining the online news of the target enterprise object; preprocessing the online news to obtain processed online news data; performing fusion processing on the enterprise data in the online news data and the target enterprise object; extracting the enterprise data of other enterprise objects except the target enterprise object in the online news data; supplementing the enterprise data of the other enterprise objects and the object relationship between the other enterprise objects and the target enterprise object to the enterprise knowledge graph of the target enterprise object.

[0014] Optionally, if the sentiment label is a positive label, the step of outputting an enterprise recommendation list based on the graph association relationship includes: if the sentiment label is a positive label, obtaining the graph distance, the number of associations, the first ranking information of the associated enterprise object, and the second ranking information of the industrial chain to which the associated enterprise object belongs between each associated enterprise object and the target enterprise object based on the graph association relationship; determining the first weight information corresponding to the graph distance, the second weight information corresponding to the number of associations, the third weight information corresponding to the first ranking information, and the fourth weight information corresponding to the second ranking information; calculating the ranking parameter of each associated enterprise object based on the graph distance and the corresponding first weight information, the number of associations and the corresponding second weight information, the first ranking information and the corresponding third weight information, and the second ranking information and the corresponding fourth weight information; outputting an enterprise recommendation list based on the ranking parameter.

[0015] Optionally, after outputting the enterprise recommendation list based on the graph association relationship, the enterprise recommendation method further includes: calculating the label score of each piece of news information about the target enterprise object that appears within a preset time period, where the label score includes: a first label score or a second label score, the first label score represents the score of a positive sentiment label, and the second label score represents the score of a negative sentiment label; accumulating the label scores of all news information to obtain a total label score; adjusting the weight information of the comment information for the associated enterprise object based on the total label score to adjust the ranking parameter of the associated enterprise object.

[0016] According to another aspect of the embodiments of the present invention, there is also provided an enterprise recommendation device based on a knowledge graph, including: a collection unit, configured to collect at least one piece of news information associated with an enterprise object, where each piece of the news information at least includes: a news lead and comment information; an analysis unit, configured to analyze an emotion tag corresponding to the news information based on the comment information; a construction unit, configured to construct a public opinion knowledge graph corresponding to a target event indicated by the news information based on the news lead; an establishment unit, configured to establish an association relationship between the public opinion knowledge graph of the target event and the enterprise knowledge graph of the target enterprise object to obtain a graph association relationship; an output unit, configured to, if the emotion tag is a positive tag, output an enterprise recommendation list based on the graph association relationship, where the recommendation list includes an associated enterprise object related to the target enterprise object in the public opinion knowledge graph and an associated enterprise object related to the target enterprise object in the enterprise knowledge graph.

[0017] Optionally, the collection unit includes: a first collection module, configured to collect a plurality of news titles with a sorting sequence greater than a preset sequence threshold to obtain a title set; a first processing module, configured to perform word segmentation processing and part-of-speech tagging processing on each news title in the title set to obtain a word set; a first traversal module, configured to traverse the noun words in the word set, and if there is a noun word of the enterprise object in the word set, collect the news lead and comment information in the target news corresponding to the target news title to obtain the news information.

[0018] Optionally, the analysis unit includes: a first extraction module, configured to extract keywords in the comment information; a first analysis module, configured to perform emotion analysis on each piece of the comment information to obtain an emotion distribution statistical chart corresponding to the news information; a first addition module, configured to add an emotion tag to the news information based on the keywords and the emotion distribution statistical chart.

[0019] Optionally, the construction unit includes: a second processing module, configured to perform information extraction processing on the news lead to obtain object information of the target enterprise object appearing in the news information, where the information extraction processing method includes at least one of the following: entity extraction, attribute extraction, relationship extraction; a first construction module, configured to construct a public opinion knowledge graph corresponding to the target event indicated by the news information based on the object information of the target enterprise object.

[0020] Optionally, the enterprise recommendation device further includes: a first association module, configured to, after constructing an opinion knowledge graph corresponding to the target event indicated by the news lead, associate the opinion knowledge graph with an opinion knowledge base to obtain an updated opinion knowledge graph, where the opinion knowledge base pre-stores opinion knowledge graphs obtained in the historical process; a first storage module, configured to store the updated opinion knowledge graph into a graph database.

[0021] Optionally, the enterprise recommendation device further includes: a first acquisition module, configured to acquire attribute information and enterprise relationship data of a target enterprise object before establishing an association relationship between the opinion knowledge graph of the target event and the enterprise knowledge graph of the target enterprise object; a second construction module, configured to construct an enterprise knowledge graph based on the attribute information and the enterprise relationship data.

[0022] Optionally, the enterprise recommendation device further includes: a second acquisition module, configured to acquire online news of the target enterprise object after constructing the enterprise knowledge graph; a third processing module, configured to preprocess the online news to obtain processed online news data; a fourth processing module, configured to perform a fusion process on enterprise data in the online news data and the target enterprise object; a second extraction module, configured to extract enterprise data of other enterprise objects except the target enterprise object in the online news data; a first supplementation module, configured to supplement the enterprise data of the other enterprise objects and the object relationship between the other enterprise objects and the target enterprise object to the enterprise knowledge graph of the target enterprise object.

[0023] Optionally, the output unit includes: a third acquisition module, configured to, if the sentiment label is a positive label, acquire, based on the graph association relationship, the graph distance, the number of associations, the first ranking information of each associated enterprise object, and the second ranking information of the industrial chain to which the associated enterprise object belongs, between each associated enterprise object and the target enterprise object; a first determination module, configured to determine first weight information corresponding to the graph distance, second weight information corresponding to the number of associations, third weight information corresponding to the first ranking information, and fourth weight information corresponding to the second ranking information; a first calculation module, configured to calculate a ranking parameter of each associated enterprise object based on the graph distance and the corresponding first weight information, the number of associations and the corresponding second weight information, the first ranking information and the corresponding third weight information, and the second ranking information and the corresponding fourth weight information; and output an enterprise recommendation list based on the ranking parameter.

[0024] Optionally, the enterprise recommendation device further includes: a second calculation module, configured to calculate a tag score for each piece of news information about the target enterprise object that appears within a preset time period after outputting an enterprise recommendation list based on the graph association relationship, where the tag score includes: a first tag score or a second tag score, the first tag score represents the score of a positive sentiment tag, and the second tag score represents the score of a negative sentiment tag; a first accumulation module, configured to accumulate the tag scores of all news information to obtain a total tag score; a first adjustment module, configured to adjust the weight information of the associated enterprise object by the comment information based on the total tag score, so as to adjust the ranking parameter of the associated enterprise object.

