TWM649741U - Similar internet celebrity matching recommendation system - Google Patents
Similar internet celebrity matching recommendation system Download PDFInfo
- Publication number
- TWM649741U TWM649741U TW112208962U TW112208962U TWM649741U TW M649741 U TWM649741 U TW M649741U TW 112208962 U TW112208962 U TW 112208962U TW 112208962 U TW112208962 U TW 112208962U TW M649741 U TWM649741 U TW M649741U
- Authority
- TW
- Taiwan
- Prior art keywords
- internet
- similar
- celebrity
- internet celebrity
- data
- Prior art date
Links
- 238000013500 data storage Methods 0.000 claims abstract description 24
- 238000007418 data mining Methods 0.000 claims abstract description 9
- 238000005516 engineering process Methods 0.000 claims description 18
- 238000013473 artificial intelligence Methods 0.000 claims description 10
- 230000003993 interaction Effects 0.000 claims description 10
- 238000013135 deep learning Methods 0.000 claims description 7
- 230000003796 beauty Effects 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 8
- IWEDIXLBFLAXBO-UHFFFAOYSA-N dicamba Chemical compound COC1=C(Cl)C=CC(Cl)=C1C(O)=O IWEDIXLBFLAXBO-UHFFFAOYSA-N 0.000 description 5
- 230000006399 behavior Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000011156 evaluation Methods 0.000 description 3
- 230000002452 interceptive effect Effects 0.000 description 3
- 238000000034 method Methods 0.000 description 2
- 238000003058 natural language processing Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
本創作揭露一種相似網紅匹配推薦系統,包含資料探勘裝置、資料儲存器、資料處理器以及計算機裝置。資料探勘裝置透過網路連接社群平台,並根據目標網紅的目標特徵,於社群平台中搜尋及擷取網紅資料;其中網紅資料具有相似特徵,相似特徵與目標特徵相似。資料儲存器連接資料探勘裝置,儲存複數位網紅資料。資料處理器連接資料儲存器,利用演算法分析比較目標特徵與相似特徵的近似程度,以產生初始網紅排名資料。計算機裝置連接資料處理器,根據加權比例加權初始網紅排名資料,以產生相似目標網紅的相似網紅分析結果。This invention discloses a similar internet celebrity matching recommendation system, including a data exploration device, a data storage, a data processor and a computer device. The data mining device connects to the social platform through the Internet, and searches and retrieves Internet celebrity information in the social platform based on the target characteristics of the target Internet celebrity; the Internet celebrity information has similar characteristics, and the similar characteristics are similar to the target characteristics. The data storage device is connected to the data exploration device and stores multiple Internet celebrity data. The data processor connects to the data storage and uses algorithms to analyze and compare the similarity between the target features and similar features to generate initial internet celebrity ranking data. The computer device is connected to the data processor and weights the initial influencer ranking data according to the weighted ratio to generate similar influencer analysis results of similar target influencers.
Description
本創作係關於一種網紅分析推薦系統,特別是關於一種相似網紅匹配推薦系統。This creation is about an Internet celebrity analysis and recommendation system, especially a similar Internet celebrity matching recommendation system.
現今網路傳播速度快速,社群媒體眾多,因而興起許多網路紅人(以下簡稱為網紅)透過各種社群媒體進行表演,以吸引粉絲的觀看。而對於粉絲而言,現今的網紅人數眾多,風格迥異,各有特色,因此,若可將相似網紅推薦給粉絲,則可進一步媒合促進網紅與粉絲之間的互動。Nowadays, the speed of Internet communication is fast and there are many social media. As a result, many Internet celebrities (hereinafter referred to as Internet celebrities) have emerged to perform through various social media to attract fans to watch. For fans, there are a large number of Internet celebrities today with different styles and characteristics. Therefore, if similar Internet celebrities can be recommended to fans, it can further promote the interaction between Internet celebrities and fans.
在現有技術中,相似網紅的推薦方法主要依賴粉絲在網紅社群媒體上所產生的互動,例如點讚、評論、分享等互動數據,進一步分析出粉絲的喜好之後,將相似的網紅推薦給粉絲。然而,由於現有技術往往忽略重要的相似性指標,例如忽略了網紅的內容風格、受眾組成等指標,因而無法全面評估各個網紅之間的相似性。In the existing technology, the recommendation method of similar Internet celebrities mainly relies on the interaction generated by fans on Internet celebrity social media, such as likes, comments, shares and other interactive data. After further analyzing the fans' preferences, similar Internet celebrities are Recommended to fans. However, because the existing technology often ignores important similarity indicators, such as the content style and audience composition of Internet celebrities, it is impossible to comprehensively assess the similarity between Internet celebrities.
