Method for generating poll results and system therefor

Abstract

A method for generating poll results according to an embodiment of the present disclosure may comprise the steps of: receiving, from a voter terminal group, a first poll result for a basic poll related to a first topic included in a first keyword; on the basis of a basic poll selection option and a selection reason of the first poll result, generating a derivative option related to a second topic associated with the first topic; generating a derivative poll different from the basic poll on the basis of the second topic and the derivative option; receiving a second poll result for the derivative poll from the voter terminal group; and generating a final poll result on the basis of the first poll result and the second poll result.

Description

Method and System for Generating Voting Results

[0001] The present disclosure relates to a method and system for generating voting results. More specifically, it relates to a method and system for generating voting results that generates a vote by reflecting public opinion, such as voter opinions and news articles, and generates a vote by reflecting voter opinions regarding said vote.

[0002] An online voting platform is a tool that allows polls to be created and managed over the internet, facilitating the efficient voting process and result tabulation for multiple people. Online voting platforms must define the purpose of the vote and, accordingly, be able to effectively reflect the opinions of voters.

[0003] Even if multiple voters select the same voting item regarding a specific voting topic, their reasons for selecting that item may differ.

[0004] Therefore, technology is required to generate various derivative votes that reflect the opinions of voters, and to analyze those opinions more specifically accordingly.

[0005] (Patent Document 1) Published Patent Application No. 10-2023-0155277 (Published Nov. 10, 2023)

[0006] The technical problem to be solved in some embodiments of the present disclosure is to provide a method and system for generating voting results that generates a vote by reflecting public opinion, such as voter opinions and news articles.

[0007] Another technical problem to be solved in some embodiments of the present disclosure is a method and system for generating voting results that generate various derived votes by reflecting the opinions of voters and, accordingly, analyze the opinions of voters more specifically.

[0008] The technical problems of the present disclosure are not limited to those mentioned above, and other unmentioned technical problems will be clearly understood by a person skilled in the art of the present disclosure from the description below.

[0009] A method for generating a voting result according to an embodiment of the present disclosure for solving the above technical problem may include, in a method performed by a computing device, a step of receiving a first voting result for a basic vote related to a first topic included in a first keyword from a group of voter terminals, wherein the first voting result includes, for the basic vote, a basic voting selection item for each voter terminal included in the group of voter terminals and a reason for selecting the basic voting selection item; a step of generating a derivative item related to a second topic related to the first topic based on the basic voting selection item and the reason for selection; a step of generating a derivative vote different from the basic vote based on the second topic and the derivative item; a step of receiving a second voting result for the derivative vote from the group of voter terminals, wherein the second voting result includes information regarding the ratio selected by each of the derivative items included in the derivative vote by the voter terminals included in the group of voter terminals; and a step of generating a final voting result based on the first voting result and the second voting result.

[0010] In one embodiment, the method further includes the step of generating the basic vote based on the content for the first keyword, wherein the first keyword is included in a keyword group composed of a plurality of keywords, and the content may include a first news article stored in a database regarding the first keyword or a voter's opinion regarding the first keyword.

[0011] In one embodiment, the step of generating the basic vote may include: selecting the first keyword among the keyword groups for which voter opinions exceeding a threshold value have been written; grouping the voter opinions and obtaining a first group having the maximum number of voter opinions; obtaining a first embedding vector for the first voter opinions included in the first group; calculating the similarity between the first embedding vector and a second embedding vector for an external news article different from the first news article, and obtaining a second news article having the maximum calculated similarity; and generating the basic vote related to the first topic based on the first voter opinions and the second news article.

[0012] In one embodiment, the step of generating the basic vote comprises: selecting the first keyword among the keyword groups for which news articles exceeding a threshold value have been written; grouping the first news articles and obtaining a first group having the maximum number of news articles; dividing the second news articles included in the first group based on line breaks to obtain a first paragraph, and dividing the third news articles included in the first group based on line breaks to obtain a second paragraph; calculating a first representativeness score for a first embedding vector for the first paragraph; calculating a second representativeness score for a second embedding vector for the second paragraph; comparing the first representativeness score with the second representativeness score; and, as a result of the comparison, if the first representativeness score is greater than the second representativeness score, selecting the first paragraph as the representative paragraph. and based on the above representative paragraph, the method includes the step of generating the above basic vote related to the above first topic, wherein the first representativeness score is the sum of the similarities of the above first paragraph to the above second paragraph, and the second representativeness score may be the sum of the similarities of the above second paragraph to the above first paragraph.

[0013] In one embodiment, the step of calculating the first representative score may include the step of calculating the first representative score according to the following mathematical formula 1.

[0014] [Mathematical Formula 1]

[0015]

[0016] In one embodiment, the step of generating the basic vote related to the first topic based on the first paragraph may include: calculating a similarity between a first embedding vector for the first paragraph and a second embedding vector for an external news article different from the first news article, and obtaining a second news article having the maximum calculated similarity; and generating the basic vote related to the first topic based on the representative paragraph and the second news article.

[0017] In one embodiment, the step of generating the derivative item may include: obtaining a first selection reason for selecting a first selection item included in the basic voting selection item; grouping the first selection reason by topic and obtaining a first group including some of the selection reasons included in the first selection reason and a second group including some of the selection reasons included in the first selection reason; and inputting the selection reason included in the first group and the first topic into a generative AI model and generating the second topic and the derivative item as a result of the input.

[0018] In one embodiment, the step of generating the derived vote may include: obtaining a first embedding vector for a selection reason included in the first group; calculating a similarity between the first embedding vector and a second embedding vector for a news article for the first keyword, and obtaining a first news article having the maximum calculated similarity; and inputting the second topic, the derived item, and the first news article into the generative AI model, and generating the derived vote as a result of the input.

[0019] In one embodiment, the derivative item includes a first derivative item and a second derivative item, and the step of generating the derivative vote may include: a step of comparing a first count, which is the amount of votes for the first derivative item, with a second count, which is the amount of votes for the second derivative item; and, as a result of the comparison, if the first count is greater than the second count, a step of placing the first derivative item above the second derivative item on the derivative vote.

[0020] In one embodiment, the derived vote includes a vote body, a first derived item, and a second derived item, and the step of generating the derived vote may include the step of extracting a first keyword from the vote body, the step of extracting a second keyword from the selection reasons included in the first group, the step of calculating the similarity between the first keyword and the second keyword and obtaining a similarity score, and the step of changing the arrangement order of the first derived item and the second derived item based on the similarity score.

[0021] A voting result generation system according to another embodiment of the present disclosure for solving the above technical problem comprises: a communication interface; a memory on which a computer program is loaded; and one or more processors on which the computer program is executed, wherein the computer program may include instructions for performing operations such as: receiving a first voting result for a basic vote related to a first topic included in a first keyword from a group of voter terminals, wherein the first voting result includes, for the basic vote, a basic voting selection item for each voter terminal included in the group of voter terminals and a reason for selecting the basic voting selection item; generating a derivative item related to a second topic related to the first topic based on the basic voting selection item and the reason for selection; generating a derivative vote different from the basic vote based on the second topic and the derivative item; receiving a second voting result for the derivative vote from the group of voter terminals, wherein the second voting result includes information regarding the ratio selected by each voter terminal included in the group of voter terminals for each derivative item included in the derivative vote; and generating a final voting result based on the first voting result and the second voting result.

[0022] A computer program stored in a computer-readable recording medium according to another embodiment of the present disclosure for solving the above technical problem may be combined with a computing device and may execute the steps of: receiving a first voting result for a basic vote related to a first topic included in a first keyword from a group of voter terminals, wherein the first voting result includes, for the basic vote, a basic voting selection item for each voter terminal included in the group of voter terminals and a reason for selecting the basic voting selection item; generating a derivative item related to a second topic related to the first topic based on the basic voting selection item and the reason for selection; generating a derivative vote different from the basic vote based on the second topic and the derivative item; receiving a second voting result for the derivative vote from the group of voter terminals, wherein the second voting result includes information regarding the ratio selected by each of the derivative items included in the derivative vote by the voter terminals included in the group of voter terminals; and generating a final voting result based on the first voting result and the second voting result.