[0025] According to another aspect of the embodiments of the present invention, there is also provided an electronic device, including: a processor; and a memory for storing executable instructions of the processor; wherein, the processor is configured to execute the enterprise recommendation method based on a knowledge graph as described in any one of the above via executing the executable instructions.

[0026] According to another aspect of the embodiments of the present invention, there is also provided a computer-readable storage medium, the computer-readable storage medium includes a stored computer program, wherein, when the computer program runs, it controls the device where the computer-readable storage medium is located to execute the enterprise recommendation method based on a knowledge graph as described in any one of the above.

[0027] In this application, at least one piece of news information associated with an enterprise object is collected, where each piece of news information includes at least: a news lead and a comment information. Based on the comment information, the sentiment tags corresponding to the news information are analyzed. Based on the news lead, a public opinion knowledge graph corresponding to the target event indicated by the news information is constructed. An association relationship is established between the public opinion knowledge graph of the target event and the enterprise knowledge graph of the target enterprise object to obtain a graph association relationship. If the sentiment tag is a positive tag, an enterprise recommendation list is output based on the graph association relationship. By analyzing the collected real-time hot news, screening out the public opinion information related to the enterprise, this application can analyze the sentiment tendency of the public opinion and construct a public opinion knowledge graph, associate according to the associated entities of the public opinion knowledge graph and the enterprise knowledge graph, and according to the association analysis result, it can provide a suitable recommended enterprise list for local investment promotion government personnel, thereby solving the technical problem in the related art that the recommended enterprise list does not meet the investment promotion requirements because the impact on other associated enterprises is not considered according to the public opinion. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The drawings described herein are used to provide a further understanding of the present invention, and constitute a part of this application. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention. In the drawings:

[0029] Figure 1 is a flowchart of an optional enterprise recommendation method based on a knowledge graph according to an embodiment of the present invention;

[0030] Figure 2 is a schematic diagram of the association between an optional public opinion knowledge graph (b) and an enterprise knowledge graph (a) according to an embodiment of the present invention;

[0031] Figure 3 is a schematic diagram of an enterprise recommendation device based on a knowledge graph according to an embodiment of the present invention. Detailed implementation manners

[0032] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0033] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above drawings are used to distinguish similar objects, and do not necessarily need to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those clearly listed steps or units, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0034] To facilitate the understanding of the present invention by those skilled in the art, some terms or nouns involved in the embodiments of the present invention are explained below:

[0035] Public opinion: refers to the social attitudes generated and held by the public as the subject towards social managers, enterprises, individuals and other various organizations and their political, social, moral and other aspects around the occurrence, development and change of intermediate social events within a certain social space.

[0036] Knowledge graph: is a series of various graphs showing the development process and structural relationship of knowledge, using visualization technology to describe knowledge resources and their carriers, mining, analyzing, constructing, drawing and displaying knowledge and their mutual connections.

[0037] The following embodiments of the present invention can be applied to various investment promotion systems / enterprise recommendation applications or other scenarios for recommending enterprises. The types of investment promotion involved include, but are not limited to: government investment promotion (for example, the government needs to transfer land for investment promotion, needs to invest for investment promotion, etc.), investment promotion of various enterprises (for example, enterprises seeking investment partners), etc. The present invention crawls hot search news in real time, screens out public opinion information related to enterprises, analyzes whether the sentiment tendency of the public opinion is positive or negative according to sentiment analysis technology, and uses the automatic construction technology of the public opinion knowledge graph to construct and store the public opinion knowledge graph in the knowledge base. In addition, according to the associated entities of the public opinion knowledge graph, an association is made with the enterprise knowledge graph. Constraints are added to the recommended results based on factors such as the distance between the candidate enterprise and the main enterprise, the number of associations between the candidate enterprise and the main enterprise, the basic situation of the candidate enterprise, and the industrial chain that the local investment promotion bureau is more interested in. The recommended or non-recommended enterprises are re-screened and sorted to provide investment promotion clues for investment promotion personnel.

[0038] Embodiment 1

[0039] According to an embodiment of the present invention, there is provided an embodiment of an enterprise recommendation method based on a knowledge graph. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0040] Figure 1 is a flowchart of an optional enterprise recommendation method based on a knowledge graph according to an embodiment of the present invention, as Figure 1 shown. The method includes the following steps:

[0041] Step S102, collect at least one piece of news information associated with an enterprise object, where each piece of news information includes at least: a news lead and comment information.

[0042] Step S104, analyze the sentiment label corresponding to the news information based on the comment information.

[0043] Step S106, construct a public opinion knowledge graph corresponding to the target event indicated by the news information based on the news lead.

[0044] Step S108, establish an association relationship between the public opinion knowledge graph of the target event and the enterprise knowledge graph of the target enterprise object to obtain a graph association relationship.

[0045] Step S110, if the sentiment label is a positive label, output a list of recommended enterprises based on the graph association relationship, where the recommended list includes the associated enterprise objects related to the target enterprise object in the public opinion knowledge graph and the associated enterprise objects related to the target enterprise object in the enterprise knowledge graph.

[0046] Through the above steps, at least one news information associated with the enterprise object can be collected. Each news information includes at least: a news lead and comment information. Based on the comment information, analyze the sentiment label corresponding to the news information. Based on the news lead, construct a public opinion knowledge graph corresponding to the target event indicated by the news information, establish an association relationship between the public opinion knowledge graph of the target event and the enterprise knowledge graph of the target enterprise object to obtain the graph association relationship. If the sentiment label is a positive label, output a list of recommended enterprises based on the graph association relationship. In the embodiment of the present invention, by analyzing the collected real-time hot news, screening out the public opinion information related to the enterprise, being able to analyze the sentiment tendency of the public opinion and construct a public opinion knowledge graph, associating the associated entities of the public opinion knowledge graph with the enterprise knowledge graph, and according to the association analysis result, being able to provide a suitable list of recommended enterprises for local investment promotion government personnel, thereby solving the technical problem in the related art that the list of recommended enterprises does not meet the investment promotion requirements due to not considering the impact on other associated enterprises according to the public opinion.

[0047] The embodiments of the present invention will be described in detail below in combination with the above steps.

[0048] Step S102, collect at least one news information associated with the enterprise object, where each news information includes at least: a news lead and comment information.

