search

hope-data-science / akc

Automatic knowledge classification based on keyword co-occurrrence network
https://hope-data-science.github.io/akc/
Other
15 stars 4 forks source link

Response to more suggestions from Richard #3

Open hope-data-science opened 4 years ago

hope-data-science commented 4 years ago
  1. Control the color of visualization.
  2. Get the word cloud from a simple group.
  3. Classification using word embedding.
hope-data-science commented 4 years ago

Response:

  1. Try :

    library(akc)

bibli_data_table %>%
keyword_clean(id = "id",keyword = "keyword") %>%
keyword_group(id = "id",keyword = "keyword") %>%
keyword_network() + ggplot2::scale_fill_viridis_d()

    I believe you can add more colors with "scalefill" from ggplot2. So I don't use it directly here.

  2. Since now you can get the raw table from the results, just try ggwordcloud and DIY. akc just provides a robust method to get a good-enough-for-publish result. From that package, you can absolutely get more options to adjust. (Still, I will add it when I am free)

  3. This is the scope for larger than version 1.0.0. akc now focuses on grouping based on community detection (as you could find in the logo). Vectorizing would be done outside of akc, but akc should be able to receive word vectors and cluster them automatically. That is in the plan.

snvv commented 1 month ago

What is the theory (paper) behind automatic clustering?

hope-data-science commented 1 month ago

Community detection in network science.