此外,現有技術在相似網紅的分析上亦容易受到數據稀疏性、數據有效性以及數據完整性的影響,導致分析結果失真,因而無法精準、理想地向粉絲推薦相似的網紅。In addition, the existing technology is also easily affected by data sparsity, data validity and data integrity in the analysis of similar Internet celebrities, resulting in distortion of the analysis results, making it impossible to accurately and ideally recommend similar Internet celebrities to fans.
據此,如何提供一種相似網紅匹配推薦系統,已成為目前急需研究的課題。Accordingly, how to provide a matching recommendation system for similar internet celebrities has become an urgent research topic.
本創作揭露一種相似網紅匹配推薦系統,包含一資料探勘裝置、一資料儲存器、一資料處理器以及一計算機裝置。資料探勘裝置透過一網路連接至少一社群平台,並根據一目標網紅的至少一目標特徵,於至少一社群平台中搜尋及擷取複數位網紅資料;其中複數位網紅資料具有至少一相似特徵,至少一相似特徵與至少一目標特徵相似。資料儲存器連接資料探勘裝置,儲存複數位網紅資料。資料處理器連接資料儲存器,利用至少一演算法分析比較至少一目標特徵與至少一相似特徵的近似程度,以產生初始網紅排名資料。計算機裝置連接資料處理器,根據一加權比例加權初始網紅排名資料,以產生相似目標網紅的相似網紅分析結果。This invention discloses a similar internet celebrity matching recommendation system, which includes a data exploration device, a data storage, a data processor and a computer device. The data mining device connects to at least one social platform through a network, and searches and retrieves data on a plurality of Internet celebrities in at least one social platform based on at least one target characteristic of a target Internet celebrity; wherein the data on the plurality of Internet celebrities has At least one similar feature is similar to at least one target feature. The data storage device is connected to the data exploration device and stores multiple Internet celebrity data. The data processor is connected to the data storage and uses at least one algorithm to analyze and compare the similarity between at least one target feature and at least one similar feature to generate initial internet celebrity ranking data. The computer device is connected to the data processor and weights the initial influencer ranking data according to a weighted ratio to generate similar influencer analysis results of similar target influencers.
承上所述,本創作提供基於多維度特徵的相似網紅匹配推薦系統,從多方面評估目標網紅與相似網紅之間的相似性,提供粉絲可更方便、快速地找到具有相似特徵的網紅,同時網紅和廣告主也可根據推薦結果進行精準推廣和合作。Following the above, this creation provides a similar internet celebrity matching recommendation system based on multi-dimensional characteristics, which evaluates the similarity between the target internet celebrity and similar internet celebrities from multiple aspects, allowing fans to more conveniently and quickly find those with similar characteristics. Internet celebrities, Internet celebrities and advertisers can also carry out precise promotion and cooperation based on the recommendation results.
請參閱圖1,其係為本創作相似網紅匹配推薦系統的方塊示意圖。相似網紅匹配推薦系統1包含一資料探勘裝置11、一資料儲存器12、一資料處理器13以及一計算機裝置14。資料探勘裝置11透過一網路連接至少一社群平台F,並根據一目標網紅的至少一目標特徵,於至少一社群平台F中搜尋及擷取複數位網紅資料。複數位網紅資料具有至少一相似特徵,至少一相似特徵與至少一目標特徵相似。資料儲存器12連接資料探勘裝置11,儲存複數位網紅資料。資料處理器13連接資料儲存器12,利用至少一演算法分析比較至少一目標特徵與至少一相似特徵的近似程度,以產生初始網紅排名資料。計算機裝置14連接資料處理器13,根據一加權比例加權初始網紅排名資料,以產生相似目標網紅的相似網紅分析結果。透過使用本創作相似網紅匹配推薦系統1,可提供如本案圖2所示的系統平台給使用的用戶,瀏覽參閱相似網紅的排名及分析結果。Please refer to Figure 1, which is a block diagram of the similar internet celebrity matching recommendation system of this creation. The similar internet celebrity
目標網紅的目標特徵包含一網紅風格特徵、一粉絲組成特徵以及一互動模式特徵,各個特徵可對應設定高低的加權比例。於本創作的實施例中,可設定網紅風格特徵以及粉絲組成特徵具有較高的加權比例,互動模式特徵具有較低的加權比例,於本創作中並不限定高低的比例。因此,相似網紅匹配推薦系統1透過設定目標網紅的目標特徵後,針對各個社群媒體中的網紅進行相似性評估分析。例如,相似網紅匹配推薦系統1透過關鍵字搜尋到目標網紅,並透過例如標籤、性別、過濾器等方式,選擇其中一個目標網紅後,產生推薦的目標網紅,以進行後續的相似網紅分析。The target characteristics of the target Internet celebrity include an Internet celebrity style characteristic, a fan composition characteristic, and an interaction mode characteristic. Each characteristic can have a high or low weighting ratio set accordingly. In the embodiment of this creation, the influencer style characteristics and fan composition characteristics can be set to have a higher weighting ratio, and the interaction mode characteristics have a lower weighting ratio. In this creation, the high and low ratios are not limited. Therefore, the similar influencer matching
