Surety is performing system maintenance this weekend. Electronic date stamps on new Prior Art Database disclosures may be delayed.
Browse Prior Art Database

Sentiment Analysis for Social Media Screening

IP.com Disclosure Number: IPCOM000236682D
Publication Date: 2014-May-08
Document File: 4 page(s) / 91K

Publishing Venue

The IP.com Prior Art Database


Disclosed is a method for using composite sentiment scoring to prevent inappropriate messages from being sent from corporate (or personal) social media accounts.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 49% of the total text.

Page 01 of 4

Sentiment Analysis for Social Media Screening

According to Twitter* and Facebook**, there are over 600 million social media posts every day. The sheer scale of the data has required companies to re-evaluate big data, and redefine approaches to communicating with customers.

Prior to 2012, most social media solutions were low scale, low investment approaches. The primary account management tools have proven to be severely lacking in scale and management capabilities. Such tools are based on managing a few accounts, tracking general sentiment, and scheduling communications. The solutions have failed to address true corporate enterprise needs, and lack the scale to be a comprehensive answer to large complex companies.

The next generation of solutions requires a full solution stack, well-defined business processes, and the software needed to handle big data on an enterprise level. A more balanced approach is needed, which that enables large volumes of outbound social media, while still controlling the messaging, and avoiding negative outbound social media. This balanced approach requires intelligent rules built into the process (both system and human) that can ensure business outcomes based interactions derived from social actions, based on sentiment drivers and tactical metrics that support the objectives.

Many major companies have faced major backlash from millions of customers due to inappropriate tweet content that has gone out. A system is needed that enables bulk tweeting from many areas of the company, but with an appropriate enterprise level filtering mechanism that effectively blocks the kinds of tweets that can cause significant public relations backlash and damage to many large brands.

The solution described herein addresses the need for enterprise level software and systems that can be used to prevent and moderate negative outbound social media communications. The invention uses a unique combination of tools to screen for inappropriate messaging. The tools presented here are used to 'catch' potentially damaging messages prior to public posting.

The first approach uses libraries of sentiment analysis to create composite scores for outbound messages. Each message to be posted is run through a rigorous scoring mechanism to determine a composite 'negativity weighting' shown below is a subset of the filters that are used.

Figure 1: Filters used for negativity weighting


Page 02 of 4

Finding negative sentiment from textual analysis is not a point of novelty. The novel contribution is a method for using composite sentiment scoring to prevent inappropriate messages from being sent from corporate (or personal) social media accounts. Using this scoring with the assistance of business outcome based rules that have varying degrees of categorizing outbound messages (e.g., "valid", "requires manual review" and "block") can allow organizations to leverage the power of mass communication whilst ensuring full alignment with the company's...