Student Projects

Exploring the properties of data applicable for training socialbot evaluation model

Duration: One Semester
Lab: HCI/IC/EPFL
Goals: Identify the properties of training data that yield the strongest correlation of automatic socialbot evaluation model with human.
Assistant: Ekaterina Svikhnushina (ekaterina DOT svikhnushina AT epfl DOT ch)
Student Name: Open
Keywords:  
Abstract:

This project aims at investigating what kind of data should be used for training a socialbot evaluation model, so that its predicted scores for bot’s responses meet the human expectations.

The student is expected to:

  • train the socialbot evaluation model with different datasets (Twitter, OpenSubtitles, Movie Dialogs)
  • frame and conduct a human experiment to collect human scores for a test set of responses
  • compare the correlations of scores predicted by different evaluation models with the human
  • retrieve the outstanding properties of the dataset that resulted in the highest correlation of automatic score with the human
Related Skills: Background or interest in data mining, machine learning, natural language processing
Suitable for: Master student. Interested student should contact Ekaterina Svikhnushina (ekaterina DOT svikhnushina AT epfl DOT ch) and Pearl Pu (pearl DOT pu AT epfl DOT ch) along with a copy of your CV.

Creation of large high-quality human dialog corpus

Duration: One Semester
Lab: HCI/IC/EPFL
Goals: Crawl multi turn dialogs from Twitter, clean and analyze the resulting dataset.
Assistant: Ekaterina Svikhnushina (ekaterina DOT svikhnushina AT epfl DOT ch)
Student Name: Open
Keywords:  
Abstract:

This project aims at creating a large (~1M samples) clean dataset of natural human dialogs, which will be applicable for development and evaluation of open-domain dialog agents

The student is expected to:

  • use Twitter API to crawl multi-turn sequences of tweets
  • develop criteria for cleaning the data
  • analyze the clean dataset on utterance and vocabulary levels (identify emotional coloring of the utterances, evaluate the percentage of obscene language in the vocabulary, etc.)
  • visualize the findings of the analysis
Related Skills: Background or interest in data mining, machine learning, natural language processing
Suitable for: Master student. Interested student should contact Ekaterina Svikhnushina (ekaterina DOT svikhnushina AT epfl DOT ch) and Pearl Pu (pearl DOT pu AT epfl DOT ch) along with a copy of your CV.

Learning to Converse with Personas

Duration: One Semester
Lab: HCI/IC/EPFL
Goals: Build a dialog model that can produce character-specific response.
Assistant: Yubo Xie (yubo DOT xie AT epfl DOT ch)
Student Name: Open
Keywords: human dialog modeling, natural language processing
Abstract:

One of the main characteristics of human dialogs is the diversity of the responses: given an input message, one could have many responses that are equally good. This could be explained by the fact that people with different backgrounds and personalities could respond differently to the same utterance. This semester project aims at modeling different personas in data-driven approaches.

The student is expected to:

  • reproduce the existing work and train the model on the provided Big Bang Theory corpus
  • evaluate the model performance
Related Skills: Basic knowledge in neural networks and natural language processing
Suitable for: Master student. Interested student should contact Yubo Xie (yubo DOT xie AT epfl DOT ch) and Pearl Pu (pearl DOT pu AT epfl DOT ch) along with a copy of your CV.

Dueling Chinese Couplets with Neural Networks

Duration: One Semester
Lab: HCI/IC/EPFL
Goals: Build a neural model that can generate Chinese couplet response.
Assistant: Yubo Xie (yubo DOT xie AT epfl DOT ch)
Student Name: Open
Keywords: Chinese couplets, sequence transduction, natural language processing
Abstract:

Chinese couplets are a reduced form of Chinese poems, composed of two lines of poetry, adhering to some rules. Dueling couplets has been a popular game among intellectuals since ancient China, where one composes the second line according to the given first line. This can be regarded as a sequence transduction problem. This semester project aims at teaching machines to compose the second lines in couplets, using data-driven approaches such as neural networks.

