System and method for measuring and predicting content dissemination in social networks

Number: 10846613
Published: 2020-11-24
Filed: 2016-12-29
Assignee: DISNEY ENTERPRISES, INC.
Inventors: Ray; Abhik, Branch; Joel, Williams; James, Topol; Zvi, Brutch; Tasneem
CPC: G06N20/20; G06N5/022; G06N20/00; G06Q10/40
Verdict: Set aside social media analytics, business/marketing ML
Source: Google Patents · FreePatentsOnline

Abstract

Methods and systems for measuring and predicting content dissemination in social networks includes computing a “virality score” for popularity of social media content, a “pattern” of diffusion of the content, and a “hype” parameter of such content without requiring a “friendship graph” or “information diffusion” structured data. It also includes an iterative, predictive model, which predicts future performance (or future rate of dissemination) of the content while mitigating a class imbalance problem inherent in predicting viral posts, and which provides updates to the model based on actual performance results.

Background

BACKGROUND(1) It is common for media companies and other entities to use social networks, e.g., such as Facebook®, Twitter®, Google+® (or Google Plus), Snapchat®, and the like, to publish (or “post”) content that is of interest to their viewing or listening audience (or the public in general). To achieve this objective, the media companies desire to measure or predict the amount (or degree) of content dissemination reflective of audience engagement for such published content.(2) One challenge is that predictive techniques that rely on structural features typically require access to “friendship graphs” of social network members, which shows, e.g., for a given member, which other members are connected to the given member. Friendship graphs are useful because the structure of such graphs can provide visibility into different patterns and rates of content dissemination among members. However, social networks do not typically expose these data for privacy reasons making it very difficult for media companies and third parties (vendors, and the like) to create and evaluate metrics on how content is disseminating.(3) Accordingly, it would be desirable to have a method and system that can measure and predict the dissemination of content in social networks.

Claims

1. A method for determining, continuously in real-time, without using a social network structure, at least one metric indicative of dissemination of social media content, comprising: receiving shares data over time, forming a shares time series using only shares data, the shares time series indicative of how the content is being shared on social media; fitting a plurality of curve fit models to the shares time series; identifying a best curve fit model from the plurality of curve fit models that best fits the shares time series; calculating a virality score metric, indicative of the rate of dissemination of the social media content, using the best curve fit model; determining a dissemination type comprising one of viral dissemination and non-viral dissemination, using the best curve fit model; wherein the plurality of curve fit models comprises at least one parametric curve with an increasing gradient, followed by a decreasing gradient, and an inflection point there-between; calculating a hype lifetime (Th) using the inflection point, as the time at which the inflection point occurs; calculating a hype lifetime percentage as a percentage of ex-post time (Txp) that the shares spent at the inflection point by the equation: (T.sub.h/ T.sub.XP)*100; and wherein the receiving, fitting, identifying, and calculating are performed continuously in real-time. 15. A method of predicting, continuously in real-time, future dissemination of social media content, comprising: Performing, continuously in real-time, an ensemble-based classification, using feature data obtained at a wait time (Tw) after the content was posted to social media, the feature data comprising at least one of: a virality score, a hype, a diffusion pattern, diversity/generality of topics, and a similarity of content with most recent few posts contents; the classification iteratively increasing a share growth factor to determine a shares growth range at a future prediction time (Tp), which is indicative of the future dissemination of the content; and wherein the increasing the share growth factor comprises increasing the share growth factor by a predetermined unique multiplier value for each iteration pass until a predicted result changes and wherein the ensemble-based classification is performed by an ensemble of predictive classifier models, each classifier model having the unique multiplier associated therewith.