網紅風格特徵包含一貼文風格特徵、一評論風格特徵以及一頭圖圖像風格特徵。貼文風格特徵包含一遊戲、一美妝時尚、一健康、一美食、一藝術和娛樂、一運動、一商業和經濟、一家庭和關係、一科技、一法律政治和社會、一旅遊、一教育工作和學習、一寵物、一生活、一宗教命理、一成人、一交通工具、一日常話題以及一氣候和環境貼文風格特徵,評論風格特徵包含一產業經濟風格特徵、一政治新聞風格特徵、一娛樂媒體風格特徵,頭圖圖像風格特徵包含一性別風格特徵、一裝扮風格特徵以及一造型風格特徵。舉例而言,若該網紅經常在社群媒體上貼出旅遊地點的景點照片以及美食店家的介紹,透過演算法擷取到的圖片及文字,資料處理器13可分析歸納出該網紅屬於旅遊風格特徵的網紅。再者,透過資料探勘裝置11擷取到該網紅的頭圖圖像是女性的圖像特徵,且資料處理器13透過演算法辨識出該網紅的頭髮造型風格特徵以及裝扮風格特徵近似於韓國藝人的風格特徵,資料處理器13可分析歸納出該網紅屬於韓國風格的網紅。相似地,其餘網紅風格特徵以此類推,於此不再贅述。Internet celebrity style features include a post style feature, a comment style feature, and a picture image style feature. Post style characteristics include - games, - beauty and fashion, - health, - food, - arts and entertainment, - sports, - business and economics, - family and relationships, - technology, - law, politics and society, - travel, - Education, work and study, one pet, one life, one religious numerology, one adult, one transportation, one daily topic and one climate and environment post style feature, the comment style feature includes an industry and economic style feature, and a political news style feature , an entertainment media style feature, the header image style feature includes a gender style feature, a dress style feature and a modeling style feature. For example, if the Internet celebrity often posts photos of tourist attractions and introductions to food stores on social media, the data processor 13 can analyze and summarize the images and text captured by the algorithm to determine whether the Internet celebrity belongs to Internet celebrities with travel style characteristics. Furthermore, the profile image of the internet celebrity captured through the data mining device 11 is a female image feature, and the data processor 13 uses an algorithm to identify that the hair style and dress style features of the internet celebrity are similar to Based on the style characteristics of Korean artists, the data processor 13 can analyze and conclude that the Internet celebrity belongs to Korean-style Internet celebrities. Similarly, the style characteristics of other Internet celebrities can be deduced in the same way and will not be described again here.
粉絲組成特徵包含性別特徵、人數特徵、地區特徵、國家特徵及年齡特徵。透過搜尋、擷取及比對相似的粉絲組成特徵,可提供用戶較為相近的網紅風格。性別特徵及年齡特徵可透過演算法收集辨識粉絲的性別及年齡區間,人數特徵可透過目標網紅的粉絲人數得知,地區特徵及國家特徵可透過演算法蒐集到粉絲的打卡地點歸類分群。舉例而言,透過資料探勘裝置11擷取到目標網紅的粉絲大部分為男性,人數大約100萬,年齡層介於20至30歲之間,國家及地區在台灣的台北市區,資料處理器13可分析歸納出目標網紅屬於台灣的女性網紅,且年齡同樣介於20至30歲之間。Fan composition characteristics include gender characteristics, number characteristics, regional characteristics, national characteristics and age characteristics. By searching, retrieving and comparing similar fan composition characteristics, users can be provided with relatively similar internet celebrity styles. Gender characteristics and age characteristics can be collected through algorithms to identify the gender and age range of fans. Population characteristics can be known through the number of fans of the target Internet celebrity. Regional characteristics and country characteristics can be collected through algorithms to classify the fans' check-in locations into groups. For example, through the data mining device 11, it is captured that most of the fans of the target Internet celebrity are male, the number is about 1 million, the age range is between 20 and 30 years old, the country and region is Taipei City, Taiwan, and the data processing Tool 13 can analyze and conclude that the target internet celebrity is a Taiwanese female internet celebrity, and her age is also between 20 and 30 years old.
