The goal was to to create a realistic bot that acted like a human. Through network theory, the bot should enter a community and get influence in Boston.
In order to do this, the persona of the bot was chosen to be a journalist. Earlier bot-testing showed that it was more likely that people would follow a female user prior to a male user.
The goal for the bot was to attract attention in Boston through 4 interventions. This turned out to be successful in the last of the interventions.
The networkX package for python can build a network graphs containing the bot’s followers and friends. Further using Gephi a visualization and filter can be applied to the graph to analyze clusters and communities.
The Natural Language ToolKit is able to tokenize text and process it with different analysis tools. Those tools include calculating the most used words, lexical diversity and sentiment value of a text.
The two machine learning methods; topic- and bot-detection are considered for this Boston bot.
Topic detection is used to discover what the followers are talking about and find topics that are upcoming in Boston. This can be used for tweeting.
The bot detection part was not implemented with usual machine learning classifiers as it was in assignment 3, since it was not necessary. The use of such classifier was not implemented since other strategies was taken into consideration. This will be reflected in the statistics for the bot.
Read the article about it here:
- Client DTU: Course 02805 - Social Graphs and Interactions
- Date januar 5, 2014
- Tags Data Mining, Machine Learning