The student is expected to:

  • find or create a Chinese couplet dataset
  • use some sequence transduction architecture (for example seq2seq) to build a model
  • evaluate the results
Related Skills: Proficiency in Chinese; basic knowledge in neural networks
Suitable for: Master student. Interested student should contact Yubo Xie (yubo DOT xie AT epfl DOT ch) and Pearl Pu (pearl DOT pu AT epfl DOT ch) along with a copy of your CV.

Dialog Segmentation in Movie and TV Subtitles

Duration: One Semester
Lab: HCI/IC/EPFL
Goals: Devise an algorithm that can automatically segment dialogs in subtitles.
Assistant: Yubo Xie (yubo DOT xie AT epfl DOT ch)
Student Name: Open
Keywords: human dialogs, text segmentation, natural language processing
Abstract:

One of the main challenges of open-domain dialog modeling is the scarcity of training data, especially for multi-turn settings. Movie and TV subtitles are naturally a good source for developing conversation corpora. Currently the biggest corpus is the OpenSubtitles dataset. However, subtitle files usually lack clear scene markers, making it difficult to extract self-contained dialogs used for training multi-turn dialog models.

The student is expected to:

  • preprocess and analyze the OpenSubtitles dataset
  • devise an automatic dialog segmentation algorithm (rule-based or data-driven)
  • evaluate the segmentation accuracy
Related Skills: Basic knowledge in natural language processing
Suitable for: Undergraduate/Master student. Interested student should contact Yubo Xie (yubo DOT xie AT epfl DOT ch) and Pearl Pu (pearl DOT pu AT epfl DOT ch) along with a copy of your CV.

Discovering behavior change profiles from energy consumption data

Duration: One Semester
Lab: HCI/IC/EPFL
Goals:

Understand and predict how energy consumption patterns change from one season to another.

Assistant: Igor Kulev (igor DOT kulev AT epfl DOT ch)
Student Name: Open
Keywords: clustering, time series modeling, machine learning, behavior change analysis
Abstract:

People have different energy consumption patterns in different seasons. This project aims at understanding and predicting how these patterns change from one season to another. This could help distribution network operators to identify suitable customer groups for demand reduction solutions and apply timely interventions. There are few novel machine learning methods developed in the HCI lab that can help to solve this problem. The goal is to adapt, apply and compare these methods in the context of energy consumption. The student will be provided with large energy consumption time series dataset.

Related Skills: Knowledge in Data Mining and/or Machine Learning; Programming skills
Suitable for: Master Student in computer science and data science. Interested student should contact Igor Kulev (igor DOT kulev AT epfl DOT ch) and Pearl Pu (pearl DOT pu AT epfl DOT ch) along with a copy of your CV.

Behaviour recommender system for active lifestyle

Duration: One Semester
Lab: HCI/IC/EPFL
Goals: Develop a mobile recommender system app.
Assistant: Kavous Salehzadeh Niksirat (kavous DOT salehzadehniksirat AT epfl DOT ch)
Student Name: Open
Keywords: recommender systems, algorithm, physical activity, lifestyle, behavior change, persuasion
Abstract:

Lack of physical activity is one of the leading risk factors that threaten global health. For instance, 46% of European citizens never exercise. Sedentary lifestyle can cause cardiovascular disease, diabetes, and mental health problems. One promising technique to increase physical activity is using recommender systems to suggest daily goals (step numbers). A recent technique developed in our lab analyses physical activity patterns of existing users to data mine these achievable goals. Nevertheless, it is unclear how to recommend them to new users’ in an optimal way. In this project, we focus on comparing existing algorithms in traditional recommender systems such as TOP-N or AVERAGE to understand the efficacy of behavior change strategies.