互動模式特徵包含網紅留言回覆頻率、網紅留言分享頻率、網紅按讚比率、網紅發文比率,透過資料探勘裝置11搜尋及擷取各種互動模式的紀錄,並由資料處理器13比對相似的互動模式特徵,可提供用戶較為相近的網紅風格。舉例而言,當用戶對目標網紅的按讚比率達到70%時,透過資料探勘裝置11搜尋及擷取到另一位網紅的按讚比率達到65%時,資料處理器13可將網紅按讚比率相近的網紅歸類為相似網紅。相似地,網紅留言回覆頻率、網紅留言分享頻率以及網紅發文比率以此類推,於此不再贅述。The characteristics of the interaction mode include the frequency of replying to the Internet celebrity's messages, the frequency of sharing the Internet celebrity's messages, the Internet celebrity's like rate, and the Internet celebrity's posting rate. Records of various interaction modes are searched and retrieved through the data mining device 11 and compared by the data processor 13 Similar interaction mode features can provide users with relatively similar Internet celebrity styles. For example, when the user's like rate for the target internet celebrity reaches 70%, and when the like rate of another internet celebrity reaches 65% through the data mining device 11, the data processor 13 can search and retrieve the like rate of the target internet celebrity. Internet celebrities with similar like ratios are classified as similar Internet celebrities. Similarly, the frequency of replying to Internet celebrities' messages, the frequency of sharing of Internet celebrities' messages, and the ratio of Internet celebrities' postings can be deduced and will not be repeated here.
於本創作中,相似網紅匹配推薦系統1利用各種演算法自動擷取、分析及轉換相似網紅的相似特徵,並利用各種演算法及加權比例為用戶產生推薦相似的網紅。於本創作中,演算法包含深度學習演算法、自然語言演算法、影像辨識演算法、爬蟲演算法。In this creation, the similar internet celebrity
自然語言演算法及爬蟲演算法係根據複數位網紅資料與目標網紅的留言、評論或者貼文中的關鍵字分析比較至少一目標特徵與至少一相似特徵的近似程度。於本創作中,資料處理器13透過爬蟲演算法結合自然語言演算法可針對網紅的貼文分析語意內容的近似程度,並於該語意分析內容近似時,將網紅貼上相似網紅的標籤(Tag),以進行內容分類。舉例而言,當目標網紅在社群媒體上經常貼出台北美食的文章,透過爬蟲演算法擷取到另一位網紅也經常在社群媒體上貼出台北美食的文章時,甚至於該位網紅與目標網紅都會互相提到(mention)對方的名字時,資料處理器13透過自然語言演算法可由語意推論出該位網紅與目標網紅是有關聯性、或者有交集存在,因此,將該位網紅歸類為相似網紅。The natural language algorithm and the crawler algorithm compare the degree of similarity between at least one target feature and at least one similar feature based on the plurality of Internet celebrity information and keyword analysis in the target Internet celebrity's messages, comments, or posts. In this creation, the data processor 13 uses a crawler algorithm combined with a natural language algorithm to analyze the similarity of the semantic content of the Internet celebrity's posts, and when the semantic analysis content is similar, the Internet celebrity posts similar Internet celebrities. Tags to classify content. For example, when the target Internet celebrity often posts articles about Taipei food on social media, and it is retrieved through the crawler algorithm that another Internet celebrity often posts articles about Taipei food on social media, even When the Internet celebrity and the target Internet celebrity mention each other's names, the data processor 13 can use the natural language algorithm to infer from the semantics that the Internet celebrity and the target Internet celebrity are related or have intersections. , therefore, classify this internet celebrity as a similar internet celebrity.