[0049] In the embodiment of the present invention, the web crawler technology can be used to crawl the network news (for example, the hot search list) every time τ to obtain the hot search title set T = {t1, t2,..., t i}, screen out the titles related to the enterprise from it, and then obtain the news lead (i.e., the summary of the news) and comment information (i.e., the public opinion set composed of all comments on the news) under the title news.

[0050] Optionally, the step of collecting at least one news information associated with the enterprise object includes: collecting multiple news titles whose sorting sequence is greater than a preset sequence threshold to obtain a title set; performing word segmentation processing and part-of-speech tagging processing on each news title in the title set to obtain a word set; traversing the noun words in the word set, and if there are noun words of the enterprise object in the word set, collect the news lead and comment information in the target news corresponding to the target news title to obtain the news information.

[0051] In the embodiments of the present invention, news titles sorted before a preset value (i.e., the sorting sequence is greater than a preset sequence threshold, for example, the hot search list shows 30 hot search titles, and only the first 20 titles are collected) are collected according to the existing sorting of online media (e.g., the hot search sorting of Weisou), and a hot search title set T = {t1, t2,..., t i} is obtained. The title sequence t i in the set T is subjected to word segmentation processing and part-of-speech tagging processing to obtain a word set W = {w1, w2,..., w j}. The noun words in the set W are traversed and aligned with the enterprise knowledge graph (i.e., it is determined whether there are noun words of enterprise objects in the word set). If there exists w i as an enterprise (i.e., there are noun words of enterprise objects in the word set), then the lead S i (i.e., the news lead) and public opinion O = {o1, o2,..., o i} (i.e., comment information) under the hot news t n (i.e., the target news corresponding to the target news title) are crawled to obtain news information.

[0052] Step S104, based on the comment information, analyze the sentiment label corresponding to the news information.

[0053] In the embodiments of the present invention, each comment in the public opinion set O can be traversed and analyzed to obtain the sentiment label of the news information, so as to determine whether the news is positive news or negative news.

[0054] Optionally, the step of analyzing the sentiment label corresponding to the news information based on the comment information includes: extracting keywords from the comment information; performing sentiment analysis on each comment information to obtain a sentiment distribution statistical chart corresponding to the news information; and adding a sentiment label to the news information based on the keywords and the sentiment distribution statistical chart.

[0055] In the embodiments of the present invention, a keyword extraction algorithm can be used to extract keywords from the public opinion set O (i.e., comment information). Each comment information in the public opinion set O can be traversed, and sentiment analysis technology can be used to perform sentiment analysis on the public opinion information o n to obtain the sentiment distribution statistical chart of the hot news t i . By combining the visual display of keywords to understand the key comments of public opinion and the statistical chart of sentiment distribution, a sentiment label f i is added to the topic t emo (i.e., news information). If the topic t i is a negative news of an enterprise, f emo = -1. If the topic t i is a positive news of an enterprise, f emo = 1.

[0056] Step S106: Based on the news lead, construct a public opinion knowledge graph corresponding to the target event indicated by the news information.

[0057] In the embodiments of the present invention, for each hot news t i (i.e., news information) updated in real time by each news application, the lead S i (i.e., news lead) is a summary of this hot news. Based on this news lead, a public opinion knowledge graph is constructed (i.e., construct a public opinion knowledge graph corresponding to the target event indicated by the news information).

[0058] Optionally, the step of constructing a public opinion knowledge graph corresponding to the target event indicated by the news information based on the news lead includes: performing information extraction processing on the news lead to obtain the object information of the target enterprise object that appears in the news information, where the information extraction processing methods include at least one of the following: entity extraction, attribute extraction, and relationship extraction; based on the object information of the target enterprise object, construct a public opinion knowledge graph corresponding to the target event indicated by the news information.

[0059] In the embodiments of the present invention, technologies such as entity extraction, attribute extraction, and relationship extraction can be used to process the lead S i (i.e., perform information extraction processing on the news lead) to obtain entity knowledge (i.e., the object information of the target enterprise object that appears in the news information, for example, person names, place names, person relationships, etc.). For an enterprise public opinion hot event, the core relationships include relevant persons, relevant places, relevant times, relevant events, the main enterprise, relevant enterprises, etc. Use knowledge graph construction technology to construct the public opinion knowledge graph KG i (i.e., the target event indicated by the news information) of this hot event. ti .

[0060] Optionally, after constructing a public opinion knowledge graph corresponding to the target event indicated by the news information based on the news lead, the enterprise recommendation method further includes: associating the public opinion knowledge graph with the public opinion knowledge base to obtain an updated public opinion knowledge graph, where the public opinion knowledge base pre-stores the public opinion knowledge graph obtained in the historical process; storing the updated public opinion knowledge graph in the graph database.

[0061] In the embodiments of the present invention, associate the obtained public opinion knowledge graph KG i of the hot event t ti with the public opinion knowledge base (which stores the previously obtained public opinion knowledge graph), that is, considering the time factor according to the characteristic that the public opinion of the enterprise changes dynamically over time, an updated enterprise public opinion knowledge graph KG op (i.e., the updated public opinion knowledge graph) can be obtained, and the enterprise public opinion knowledge graph KGop Store it in the graph database.

[0062] Step S108, establish an association relationship between the public opinion knowledge graph of the target event and the enterprise knowledge graph of the target enterprise object to obtain the graph association relationship.

[0063] In the embodiment of the present invention, the public opinion knowledge graph KG of the hot event t i (i.e., the target event) ti is associated with the enterprise knowledge graph KG En (i.e., the enterprise knowledge graph of the target enterprise object). That is, the main enterprise involved in the public opinion knowledge graph KG ti can be aligned with the enterprise in the enterprise knowledge graph KG En (i.e., the target enterprise object), or other attributes involved in the public opinion knowledge graph KG ti (such as related persons) can be used as auxiliary attributes for alignment to obtain the graph association relationship.

[0064] Optionally, before establishing the association relationship between the public opinion knowledge graph of the target event and the enterprise knowledge graph of the target enterprise object, the enterprise recommendation method further includes: obtaining the attribute information of the target enterprise object and the enterprise relationship data; constructing an enterprise knowledge graph based on the attribute information and the enterprise relationship data.

[0065] In the embodiment of the present invention, the attribute information of the target enterprise object can be obtained first. The attribute information can include: enterprise name, unified social credit code, registration date, enterprise type, legal representative, registered capital, business scope, province where located, region, registered address, etc. Then, the enterprise data website API is called according to the enterprise name or the unified social credit code to obtain the enterprise relationship data associated with the enterprise (such as branches, shareholdings, external investments, etc.). Then, based on the attribute information and the enterprise relationship data, the enterprise knowledge graph KG En .

