Classification Approaches for Social Media Text

Course organization

Instructor:	Tatjana Scheffler
Time:	fridays, 12-14:00 summer term 2017
Place:	Golm, Haus 14, Raum 009
Modules:	AM3b (B.Sc.) AM11, AM12 (M.Sc.)
Moodle:	Please register to the class's Moodle page.

Requirements

readings
active participation
presentation of either a classification algorithm or a task
Grade: Final Project -- Either a programming project implementing a classification approach based on some social media data, or a theoretical/concept paper discussion different approaches to the same problem, giving a literature overview, etc.

Course description

In this class, we study classification and clustering approaches for social media. Social media data typically comes with text but also other metadata (user information, geo-tags, network structure, etc.) which can be exploited for classification.

We will present different classification algorithms including:

• naive Bayes
• support vector machines (SVM)
• k-nearest-neighbor
• decision trees
• neural network approaches

We will work with concrete implementations of these algorithms to try them out on Twitter data. The class will include a practical component where we collaboratively build a classifier, probably for detecting hate speech.

We will discuss tasks in social media classification such as:

• sentiment
• spam
• impartiality
• hate speech
• bot or not
• on/off topic
• user characteristics: gender, age, occupation, social class
• ... and others

For practical matters, we will also introduce toolkits such as WEKA or RapidMiner if requested by participants, but will mainly use Python and scikit-learn

Datasets

Available datasets for final projects.

Readings

Listed below with each topic. For a detailed overview on classification approaches in general, see:

• Aggarwal, C. C. and Zhai, C. (2012). A survey of text classification algorithms. In Aggarwal, C. C. and Zhai, C. (Eds.), Mining text data, pp. 163–222. Springer.

Schedule