edublancas / song-lyrics

Exploratory Analysis of 200K+ song lyrics from the 1 million songs dataset
https://blancas.io/song-lyrics/
MIT License
5 stars 1 forks source link

Explaining sentiment score and topic scores #36

Closed edublancas closed 6 years ago

edublancas commented 6 years ago

We are not explaining what sentiment score and topic scores in the interactive component, should we add a small legend somewhere to do so?

valmikkpatel commented 6 years ago

We could add it. There will be too much text to add it on the page directly but I could try using a tooltip.

What should I write there?

aaronsadholz commented 6 years ago

If we opt to add this, here is a one sentence description of topic score:

Each artist's topic scores range from 0 to 1, and represent the frequency the artist references that topic relative to other artists.

edublancas commented 6 years ago

@aaronsadholz @jose-alvarado-guzman how can we explain the scores in one sentence?

edublancas commented 6 years ago

@valmikkpatel maybe add it at the bottom? if you go for the tooltip route, just make sure to hint the user that they can get that by hovering

valmikkpatel commented 6 years ago

@edublancas can you also give me a one-liner about how similar artists are found?

edublancas commented 6 years ago

We used word embeddings to represent the bag of words as dense vectors, then took the mean vector per artist and computed pairwise distances. Most similar artists are the closest artists using euclidean distance.

valmikkpatel commented 6 years ago

Done