影像辨識演算法根據複數位網紅資料的網紅圖像以及目標網紅的目標圖像分析比較至少一相似特徵與至少一目標特徵的該近似程度。例如,影像辨識演算法辨識出網紅圖像中網紅髮型為金黃色髮型的特徵,目標網紅的目標圖像同樣為金黃色髮型的特徵時,資料處理器13透過影像辨識演算法可將該位網紅歸類為相似目標網紅。或者,透過影像辨識演算法辨識出當目標網紅以及該網紅在社群媒體上都有標記(Hashtag)同一個人或是標記同一群人時,資料處理器13透過影像辨識演算法可將該位網紅歸類為相似目標網紅。The image recognition algorithm analyzes and compares the degree of similarity between at least one similar feature and at least one target feature based on the internet celebrity images of the plurality of internet celebrity data and the target image of the target internet celebrity. For example, the image recognition algorithm identifies the characteristics of the Internet celebrity's hair style in the Internet celebrity image, and the target image of the target Internet celebrity also has the characteristics of the Internet celebrity's hair style. The data processor 13 can use the image recognition algorithm to identify the characteristics of the Internet celebrity's hair style. This influencer is classified as a similar target influencer. Alternatively, when the target Internet celebrity and the Internet celebrity have tagged (Hashtag) the same person or tagged the same group of people on social media, the data processor 13 can use the image recognition algorithm to identify the target Internet celebrity through the image recognition algorithm. The influencers are classified as similar target influencers.
深度學習演算法係根據網紅留言回覆頻率、網紅留言分享頻率及網紅按讚比率的高低,分析比較至少一目標特徵與至少一相似特徵的近似程度。例如,上述的網紅按讚比率可透過深度學習演算法進行相似度的比較,根據學習的比例高低判斷網紅的近似程度。或者,透過分析該網紅的互動表現特徵進行判斷。例如,若目標網紅與相似網紅的粉絲組成特徵皆有約1百萬的粉絲,但目標網紅在社群媒體上的影響力以及互動表現主要是針對旅遊族群的粉絲,而另一位網紅的在社群媒體上的影響力以及評論(comment)透過自然語言處理分析出其評論主要是針對政治族群的粉絲,因此,透過分析該網紅的互動表現特徵可判斷出該位網紅並不近似目標網紅。The deep learning algorithm analyzes and compares the degree of similarity between at least one target feature and at least one similar feature based on the frequency of replying to the internet celebrity's messages, the frequency of sharing the internet celebrity's messages, and the rate of the internet celebrity's likes. For example, the above-mentioned internet celebrity like ratio can be compared for similarity through a deep learning algorithm, and the degree of similarity of the internet celebrity can be judged based on the learning ratio. Or, judge by analyzing the interactive performance characteristics of the Internet celebrity. For example, if the fan composition characteristics of the target Internet celebrity and similar Internet celebrities both have about 1 million fans, but the influence and interaction performance of the target Internet celebrity on social media are mainly targeted at fans of the travel group, while another The influence and comments of Internet celebrities on social media are analyzed through natural language processing. It is found that their comments are mainly targeted at fans of political groups. Therefore, the Internet celebrity can be judged by analyzing the interactive performance characteristics of the Internet celebrity. It is not similar to the target influencer.
此外,本創作可利用上述各種演算法轉換各種相似特徵為對應的一絕對數值或相對數值,以便於由數值大小進行後續的相似性評估。於本創作中,絕對數值例如網紅貼文的按讚人數,相對數值例如網紅留言回覆頻率。再者,近似的程度亦可針對各種相似特徵的組合進行分析判斷,並不限於單獨以單一相似特徵進行匹配比較。In addition, this creation can use the above-mentioned various algorithms to convert various similar features into a corresponding absolute value or relative value, so as to facilitate subsequent similarity evaluation based on the numerical value. In this creation, the absolute value is such as the number of likes of the Internet celebrity's post, and the relative value is such as the frequency of replying to the Internet celebrity's message. Furthermore, the degree of similarity can also be analyzed and judged based on the combination of various similar features, and is not limited to matching and comparison with a single similar feature alone.
