Browse Prior Art Database

Method and System for Modelling Content Complexity, Consistency and Effect on User Engagement

IP.com Disclosure Number: IPCOM000246822D
Publication Date: 2016-Jul-04
Document File: 5 page(s) / 369K

Publishing Venue

The IP.com Prior Art Database

Related People

Suleyman Cetintas: INVENTOR [+2]

Abstract

A method and system is disclosed for modelling content complexity, similarity, and consistency of social media content and effect of the content on user engagement. In order to model the content complexity and consistency, the method and system utilizes state-of-the-art deep learning approaches such as, but not limited to, deep convolutional neural networks (DCNN) to analyze visual content and deep neural language models such as word2vec in order to study textual content.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 33% of the total text.

Method and System for Modelling Content Complexity, Consistency and Effect on User Engagement

Abstract

A method and system is disclosed for modelling content complexity, similarity, and consistency of social media content and effect of the content on user engagement.  In order to model the content complexity and consistency, the method and system utilizes state-of-the-art deep learning approaches such as, but not limited to, deep convolutional neural networks (DCNN) to analyze visual content and deep neural language models such as word2vec in order to study textual content.

Description

Disclosed is a method and system for modelling content complexity, similarity, and consistency of social media content and effect of the content on user engagement.  Specifically, the method and system focuses on modelling social media content using data from company-generated posts in blogs.  A blog post consists of both visual content such as images and textual content such as text and tags.  In order to model the content complexity and consistency, the method and system utilizes state-of-the-art deep learning approaches such as, but not limited to, deep convolutional neural networks (DCNN) to analyze the visual content and deep neural language models such as word2vec in order to study the textual content.  The utilization of these deep learning approaches for content analysis is further described in detail in accordance with various embodiments of the invention.

The method and system, thus, measures a semantic content complexity and generates a measure between the visual and textual content to assess the coherence of a given post and also generates a consistency measure between a post of a blog and the blog itself to assess the coherence of a blog.  Further, the method and system analyzes effect of the visual and textual content on user engagement.

In accordance with an embodiment, firstly the method and system models visual feature complexity of visual content from posts using a normalized compressed file size of images as the feature complexity, since images of company-generated posts are generated in various resolutions and formats with different compression behaviors.  In order to obtain a consistent measure of feature complexity, the method and system normalizes the compressed file size of an image by its resolution (number of pixels) and compression quality.

Further, in order to extract useful and meaningful semantic features from images, the method and system employs the deep CNN model that typically consists of multiple layers, where each layer transforms the representation from the previous layer starting with the raw input image into a more abstract representation.  The objective of the model is to accurately classify objects that appear in the image through the composition of such multiple transformation.

Specifically, a typical convolutional neural network (CNN) consists of multiple convolutional layers followed by a fully connected layer...