[0023] Through some embodiments of the present disclosure, a method and system for generating voting results that generate votes by reflecting public opinion, such as voter opinions and news articles, can be provided.

[0024] Through some embodiments of the present disclosure, a method and system for generating voting results can be provided, which generate various derivative votes by reflecting the opinions of voters and, accordingly, analyze the opinions of voters more specifically.

[0025] The effects of the invention disclosed herein are not limited to those mentioned above, and other unmentioned effects of the invention will be clearly understood by a person skilled in the art from the description below.

[0026] FIG. 1 is a system configuration diagram showing the configuration and operation of a voting result generation system according to one embodiment of the present disclosure.

[0027] FIG. 2 is a flowchart for explaining a method for generating voting results according to another embodiment of the present disclosure.

[0028] FIG. 3 is an illustrative diagram for explaining a keyword group referenced by some embodiments of the present disclosure.

[0029] FIG. 4 is an illustrative diagram for explaining a basic vote referenced by some embodiments of the present disclosure.

[0030] FIG. 5 is an illustrative diagram for explaining the voting results for a basic vote according to another embodiment of the present disclosure.

[0031] FIG. 6 is a detailed flowchart for explaining a method for generating voting results according to another embodiment of the present disclosure, described with reference to FIG. 2.

[0032] FIG. 7 is a detailed flowchart for explaining a method for generating voting results according to another embodiment of the present disclosure, described with reference to FIG. 2.

[0033] FIG. 8 is a detailed flowchart for explaining a method for generating voting results according to another embodiment of the present disclosure, described with reference to FIG. 2.

[0034] FIG. 9 is a detailed flowchart for explaining a method for generating voting results according to another embodiment of the present disclosure, described with reference to FIG. 2.

[0035] FIGS. 10 to 12 are illustrative diagrams for explaining a derivative vote and a voting result for a derivative vote, which are referenced by some embodiments of the present disclosure.

[0036] FIG. 13 is a detailed flowchart for explaining a method for generating voting results according to another embodiment of the present disclosure, described with reference to FIG. 2.

[0037] FIG. 14 is a detailed flowchart for explaining a method for generating voting results according to another embodiment of the present disclosure, described with reference to FIG. 2.

[0038] FIG. 15 is an illustrative diagram for explaining the final voting results referenced by some embodiments of the present disclosure.

[0039] FIG. 16 is a hardware configuration diagram of a computing device described in some embodiments of the present disclosure.

[0040] Hereinafter, preferred embodiments of the present disclosure will be described in detail with reference to the attached drawings. The advantages and features of the present invention and the methods for achieving them will become clear by referring to the embodiments described below in detail together with the attached drawings. However, the technical concept of the present invention is not limited to the following embodiments but can be implemented in various different forms. The following embodiments are provided merely to complete the technical concept of the present invention and to fully inform those skilled in the art of the scope of the present invention, and the technical concept of the present invention is defined only by the scope of the claims.

[0041] In describing the present disclosure, if it is determined that a detailed description of related known configurations or functions could obscure the essence of the invention, such detailed description is omitted.

[0042] Unless otherwise defined, terms used in the following embodiments (including technical and scientific terms) may be used in a meaning commonly understood by those skilled in the art to which this disclosure pertains, but this may vary depending on the intent of those skilled in the art, case law, the emergence of new technology, etc. The terms used in this disclosure are for describing the embodiments and are not intended to limit the scope of this disclosure.

[0043] In the following embodiments, singular expressions include plural concepts unless the context clearly specifies them as singular. Additionally, plural expressions include singular concepts unless the context clearly specifies them as plural.

[0044] In addition, terms such as first, second, A, B, (a), (b), etc. used in the following embodiments are used merely to distinguish one component from another, and the essence, order, or sequence of the said component is not limited by such terms.

[0045] Various embodiments of the present disclosure will be described below with reference to the attached drawings.

[0046] Hereinafter, with reference to FIG. 1, the configuration and operation of a voting result generation system according to one embodiment of the present disclosure will be described. FIG. 1 is a system configuration diagram showing the configuration and operation of a voting result generation system according to one embodiment of the present disclosure.

[0047] Referring to FIG. 1, a voting result generation system may be configured to include a voting system (10), a voter terminal group (20), and an external database (30). However, the scope of the present disclosure is not limited thereto. In some cases, the voting result generation system may be configured to include additional modules / devices / systems not shown in FIG. 1. Alternatively, the voting result generation system may be configured in a form in which at least some of the components (10, 20, and 30) shown in FIG. 1 are excluded.

[0048] The voting system (10) may be configured to include a service server (11), a database (12), and a generative AI model (13). However, the scope of the present disclosure is not limited thereto. In some cases, the voting system (10) may be configured to include additional modules / devices / systems not shown in FIG. 1. Alternatively, the voting system (10) may be configured in a form in which at least some of the components (11, 12 and 13) shown in FIG. 1 are excluded.

[0049] The service server (11) can generate a basic vote related to a specific topic included in a specific keyword based on content regarding various keywords (e.g., user opinions of the voting system (10), news articles, etc.). The service server (11) can generate a basic vote using news articles and voter opinions stored in the database (12). Keywords are described in detail with reference to FIG. 3. The method for generating a basic vote is described in detail with reference to FIG. 4, FIG. 6, and FIG. 7.

[0050] The voter terminal group (20) may be configured to include a first voter terminal (21) and a second voter terminal (22). However, the scope of the present disclosure is not limited thereto, and the voter terminal group (20) may include three or more voter terminals. Each voter terminal included in the voter terminal group (20) may be a terminal of a user using the voting system (10).

[0051] Each voter terminal included in the voter terminal group (20) can transmit voting results for the basic vote and the derived vote generated by the voting system (10) to the voting system (10). For the basic vote, each voter terminal included in the voter terminal group (20) can transmit voting items and the reasons for selecting the corresponding voting items to the voting system (10). The reasons for selection may be recorded in the comment window for the basic vote, but the scope of the present disclosure is not limited thereto. The basic vote may be the initial vote generated by the service server (11), and the derived vote may be a vote generated by the service server (11) based on the basic vote.

[0052] The service server (11) can generate a derived item regarding a topic related to the topic of the basic vote based on the voting results for the basic vote received from each voter terminal included in the voter terminal group (20). For example, the topic related to the topic of the basic vote can be the voting title of the derived vote, and the derived item can be the voting item of the derived vote. The method for generating the derived item will be explained in detail with reference to FIG. 8.

[0053] The service server (11) can generate a derivative vote different from the basic vote based on the topic related to the topic of the basic vote and the derivative item. The service server (11) can generate the derivative vote by inputting the topic related to the topic of the basic vote and the derivative item, etc., into a generative AI model (13).

[0054] The service server (11) can generate a derivative vote different from the basic vote based on the topic related to the topic of the basic vote and the derivative item. The method for generating the derivative vote will be explained in detail with reference to FIGS. 9 to 14.

[0055] The service server (11) can receive voting results for the derivative vote from each voter terminal included in the voter terminal group (20). The voting results for the derivative vote may include information regarding the ratio selected by the voter terminals included in the voter terminal group (20) for each derivative item included in the derivative vote.

[0056] The service server (11) can generate a final vote result based on the voting result for the basic vote and the voting result for the derived vote. In this case, the service server (11) can generate the final vote result by inputting the voting result for the basic vote and the voting result for the derived vote into a generative AI model (13). The final vote result is described in detail with reference to FIG. 14.

[0057] Meanwhile, the external database (30) is a knowledge base outside the voting system (10) and may be a database maintained and managed by an entity separate from the voting system (10). The service server (11) can perform a search on an external search engine, etc., and obtain content such as news articles from the external database (30). This allows for additional reference to knowledge stored in the external database (30), thereby ensuring diversity of information.

[0058] For the sake of ease of understanding, the following explanation will continue by assuming that all steps / operations of the methods described below are performed on the service server (11) described above. Therefore, if the subject of a specific step / operation is omitted, it can be understood that it is performed on the service server (11). However, in an actual environment, some steps / operations of the methods described below may be performed on other computing devices.

[0059] Hereinafter, a method for generating a voting result according to one embodiment of the present disclosure will be described with reference to FIG. 2. FIG. 2 is a flowchart for describing a method for generating a voting result according to another embodiment of the present disclosure.

