direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Sie sind hier

TU Berlin

Page Content

There is no English translation for this web page.

Studienprojekt Quality & Usability

Project: (Möller; 4/6 SWS, 6/9 LP; jeweils im SoSe/WiSe)

LV-Number: 6LP Projekt: 0434 L 918 / 9LP Projekt: 0434 L 919

Language: Deutsch/Englisch

Contact person: Neslihan Büyükdemircioglu

Content

Topics related to current research focus of QU-Lab are offered to teach the necessary basics and pracitcal applications. Please see the topics below.

Introductory Session

19.10.2018, Fr. 10-13, EB301

Here are the slides from intro session.

STUDENT ASSIGNMENTS:

Assignment list (Study Project Quality and Usability) 

Assignment list (Interdisziplinäres Medienprojekt) 

Registration

1) Read the project descriptions and check out the slots for project meetings under "Topics".

2) On 19.10.: Attend the introductory session for all of the projects (mandatory). In this session, all of the topics will be presented by the supervisors and you will get a link to a google form which you need to fill to apply for topics.

* Note that the introductory session's date and time is not the slot of the next project meetings. 

3) From 19.10 until 23.10.: Fill in the google form that we will provide in the introductory session on 19.10 with your preference for a project: first choice, second choice and third choice.  

3) On 24.10: you will be assigned to your selected project if there are spots left. 

Preference is given to those who have the project's required skills and couldn`t get a place in the last semesters. 

4) Until 07.11.: If you *are not* on the waiting list: register for the corresponding modul via QISPOS or Prüfungsamt:

Interdisziplinäres Medienprojekt (10LP), POS-Nummer: 6640

Studienprojekt Quality & Usability (9 LP), POS-Nummer: 2200

Studienprojekt Quality & Usability (6 LP), POS-Nummer: 2080

Note that there are two modules with the same title but different numbers of LP/CP in QISPOS

Please be aware that the effort (time to spent) is a third higher for the 9 LP module than for the 6 LP module.

You can also register for the 10LP/CP module "Interdisziplinäres Medienprojekt", if you need it for your course of study (e.g. Medieninformatik).

Note:

If you are on the waiting list: we will notify you after 07.11, if there is a spot left in your preferred project.

You cannot participate in the project if you do not attend the corresponding introductory session on 19.10.

You cannot participate in the project if you have been assigned to a project but do not register via QISPOS or Prüfungsamt (+ send confirmation) by 07.11

Topics

1. TOPIC: Chatbot Technologies (FULL)

  • Supervisors: María González García (DFKI), Stefan Hillmann (TU Berlin)
  • Language: English, German
  • 6 or 9 ECTS points
  • 5 to 12 participants
  • date: initially Thursdays 9 to 10:00

Hint: The weekly meeting date is just an indication. The actual meetings will be organized depending on the constraints and needs of participants and supervisors.

In this project we plan to extend our chatbot framework with additional technologies and external services. The chatbot framework is a client server application. This client is written in HTML / CSS / Javascript / NodeJS, while the backend is written in Java. We will focus mostly on the implementation of backend services. Depending on the number of students of the course, our plan is that one or more of the following subprojects are implemented in smaller groups of 1-3 students. Depending on the size of the groups, we might also combine two subprojects for one group. Some projects require knowledge or interest in learning Natural Language Processing. The chatbot can talk and understand either in English or German or both.The subprojects that we plan to be implemented are: 

  1. Add route planning abilities to the chatbot. E.g. “When do I need to leave TU if I want to be at 3 o’clock at Wannsee?”. Or “How do I get to Hauptbahnhof?”. This task also requires frontend work.
  2. Add weather reporting abilities to the chatbot. E.g. “What will the weather be like tomorrow?”. This task also requires frontend work.
  3. Integrate external databases / APIs like Yahoo answers, Linked Open Data Cloud (DBPedia) or Quora to the chatbot. This can be used to answer questions like “What is the population of Madrid?”, “How do I take care of my leather jacket?”, “Who is the president of Nicaragua?” or “How do I cook a vegetable soup?”.
  4. Automatically parse FAQ lists in the chatbot.
  5. Add synonym detection through Semantic Parse Trees. This project requires primary knowledge in Natural Language Processing.
  6. Create a Natural Language Processing Pipeline which can detect locations and temporal expressions. This task is used for tasks 1 and 2. It can, for example, detect the temporal expression “tomorrow” in the sentence “What will the weather be like tomorrow?” and therefore, create structured information from unstructured textual data.

Project requirements (desired student skills):programming experience in Java

  • optional: knowledge in HTML / CSS /Javascript / NodeJS
  • ideally prior experience in natural language processing or chatbot development
  • interest to implement Java code in a small team
  • the will to meet with colleagues from DFKI (Berlin, Alt-Moabit 92, 10 minutes from TU Berlin by bike or bus) from time to time.

