In Need of Ideas? How About:

1. Choose a paper from the above list and try to verify its results. This would involve reading the paper for understanding, and then trying to repeat some of the work they have written up.

1. Read the paper. Find at least two of the papers referenced by this one and read those too.
2. [Part 1] Write up a summary (this forms the basis for your proposal). Descibe what parts you will repeat and why. Explain what you hope to learn / verify from this.
3. [Part 2] Emulate ("copy" by repeating) the work written up in the paper. Include testing and data analysis. Note that you will have to write & run some or all of the programs discussed in the paper.
4. [Part 3] Discuss the paper:
   1. Do you agree with what it says? Why?
   2. What was good about the work / paper/ Why?
   3. What was problematic? Why?
   4. Do you think they achieved their goals? Why?
   5. How do you think they could have done it better?

Note if you cannot verify the results, there are two possible reasons:

1. You have made a mistake.
2. The authors of the paper have made a mistake.

The first is more likely, but if you have checked and double-checked your results, then try double checking theirs.

2. Experiment with ways to "reliably" distinguish music from non-music. You will first have to come up with a clear definition of music. Start with something you can easily distinguish such as popular songs. Stick to one file format (like .wav, or MIDI).

3. Experiment with document searches using salient terms. The basic idea is to examine common terms between documents: if they share enough terms, the argument is that the documents are about the same kinds of things. Issues will include finding ways to do this efficiently (imagine how this technique might be useful when applied to thousands of documents).

4. Video: isolate one or two elements of a video file and use these to build a "profile" of the file. See if this allows for any kind of meaningful search.

5.