Method of predicting social article influence and device using the same

Patent metadata

Field	Value
Publication number	US-2017140291-A1
Application number	US-201514966563-A
Country	US
Kind code	A1
Filing date	Dec 11, 2015
Priority date	Nov 18, 2015
Publication date	May 18, 2017
Grant date	—

Abstract

A method of predicting a social article influence includes following steps: a) analyzing an issue on the social article to classify the social article into a content domain; b) calculating a term weight to obtain a basic influence; c) collecting an author's at least one author's historical article, calculating a first influence average value, and subtracting the first influence average value and a first reference influence average value to obtain an author's general influence correction amount; d) selecting the author's at least one author-domain historical article in the content domain, calculating a second influence average value, and subtracting the second influence average value and a second reference influence average value to obtain an author's domain influence correction amount; and e) calculating the prediction influence, the author's general influence correction amount, and the author's domain influence correction amount.

First claim

What is claimed is:

1 . A method of predicting a social article influence which is adapted to predict a prediction influence of the social article, wherein the prediction influence is an estimate value of an interaction amount obtained from the social article, comprising following steps:
a) analyzing an issue on the social article to classify the social article into a content domain that the social article belongs; b) calculating a term weight of a plurality of terms in the social article to obtain a basic influence according to the term weight, wherein the basic influence is an initial estimate value of the interaction amount obtained from the social article being published after a first predetermined time; c) collecting an author's at least one author's historical article in the social article, calculating a first influence average value according to the interaction amount of the author's historical article, and subtracting the first influence average value and a first reference influence average value to obtain an author's general influence correction amount, wherein the first reference influence average value is an average value of the interaction amount of all of the historical articles; d) selecting the author's at least one author-domain historical article in the content domain, calculating a second influence average value according to the interaction amount of the author-domain historical article, and subtracting the second influence average value and a second reference influence average value to obtain an author's domain influence correction amount, wherein the second reference influence average value is an average value of the interaction amount of all of the historical articles in the content domain; and e) calculating the prediction influence according to the basic influence, the author's general influence correction amount, and the author's domain influence correction amount.

2 . The method of predicting the social article influence as claimed in claim 1 , wherein the step a) further comprises following steps:
a-1) analyzing the terms of the social article to determine the terms belonging to an emotional term or a negative term, wherein the emotional term is divided into a positive emotional term and a negative emotional term; a-2) sorting the emotional term or the negative term in the social article to determine an emotional polarity of the social article; and a-3) determining an emotional strength according to appearance proportions of the positive emotional term or the negative emotional term in the social article.

3 . The method of predicting the social article influence as claimed in claim 1 , wherein the step b) calculates the term weight of the terms according to a method of term frequency-inverse document frequency to arrange the term weight of the terms, so as to obtain a plurality of representative terms of the social article.

4 . The method of predicting the social article influence as claimed in claim 3 , wherein the step b) further comprises following steps:
b-1) obtaining at least one reference historical article having a representative term which is at least one of the representative terms of the social article within a second predetermined time before publishing the social article; b-2) calculating an average influence of the representative terms of the social article before publishing the social article according to the interaction amount of the reference historical article; and b-3) obtaining the basic influence of the social article by calculating a weighted average of the term weight of the representative terms and the corresponding average influence.

5 . The method of predicting the social article influence as claimed in claim 1 , further comprising following steps:
f) determining at least one news events that the social article belongs according to the term weight of the terms, and calculating an event influence correction amount according to the news events, wherein the news events includes at least one news item which has an identical feature with the social article; and g) calculating the prediction influence according to the basic influence, the author's general influence correction amount, the author's domain influence correction amount, and the event influence correction amount.

6 . The method of predicting the social article influence as claimed in claim 5 , wherein the step f) further comprises following steps:
f-1) respectively calculating the term weight of the plurality of terms in the at least one news item; f-2) determining an article similarity between the at least one news item according to the term weight; f-3) comparing a published time between the at least one news item to calculate a published interval between the at least one news item; and f-4) sorting the article similarity being larger than or equal to a minimum similarity and the published interval being lower than a minimum interval in the at least one news item into a news event.

7 . The method of predicting the social article influence as claimed in claim 6 , wherein the step f) further comprises following steps:
f-5) obtaining a centroid vector of the news event by summing and averaging the term weight of all of the news items in the news event; f-6) calculating an event similarity between the social article and the news event according to the term weight of the social article and the centroid vector of the news event, obtaining a time interval by comparing a difference of a published time between the social article being published and the news item being latest published in the news event, and selecting the news event related to the social article according to the event similarity and the time interval; f-7) obtaining at least one historical article related to the news event within a third predetermined time before publishing the social article, and calculating a third influence average value of the news event according to the interaction amount of the related historical article; f-8) subtracting the third influence average value of the news event and a third reference influence average value to obtain an influence correction amount of the news event, wherein the third reference influence average value is an average value of the interaction amount of all of the historical articles within the third predetermined time before publishing the social article; and f-9) obtaining the event influence correction amount of the social article by calculating a weighted average of the event similarity and the influence correction amount.

8 . The method of predicting the social article influence as claimed in claim 5 , wherein the step g) further comprises following steps:
g-1) when the author's general influence correction amount being larger than the author's domain influence correction amount, summing the basic influence, the author's general influence correction amount, and the event influence correction amount into as the prediction influence; and g-2) when the author's general influence correction amount being smaller than the author's domain influence correction amount, summing the basic influence, the author's domain influence correction amount, and the event influence correction amount into as the prediction influence.

9 . The method of predicting the social article influence as claimed in claim 5 , wherein the step g) further comprises following steps:
g-1′) building a regression model, composing the author's general influence correction amount, the author's domain influence correction amount, and the event influence correction amount into an independent variable value, and predicting a deviation between the basic influence and an actual influence to obtain the prediction influence.

Assignees

• Inst Information Ind

Inventors

• Wu Chia-Chi
• Lin Ting-Chia
• Lin Grace
• Chan Ya-Hui

Classifications

• G06Q10/40

  Business processes related to social networking or social networking services · CPC title

• G06N7/01 Primary

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N7/005 Primary

  Physics · mapped topic

• H04L67/22

  Electricity · mapped topic

• G06F16/2462

  Approximate or statistical queries · CPC title