[0066] Optionally, after constructing the enterprise knowledge graph, the enterprise recommendation method further includes: obtaining the online news of the target enterprise object; preprocessing the online news to obtain the processed online news data; performing a fusion process on the enterprise data in the online news data with the target enterprise object; extracting the enterprise data of other enterprise objects except the target enterprise object in the online news data; supplementing the enterprise data of other enterprise objects and the object relationship between other enterprise objects and the target enterprise object to the enterprise knowledge graph of the target enterprise object.

[0067] In an embodiment of the present invention, the enterprise knowledge graph can be expanded using online news, that is, using web crawler technology to crawl news data (i.e., online news) related to the enterprise (i.e., the target enterprise object), obtaining relevant online news, and then preprocessing the online news (for example, data cleaning, etc.) to obtain normalized data (i.e., processed online news data). The content mentioned in the online news data (such as enterprise name, legal representative, etc.) is aligned with the target enterprise object (i.e., fusion processing), and the names of the persons and enterprise names mentioned in the news other than the target enterprise (i.e., enterprise data of other enterprise objects) are extracted and stored as implicit relationships of the enterprise to expand the enterprise knowledge graph (i.e., supplement it to the enterprise knowledge graph of the target enterprise object).

[0068] Step S110, if the sentiment label is a positive label, then output an enterprise recommendation list based on the graph association relationship, where the recommendation list includes associated enterprise objects related to the target enterprise object in the public opinion knowledge graph and associated enterprise objects related to the target enterprise object in the enterprise knowledge graph.

[0069] In an embodiment of the present invention, according to the sentiment tendency indicated by the sentiment label, if it is positive, then perform an association analysis on the public opinion knowledge graph KG ti and the enterprise knowledge graph KG En After that, it can provide investment promotion clues (i.e., provide an enterprise recommendation list) for investment promotion. The recommendation list can include associated enterprise objects related to the target enterprise object in the public opinion knowledge graph and associated enterprise objects related to the target enterprise object in the enterprise knowledge graph (such as foreign investment, branch offices, etc.).

[0070] Optionally, the step of outputting an enterprise recommendation list based on the graph association relationship if the sentiment label is a positive label includes: if the sentiment label is a positive label, obtain the graph distance, the number of associations, the first ranking information of the associated enterprise object, and the second ranking information of the industrial chain to which the associated enterprise object belongs between each associated enterprise object and the target enterprise object based on the graph association relationship; determine the first weight information corresponding to the graph distance, the second weight information corresponding to the number of associations, the third weight information corresponding to the first ranking information, and the fourth weight information corresponding to the second ranking information; calculate the ranking parameter of each associated enterprise object based on the graph distance and the corresponding first weight information, the number of associations and the corresponding second weight information, the first ranking information and the corresponding third weight information, and the second ranking information and the corresponding fourth weight information; output an enterprise recommendation list based on the ranking parameter.

[0071] In an embodiment of the present invention, the sentiment label f i corresponding to the news information for the target event t emoWhether the value is -1 or 1 is used to distinguish whether the sentiment polarity of this piece of public opinion is negative or positive. If the public opinion about the enterprise is positive news, then the related enterprises in the public opinion knowledge graph KG op and the related enterprises in the related enterprise knowledge graph (such as foreign investments, branch offices, etc.) can be recommended. Among them, the recommended enterprises can be sorted according to the following aspects, including but not limited to: the distance (i.e., the graph distance) x1 between the candidate enterprise (i.e., each related enterprise object obtained from the graph association relationship) and the main enterprise (i.e., the target enterprise object), the number of associations x2 between the candidate enterprise and the main enterprise, the basic situation ranking of the candidate enterprise (i.e., the first ranking information, for example, the ranking in a certain publicly recognized ranking list) x3, the industrial chain ranking that the local investment promotion bureau is more interested in (i.e., the second ranking information) x4, etc. For candidate enterprises, the shorter the distance from the main enterprise, the more recommended; the more associations, the more recommended; the higher the basic situation of the enterprise, the more recommended; and the higher the industrial chain ranking, the more recommended. Therefore, the recommended enterprise ranking can adopt the following formula: RANK = MAX(βx2 - αx1 - γx3 - δx4), where α, β, γ, and δ respectively represent different weights (i.e., the first weight information corresponding to the graph distance, the second weight information corresponding to the number of associations, the third weight information corresponding to the first ranking information, and the fourth weight information corresponding to the second ranking information). The specific weight settings can be flexibly determined according to the different attributes valued by each investment promotion bureau, and then the recommended list is output.

[0072] Another option is that if the public opinion about the enterprise is negative news, then the public opinion knowledge graph KG opAssociated enterprises in it and associated enterprises in the associated enterprise knowledge graph (such as external investments, branches, etc.) are not recommended. Among them, the non-recommended enterprises can be ranked according to the following aspects, including but not limited to: the distance (i.e., graph distance) x1 between the candidate enterprise (i.e., each associated enterprise object obtained from the graph association relationship) and the main enterprise (i.e., the target enterprise object), the number of associations x2 between the candidate enterprise and the main enterprise, the basic situation ranking of the candidate enterprise (i.e., the first ranking information, for example, the ranking in a certain publicly recognized ranking list) x3, the industrial chain ranking (i.e., the second ranking information) x4 that the local investment promotion bureau is more interested in, etc. For candidate enterprises, the shorter the distance from the main enterprise, the less recommended; the more associations, the less recommended; the lower the basic situation of the enterprise, the less recommended; and the lower the industrial chain ranking, the less recommended. Therefore, the non-recommended enterprise ranking can adopt the following formula: RANK = MAX(βx2 - αx1 + γx3 + δx4), where α, β, γ, and δ respectively represent different weights (i.e., the first weight information corresponding to the graph distance, the second weight information corresponding to the number of associations, the third weight information corresponding to the first ranking information, and the fourth weight information corresponding to the second ranking information). The specific weight settings can be flexibly determined according to the different attributes valued by each investment promotion bureau, and then the non-recommended list is output.

