Seminar in Intelligent Information Systems 

Social Media Mining
367.112 (2SE), SS 2017

Overall Seminar Topic:

Seminar topics will fall into the overall area of Social Media Mining, with an application focus to crisis and emergency management - see our list of CrowdSA topics for getting an idea on potential topics. The main focus of this semester's course will be on tools and techniques for detecting real-world events and extracting and analyzing crisis-relevant content (text, images, videos) from social media platforms (e.g., Twitter, Instagram, Facebook).

Areas: Natural Language Processing, (Spatio-temporal) Data Mining, Text Mining, Topic Detection & Tracking, Stream Processing, Community Analysis

Note: Upon interest, the selected thesis topic can be extended to an implementation project for the Project in Intelligent Information Systems - please notify lecturer if interested!

Goals:

In the course of this seminar, students should acquire a solid knowledge on the scientific state of the art on the chosen topic and summarize their findings in the form of a seminar thesis (15-20 pages in length), by

  • searching and studying scientific papers, providing the basis for
  • elaborating a general overview on the chosen topic (definition, state of the art, algorithms, challenges, etc.)
  • where applicable: presenting and evaluating systems in the scope of the chosen topic, including both commercial tools as well as research prototypes, if available; give a demonstration of these systems in your presentation; if no systems are available for your topic, provide a demonstration of algorithms or similar (hands-on part!)

Aim: Get a thorough overview and understanding of the selected topic.

Students will share their findings in a 20 min. presentation, held towards the end of the semester.

Topics:

Students can select a topic of interest from a topic list presented in the first course meeting.

Some examples from previous semesters include:

  • humor & irony detection in tweets
  • predictive modeling with social media data
  • crowd-sourcing social media analysis
  • who is tweeting? user account classification
  • generating natural language descriptions for images
  • multi-lingual information extraction
  • stance detection (detecting the writer's opinion)
  • natural language understanding, FrameNet, semantic role labeling
  • geo-information extraction
  • ontology learning

Organizational:

Schedule of the course and available seminar topics will be discussed in the first course meeting (presence obligatory, date will be announced in KUSSS and via e-mail).

Outline:

Dates will be fixed in the first introductory meeting, please check back for updates!

  • Introductory Meeting: 8th of March, 2017, 10:00 - 11:00
    • organizational issues
    • short introduction into the general topic
    • seminar guidelines
    • presentation and assignment of topics
  • 1st (individual) meeting: THU, 6th of April, 2017, between 10:00 - 15:00
    • individual progress discussion with lecturers, feedback by lecturers
  • 2nd (individual) meeting: THU, 11th of May, between 10:00 - 15:00
    • individual progress discussion with lecturers, feedback by lecturers
  • Presentation Day: MON, 12th of June, 2017, 10:00 - 16:00
    • presentation of findings to colleagues & lecturers

Grading:

Seminar work (70%), presentation (20%) and overall engagement (10%) (meetings, discussion)