 

2. TOPIC: Foundations of Face Recognition (FULL)

  • Supervisors: Patrick Ehrenbrink, Stefan Hillmann
  • Language: English, German
  • ECTS Points: 6 or 9
  • Participants: up to 10
  • Thursdays 10:15-11:45

This project will deal with basic techniques of face recognition. We will apply and benchmark different techniques for detecting and, based on the project's pace, recognizing faces. Students will learn the principles behind face detection and recognition and will implement basic programs that can detect faces in images and videos. We will also learn how to generate the necessary image corpora and how to generate feature-sets that can be used for recognizing faces. 
The project also involves benchmarking face detection and recognition techniques on faces of elderly persons. The tasks that are to be performed during the project include:

  • generating image corpora
  • applying face detection on images
  • benchmarking different image-processing techniques
  • [optional] applying face detection on videos
  • [optional] generating features sets for faces detection

The project requires the participating students to write some code in python. You should also be comfortable with using the Shell of your operating system. 
The project will based on openCV and openBR. Make sure both are compatible to your computer before participating in the project.

 

3. TOPIC: Dialog System Technology Challenge 7 (4 Places open)

    • Supervisors: Stefan Hillmann, Thilo Michael
    • 5 to 12 participants
    • 9 ECTS points
    • Tuesdays, 9:00–10:00 am in room TEL 209
    • Language German and English

    In this project you will work on the Sentence Selection track [1] of the Dialog System Technology Challenges (DSTC7) [2]. The goal of the sentence selection challenge is to predict the possible next sentences from a list of given sentences in a conversation between two parties. Such a module can be used to build a topic-specific chatbot (which is not part of this project).
    As the real DSTC7 has already started the system(s) implemented in this project cannot directly participate in the official challenge. However, we can compare the results of your system to the results published by DSTC7.
    The organizers of the challenge provide us with data sets of recorded conversations (in form of written text) which can be used to train and evaluate a sentence selection module. Such a module can be a simple statistical model but also an advanced neural network. We would prefer to implement a rather simple approach (as baseline) and a rather advanced/complex approach in frame of this project.
    Ideally, we have about 12 participants in order two build groups of 6 members, each implementing their own approach.

    [1] https://ibm.github.io/dstc7-noesis/public/index.html
    [2] http://workshop.colips.org/dstc7/call.html

    Project requirements (desired student skills):

    • prior experiences in Python and /or natural language processing
    • experience with GIT and scientific libraries (e.g. numPy) are a plus
    • interest in the conception and implementation of a predictive model in a team
    • the conversation data are dialogs in English, thus you need English skills to understand the data and to read related literature

     

    4. TOPIC: Environmental Noise Loudness Estimation for Crowdsourcing Systems (CANCELED DUE TO LACK OF INTEREST)

    • Supervisor: Rafael Zequeira Jiménez
    • Contact email:
    • 6 ECTS points
    • Maximum number of participants: 15
    • Intro session slides: here

    When and where the weekly meetings will take place:
    Day of the week: Wednesdays
    Time from: 11:00
    Time until: 12:00
    Building: TEL-20
    Room: Auditorium 1
    Project language: English

    Project description: 

    The goal of this project is to create a labeled database regarding the level of noise. Such noises will be those most commonly present in home Crowdsourcing environments. This data will be used to train a machine learning classifier, and/or a neural network, to estimate loudness given audio files. The developed tool will be useful for quality control within speech quality assessment studies.

    Project requirements (desired student skills):

    • Skills in machine learning
    • Neural Networks
    • Python, NodeJS, JavaScript, Matlab, R, HTML
    • Statistics
    • TensorFlow, Keras

     

    5. TOPIC: Applied Statistics for HCI (CANCELED DUE TO LACK OF INTEREST)

    • Supervisors: Babak Naderi, Steven Schmidt
    • Language: English
    • 6CP Project (POS-Nummer: 2080)
    • Fridays 10:00- 12:00 (TEL 209)
    • Limit: 16 participants

    Do you love statistics? Either you answer yes or no, this project could be interesting for you.

    The aim of this project is to introduce statistical methods, which are commonly used in the Human-computer Interaction (HCI) research, to students from practical perspective. At the end of this course, students should be able to select appropriate statistical methods, for the given research question, and understand the implications and limitations of various methods.

    This course will not contain heavy details about math, rather we focus on how and when to use methods already provided by statistical packages.

    We cover following topics: basics on research and experiment design, data visualization, data screening, testing associations, prediction models and common statistical tests.

    During this course, we introduce each topic and provide couple of datasets from real research questions. You will apply the concept on each dataset using a statistical package (Python, R, MATLAB, SPSS), report the results, and write a wiki guide using the samples. Students work in groups, and deliver home works each week. A final project will also be assigned to each group, which its result should be presented at the last session.

    This course is recommended to students who want to write a thesis at our chair or are interested to scientific carrier. Note that, although we take examples from HCI domain, the topics covered in this course are fundamental statistical methods useful in many applications in computer science.

    Zusatzinformationen / Extras

    Quick Access:

    Schnellnavigation zur Seite über Nummerneingabe

    Auxiliary Functions