[0073] Optionally, after outputting the enterprise recommendation list based on the graph association relationship, the enterprise recommendation method further includes: calculating the label score of each news information among multiple news information about the target enterprise object that appears within a preset time period, where the label score includes: the first label score or the second label score, the first label score represents the score of the positive sentiment label, and the second label score represents the score of the negative sentiment label; accumulating the label scores of all news information to obtain the total label score; and adjusting the weight information of the associated enterprise object for the review information based on the total label score to adjust the ranking parameter of the associated enterprise object.

[0074] In the embodiment of the present invention, if the enterprise (i.e., the target enterprise object) has multiple hot public opinion events (i.e., news information) within a period of time (i.e., within the preset time period), there are positive public opinions (indicating that the sentiment label is a positive label, i.e., the first label) and negative public opinions (indicating that the sentiment label is a negative label, i.e., the second label), then consider the superimposed effect of the negative public opinion and the positive public opinion on the candidate enterprise, that is, accumulate and calculate the score of the positive sentiment label and the score of the negative sentiment label, and adjust the weight information of the associated enterprise object for the review information according to the accumulated total label score to rank the recommended enterprises (i.e., adjust the ranking parameter of the associated enterprise object).

[0075] In the embodiments of the present invention, after the public opinion knowledge graph and the enterprise knowledge graph are associated and analyzed, the knowledge graph can intuitively represent the relationships between enterprises, making the knowledge extensible. Moreover, after using the correlation analysis of the knowledge graph, it can provide investment promotion clues (such as recommended lists) for the personnel of the investment promotion bureau. The knowledge graph makes the knowledge inferable, and while providing investment promotion clues, it makes the recommended enterprise list interpretable.

[0076] Embodiment 2

[0077] The embodiments of the present invention propose an investment promotion recommendation method based on the association of the public opinion knowledge graph, which can crawl the hot search news in real time, screen out the public opinion information related to enterprises, analyze whether the sentiment tendency of the public opinion is positive or negative according to sentiment analysis technology, and use the automatic construction technology of the public opinion knowledge graph to construct and store the public opinion knowledge graph in the knowledge base. In addition, according to the associated entities of the public opinion knowledge graph and the enterprise knowledge graph, constraints are added to the recommendation results according to factors such as the distance between the candidate enterprise and the main enterprise, the number of associations between the candidate enterprise and the main enterprise, the basic situation of the candidate enterprise, and the industrial chain that the local investment promotion department is interested in, and the recommended or non-recommended enterprises are re-screened and sorted to provide investment promotion clues for the local investment promotion government personnel. The specific steps are as follows:

[0078] Step 1: Information collection;

[0079] Step 1.1: Use web crawler technology to crawl the news hot search list every time interval τ to obtain the hot search title set T = {t1, t2,..., t i};

[0080] Step 1.2: Perform word segmentation and annotation on the title sequence t i in the set T to obtain the word set W = {w1, w2,..., w j}. Traverse the noun words in the set W and align them with the background enterprise knowledge graph. If there is w i as an enterprise, then crawl the lead S i under the hot news t i and the public opinion O = {o1, o2,..., o n};

[0081] Step 2: Multi-dimensional data analysis of public opinion;

[0082] Step 2.1: Use the keyword extraction algorithm to extract keywords from the public opinion set O obtained in Step 1;

[0083] Step 2.2: Traverse each comment information in the public opinion set O, and use sentiment analysis technology to analyze the public opinion information o nConduct sentiment analysis and obtain the sentiment distribution statistical chart of the hot news item t i ;

[0084] Step 2.3: Combine the visual display of keywords to understand the key comments of public opinion and the statistical chart of sentiment distribution, and assign a sentiment label f i to the topic t emo . If the topic t i is negative news about an enterprise, f emo = -1. If the topic t i is positive news about an enterprise, f emo = 1;

[0085] Step 3: Construction of the public opinion knowledge graph;

[0086] Step 3.1: The lead S i under each hot news item t i is a summary of the hot news item. Based on this lead S i , construct the public opinion knowledge graph.

[0087] Step 3.1.1: Use entity extraction, attribute extraction, and relationship extraction technologies to process the lead S i to obtain knowledge. For an enterprise public opinion hot event, the core relationships include relevant people, relevant locations, relevant times, relevant events, the main enterprise, related enterprises, etc. And use the knowledge graph construction technology to construct the public opinion knowledge graph KG i of this hot event t ti ;

[0088] Step 3.2: Associate the obtained public opinion knowledge graph KG i of the hot event t ti with the public opinion knowledge base, and store the enterprise public opinion knowledge graph KG op in the graph database.

[0089] Step 4: Construction of the enterprise knowledge graph;

[0090] Step 4.1: Obtain enterprise attribute information, and the following enterprise attribute information can be obtained: enterprise name, unified social credit code, registration date, enterprise type, legal representative, registered capital, business scope, province where located, region, registered address, etc.;

[0091] Step 4.2: Obtain enterprise relationships. According to the enterprise name or unified social credit code, call the enterprise data website API to obtain enterprise data associated with this enterprise, such as branches, shareholdings, external investments, etc., and construct the enterprise knowledge graph KG En ;

[0092] Step 4.3: Expand the enterprise knowledge graph using network data;

[0093] Step 4.3.1: Use web crawler technology to crawl news data related to the enterprise;

[0094] Step 4.3.2: Preprocess the news data, such as data cleaning, etc. After obtaining relatively normalized data, align the content mentioned in the news, such as enterprise names, legal representatives, etc., with the original target enterprise;

[0095] Step 4.3.3: Extract the names of persons and enterprise names mentioned in the news other than the original target enterprise, store them as implicit relationships of the enterprise, and expand the enterprise knowledge graph;

[0096] Step 5: Associate the enterprise knowledge graph;

[0097] Step 5.1: Associate the public opinion knowledge graph KG i of the hot event t ti with the enterprise knowledge graph KG En ;

[0098] Step 5.1.1: Align the main enterprise involved in the public opinion knowledge graph KG ti with the enterprises in the enterprise knowledge graph KG En . Other attributes involved in the public opinion knowledge graph KG ti , such as relevant persons, can be aligned as auxiliary attributes. As Figure 2 shown, Figure 2 in (a) is the enterprise knowledge graph of Company A, including: company name, address, number of people, slogan, trademark, time, funds, relevant persons, etc., Figure 2 in (b) is the public opinion knowledge graph of news about Company A, including: comment keywords, funds, time, events, relevant persons, provinces, relevant enterprises, similar events of relevant enterprises, etc. Align with Company A as the main enterprise,

[0099] and align with the relevant person "Mr. Wu" as an auxiliary attribute.