[0060] Referring to FIG. 2, the service server (11) can generate a basic vote related to a first topic included in the first keyword based on the content for the first keyword (S100). The first keyword is included in a keyword group consisting of multiple keywords, and the content may include a first news article stored in the database (12) for the first keyword or a voter opinion for the first keyword.

[0061] Hereinafter, with reference to FIG. 3, keyword groups referenced by some embodiments of the present disclosure are described. FIG. 3 is an illustrative diagram for explaining keyword groups referenced by some embodiments of the present disclosure.

[0062] Referring to FIG. 3, a keyword group (300) is illustrated. The keyword group (300) may include multiple keywords such as policy, interest rate, president, international, artificial intelligence, science, baseball, movie, and travel. However, the scope of the present disclosure is not limited thereto, and the keyword group (30) may include various keywords.

[0063] Referring to FIG. 3, the first keyword (301) is a keyword included in the keyword group (300) and may be travel. The first keyword (301) may include a travel destination, which is the first topic (302). However, the scope of the present disclosure is not limited thereto, and the first keyword (301) may include various topics other than the first topic (302), such as a desired travel period.

[0064] In this disclosure, for the convenience of explanation, the explanation will proceed on the premise that the first keyword is travel and the first subject is a travel destination.

[0065] Hereinafter, with reference to FIG. 4, a basic vote referenced by some embodiments of the present disclosure is described. FIG. 4 is an illustrative diagram for explaining a basic vote referenced by some embodiments of the present disclosure.

[0066] Referring to FIG. 4, a basic vote (40) is illustrated. The basic vote (40) may be composed of a vote title (41), a vote body (42), and a vote item (=basic vote selection item) (43).

[0067] The voting title (41) may be a topic related to the first topic included in the first keyword. For example, the voting title (41) may be related to the first topic (302), which is a travel destination, and may be determined as a title related to the 'most preferred overseas travel destination'.

[0068] The voting body (42) serves to support the voting title (41) and may contain a detailed explanation of the purpose of the vote and the voting title (41). The voting body (42) may be automatically generated based on news articles stored in the database (12) regarding the first keyword, voter opinions regarding the first keyword, or news articles obtained from an external database (13).

[0069] Voting items (43) are basic voting selection items and may be items that a voter can select. A voter may select one of the voting items (43). Voting items (43) may be generated based on news articles stored in the database (12) regarding the first keyword, travel, or voter opinions regarding the first keyword (e.g., comments on the news articles, etc.).

[0070] The method for generating a basic vote will be explained in detail with reference to Figs. 6 and 7.

[0071] Again, this will be explained with reference to Fig. 2.

[0072] When the above basic vote is generated (S100), the service server (11) can receive a first voting result for the basic vote related to the first topic included in the first keyword from the voter terminal group (20) (S200). The first voting result may include, for the basic vote, a basic voting selection item for each voter terminal included in the voter terminal group (20) and a reason for selecting the basic voting selection item.

[0073] Hereinafter, with reference to FIG. 5, the first voting result for a basic vote according to another embodiment of the present disclosure will be described. FIG. 5 is an illustrative diagram for explaining the voting result for a basic vote according to another embodiment of the present disclosure.

[0074] Referring to FIG. 5, the first voting result may include, for each basic voting selection item, information (50) regarding the ratio of basic voting selection items selected by voter terminals included in the voter terminal group (20), and information (51) regarding each voter terminal's basic voting selection item and the reason for selecting the basic voting selection item.

[0075] Information (51) regarding the basic voting selection item for each voter terminal and the reason for selecting the basic voting selection item may include the first voter voting result (52). The first voter voting result (52) may include the first basic voting selection item (52a) selected by the first voter and the first reason for selecting the first basic voting selection item (52a) (52b). The first basic voting selection item (52a) may be 'A. Asia', and the first reason for selection (52b) may be a comment saying "Thailand in summer". In FIG. 5, 'first voter', 'second voter', etc. are shown, but the ID or nickname of the first voter may be shown instead of the 'first voter'.

[0076] The service server (11) can receive various voting results (53, 54, 55, 56, 57, 58, 59) for a basic vote from a plurality of voter terminals included in the voter terminal group (20), such as the first voter voting result (52).

[0077] Again, this will be explained with reference to Fig. 2.

[0078] When the first voting result for the above basic vote is received (S200), the service server (11) may generate a derivative item regarding a second topic related to the first topic based on the above basic vote selection item and the reason for selecting the above basic vote selection item (S300). The second topic may be a topic included in the first keyword, like the first topic, and may be a topic derived from the reason for selecting the above basic vote selection item. A method for generating a derivative item regarding the second topic will be explained in detail with reference to FIG. 8.

[0079] Subsequently, the service server (11) may generate a derivative vote different from the basic vote based on the second topic and the derivative item (S400). The derivative vote may be composed of a vote title, a vote body, and a vote item, similar to the basic vote. The vote title may relate to the second topic. The vote body may support the vote title and may contain a detailed explanation of the purpose of the vote and the vote title. The vote item may be the derivative item and may be an item that the voter can select. The method for generating the derivative vote will be explained in detail with reference to FIGS. 9 through 13.

[0080] Subsequently, the service server (11) may receive a second voting result for the derived vote from the voter terminal group (20) (S500). The second voting result may include information regarding the ratio selected by the voter terminals included in the voter terminal group (20) for each derived item included in the derived vote. The second voting result will be explained in detail with reference to FIGS. 10 to 12.

[0081] Subsequently, the service server (11) can generate a final voting result based on the first voting result and the second voting result (S600). The service server (11) can generate the final voting result by automatically generating a prompt to generate a final analysis result using the first voting result and the second voting result, and by inputting the first voting result and the second voting result together with the prompt into a generative AI model (13). The final voting result is explained in detail with reference to FIG. 14.

[0082] According to the present embodiment, a poll can be generated by reflecting public opinion, such as voter opinions and news articles, regarding a topic included in a specific keyword. Furthermore, according to the present embodiment, various derivative polls can be generated by reflecting voter opinions, and voter opinions can be analyzed more specifically accordingly. In other words, according to the present embodiment, there is an advantage in that a final poll result can be automatically generated that more faithfully reflects the public opinion of voters (=public opinion) regarding a topic included in a specific keyword.

[0083] Accordingly, according to the present embodiment, there is an advantage in that various information is provided to users using the voting system (10), and the user experience can be enhanced according to the diversity of information.

[0084] Hereinafter, with reference to FIG. 6, a method for generating a basic vote according to one embodiment of the present disclosure will be described. FIG. 6 is a detailed flowchart for explaining a method for generating a vote result according to another embodiment of the present disclosure, which was described with reference to FIG. 2.

[0085] Referring to FIG. 6, the service server (11) can select a first keyword from among a keyword group for which voter opinions exceeding a threshold value have been written (S110). The threshold value may be set differently depending on the situation. For example, if the first keyword is 'travel', 'travel' may be selected as the first keyword if 1,000 or more voter opinions, such as comments on news articles related to topics included in 'travel' (e.g., travel destinations, etc.), have been written. However, the scope of the present disclosure is not limited thereto.

[0086] Afterward, the service server (11) can group the voter opinions and obtain a first group having the maximum number of voter opinions (S111). Below, an embodiment regarding the method of grouping the voter opinions is described.

[0087] In one embodiment, the service server (11) can generate a voter opinion embedding vector by converting each of the voter opinions into a vector using a BERT (Bidirectional Encoder Representations from Transformers) model that has learned the similarity between documents. The BERT model can be trained using the SimCSE (Simple Contrastive Sentence Embedding) technique. The BERT model is a model based on the Transformer architecture and is efficient for natural language understanding (NLU) tasks. The SimCSE technique is a technique for sentence embedding designed to effectively capture semantic similarity between sentences. SimCSE demonstrates highly effective performance in the method of learning sentence embeddings using a large amount of sentence data. Since the BERT model and the SimCSE technique are matters that a person skilled in the art to which the invention according to the present disclosure belongs would already be familiar with, a detailed description is omitted.

