This workshop will guide participants through the first steps for doing data analysis, specifically text mining with Weka. Weka is an open source machine-learning tool. We will be replicating the work of Mike Thelwall in his paper on Gender bias in machine learning for sentiment analysis (https://wlv.openrepository.com/handle/2436/620690)
Before getting into the hands-on text mining exercise, we will present a brief introduction to AI and machine learning, as well as the notion of algorithmic bias; what it is, how is introduced and it’s repercussions.
By the end of the workshop participants will have applied a sentiment analysis technique to a gender segregated data set and be able to determine its effect on the resulting predictive model.
IMPORTANT NOTE
Before the workshop, students are strongly encouraged to install Weka and download the data.
Understand basic notions of AI: machine learning and predictive models.
Understand what is algorithmic bias, how is introduced in machine learning models and its possible repercussions.
Transform text into word vectors (Bag of Words Approach) as a technique to perform text-mining tasks.
Create a model for sentiment prediction using a machine learning approach based on a training corpus of real-life textual data (Tripadvisor comments on hotels and restaurants).
Evaluate the model and compare the performance with different gender biased training corpuses.