twitchNLP

Toolkit to do NLP analysis of twitch comments streams.

Installation

You can install the development version of twitchNLP from GitHub with:

# install.packages("devtools")
devtools::install_github("badbayesian/twitchNLP")

You'll need to also generate the conda enviroment twitchNLP with:

conda env create -f enviroment.yml

Downloading twitch comments and structure.

You can download twitch stream comments here, be nice!

For this program to understand the chat logs, the structure needs to be /<streamer>/<date of stream>/chat/*.html e.g. data/dylanburnstv/Friday, November 26, 2021/chat/page_1.html. There may be multiple pages of html files corresponding to the number of comments. Similarly, the first line in /<streamer>/<date of stream>/chat/README.md will be considered as the stream title for any plots.

Example

This is a simple example for processing a single downloaded stream and all the streams downloaded from a streamer.

library(furrr)
plan(multisession)

library(progressr)
handlers("pbmcapply", "beepr")

library(twitchNLP)

# From a given twitch stream chat logs, we can generate a number of different plots
dylan_nov_26_path <- "data/dylanburnstv/Friday, November 26, 2021/chat/"
df <- load_stream_chat(dylan_nov_26_path)
single_stream_plots <- gen_all_plots(df, path = dylan_nov_26_path, window_size = 5, interactive = TRUE)

# Similarly, from a given streamer, we can generate the previous plots for each stream
streamer_plots <- analysis_of_streamer("data/dylanburnstv")