[0088] Subsequently, the service server (11) can generate clusters using the K-means algorithm with the voter opinion embedding vectors. In this case, the number of clusters can be determined through the silhouette coefficient. The K-means algorithm is a clustering algorithm that can form clusters by selecting and modifying the center points of the data. The silhouette coefficient is a factor for evaluating the performance of clustering based on how much the data within a cluster is aggregated (gathered at the center point) and how far it is from other clusters. Since the K-means algorithm and the silhouette coefficient are matters that a person skilled in the art to which the invention according to the present disclosure belongs would already be familiar with, a detailed explanation is omitted.

[0089] Due to the characteristics of the K-means algorithm, there may be differences in the shape and number of clusters formed depending on the initial centroid selection. Therefore, the service server (11) can obtain a clustering result with the highest silhouette coefficient by repeatedly performing clustering on the voter opinion embedding vectors using the K-means algorithm so that more accurate clustering can be performed. In this way, the service server (11) can group voter opinions into multiple clusters.

[0090] Subsequently, the service server (11) can use a T5-based summary model to obtain items regarding voter opinions representing each cluster for voter opinions within each cluster, and obtain a first group having the maximum number of voter opinions among them. The T5-based summary model is a Transformer-based model that uses both an encoder and a decoder.

[0091] Meanwhile, in another embodiment, the service server (11) can automatically generate the voter opinion, the prompt for generating an item for grouping the voter opinion, and automatically input the generated prompt into the LLM (Large Language Model).

[0092] Afterward, the service server (11) can obtain the results of grouping the voter opinions from the LLM. The service server (11) can use the grouped results to obtain items regarding voter opinions for each group and obtain a first group having the maximum number of voter opinions among them.

[0093] However, it should be noted that the scope of the present disclosure is not limited thereto, and that the embodiments regarding the method of grouping voter opinions are examples.

[0094] Again, this will be explained with reference to Fig. 6.

[0095] When the first group is obtained (S111), the service server (11) can obtain a first embedding vector for the first voter opinion included in the first group (S112). The first voter opinion is one of the voter opinions included in the first group, and since the first group is already in a clustered state, it can be randomly selected from the first group. As previously explained, the first embedding vector for the first voter opinion may be a voter opinion embedding vector generated by converting each of the voter opinions into a vector using the BERT (Bidirectional Encoder Representations from Transformers) model. That is, the service server (11) can obtain the voter opinion embedding vector generated during the grouping process of the voter opinions as the first embedding vector.

[0096] Subsequently, the service server (11) calculates the similarity between the first embedding vector and the second embedding vector for an external news article that is different from the first news article stored in the database (12) for the first embedding vector and the first keyword, and can obtain the second news article with the maximum calculated similarity (S113). The service server (11) can obtain the external news article by performing a search for the first topic included in the first keyword in the external database (30). The second embedding vector can be generated by converting each of the external news articles into a vector using the BERT (Bidirectional Encoder Representations from Transformers) model, as explained in the previous process of grouping voter opinions. The service server (11) calculates the similarity between the first embedding vector and the second embedding vector and can obtain the external news article most similar to the first voter opinion included in the first group as the second news article.

[0097] Subsequently, the service server (11) can generate a basic vote related to the first topic based on the first voter's opinion and the second news article (S114). The service server (11) can generate a basic vote by automatically generating a prompt to generate a vote using the first voter's opinion and the second news article, and by inputting the first voter's opinion and the second news article together with the prompt into a generative AI model (13).

[0098] For example, if the first topic is about 'travel destinations' and the first voter's opinion is 'preferred travel destination is Asia', the above basic vote may be about 'preferred overseas travel destinations' as illustrated in FIG. 4. The voting text of the basic vote may be written based on the above second news article. The voting items of the basic vote may be generated by continent, such as 'Asia', 'Europe', 'North America', 'Africa', 'Others', etc., reflecting the above first voter's opinion.

[0099] According to the present embodiment, a basic vote can be generated by reflecting public opinion, such as voter opinions and news articles, regarding a topic included in a specific keyword. That is, according to the present embodiment, a basic vote can be generated that more faithfully reflects the opinions of voters (=public opinion) regarding a topic included in a specific keyword.

[0100] Accordingly, according to the present embodiment, there is an advantage in that it provides a user using the voting system (10) with votes from various perspectives regarding topics included in a specific keyword, and the user experience can be enhanced due to the diversity of the votes.

[0101] Hereinafter, with reference to FIG. 7, a method for generating a basic vote according to another embodiment of the present disclosure will be described. FIG. 7 is a detailed flowchart for explaining a method for generating a vote result according to another embodiment of the present disclosure, which was described with reference to FIG. 2.

[0102] Referring to FIG. 7, the service server (11) can select a first keyword from among a keyword group for which news articles exceeding a threshold value have been written (S120). The threshold value may be set differently depending on the situation. For example, if the first keyword is 'travel', 'travel' may be selected as the first keyword when 1,000 or more voter opinions, such as comments on news articles related to topics included in 'travel' (e.g., travel destinations, etc.) have been written. However, the scope of the present disclosure is not limited thereto.

[0103] Afterwards, the service server (11) can group the first news articles stored in the database (12) for the first keyword and obtain the first group having the maximum number of news articles (S121).

[0104] In one embodiment, the service server (11) can generate a first news article embedding vector by converting each of the first news articles into a vector using a BERT (Bidirectional Encoder Representations from Transformers) model that has learned the similarity between documents. The BERT model can be trained using the SimCSE (Simple Contrastive Sentence Embedding) technique. As the BERT model and the SimCSE technique have been described above with reference to FIG. 6, a detailed description is omitted.

[0105] Afterward, the service server (11) can generate clusters using the K-means algorithm for the first news article embedding vector. In this case, the number of clusters can be determined through the silhouette coefficient. The K-means algorithm and the silhouette coefficient are as described above with reference to FIG. 6, so a detailed explanation is omitted.

[0106] Due to the characteristics of the K-means algorithm, there may be differences in the shape and number of clusters formed depending on the initial centroid selection. Therefore, the service server (11) can obtain a clustering result having the highest silhouette coefficient by repeatedly performing clustering on the first news article embedding vector using the K-means algorithm so that more accurate clustering can be performed. In this way, the service server (11) can group the first news article into multiple clusters.

[0107] Subsequently, the service server (11) can use a T5-based summary model to obtain an item regarding a news article representing each cluster for a first news article within each cluster, and obtain a first group having the maximum number of news articles among them. The T5-based summary model is a Transformer-based model that uses both an encoder and a decoder.

[0108] Meanwhile, in another embodiment, the service server (11) can automatically generate the first news article, the first news article, the prompt for generating an item for grouping the first news article, and automatically input the generated prompt into the LLM (Large Language Model).

[0109] Afterward, the service server (11) can obtain the results of the first news article being grouped from the LLM. The service server (11) can use the grouped results to obtain items regarding news articles for each group and obtain the first group having the maximum number of news articles among them.

[0110] However, it should be noted that the scope of the present disclosure is not limited thereto, and the embodiments regarding the method of grouping the first news article are examples.

[0111] Again, this will be explained with reference to Fig. 7.

[0112] When the first group is obtained (S121), the service server (11) may obtain a first paragraph by dividing a second news article included in the first group based on a line break, and obtain a second paragraph by dividing a third news article included in the first group based on a line break (S122). The second news article and the third news article may be included in the first news article. The second news article and the third news article may each include multiple paragraphs in addition to the first paragraph and the second paragraph, but for convenience of explanation, the present disclosure describes the second news article and the third news article on the premise that they each include the first paragraph and the second paragraph.

[0113] Subsequently, the service server (11) can calculate a first representativeness score for the first embedding vector for the first paragraph and calculate a second representativeness score for the second embedding vector for the second paragraph (S123). The first embedding vector and the second embedding vector can be generated by converting each of the first paragraph and the second paragraph into a vector using the BERT (Bidirectional Encoder Representations from Transformers) model, as previously explained in the grouping process of the first news article.

[0114] The above first representativeness score is the sum of the similarities of the above first paragraph with respect to the above second paragraph, and the above second representativeness score may be the sum of the similarities of the above second paragraph with respect to the above first paragraph. Below, a method for calculating the above first representativeness score and the above second representativeness score is described.