[0100] Step 6: Recommend (or not recommend) enterprises;

[0101] Step 6.1: Conduct correlation analysis based on the public opinion knowledge graph KG ti and the enterprise knowledge graph KG En to provide investment promotion clues for investment promotion;

[0102] Step 6.1.1: According to the sentiment label f i of the public opinion topic t emoThe value of -1 or 1 is used to distinguish whether the sentiment polarity of this piece of public opinion is negative or positive;

[0103] Step 6.1.2: If the public opinion about the enterprise is positive news, then the associated enterprises in the public opinion knowledge graph KG op and the associated enterprises in the associated enterprise knowledge graph (such as foreign investments, branch offices, etc.) can be

[0104] recommended;

[0105] Step 6.1.2.1: The TOPN enterprises recommended in Step 6.1.2 can be sorted according to the following aspects, including but not limited to the distance (number of hops in the knowledge graph) x1 between the candidate enterprise and the main enterprise, the number of associations x2 between the candidate enterprise and the main enterprise, the basic situation ranking x3 of the candidate enterprise, the industrial chain ranking x4 that the local investment promotion bureau is more interested in, etc.

[0106] RANK = MAX(βx2 - αx1 - γx3 - δx4) (1);

[0107] For candidate enterprises, the shorter the distance from the main enterprise, the more recommended; the more associations, the more recommended; the higher the basic situation of the enterprise, the more recommended; and the higher the industrial chain ranking, the more recommended. Therefore, the recommended enterprise

[0108] ranking is as shown in formula (1);

[0109] In addition, each investment promotion bureau values different attributes. α, β, γ, and δ are respectively used to represent different factors considered in the ranking, which are flexibly determined by the local investment promotion bureau;

[0110] Step 6.1.3: If the public opinion about the enterprise is negative news, then the associated enterprises in the public opinion knowledge graph KG op and the associated enterprises in the associated enterprise knowledge graph (such as foreign investments, branch offices, etc.) are not recommended;

[0111] Step 6.1.3.1: The TOPN enterprises not recommended in Step 6.1.3 can be sorted according to the following aspects, including but not limited to the distance (number of hops in the knowledge graph) x1 between the candidate enterprise and the main enterprise, the number of associations x2 between the candidate enterprise and the main enterprise, the basic situation ranking x3 of the candidate enterprise, the local investment promotion bureau

[0112] is more interested in the industrial chain ranking x4, etc.;

[0113] RANK = MAX(βx2 - αx1 + γx3 + δx4) (2);

[0114] For candidate enterprises, the shorter the distance from the main enterprise, the less recommended; the more associations, the less recommended; the more backward the basic situation of the enterprise, the less recommended; and the more backward the industrial chain ranking, the less recommended. Therefore, the recommended enterprise ranking is as shown in Equation (2).

[0115] Similar to Step 6.1.2.1, each investment promotion bureau values different attributes, and α, β, γ, and δ are respectively used to represent different factors considered in the ranking, which are flexibly determined by the local investment promotion bureau.

[0116] Step 6.2: Since the public opinion of enterprises changes dynamically over time, the time factor can also be considered, and then correlation analysis is carried out based on the public opinion knowledge graph and the enterprise knowledge graph to provide investment promotion clues for investment promotion.

[0117] Step 6.2.1: If a candidate enterprise has had multiple hot public opinion events during a period of time, including both positive and negative public opinions, then the superposition effect of the influence of negative and positive public opinions on the candidate enterprise can be considered, and the weight of the public opinion in the most recent period is relatively large to rank the recommended (or non-recommended) enterprises.

[0118] The embodiments of the present invention have the following beneficial effects:

[0119] (1) After correlatively analyzing the public opinion knowledge graph and the enterprise knowledge graph, the knowledge graph can intuitively represent the relationships between enterprises, making the knowledge extensible.

[0120] (2) After using the correlation analysis of the knowledge graph, investment promotion clues can be discovered for investment promotion personnel.

[0121] (3) The knowledge graph makes the knowledge inferable.

[0122] (4) While providing investment promotion clues, the list of recommended (or non-recommended) enterprises is made interpretable.

[0123] Embodiment III

[0124] A knowledge-graph-based enterprise recommendation device provided in this embodiment includes multiple implementation units, and each implementation unit corresponds to each implementation step in Embodiment I above.

[0125] Figure 3 is a schematic diagram of a knowledge-graph-based enterprise recommendation device according to an embodiment of the present invention, as Figure 3 shown. The recommendation device may include: a collection unit 30, an analysis unit 32, a construction unit 34, an establishment unit 36, and an output unit 38, where

[0126] The acquisition unit 30 is used to acquire at least one piece of news information associated with an enterprise object, where each piece of news information includes at least: a news lead and comment information;

[0127] The analysis unit 32 is used to analyze the sentiment label corresponding to the news information based on the comment information;

[0128] The construction unit 34 is used to construct a public opinion knowledge graph corresponding to the target event indicated by the news information based on the news lead;

[0129] The establishment unit 36 is used to establish an association relationship between the public opinion knowledge graph of the target event and the enterprise knowledge graph of the target enterprise object to obtain a graph association relationship;

[0130] The output unit 38 is used to output an enterprise recommendation list based on the graph association relationship if the sentiment label is a positive label, where the recommendation list includes the associated enterprise objects related to the target enterprise object in the public opinion knowledge graph and the associated enterprise objects related to the target enterprise object in the enterprise knowledge graph.

[0131] The above-mentioned recommendation device can acquire at least one piece of news information associated with an enterprise object through the acquisition unit 30, where each piece of news information includes at least: a news lead and comment information. Based on the comment information, the analysis unit 32 analyzes the sentiment label corresponding to the news information. Based on the news lead, the construction unit 34 constructs a public opinion knowledge graph corresponding to the target event indicated by the news information. The establishment unit 36 establishes an association relationship between the public opinion knowledge graph of the target event and the enterprise knowledge graph of the target enterprise object to obtain a graph association relationship. The output unit 38 outputs an enterprise recommendation list based on the graph association relationship when the sentiment label is a positive label. In the embodiment of the present invention, by analyzing the collected real-time hot news, screening out the public opinion information related to the enterprise, being able to analyze the sentiment tendency of the public opinion and construct a public opinion knowledge graph, associating according to the associated entities of the public opinion knowledge graph and the enterprise knowledge graph, and according to the association analysis result, being able to provide a suitable recommended enterprise list for local investment promotion government personnel, thereby solving the technical problem in the related art that the recommended enterprise list does not meet the investment promotion requirements due to not considering the impact on other associated enterprises according to the public opinion. Optionally, the acquisition unit includes: a first acquisition module, which is used to acquire a plurality of news titles whose sorting sequence is greater than a preset sequence threshold to obtain a title set; a first processing module, which is used to perform word segmentation processing and part-of-speech tagging processing on each news title in the title set to obtain a word set; a first traversal module, which is used to traverse the noun words in the word set. If there are noun words of enterprise objects in the word set, it acquires the news lead and comment information in the target news corresponding to the target news title to obtain news information.

