Thesis Topics

My research interests include Information Retrieval, Music Information Retrieval, recommender technology, pattern matching, compression computer music technology, audio processing, informetrics, genomic data-searching techniques.

Below are a list of suggested topics. If you have an idea that is not on this list but in one of the areas listed above, feel free to see me and discuss its potential as a thesis/project topic.

  1. Music Recommender Systems
    Using an existing prototype music recommender written in PHP, Postgresql, and C++, develop some algorithms for finding music that users will like, based on their preferences and other evidence. You will run experiments to test the effectiveness of the algorithms. Novel work will include incorporating music audio features into the recommendation process.
  2. Music Information Retrieval Techniques
  3. A Web-based Graded Reader for Language Learning
    This thesis involves implementing a prototype application that allows a user to select reading material from the Web based on their knowledge of the language that they are learning.

    The prototype should provide a method of determining the user's level of language knowledge (using vocabulary lists based on word frequency), and return a ranked list of documents that match the user's knowledge.

    For regular users, the application should track what they've read as well as their knowledge, so that new documents can be presented.
      Other aspects of the topic include implementing several solutions to indexing documents for retrieval based on reading difficulty. Reading difficulty can be approximated using vocabulary measures and sentence length. Documents that are easy to read are likely to be those that are largely made up of common words and short sentences. However, search engines usually don't index common words due to the resulting large index size.

  4. Speech Retrieval for Sung Documents
    People often want to locate songs that they have heard. Often a query based on the lyrics will be successful on the Web. However, not all songs have their lyrics available. Also, recordings are sometimes not very clear and people may mishear the text, resulting in search phrases that sound similar but not matching the original text. This thesis topic involves applying speech recognition and speech retrieval techniques to index songs based on their phonemes for retrieval with text or sung queries.