Re-evaluate package selection

[mjhendrickson]

Re-evaluate the selection of the saotd package for sentiment analysis from https://github.com/mjhendrickson/rtweet-sentiment-analysis/issues/2

saotd keys off of topics or hashtags from tweets, using it's own tweet_acquire function. I was able to step through the vignette, but became stuck on the sentiment analysis portion.

See here for more https://github.com/mjhendrickson/rtweet-sentiment-analysis/blob/master/Exploration%20Code/saotd_exploration.R

Looking at https://rdrr.io/cran/saotd/man/tweet_acquire.html shows the initial selection. I had been using my own data from rtweet.

mh <- get_timeline('mjhendrickson', n = 5000, include_rts = FALSE)

[mjhendrickson]

Initial thoughts are to re-review tidytext https://www.tidytextmining.com/, as well as this recent Tidy Tuesday video from Julia Silge https://twitter.com/juliasilge/status/1258111917098332160?s=20.

Tweet video: https://youtu.be/whE85O1XCkg Tweet code: https://juliasilge.com/blog/animal-crossing/

[mjhendrickson]

Consider quanteda as well. https://quanteda.io/articles/pkgdown/quickstart.html Could pair with sentimenr.

[mjhendrickson]

Evaluated sentimentr, possibly paired with quanteda.

Original sentimentr evaluation: https://github.com/mjhendrickson/rtweet-sentiment-analysis/issues/2#issuecomment-623240094

Seems to be working better, but still struggling with splitting tweets into sentences vs analyzing by full tweets. Failing on the out command in the vignette.

library(tidyverse)
library(rtweet)
library(sentimentr)

mh <- get_timeline('mjhendrickson', n = 5000)
mh_sentences <- get_sentences(mh$text)

sentiment(mh_sentences)
sentiment_by(mh_sentences)

out <- with(
    mh_sentences,
    sentiment(
        get_sentences(mh_text),
        list()
    )
)

plot(out)