[0132] Optionally, the analysis unit includes: a first extraction module for extracting keywords from the comment information; a first analysis module for performing sentiment analysis on each piece of comment information to obtain a sentiment distribution statistical graph corresponding to the news information; and a first addition module for adding sentiment tags to the news information based on the keywords and the sentiment distribution statistical graph.

[0133] Optionally, the construction unit includes: a second processing module for performing information extraction processing on the news lead to obtain the object information of the target enterprise object that appears in the news information, where the information extraction processing method includes at least one of the following: entity extraction, attribute extraction, and relationship extraction; and a first construction module for constructing an opinion knowledge graph corresponding to the target event indicated by the news information based on the object information of the target enterprise object.

[0134] Optionally, the enterprise recommendation device further includes: a first association module for associating the opinion knowledge graph with the opinion knowledge base after constructing the opinion knowledge graph corresponding to the target event indicated by the news information based on the news lead, to obtain an updated opinion knowledge graph, where the opinion knowledge base pre-stores the opinion knowledge graph obtained in the historical process; and a first storage module for storing the updated opinion knowledge graph into the graph database.

[0135] Optionally, the enterprise recommendation device further includes: a first acquisition module for acquiring the attribute information and enterprise relationship data of the target enterprise object before establishing the association relationship between the opinion knowledge graph of the target event and the enterprise knowledge graph of the target enterprise object; and a second construction module for constructing an enterprise knowledge graph based on the attribute information and the enterprise relationship data.

[0136] Optionally, the enterprise recommendation device further includes: a second acquisition module for acquiring the online news of the target enterprise object after constructing the enterprise knowledge graph; a third processing module for preprocessing the online news to obtain processed online news data; a fourth processing module for performing fusion processing on the enterprise data in the online news data with the target enterprise object; a second extraction module for extracting the enterprise data of other enterprise objects except the target enterprise object in the online news data; and a first supplement module for supplementing the enterprise data of other enterprise objects and the object relationship between other enterprise objects and the target enterprise object to the enterprise knowledge graph of the target enterprise object.

[0137] Optionally, the output unit includes: a third acquisition module, configured to, if the sentiment label is a positive label, obtain the graph distance, the number of associations, the first ranking information of each associated enterprise object, and the second ranking information of the industrial chain to which the associated enterprise object belongs between each associated enterprise object and the target enterprise object based on the graph association relationship; a first determination module, configured to determine the first weight information corresponding to the graph distance, the second weight information corresponding to the number of associations, the third weight information corresponding to the first ranking information, and the fourth weight information corresponding to the second ranking information; a first calculation module, configured to calculate the ranking parameter of each associated enterprise object based on the graph distance and the corresponding first weight information, the number of associations and the corresponding second weight information, the first ranking information and the corresponding third weight information, and the second ranking information and the corresponding fourth weight information; and output a list of recommended enterprises based on the ranking parameter.

[0138] Optionally, the enterprise recommendation device further includes: a second calculation module, configured to calculate the label score of each piece of news information about the target enterprise object that appears within a preset time period after outputting the list of recommended enterprises based on the graph association relationship, where the label score includes: a first label score or a second label score, the first label score represents the score of a positive sentiment label, and the second label score represents the score of a negative sentiment label; a first accumulation module, configured to accumulate the label scores of all news information to obtain a total label score; and a first adjustment module, configured to adjust the weight information of the associated enterprise object for the comment information based on the total label score to adjust the ranking parameter of the associated enterprise object.

[0139] The above control device may further include a processor and a memory. The above acquisition unit 30, analysis unit 32, construction unit 34, establishment unit 36, output unit 38, etc. are all stored in the memory as program units, and the corresponding functions are implemented by the processor executing the above program units stored in the memory.

[0140] The above processor includes a kernel, and the kernel retrieves the corresponding program unit from the memory. One or more kernels may be provided, and the list of recommended enterprises is output based on the graph association relationship by adjusting the kernel parameters.

[0141] The above memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of, for example, read-only memory (ROM) or flash memory (flash RAM), and the memory includes at least one storage chip.

[0142] The present application also provides a computer program product which, when executed on a data processing device, is adapted to execute a program initialized with the following method steps: collecting at least one piece of news information associated with an enterprise object, where each piece of news information includes at least: a news lead and comment information; analyzing an emotional label corresponding to the news information based on the comment information; constructing a public opinion knowledge graph corresponding to the target event indicated by the news information based on the news lead; establishing an association relationship between the public opinion knowledge graph of the target event and the enterprise knowledge graph of the target enterprise object to obtain a graph association relationship; and if the emotional label is a positive label, outputting an enterprise recommendation list based on the graph association relationship.

[0143] According to another aspect of the embodiments of the present invention, there is also provided an electronic device, including: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the enterprise recommendation method based on a knowledge graph according to any one of the above via executing the executable instructions.

[0144] According to another aspect of the embodiments of the present invention, there is also provided a computer-readable storage medium which includes a stored computer program, wherein when the computer program runs, it controls the device where the computer-readable storage medium is located to execute the enterprise recommendation method based on a knowledge graph according to any one of the above.

[0145] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0146] In the above embodiments of the present invention, the descriptions of the respective embodiments have their own emphases. For parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0147] In several embodiments provided by the present application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only illustrative. For example, the division of the units can be a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of the units or modules can be in an electrical or other form.

[0148] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0149] In addition, in each embodiment of the present invention, each functional unit may be integrated into a processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of a software functional unit.

[0150] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it may be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, may be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention. The foregoing storage medium includes: various media such as USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks, or optical discs that can store program codes.

