Technology Opportunities

Available Knowhow

Including content features minimizes prediction error

Highlights

• Current social media research focuses on temporal trends of information flow and on the topology of the social graph that facilitates the propagation of information.
• The content of the Tweets or ideas has been used to trace the evolution of an idea as it propagates, but not to predict the spread of ideas.

Our Innovation

A hybrid approach that combines the content, the context, global temporal features and graph topology gives the best results for predicting the spread of ideas in social media, while maintaining computational efficiency.

Table comparing the acceptance prediction of basic models and the hybrid model:

MSE (mean square error) is shown over various numbers of weeks

Even though the content attributes of the hashtag (HT) and the tweet context (TW) show only marginal improvement over the baseline, combining both content types presents a much smaller error

Key Features

• This is the first group to analyse the content of the hashtag itself.
• Analysis of the hashtag and the Tweet content is carried out according to dimensions such as sentiment, psychological effect on reader, part of speech, as well as how hard it is to process (too long/too short, cryptic).
• Research used the spread of a hashtag which is an indication of the acceptance/spread of a new idea The Tweet content was used as part of the hybrid model.
• Avoids costly graph-based algorithms
• Can be used as a general guideline for marketers who want to introduce a new hashtag.
• Analysing the hashtag itself allows early detection of potential upcoming trends

The Opportunity

• Better prediction of how ideas spread enables improved marketing of goods or ideas.