[0115] In one embodiment, the service server (11) can calculate the first representative score according to the following mathematical formula 1.

[0116] [Mathematical Formula 1]

[0117]

[0118] In the above mathematical formula 1, n represents the number of paragraphs, and X is a matrix representing the embedding vectors of each paragraph as rows. X T is the transpose matrix of X. i and j are paragraph numbers, respectively. That is, the above mathematical formula 1 may be an operation that iterates using i as the reference paragraph to calculate similarity with the remaining paragraphs excluding the i-th paragraph, and then adds up all the calculated similarities. X·X T The operation is It is the same as. This is X i and X j It is equivalent to calculating the cosine similarity.

[0119] For example, the above first representativeness score may be the sum of all calculated cosine similarity values ​​after calculating the cosine similarity between X1 and X2. X1 is the first embedding vector and X2 is the second embedding vector. Let X1 be [0.23 0.54 0.12 0.95 0.32] and X2 be [0.88 0.43 0.67 0.15 0.72]. In this case, (X·X T ) 11 The value of is 1, and (X·X T ) 12 The value of is 0.547. Therefore, the first representativeness score is 1.547.

[0120] The service server (11) can calculate the second representative score in the same manner as in the preceding embodiment.

[0121] It should be noted that for the convenience of explanation, the present disclosure uses a total of two paragraphs, the first and second paragraphs, as examples.

[0122] Again, this will be explained with reference to Fig. 7.

[0123] When the first representative score and the second representative score are calculated (S124), the service server (11) can compare the first representative score and the second representative score (S124).

[0124] As a result of the above comparison, if the first representativeness score is greater than the second representativeness score, the service server (11) can select the first paragraph as the representative paragraph (S125). On the other hand, if the second representativeness score is greater than the first representativeness score as a result of the above comparison, the service server (11) can select the second paragraph as the representative paragraph.

[0125] The service server (11) compares the representativeness scores for each paragraph and selects the paragraph with the highest representativeness score as the representative paragraph, thereby selecting the paragraph that best expresses the topic of the first group among all the paragraphs included in the first group of news articles as the representative paragraph. Therefore, the topic of the basic vote related to the first topic included in the first keyword can be comprehensively selected.

[0126] Afterward, the service server (11) can generate a basic vote related to the first topic based on the representative paragraph (S126). The service server (11) can generate a basic vote by automatically generating a prompt to generate a vote using the representative paragraph and inputting the representative paragraph along with the prompt into a generative AI model (13).

[0127] For example, if the first topic is about 'travel destinations' and the representative paragraph is 'It is reported that there are various overseas travel destinations preferred by office workers during the vacation season. As a result of the tally, Asia ranked first with 40%...', the above basic vote may be about 'preferred overseas travel destinations' as illustrated in Fig. 4. The voting body of the basic vote may be written based on the above representative paragraph. The voting items of the basic vote may be generated by continent, such as 'Asia', 'Europe', 'North America', 'Africa', 'Others', etc., reflecting the above representative paragraph.

[0128] According to the present embodiment, a basic poll can be generated by reflecting public opinion, such as news articles, regarding a topic included in a specific keyword. That is, according to the present embodiment, a basic poll that more faithfully reflects public opinion regarding a topic included in a specific keyword can be generated.

[0129] Accordingly, according to the present embodiment, there is an advantage in that it provides a user using the voting system (10) with votes from various perspectives regarding topics included in a specific keyword, and the user experience can be enhanced due to the diversity of the votes.

[0130] Meanwhile, in one embodiment, the service server (11) calculates the similarity between the first embedding vector for the first paragraph and the second embedding vector for an external news article that is different from the first news article stored in the database (12) for the first keyword, and can obtain the second news article with the maximum calculated similarity. The service server (11) can obtain the external news article by performing a search for the first topic included in the first keyword in the external database (30). The second embedding vector can be generated by converting each of the second news articles into a vector using the BERT (Bidirectional Encoder Representations from Transformers) model, as previously explained in the grouping process of the first news article. The service server (11) calculates the similarity between the first embedding vector and the second embedding vector and can obtain the external news article that is most similar to the first news article included in the first group as the second news article.

[0131] Subsequently, the service server (11) can generate a basic vote related to the first topic based on the representative paragraph and the second news article. The service server (11) can generate a basic vote by automatically generating a prompt to generate a vote using the representative paragraph and the second news article, and by inputting the representative paragraph and the second news article together with the prompt into a generative AI model (13).

[0132] For example, if the first topic is about 'travel destinations' and the representative paragraph is 'It is reported that there are various overseas travel destinations preferred by office workers during the vacation season. As a result of the tally, Asia ranked first with 40%...', the above basic vote may be about 'preferred overseas travel destinations' as illustrated in Fig. 4. The voting body of the basic vote may be written based on the above second news article. The voting items of the basic vote may be generated by continent, such as 'Asia', 'Europe', 'North America', 'Africa', 'Others', etc., reflecting the above representative paragraph.

[0133] According to the present embodiment, a search related to a specific topic can be automatically performed in an external database (30), and a vote can be generated based on the search results. Therefore, according to the present embodiment, there is an advantage in that the diversity and reliability of information can be guaranteed through additional information such as external search results.

[0134] Hereinafter, with reference to FIG. 8, a method for generating a derivative item according to one embodiment of the present disclosure will be described. FIG. 8 is a detailed flowchart example for explaining a method for generating a voting result according to another embodiment of the present disclosure, which was described with reference to FIG. 2.

[0135] Referring to FIG. 8, the service server (11) can obtain a first selection reason for selecting a first selection item included in the basic voting selection items (S210). For example, referring to FIG. 4, the first selection item may be 'A. Asia'. Referring to FIG. 5, the first selection reason may be a first selection reason (52b) for selecting a first basic voting selection item (52a) included in the first voter voting result (52). The first selection reason may include all selection reasons (53b, 54b, 55b, 56b, 57b, 58b, 59b) included in the basic voting selection items (53a, 54a, 55a, 56a, 57a, 58a, 59a) of each voter terminal and the information (51) regarding the selection reason for selecting the basic voting selection items.

[0136] Afterwards, the service server (11) can group the first selection reason by topic and obtain a first group including some of the selection reasons included in the first selection reason and a second group including some of the selection reasons included in the first selection reason (S211).

[0137] In one embodiment, the service server (11) can generate a first selection reason embedding vector by converting each of the first selection reasons into a vector using a BERT (Bidirectional Encoder Representations from Transformers) model that has learned the similarity between documents. The BERT model can be trained using the SimCSE (Simple Contrastive Sentence Embedding) technique. As the BERT model and the SimCSE technique have been described above with reference to FIG. 6, a detailed description is omitted.

[0138] Afterward, the service server (11) can generate clusters using the K-means algorithm with the first selection reason embedding vector. In this case, the number of clusters can be determined through the silhouette coefficient. The K-means algorithm and the silhouette coefficient are as described above with reference to FIG. 6, so a detailed explanation is omitted.

[0139] Due to the characteristics of the K-means algorithm, there may be differences in the shape and number of clusters formed depending on the initial centroid selection. Therefore, the service server (11) can obtain a clustering result having the highest silhouette coefficient by repeatedly performing clustering on the first selection reason embedding vector using the K-means algorithm so that more accurate clustering can be performed. In this way, the service server (11) can group the first selection reason into multiple clusters.

[0140] Subsequently, the service server (11) can use a T5-based summary model to obtain items regarding the selection reasons representing each cluster for the first selection reasons within each cluster, and obtain a first group having the maximum number of selection reasons among them. The T5-based summary model is a Transformer-based model that uses both an encoder and a decoder.

[0141] Meanwhile, in another embodiment, the service server (11) can automatically generate the first selection reason and a prompt for generating an item for grouping the first selection reason, and can automatically input the generated prompt into the LLM (Large Language Model).

[0142] Subsequently, the service server (11) can obtain a result in which the first selection reason is grouped from the LLM. The service server (11) can use the grouped result to obtain an item regarding the selection reason for each group, and among them, obtain a first group having the maximum number of selection reasons.

[0143] However, it should be noted that the scope of the present disclosure is not limited thereto, and the embodiments regarding the method of grouping the first selection reason are examples.