The student is expected to:

  • survey the state-of-the-art methods in recommender systems
  • Design and test a mobile behaviour recommender system app
Related Skills: recommender system algorithms, time series data analysis
Suitable for: Master Student. Interested student should contact Kavous Salehzadeh (kavous DOT salehzadehniksirat AT epfl DOT ch) and Pearl Pu (pearl DOT pu AT epfl DOT ch) along with a copy of your CV.

Qualitative Study of Inter-generational Social Incentives for Fitness Apps

Duration: One Semester
Lab: HCI/IC/EPFL
Goals: Conduct a qualitative study.
Assistant: Kavous Salehzadeh Niksirat (kavous DOT salehzadehniksirat AT epfl DOT ch)
Student Name: Open
Keywords: Physical activity, behavior change, persuasion, health care applications, partner
Abstract:

Although behaviour change is hard, especially for keeping a regular exercise routine, earlier research showed that having a workout companion (an exercise partner) can increase engagement and adherence. What if this partner is a family member from a different generation? In this project, we ask you to perform a qualitative user study to understand the purpose, the pains, gains, and preferences from a small group of inter-generational users.

The student is expected to:

  • Investigate and understand the state-of-the-art in behavior change, persuasion techniques, and Digital health technology
  • Recruit and interview 6-10 pairs of inter-generational users based on an interview/observation guide we have developed
  • Collect users’ reactions, feelings, preferences
  • Develop a set of design guidelines to further improve the app
Related Skills: HCI, social and personal skills (interacting with senior adults), data analysis
Suitable for: Master Student who followed the HCI course (CS 486). Interested student should contact Kavous Salehzadeh (kavous DOT salehzadehniksirat AT epfl DOT ch) and Pearl Pu (pearl DOT pu AT epfl DOT ch) along with a copy of your CV.

Survey of recent interface and interaction tools

Duration: One Semester
Lab: HCI/IC/EPFL
Goals: The aim of this project is to survey and try some of the most recent interface tools on the market.
Assistant: Kavous Salehzadeh Niksirat (kavous DOT salehzadehniksirat AT epfl DOT ch)
Student Name: Open
Keywords: HCI, VR, AR, AI, IoT
Abstract:

Each year the HCI course encourages our students to create and design an awesome app. But many students are not aware of the recent development in HCI, VR, AR, AI, and brain interface field.

The student is expected to:

  • Investigate and use some existing interface tools
  • Create a comprehensive and thorough report
  • Include usage examples to illustrate design principles
Related Skills: HCI, communication skills (writing and surveying)
Suitable for: Master Student who followed the HCI course (CS 486). Interested student should contact Kavous Salehzadeh (kavous DOT salehzadehniksirat AT epfl DOT ch) and Pearl Pu (pearl DOT pu AT epfl DOT ch) along with a copy of your CV.

Voice or dialog interface for computer aided design tools

Duration: One Semester
Lab: HCI/IC/EPFL
Goals: The aim of this project is to provide NI (natural interface) or mixed interface to BIM visualization tools.
Assistant: Kavous Salehzadeh Niksirat (kavous DOT salehzadehniksirat AT epfl DOT ch)
Student Name: Open
Keywords: voice and dialog interface for CAD tools
Abstract:

BIM visualization tools allow users to walk through a house or a building in a virtual or augmented digital world. However, the current tool, using mainly graphical user interfaces (GUI), is difficult for most lay users. Replacing it with voice and dialog interfaces (natural interfaces) or integrating them with GUI can potentially overcome this challenge.

The student is expected to:

  • Investigate and use some existing BIM visualization tools
  • Integrate DialogFlow (goole’s dialog engine) to GUI
  • Integrate voice activation
Related Skills: HCI; basic knowledge in voice and dialog interface, e.g., DialogFlow (Google)
Suitable for: Master Student. Interested student should contact Kavous Salehzadeh (kavous DOT salehzadehniksirat AT epfl DOT ch) and Pearl Pu (pearl DOT pu AT epfl DOT ch) along with a copy of your CV.