Fragment filtering

[wmilkowska]

No description provided.

[roszcz]

I'm not convinced that this filters out fragments with long pauses - this condition seems to be true for every variant that is longer than 2s :)

[roszcz]

I think it would be better to separate this logic from the prepare_fragments function. If we want the dashboard to be helpful in reviewing different fragmentation methods, a structure like this would be better:

fragments = prepare_fragments(df, piece, n)

filtering_method = st.select_box(
  label="Filtering method",
  options=["Top 5", "Fastes", "Longest"],
)

if filtering_method == "Top 5":
  filtered_fragments = top_five_filtering(fragments)
elif filtering_method == "Fastes":
  filtered_fragments = get_fastests_filtering(fragments)
...

for fragment in filtered_fragments:
  # Dsiplay
  ...

That way, you'll also be forced to come up with names for different filtering strategies, which will be useful in reviewing :)

[wmilkowska]

For now I've set the n_clustering parameter to a fixed amount (16), and now you have the option to choose between displaying the first five fragments, the longest and the fastest fragments. Also, now if you want to test it on different files, you don't have to rerun Streamlit every time:)