[0144] For example, referring to FIG. 5, the selection reason (52b) included in the first voter's voting result (52), the selection reason (55b) included in the fourth voter's selection result (55), the selection reason (56b) included in the fifth voter's selection result (56), and the selection reason (58b) included in the seventh voter's selection result (58) can all be classified as selection reasons regarding preferred overseas travel destinations within Asia by season. Therefore, the above selection reasons (52b, 55b, 56b, 58b) can be included in the first group regarding preferred overseas travel destinations by season.

[0145] The reasons for selection (53b) included in the second voter's voting result (53), the reasons for selection (54b) included in the third voter's voting result (54), and the reasons for selection (59b) included in the eighth voter's selection result (59) can all be classified as reasons for selection regarding preferred overseas travel destinations within Asia based on the length of the holiday. Therefore, the above reasons for selection (53b, 54b, 59b) can be included in the second group regarding preferred overseas travel destinations within Asia based on the length of the holiday.

[0146] The reason for selection (57b) included in the result of the sixth voter vote (57) can be classified as a reason for selection regarding preferred overseas travel destinations within Asia based on budget. Therefore, the reason for selection (57b) can be included in the third group regarding preferred overseas travel destinations within Asia based on budget.

[0147] Again, this will be explained with reference to Fig. 8.

[0148] When the first group is obtained (S211), the service server (11) inputs the selection reason and the first topic included in the first group into a generative AI model and can generate the second topic and the derived item as a result of the input (S212). The service server (11) can automatically generate a prompt to generate a second topic related to the first topic and a derived item regarding the second topic using one of the selection reasons included in the first group and the first topic, and can generate the second topic and the derived item by inputting one of the selection reasons included in the first group and the first topic into the generative AI model (13) along with the prompt. The method for generating the second topic and the derived item is described in detail with reference to FIGS. 10 to 12.

[0149] Hereinafter, with reference to FIG. 9, a method for generating a derived vote according to one embodiment of the present disclosure will be described. FIG. 9 is a detailed flowchart example for explaining a method for generating a vote result according to another embodiment of the present disclosure, which was described with reference to FIG. 2.

[0150] Referring to FIG. 9, the service server (11) can obtain a first embedding vector for the selection reasons included in the first group (S310). The first embedding vector can be generated by converting each of the selection reasons included in the first group into a vector using a BERT (Bidirectional Encoder Representations from Transformers) model, as described in the grouping process of the first selection reasons described earlier with reference to FIG. 8.

[0151] Subsequently, the service server (11) calculates the similarity between the first embedding vector and the second embedding vector for the news article for the first keyword, and can obtain the first news article with the maximum calculated similarity (S311). The news article for the first keyword may be a news article stored in the database (12) for the first keyword or an external news article obtained by performing a search for the first topic included in the first keyword in the external database (30). The second embedding vector may be generated by converting each news article included in the first keyword into a vector using a BERT (Bidirectional Encoder Representations from Transformers) model, as described in the grouping process of the first selection reason explained earlier with reference to FIG. 8. The service server (11) calculates the similarity between the first embedding vector and the second embedding vector, and can obtain the news article most similar to the selection reason included in the first group as the first news article.

[0152] Afterward, the service server (11) inputs the second topic, the derivative item, and the first news article into the generative AI model (13) and can generate a derivative vote as a result of the input (S312). The service server (11) can automatically generate a prompt to generate a vote using the second topic, the derivative item, and the first news article, and can generate a derivative vote by inputting the second topic, the derivative item, and the first news article into the generative AI model (13) along with the prompt.

[0153] Hereinafter, with reference to FIGS. 10 to 12, a method for generating a second subject and a derivative item according to some embodiments of the present disclosure will be described. FIGS. 10 to 12 are illustrative drawings for explaining a derivative vote and a voting result for a derivative vote that are referenced by some embodiments of the present disclosure.

[0154] Referring to FIG. 10, a first derivative vote (100) and a voting result (101) for the first derivative vote are shown.

[0155] The first derived vote (100) may include a vote title, a vote body, and a vote item. The vote title may be related to the second topic described with reference to FIG. 8, and the vote item may be a reason for selection included in the first group described with reference to FIG. 8. The vote body may be generated based on the first news article described with reference to FIG. 9.

[0156] For example, referring to FIG. 5, the selection reason (52b) included in the first voter's voting result (52), the selection reason (55b) included in the fourth voter's selection result (55), the selection reason (56b) included in the fifth voter's selection result (56), and the selection reason (58b) included in the seventh voter's selection result (58) can all be classified as selection reasons regarding preferred overseas travel destinations within Asia by season. Therefore, the above selection reasons (52b, 55b, 56b, 58b) can be included in the first group regarding preferred overseas travel destinations by season.

[0157] In this case, the second topic may be 'preferred Asian travel destinations in summer', and the derived item may be generated based on the Asian travel destinations included in the reasons for selection (52b, 55b, 56b, 58b).

[0158] The service server (11) can automatically generate a prompt to generate a vote using the second topic, the derivative item, and the acquired first news article, and can generate a derivative vote (100) by inputting the second topic, the derivative item, and the first news article together with the prompt into a generative AI model (13).

[0159] Referring to FIG. 11, a second derivative vote (110) and a voting result (111) for the second derivative vote are shown.

[0160] The second derivative vote (110) may include a vote title, a vote body, and a vote item. The vote title may be related to the second topic, and the vote item may be a reason for selection included in the first group.

[0161] For example, the selection reason (53b) included in the second voter's voting result (53), the selection reason (54b) included in the third voter's voting result (54), and the selection reason (59b) included in the eighth voter's selection result (59) can all be classified as selection reasons regarding preferred overseas travel destinations within Asia based on the length of the holiday. Accordingly, the above selection reasons (53b, 54b, 59b) can be included in the second group regarding preferred overseas travel destinations within Asia based on the length of the holiday.

[0162] In this case, the second topic may be 'preferred overseas travel destinations within Asia if the holiday is short', and the derivative item may be generated based on the Asian travel destinations included in the reasons for selection (53b, 54b, 59b).

[0163] The service server (11) can automatically generate a prompt to generate a vote using the second topic, the derivative item, and the acquired first news article, and can generate a derivative vote (110) by inputting the second topic, the derivative item, and the first news article together with the prompt into a generative AI model (13).

[0164] Referring to FIG. 12, the third derivative vote (120) and the voting result (121) for the third derivative vote are illustrated.

[0165] The third derivative vote (120) may include a vote title, a vote body, and a vote item. The vote title may be related to the second topic, and the vote item may be a reason for selection included in the first group.

[0166] For example, referring to FIG. 5, the selection reason (52b) included in the first voter's voting result (52), the selection reason (55b) included in the fourth voter's selection result (55), the selection reason (56b) included in the fifth voter's selection result (56), and the selection reason (58b) included in the seventh voter's selection result (58) can all be classified as selection reasons regarding preferred overseas travel destinations within Asia by season. Therefore, the above selection reasons (52b, 55b, 56b, 58b) can be included in the first group regarding preferred overseas travel destinations by season.

[0167] In this case, the second topic may be about the 'time of vacation', and the derivative item may be generated based on the season included in the reason for selection (53b, 54b, 59b).

[0168] The service server (11) can automatically generate a prompt to generate a vote using the second topic, the derivative item, and the acquired first news article, and can generate a derivative vote (120) by inputting the second topic, the derivative item, and the first news article together with the prompt into a generative AI model (13).

[0169] According to the present embodiment, a derived vote can be generated by reflecting public opinion, such as voter opinions and news articles, regarding a topic included in a specific keyword. That is, according to the present embodiment, a derived vote can be generated that more faithfully reflects the opinions of voters (=public opinion) regarding a topic included in a specific keyword.

[0170] Accordingly, according to the present embodiment, there is an advantage in that it provides a user using the voting system (10) with votes from various perspectives regarding topics included in a specific keyword, and the user experience can be enhanced due to the diversity of the votes.

[0171] Hereinafter, a method for generating derived votes according to another embodiment of the present disclosure is described with reference to FIG. 13. FIG. 13 is a detailed flowchart for describing a method for generating voting results according to another embodiment of the present disclosure, described with reference to FIG. 2.

