HueHueBot

Huehuebot is a bot studying psychology which, also, loves food and nature. The idea of the bot started as a project for “The Computational Foundations of Linguistic Creativity” course at University of Helsinki. The aim of the course was to use provided n-grams and color maps
by Dr. Tony Veale to creatively name colors. The creativity I chose for my bot was giving him the ability to see and relate colors to food and natural as humans do. The bot selectively decides when and what to tweet. Additionally, the bot used to predict the personality of twitter users based on the colors they re-tweeted from @everycolorbot by using the personalities associated with colors. After the course finished, I was pleasantly surprised by the re-tweets, followers and favorites my bot gained. Hence, I decided to evolve my twitter bot and improve its personality prediction by analyzing user’s profiles. The creativity in this amendment, I believe, is that humans would find it interesting if their personalities were predicted correctly by a machine as close to how a psychologist would, hence deem it creative.

What the Bot analyzes:

Profile

Analyzing a profile consisted of determining the number of friends, followers, and tweets, whether the account is verified, the color of the user’s profile in case if was not the default color, and analyzing the description of the profile. The usage of friends, followers, tweets counts was to predict how lovely is the user to others. As the color-to-personality functionality was already implemented, analyzing the colors selected in a profile will increase the accuracy of predicting the personality. Analyzing the user’s description, the most important part of profile analysis, was employed to discover how the user describes himself/herself, and the profession of user. This was achieved by finding the adjectives and nouns used in the description of the user. Only positive adjectives will be mapped and used. Determining the profession of a user starts by finding the category of all the nouns by using Thesaurus Rex. If a noun was in the category of “people”, it will be considered as a profession and the attributes describing it will be stored. Thereafter, the bot looks up the actions that can be performed by the profession using Metaphor Eyes, and finds famous people who have the same profession using Freebase. Lastly, the bot uses prfekt’s classifications to predict the Briggs Myers‘ personality type of the user, and CelebrityTypes to find celebrities sharing the same personality type. By the end of this analysis, the bot will have knowledge regarding what the user is and does, his/her personality type, and what well-known person is similar to the user.

Tweets

In this section, the bot analyzes only English tweets’ text (at the moment). The analysis relies heavily on the Pattern  resource which was used for all text analysis and web parsing in the bot. The bot searches for adjectives, hashtags, and mentions used in tweets. It, also, finds the tweets’ moods (facts, commands, opinions, or conjectures) and positiveness. The information obtained will be utilized in addition to profile’s information to result in better personality predictions. Sardonicus is employed in this process to find nouns which share the most adjectives describing the user. Also, the bot attempts to discover what is the most liked thing by the user and its category by finding objects in sentences and calling Freebase application programming interface to obtain its category. The most used object in the most repeated category is considered the most liked object by the user.

Generating Tweets

In this critical and challenging process, the information is ordered by its accuracy, how believable and creative it would be considered from the user’s perspective. Therefore, the first information would be used is the profession of the user if it was obtained as it directly describes the user, followed by the personality mapped to the profile’s color. Then, the predicted personality type. In case if a liked object is found, it will be appended to the original generated tweet.

---

Computational Creativity

Since my childhood, creativity and computers have always attracted me. Therefore, my attention was stolen by computational creativity the second I knew about it. It is fascinating to see computers generate creative outputs without taking any decisions from humans.

Useful Resources

• http://robotcomix.com/
• http://prosecco-network.eu/webservices
• https://www.cs.helsinki.fi/node/78271
• http://www.clips.ua.ac.be/resources
• http://afflatus.ucd.ie/article.do?action=list&categoryId=4