Browse Prior Art Database

Predictive Analytics Based Online Forum Moderator

IP.com Disclosure Number: IPCOM000225786D
Publication Date: 2013-Mar-05
Document File: 7 page(s) / 121K

Publishing Venue

The IP.com Prior Art Database

Abstract

Described is an invention that seeks to overcome the problem of non-pertinent comment injection by providing a computer system that measures relevancy of comments and rejects those that are not relevant to the topic itself or to one of its responses. The system also ensures that the original topic itself adheres to forum policies as well.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 25% of the total text.

Page 01 of 7

Predictive Analytics Based Online Forum Moderator

Topics published on any online forum can easily get off topic and lose direction without a human moderator intervention because of non-relevant comment injection. This invention proposes a system for automatically controlling comments on forum posting, comprising the steps of:

Analyzing the textual content of the initial post and generating a set of relevant concepts from it


Generating a predictive model comprising of weighted concept tree(s) from the relevant

concepts and a set of ad hoc rules for the purpose of determining adherence to the original topic


Analyzing the textual content of an arriving comment that is a candidate to be added to the

thread as a response and generating a set of relevant concept tree(s) from it


Measuring relevancy of each arriving comment by applying a predictive model which

compares weighted concept tree(s) from both the initial posts and arriving comment


Accepting or rejecting the arriving comment based on a set threshold of the relevancy score


Updating the aforementioned predictive model and associated rules for the forum thread for

each accepted comment to accommodate for new key concepts being discussed in the accepted comment

Various embodiments of online forums are suitable for this invention, such as:

Online news organizations allowing reader/subscriber responses to its published articles. [1]


Domain specific forums allowing subscriber query and responses on various topics of

interest. [2]

Relevant Concept Extraction

    The initial post that creates the topic and each subsequent responses to the initial post is processed by a "relevant concept extractor" module (see Figure 1 below). This module is configured to identify the specific domain where an online forum is going to be employed. The configuration step connects the system to various corpus of data dictionaries that might be available as part of the system or that may be harvested from the worldwide web. An example of such online dictionary that the system utilizes is Open Dictionary Project. [3] The system is operational only after its domain identity has been established. The system not only prevents extraction of stop words, but it prevents concepts extraction that are not within the realm of the identified domain. This module enforces that an online forum primarily designated for publishing topics related to Sports cannot be subverted to publish a topic about automobiles, unless that topic somehow also relates to the broader category of Sports. The Relevant Concept Extraction module provides an ordered list of concepts. The list is ordered based on the appearances of each concept and may contain duplicates if the same concept appears at various locations in the topic.

1


Page 02 of 7

Figure 1

Validation Module

    The present invention handles multiple disjointed chain of thoughts (unrelated concepts or themes) that may be present within the same topic (or within the same response to a topic)...