[0172] Referring to FIG. 13, the derived item may include a first derived item and a second derived item. The service server (11) can compare a first count, which is the voting amount for the first derived item, with a second count, which is the voting amount for the second derived item (S320).

[0173] As a result of the comparison of S320, if the first count is greater than the second count, the service server (11) may place the first derived item above the second derived item on the derived vote (S321).

[0174] Referring to Fig. 12, the voting result (121) for the third derivative vote is shown.

[0175] Referring to the voting results (121), the percentage of people who selected ‘A. Spring’ was 10%, the percentage who selected ‘B. Summer’ was 57%, the percentage who selected ‘C. Autumn’ was 10%, and the percentage who selected ‘D. Winter’ was 23%.

[0176] In this case, the service server (11) can change the arrangement of the voting items included in the third derivative vote according to the amount of votes for each derivative item (122, 123, 124, 125). That is, since the amount of votes for the second derivative item (123) is the largest at 57%, the service server (11) can change the arrangement so that the second derivative item (123) is positioned above the first derivative item (122), the third derivative item (124), and the fourth derivative item (125) in the voting items on the derivative vote (120). Since the amount of votes for the fourth derivative item (125) is the second largest at 23%, the service server (11) can change the arrangement so that the fourth derivative item (125) is positioned above the first derivative item (122) and the third derivative item (124) in the voting items on the derivative vote (120).

[0177] Hereinafter, a method for generating derived votes according to another embodiment of the present disclosure is described with reference to FIG. 14. FIG. 14 is a detailed flowchart for describing a method for generating voting results according to another embodiment of the present disclosure, described with reference to FIG. 2.

[0178] Referring to FIG. 14, the derivative vote may include a vote body, a first derivative item, and a second derivative item. The vote body may be the vote body included in each derivative vote (100, 110, 120) of FIG. 10 to 12, and the first derivative item and the second derivative item may be the vote items included in each derivative vote (100, 110, 120) of FIG. 10 to 12.

[0179] The service server (11) can extract a first keyword from the voting text (S420). In this case, there may be multiple first keywords. The service server (11) can extract the first keyword from the voting text using a rule-based, statistical-based, machine learning, and deep learning method. A person skilled in the art to which the invention according to the present disclosure belongs would already be familiar with the method of extracting keywords from text, so a detailed explanation is omitted.

[0180] Subsequently, the service server (11) can extract a second keyword from each of the selection reasons included in the first group (S421). That is, the service server (11) can extract a second keyword from each of the selection reasons included in the first group grouped by a specific topic. At this time, there may be multiple second keywords. Since the method for extracting the second keyword is the same as described with reference to step S420, a detailed explanation is omitted for ease of understanding.

[0181] Afterwards, the service server (11) can calculate the similarity between the first keyword and the second keyword and obtain a similarity score (S422). The service server (11) can calculate the similarity score between the first keyword and the second keyword according to the following mathematical formula.

[0182] [Mathematical Formula 2]

[0183]

[0184] (N: Number of second keywords extracted from each selection reason included in the first group)

[0185] M: Number of first keywords extracted from the vote text

[0186] kv: 1st keyword

[0187] kc: 2nd keyword

[0188] sim(kv, kc): Cosine similarity between the first keyword and the second keyword

[0189] ∑sim(kv, kc): Sum of similarities for the first keyword)

[0190] That is, the service server (11) can calculate the significance of the comment as the similarity score based on the keyword similarity between the voting text and the selection reason (e.g., comment) included in the first group.

[0191] The service server (11) can calculate a numerical score (similarity score) indicating how much the selection reason included in the first group is related to the voting text through the above mathematical formula 2.

[0192] The service server (11) can calculate the similarity between each first keyword and each second keyword. In this case, the service server (11) can vectorize each first keyword and each second keyword using the Word2Vec technique and calculate the similarity between the first keyword and the second keyword through cosine similarity.

[0193] The service server (11) can select the highest maximum similarity value among the similarities with the second keyword for each first keyword and calculate ∑sim(kv, kc) by summing the maximum similarity values.

[0194] The above similarity score can serve as a criterion for evaluating how highly relevant the selection reasons included in the first group are to the voting text. A higher similarity score indicates that the selection reason is more closely related to the derived vote.

[0195] Subsequently, the service server (11) can change the arrangement order of the first derived item and the second derived item based on the calculated similarity score (S423). For example, let's assume that the first derived item is generated based on the first comment among the selection reasons included in the first group, and the second derived item is generated based on the second comment among the selection reasons included in the first group. If, as a result of performing S422, the similarity score of the first comment is greater than the similarity score of the second comment, the service server (11) can determine that the first derived item is more relevant than the second derived item for the derived vote. Accordingly, the service server (11) can place the first derived item above the second derived item in the voting items within the derived vote.

[0196] According to the present embodiment, for the generated derived items, the arrangement order of the derived items included in the derived vote can be determined based on the relationship between keywords extracted from the reason for selection (e.g., comments) and keywords extracted from the body of the vote. Therefore, according to the present embodiment, there is an advantage in that a more improved user interface screen can be provided to voters by placing derived items highly relevant to the derived vote at the top.

[0197] Hereinafter, with reference to FIG. 15, the final voting results referenced by some embodiments of the present disclosure will be described. FIG. 15 is an illustrative diagram for explaining the final voting results referenced by some embodiments of the present disclosure.

[0198] Referring to FIG. 15, a final voting result (150) generated based on the voting results (101, 111, 121) for each derived vote (100, 110, 120) of FIG. 10 to 12 is shown.

[0199] The voting results (101) for the first derivative vote (100) are 35% for 'A. Thailand', 14% for 'B. Zhangjiajie', 22% for 'C. Boracay', 14% for 'D. Hokkaido', and 15% for 'E. Bali'. The voting results (111) for the second derivative vote (110) are 44% for 'A. Thailand', 12% for 'B. Zhangjiajie', 12% for 'C. Boracay', 20% for 'D. Hokkaido', and 12% for 'E. Bali'. The voting results (121) for the third derivative vote (120) are 10% for 'A. Spring', 57% for 'B. Summer', 10% for 'C. Autumn', and 23% for 'D. Winter'.

[0200] The service server (11) can automatically generate a prompt to generate a final analysis result using the voting result for the basic vote and the voting result for the derivative vote, and can generate the final voting result by inputting the voting result for the basic vote and the voting result for the derivative vote together with the prompt into a generative AI model (13).

[0201] Referring to FIG. 15, the service server (11) can automatically generate a prompt to generate a final analysis result using the voting results (101, 111, 121) for the derived vote (100, 110, 120), and can generate a final voting result (150) by inputting the voting results (101, 111, 121) for the derived vote (100, 110, 120) together with the prompt into a generative AI model (13).

[0202] FIG. 16 is a hardware configuration diagram of a computing device according to some embodiments of the present disclosure. The computing device (1000) of FIG. 15 may include one or more processors (1100), a system bus (1600), a communication interface (1200), a memory (1400) for loading a computer program (1500) executed by the processor (1100), and a storage (1300) for storing the computer program (1500).

[0203] The computing system (1000) of FIG. 16 may present a hardware structure of one or more computing systems that constitute a service server (11) described with reference to FIG. 1, for example.

[0204] The processor (1100) controls the overall operation of each component of the computing system (1000). The processor (1100) may perform operations on at least one application or program for executing methods / operations according to various embodiments of the present disclosure. Memory (1400) stores various data, instructions and / or information. Memory (1400) may load one or more computer programs (1500) from storage (1300) to execute methods / operations according to various embodiments of the present disclosure. Storage (1300) may store one or more computer programs (1500) non-temporarily.

[0205] A computer program (1500) may include one or more instructions in which methods / operations according to various embodiments of the present disclosure are implemented. When the computer program (1500) is loaded into memory (1400), a processor (1100) may perform methods / operations according to various embodiments of the present disclosure by executing the one or more instructions.