資料處理器13利用上述的演算法分析比較目標特徵與相似特徵的近似程度,並產生初始網紅排名資料至計算機裝置。計算機裝置14根據複數位用戶產生的一事件訊息產生加權比例。於本創作中,事件訊息包含複數位用戶對初始網紅排名資料中的一相似網紅點擊訊息、一收藏網紅訊息、一訂閱網紅訊息以及一商品交易訊息。於本創作中,計算機裝置14藉由一人工智慧演算法根據相似網紅點擊訊息、收藏網紅訊息、訂閱網紅訊息或者商品交易訊息的多寡等事件訊息產生一投票比重,並根據投票比重學習分析產生加權比例。The data processor 13 uses the above algorithm to analyze and compare the degree of similarity between the target feature and similar features, and generates initial internet celebrity ranking data to the computer device. The
請參閱圖2,其係為本創作初始網紅排名資料示意圖。初始網紅排名資料中,以J網紅為目標網紅,共分析出A網紅至E網紅5位網紅,提供推薦給用戶。然而,對於用戶而言,每個人喜好網紅的原因不一,因此,計算機裝置14藉由人工智慧技術學習分析複數位用戶產生的事件訊息,以調整產生的加權比例。例如,若資料處理器13產生推薦的網紅名單排名,由左至右,第一名為A網紅,依序排列,第五名為E網紅,而當許多用戶認為B網紅是他們心中更喜愛、更相似於J網紅的人選,因此,當用戶點擊B網紅的連結察看時,透過例如網頁追蹤技術,可追蹤統計出相似網紅點擊訊息的數值,並傳送至相似網紅匹配推薦系統1。進一步而言,當使用者在本創作的相似網紅匹配推薦系統1平台的網頁介面上進行點擊以及收藏相似網紅等各種行為而產生事件訊息(event)時,由網站前端、後端發送事件訊息至相似網紅匹配推薦系統1的資料庫中,相似網紅匹配推薦系統1透過演算法自動讀取、統計、追蹤及優化該些事件訊息,以追蹤統計出相似網紅點擊訊息的數值,其中事件訊息係由相似網紅匹配推薦系統1所自行定義,包含使用者資訊、行為資訊、前後脈絡資訊、用戶帳戶資訊、來源以及登入資訊等各種事件訊息。當相似網紅點擊訊息的數值大於預設數量門檻值時,表示B網紅實際上是許多用戶所喜愛的網紅,因此,人工智慧技術調整加權比例,加重計分B網紅的分數,使得B網紅的排名可往上提升,例如提升至前一個名次的位置。相似地,若有許多用戶將C網紅收藏,作為喜愛的網紅名單,相較於點擊訊息,透過統計收藏網紅訊息應可獲得更高的加權比例,人工智慧技術調整網紅的排名名單。相似地,若用戶透過D網紅的產品介紹,因而產生商品交易的行為,產生商品交易訊息,則D網紅在網紅排名名單上應可獲得最高的加權比例,因此,人工智慧技術據此調整D網紅的排名名單。有關加權比例與其它用戶產生的事件訊息對應關係以此類推,與此不在贅述。Please refer to Figure 2, which is a schematic diagram of the initial internet celebrity ranking data for this creation. In the initial Internet celebrity ranking data, with J Internet celebrity as the target Internet celebrity, a total of 5 Internet celebrities from A to E were analyzed and recommended to users. However, for users, everyone has different reasons for liking Internet celebrities. Therefore, the
此外,於本創作的實施例中,計算機裝置14可根據各種事件訊息產生不同的投票比重,並根據投票比重學習分析產生加權比例。以上述的收藏網紅訊息為例,當許多用戶在相似網紅匹配推薦系統1上點擊B網紅,作為喜愛的網紅名單,人工智慧技術可據此學習產生0.4的加權比重。當許多用戶收藏C網紅,作為喜愛的網紅名單,人工智慧技術可據此學習產生0.6的加權比重。當許多用戶透過D網紅的產品介紹因而購買相關商品時,人工智慧技術可據此學習產生0.9的加權比重。相較於上述透過設定預設數量門檻值,並判斷各種事件訊息的數值是否大於預設數量門檻值的情況,透過人工智慧技術可更彈性地對應各種情況學習產生加權比重。In addition, in the embodiment of the present invention, the
此外,為了避免同一位用戶在短時間內重複點擊同一位網紅,造成相似網紅點擊訊息的數值增加,可透過上述的網頁追蹤技術追蹤分析用戶的帳戶、來源以及登入途徑等資訊,藉以排除重複點擊的行為。In addition, in order to prevent the same user from repeatedly clicking on the same influencer within a short period of time, resulting in an increase in the number of similar influencer click messages, the above-mentioned web tracking technology can be used to track and analyze the user's account, source, login method and other information to eliminate Repeated click behavior.