[0151] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. An enterprise recommendation method based on a knowledge graph, characterized in that, Including: Collecting at least one piece of news information associated with an enterprise object, where each piece of the news information includes at least: a news lead and comment information; Analyzing an emotion label corresponding to the news information based on the comment information, where the emotion label is used to characterize whether the news information is negative news or positive news about the enterprise object; Constructing a public opinion knowledge graph corresponding to the target event indicated by the news information based on the news lead; Establishing an association relationship between the public opinion knowledge graph of the target event and the enterprise knowledge graph of the target enterprise object to obtain a graph association relationship; If the emotion label is a positive label, outputting an enterprise recommendation list based on the graph association relationship, where the recommendation list includes associated enterprise objects related to the target enterprise object in the public opinion knowledge graph and associated enterprise objects related to the target enterprise object in the enterprise knowledge graph; The step of outputting the enterprise recommendation list based on the graph association relationship includes: if the emotion label is a positive label, obtaining the graph distance, the number of associations, the first ranking information of the associated enterprise object, and the second ranking information of the industrial chain to which the associated enterprise object belongs between each associated enterprise object and the target enterprise object based on the graph association relationship; determining the first weight information corresponding to the graph distance, the second weight information corresponding to the number of associations, the third weight information corresponding to the first ranking information, and the fourth weight information corresponding to the second ranking information; calculating the ranking parameter of each associated enterprise object based on the graph distance and the corresponding first weight information, the number of associations and the corresponding second weight information, the first ranking information and the corresponding third weight information, and the second ranking information and the corresponding fourth weight information; outputting an enterprise recommendation list based on the ranking parameter.

2. The recommendation method according to claim 1, wherein The step of collecting at least one piece of news information associated with an enterprise object includes: Collecting multiple news titles with a sorting sequence greater than a preset sequence threshold to obtain a title set; Performing word segmentation processing and part-of-speech tagging processing on each news title in the title set to obtain a word set; Traversing the noun words in the word set, and if there is a noun word of the enterprise object in the word set, collecting the news lead and comment information in the target news corresponding to the target news title to obtain the news information.

3. The recommendation method according to claim 1, wherein The step of analyzing an emotion label corresponding to the news information based on the comment information includes: Extracting keywords from the comment information; Performing emotion analysis on each comment information to obtain an emotion distribution statistical chart corresponding to the news information; Adding an emotion label to the news information based on the keywords and the emotion distribution statistical chart.

4. The recommendation method according to claim 1, wherein The step of constructing a public opinion knowledge graph corresponding to the target event indicated by the news information based on the news lead includes: Perform information extraction processing on the news lead to obtain the object information of the target enterprise object that appears in the news information. Among them, the information extraction processing methods include at least one of the following: entity extraction, attribute extraction, and relationship extraction; Based on the object information of the target enterprise object, construct an opinion knowledge graph corresponding to the target event indicated by the news information.

5. The recommendation method according to claim 4, wherein After constructing an opinion knowledge graph corresponding to the target event indicated by the news information based on the news lead, the enterprise recommendation method further includes: Associate the opinion knowledge graph with an opinion knowledge base to obtain an updated opinion knowledge graph, where the opinion knowledge base pre-stores the opinion knowledge graph obtained in the historical process; Store the updated opinion knowledge graph in a graph database.

6. The recommendation method according to claim 1, characterized in that Before establishing the association relationship between the opinion knowledge graph of the target event and the enterprise knowledge graph of the target enterprise object, the enterprise recommendation method further includes: Obtain the attribute information and enterprise relationship data of the target enterprise object; Based on the attribute information and the enterprise relationship data, construct an enterprise knowledge graph.

7. The recommendation method according to claim 6, characterized in that, After constructing the enterprise knowledge graph, the enterprise recommendation method further includes: Obtain the online news of the target enterprise object; Preprocess the online news to obtain processed online news data; Fuse the enterprise data in the online news data with the target enterprise object; Extract the enterprise data of other enterprise objects except the target enterprise object in the online news data; Supplement the enterprise data of the other enterprise objects and the object relationship between the other enterprise objects and the target enterprise object to the enterprise knowledge graph of the target enterprise object.

8. The recommendation method according to claim 1, characterized in that After outputting an enterprise recommendation list based on the graph association relationship, the enterprise recommendation method further includes: Calculate the label score of each news information in multiple news information about the target enterprise object that appears within a preset time period. Among them, the label score includes: a first label score or a second label score. The first label score represents the score of a positive sentiment label, and the second label score represents the score of a negative sentiment label; Accumulate the label scores of all news information to obtain a total label score; Based on the total label score, adjust the weight information of the associated enterprise object in the comment information to adjust the ranking parameter of the associated enterprise object.

9. An enterprise recommendation device based on a knowledge graph, characterized in that Include: A collection unit for collecting at least one news information associated with an enterprise object, where each news information includes at least: a news lead and comment information; An analysis unit for analyzing the sentiment label corresponding to the news information based on the comment information, where the sentiment label is used to represent that the news information is negative news or positive news of the enterprise object; A construction unit for constructing an opinion knowledge graph corresponding to the target event indicated by the news information based on the news lead; An establishment unit for establishing an association relationship between the opinion knowledge graph of the target event and the enterprise knowledge graph of the target enterprise object to obtain a graph association relationship; An output unit, configured to output a list of enterprise recommendations based on the graph association relationship if the sentiment label is a positive label, where the list of recommendations includes associated enterprise objects related to the target enterprise object in the public opinion knowledge graph and associated enterprise objects related to the target enterprise object in the enterprise knowledge graph; The output unit includes: a third acquisition module, configured to, if the sentiment label is a positive label, acquire the graph distance, the number of associations, the first ranking information of the associated enterprise object, and the second ranking information of the industrial chain to which the associated enterprise object belongs between each associated enterprise object and the target enterprise object based on the graph association relationship; a first determination module, configured to determine first weight information corresponding to the graph distance, second weight information corresponding to the number of associations, third weight information corresponding to the first ranking information, and fourth weight information corresponding to the second ranking information; a first calculation module, configured to calculate a ranking parameter of each associated enterprise object based on the graph distance and the corresponding first weight information, the number of associations and the corresponding second weight information, the first ranking information and the corresponding third weight information, and the second ranking information and the corresponding fourth weight information; and output a list of enterprise recommendations based on the ranking parameter.

10. An electronic device, characterized in that, Comprising: A processor; And A memory, configured to store executable instructions of the processor; Wherein, the processor is configured to execute the enterprise recommendation method based on a knowledge graph according to any one of claims 1 to 8 by executing the executable instructions.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein when the computer program runs, it controls the device where the computer-readable storage medium is located to execute the enterprise recommendation method based on a knowledge graph according to any one of claims 1 to 8.

Images