[0206] In one embodiment, a computer program (1500) may include instructions for performing operations such as receiving a first voting result for a basic vote related to a first topic included in a first keyword from a group of voter terminals, wherein the first voting result includes, for the basic vote, a basic voting selection item for each voter terminal included in the group of voter terminals and a reason for selecting the basic voting selection item; generating a derivative item for a second topic related to the first topic based on the basic voting selection item and the reason for selection; generating a derivative vote different from the basic vote based on the second topic and the derivative item; receiving a second voting result for the derivative vote from the group of voter terminals, wherein the second voting result includes information regarding the ratio selected by each of the derivative items included in the derivative vote by the voter terminals included in the group of voter terminals; and generating a final voting result based on the first voting result and the second voting result.

[0207] Various embodiments of the present disclosure and effects according to those embodiments have been described with reference to FIGS. 1 to 16. The effects according to the technical concept of the present disclosure are not limited to those mentioned above, and other unmentioned effects will be clearly understood by a person skilled in the art from the description below.

[0208] Furthermore, just because the above embodiments describe a plurality of components being combined into one or operating in combination, the technical concept of the present disclosure is not necessarily limited to these embodiments. That is, within the scope of the purpose of the technical concept of the present disclosure, all such components may be selectively combined into one or more combinations to operate.

[0209] The technical concept of the present disclosure described above may be implemented as computer-readable code on a computer-readable medium. A computer program recorded on a computer-readable recording medium may be transmitted to another computing device via a network such as the Internet and installed on said other computing device, thereby being used on said other computing device.

Claims

1. In a method performed by a computing device, A step of receiving a first voting result for a basic vote related to a first topic included in a first keyword from a group of voter terminals, wherein the first voting result includes, for the basic vote, a basic voting selection item for each voter terminal included in the group of voter terminals and a reason for selecting the basic voting selection item; A step of generating derivative items regarding a second topic related to the first topic based on the above basic voting selection items and the above reasons for selection; A step of generating a derivative vote different from the basic vote based on the second topic and the derivative item above; A step of receiving a second voting result for the derivative vote from the voter terminal group, wherein the second voting result includes information regarding the ratio selected by the voter terminal included in the voter terminal group for each of the derivative items included in the derivative vote; and A method comprising the step of generating a final voting result based on the first voting result and the second voting result. Method for generating voting results.

2. In Paragraph 1, The method further includes the step of generating the basic vote based on the content for the first keyword, wherein The above first keyword is, It is included in a keyword group consisting of multiple keywords, and The above content is, A first news article stored in a database regarding the first keyword or a voter opinion regarding the first keyword, Method for generating voting results.

3. In Paragraph 2, The step of generating the above basic vote is, A step of selecting the first keyword among the above keyword groups for which voter opinions exceeding a threshold have been written; A step of grouping the above-mentioned voter opinions and obtaining a first group having the maximum number of voter opinions; A step of obtaining a first embedding vector for the first voter opinion included in the first group; A step of calculating the similarity between the first embedding vector and the second embedding vector for an external news article different from the first news article, and obtaining the second news article having the maximum calculated similarity; and A method comprising the step of generating the basic vote related to the first topic based on the first voter's opinion and the second news article. Method for generating voting results.

4. In Paragraph 2, The step of generating the above basic vote is, A step of selecting the first keyword among the above keyword group for which news articles exceeding a threshold have been written; A step of grouping the above-mentioned first news articles and obtaining a first group having the maximum number of news articles; A step of obtaining a first paragraph by dividing a second news article included in the first group based on a newline, and obtaining a second paragraph by dividing a third news article included in the first group based on a newline; A step of calculating a first representativeness score for a first embedding vector for the above first paragraph; A step of calculating a second representativeness score for a second embedding vector for the above second paragraph; A step of comparing the first representative score and the second representative score; A step of selecting the first paragraph as the representative paragraph when, as a result of the above comparison, the first representativeness score is greater than the second representativeness score; and Based on the above representative paragraph, the method includes the step of generating the above basic vote related to the above first topic, wherein The above first representativeness score is, It is the sum of the similarities between the above-mentioned first paragraph and the above-mentioned second paragraph, and The above second representativeness score is, The sum of the similarities of the above second paragraph to the above first paragraph, Method for generating voting results.

5. In Paragraph 4, The step of calculating the first representativeness score above is, A step comprising calculating the first representative score according to the following mathematical formula 1, Method for generating voting results. [Mathematical Formula 1] (n: number of paragraphs X: A matrix representing the embedding vectors of each paragraph as rows i, j: Paragraph number) 6. In Paragraph 4, Based on the first paragraph above, the step of generating the basic vote related to the first topic is, A step of calculating the similarity between a first embedding vector for the above-mentioned first paragraph and a second embedding vector for an external news article different from the above-mentioned first news article, and obtaining a second news article having the maximum calculated similarity; and Based on the above representative paragraph and the above second news article, the method comprising the step of generating the above basic vote related to the above first topic, Method for generating voting results.

7. In Paragraph 1, The step of generating the above-mentioned derived item is, A step of obtaining a first selection reason for selecting a first selection item included in the above basic voting selection items; A step of grouping the first selection reasons by topic and obtaining a first group including some of the selection reasons included in the first selection reasons and a second group including some of the selection reasons included in the first selection reasons; and A method comprising the step of inputting the selection reasons included in the first group and the first topic into a generative AI model, and generating the second topic and the derived item as a result of the input. Method for generating voting results.

8. In Paragraph 7, The step of generating the above-mentioned derived vote is, A step of obtaining a first embedding vector for the selection reason included in the first group above; A step of calculating the similarity between the first embedding vector and the second embedding vector for the news article for the first keyword, and obtaining the first news article having the maximum calculated similarity; and The method comprises the step of inputting the above-mentioned second topic, the above-mentioned derivative item, and the above-mentioned first news article into the above-mentioned generative AI model, and generating the above-mentioned derivative vote as a result of the input. Method for generating voting results.

9. In Paragraph 7, The above derived item is, Includes a first derivative item and a second derivative item, The step of generating the above-mentioned derived vote is, A step of comparing a first count, which is the voting amount for the first derivative item, and a second count, which is the voting amount for the second derivative item; and As a result of the above comparison, if the first count is greater than the second count, the method includes the step of placing the first derivative item above the second derivative item on the derivative vote. Method for generating voting results.

10. In Paragraph 7, The above derivative vote is, Includes the voting body, the first derivative item and the second derivative item, and The step of generating the above-mentioned derived vote is, A step of extracting a first keyword from the above voting text; A step of extracting a second keyword from the selection reasons included in the first group above; A step of calculating the similarity between the first keyword and the second keyword and obtaining a similarity score; and A step comprising changing the arrangement order of the first derived item and the second derived item based on the similarity score above. Method for generating voting results.

11. Communication interface; Memory where a computer program is loaded; and The computer program described above includes one or more processors on which it is executed, The above computer program is, Operation of receiving a first voting result for a basic vote related to a first topic included in a first keyword from a group of voter terminals, wherein the first voting result includes, for the basic vote, a basic voting selection item for each voter terminal included in the group of voter terminals and a reason for selecting the basic voting selection item; An action of generating a derivative item regarding a second topic related to the first topic based on the above basic voting selection item and the above reason for selection; An action of generating a derivative vote different from the basic vote based on the above second topic and the above derivative item; Operation of receiving a second voting result for the derivative vote from the above-mentioned voter terminal group, wherein the second voting result includes information regarding the ratio selected by the voter terminal included in the voter terminal group for each of the derivative items included in the derivative vote; and Instructions for performing an operation to generate a final vote result based on the first vote result and the second vote result, Voting result generation system.

12. Combined with a computing device, A step of receiving a first voting result for a basic vote related to a first topic included in a first keyword from a group of voter terminals, wherein the first voting result includes, for the basic vote, a basic voting selection item for each voter terminal included in the group of voter terminals and a reason for selecting the basic voting selection item; A step of generating derivative items regarding a second topic related to the first topic based on the above basic voting selection items and the above reasons for selection; A step of generating a derivative vote different from the basic vote based on the second topic and the derivative item above; A step of receiving a second voting result for the derivative vote from the voter terminal group, wherein the second voting result includes information regarding the ratio selected by the voter terminal included in the voter terminal group for each of the derivative items included in the derivative vote; and A step of generating a final vote result based on the first vote result and the second vote result. A computer program stored on a computer-readable recording medium.

Images