請參閱圖3,其係為本創作相似網紅匹配推薦系統包含資料倉儲裝置及資料搜尋裝置的方塊示意圖。資料倉儲裝置15連接計算機裝置14,藉由一資料倉儲技術建立相似網紅分析結果對應目標網紅及複數位相似網紅的網紅資料索引指標(index)。資料搜尋裝置16連接資料儲存器12,根據索引指標搜尋目標網紅與複數位相似網紅。資料搜尋裝置16可利用搜尋引擎技術搜尋目標網紅與複數位相似網紅,並透過後續的資料顯示器17輸出相似網紅分析結果。例如,根據相似網紅分析結果,某位目標網紅的目標特徵為美國地區、年齡層介於30至40歲的女性,透過索引指標進行搜尋時,可針對目標特徵與相似特徵的關係更快速地搜尋出推薦的網紅,進一步提升搜尋效率,並推薦給喜愛的用戶。Please refer to Figure 3, which is a block diagram of the similar internet celebrity matching recommendation system including a data storage device and a data search device. The
請參閱圖4,圖4係為本創作相似網紅匹配推薦系統包含資料顯示器的方塊示意圖。資料顯示器17透過應用程式介面(API)連接計算機裝置14,並輸出如圖4所示的相似網紅分析結果。使用者可透過智慧型裝置接收相似網紅分析結果,或是連接到相似網紅匹配推薦系統1的產品前台,以接收相似網紅分析結果。Please refer to Figure 4. Figure 4 is a schematic block diagram of the similar internet celebrity matching recommendation system including a data display. The data display 17 is connected to the
需注意的是,於本創作實施例中,相似網紅匹配推薦系統1中的各個裝置及元件實際上係對應於各種硬體,例如資料探勘裝置11係為一資料搜尋引擎的硬體裝置,資料儲存器12係為一虛擬伺服器、一資料暫存器或者一揮發性記憶體的硬體,資料處理器13係為一中央處理器或一雲端伺服器的硬體,計算機裝置14係為一資料計算處理的硬體,資料倉儲裝置15係為一雲端伺服器、一非揮發性記憶體或一硬碟的硬體,資料搜尋裝置16係為一資料處理的裝置或者一資料搜尋的裝置,資料顯示器17係為一顯示器的硬體,於本創作中並不限定。It should be noted that in this creative embodiment, each device and component in the similar internet celebrity
綜上所述,本創作提供基於多維度特徵的相似網紅匹配推薦系統,從多方面評估目標網紅與相似網紅之間的相似性,提供用戶可更方便、快速地找到具有相似特徵的網紅,同時網紅和廣告主也可根據推薦結果及排序進行精準推廣和合作,將廣告訊息發送給在相同社群平台中的用戶。此外,本創作利用深度學習、自然語言處理、影像辨識、爬蟲等演算法自動擷取各種特徵,以產生分析的結果,並透過機器學習演算法根據分析結果中的文字及圖像學習計算調整加權分數,以產生推薦相似的網紅,以及透過用戶回應的各種事件訊息驗證本創作推薦的相似網紅與用戶感興趣、期待、喜愛的網紅兩者之間是否互相匹配,以提升本創作推薦相似網紅的準確度。To sum up, this creation provides a similar internet celebrity matching recommendation system based on multi-dimensional features, which evaluates the similarity between the target internet celebrity and similar internet celebrities from many aspects, allowing users to more conveniently and quickly find those with similar characteristics. Internet celebrities, Internet celebrities and advertisers can also carry out precise promotion and cooperation based on the recommendation results and ranking, and send advertising messages to users on the same social platform. In addition, this creation uses algorithms such as deep learning, natural language processing, image recognition, and crawlers to automatically extract various features to generate analysis results, and uses machine learning algorithms to calculate and adjust weights based on text and images in the analysis results. scores to generate recommended similar internet celebrities, and to verify whether similar internet celebrities recommended by this creation match the internet celebrities recommended by this creation and the internet celebrities that the user is interested in, expects and likes through various event messages responded by the user, so as to improve the recommendation of this creation Accuracy of similar influencers.
此外,本創作在相似性評估階段,相似網紅匹配推薦系統利用各種演算法計算網紅之間的相似性指標,並根據指標對網紅進行排序。此外,相似網紅匹配推薦系統還可以根據用戶的偏好和歷史行為進行個性化推薦。最後,在推薦階段,相似網紅匹配推薦系統將根據相似性指標和用戶偏好,為用戶推薦一系列相似的網紅,用戶可以根據推薦結果瀏覽和關注相似的網紅,以便更好地探索網紅生態。In addition, in the similarity evaluation stage of this creation, the similar Internet celebrity matching recommendation system uses various algorithms to calculate the similarity indicators between Internet celebrities and ranks the Internet celebrities according to the indicators. In addition, the similar influencer matching recommendation system can also make personalized recommendations based on the user's preferences and historical behavior. Finally, in the recommendation stage, the similar Internet celebrity matching recommendation system will recommend a series of similar Internet celebrities to users based on similarity indicators and user preferences. Users can browse and follow similar Internet celebrities based on the recommendation results to better explore the Internet. Red ecology.
1:相似網紅匹配推薦系統 11:資料探勘裝置 12:資料儲存器 13:資料處理器 14:計算機裝置 15:資料倉儲裝置 16:資料搜尋裝置 17:資料顯示器 F:社群平台1: Similar internet celebrity matching recommendation system 11: Data exploration device 12:Data storage 13:Data processor 14:Computer device 15:Data storage device 16:Data search device 17:Data display F:Social platform
圖1係為本創作相似網紅匹配推薦系統的方塊示意圖; 圖2係為本創作初始網紅排名資料示意圖; 圖3係為本創作相似網紅匹配推薦系統包含資料倉儲裝置及資料搜尋裝置的方塊示意圖;以及 圖4係為本創作相似網紅匹配推薦系統包含資料顯示器的方塊示意圖。 Figure 1 is a block diagram of the similar internet celebrity matching recommendation system created in this work; Figure 2 is a schematic diagram of the initial internet celebrity ranking data for this creation; Figure 3 is a block diagram of the similar internet celebrity matching recommendation system including a data storage device and a data search device; and Figure 4 is a schematic block diagram of the similar internet celebrity matching recommendation system including a data display.
1:相似網紅匹配推薦系統 1: Similar internet celebrity matching recommendation system
11:資料探勘裝置 11: Data exploration device
12:資料儲存器 12:Data storage
13:資料處理器 13:Data processor
14:計算機裝置 14:Computer device
F:社群平台 F:Social platform
Claims (15)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW112208962U TWM649741U (en) | 2023-08-23 | 2023-08-23 | Similar internet celebrity matching recommendation system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW112208962U TWM649741U (en) | 2023-08-23 | 2023-08-23 | Similar internet celebrity matching recommendation system |
Publications (1)
Publication Number | Publication Date |
---|---|
TWM649741U true TWM649741U (en) | 2023-12-21 |
Family
ID=90041102
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
TW112208962U TWM649741U (en) | 2023-08-23 | 2023-08-23 | Similar internet celebrity matching recommendation system |
Country Status (1)
Country | Link |
---|---|
TW (1) | TWM649741U (en) |
-
2023
- 2023-08-23 TW TW112208962U patent/TWM649741U/en unknown
Similar Documents
Publication | Publication Date | Title |
---|---|---|
KR102075833B1 (en) | Curation method and system for recommending of art contents | |
Cheng et al. | Personalized click prediction in sponsored search | |
CN109690529B (en) | Compiling documents into a timeline by event | |
CN113158023B (en) | Public digital life accurate classification service method based on mixed recommendation algorithm | |
US10360623B2 (en) | Visually generated consumer product presentation | |
CN110942337A (en) | Accurate tourism marketing method based on internet big data | |
US20120066073A1 (en) | User interest analysis systems and methods | |
Zhang et al. | Multimodal marketing intent analysis for effective targeted advertising | |
KR20140026932A (en) | System and method providing a suited shopping information by analyzing the propensity of an user | |
US20150193540A1 (en) | Content ranking based on user features in content | |
CN111651678B (en) | Personalized recommendation method based on knowledge graph | |
JP6767342B2 (en) | Search device, search method and search program | |
WO2017107010A1 (en) | Information analysis system and method based on event regression test | |
US20180239790A1 (en) | Provision device, provision method and non-transitory computer readable storage medium | |
Kumar | Fake review detection using behavioral and contextual features | |
CN114841760B (en) | Advertisement recommendation management method and system based on audience behavior characteristic analysis | |
Li et al. | A personalized recommendation algorithm for college books based on user interest | |
Özyirmidokuz et al. | Analyzing customer complaints: a web text mining application | |
Omura et al. | Ad Recommendation utilizing user behavior in the physical space to represent their latent interest | |
TWM649741U (en) | Similar internet celebrity matching recommendation system | |
CN116521937A (en) | Video form generation method, device, equipment, storage medium and program product | |
Iwański et al. | Application of the Information Bottleneck method to discover user profiles in a Web store | |
CN114201680A (en) | Method for recommending marketing product content to user | |
Zhu et al. | Identifying and modeling the dynamic evolution of niche preferences | |
CN114564653A (en) | Information recommendation method and device, server and storage medium |