Any hint why the two emoji approaches are different and in what circumstance which one is better?!

[jguo1002]

In the doc there are two approaches to deal with emojis:

## Emojis
## Not run:
## Load R twitter data
x <- read.delim(system.file("docs/r_tweets.txt", package = "textclean"),
stringsAsFactors = FALSE)
x
library(dplyr); library(magrittr)
## There are 2 approaches

## Approach 1: Replace with words
x %>%
mutate(Tweet = replace_emoji(Tweet)) %$%
sentiment(Tweet)

## Approach 2: Replace with identifier token
combined_emoji <- update_polarity_table(
lexicon::hash_sentiment_jockers_rinker,
x = lexicon::hash_sentiment_emojis
)
x %>%
mutate(Tweet = replace_emoji_identifier(Tweet)) %$%
sentiment(Tweet, polarity_dt = combined_emoji)
## End(Not run)

The result is different.

Is there any hint about why the results are different and in what circumstance which one is better? Thanks!

[trinker]

I would not expect the results to be the same. One is changing the underlying text (thus changing word counts) while the other uses the emojis as matches. So the difference is in the word counts which is used in computing the polarity.

Additionally, in the second example you give you replace the words as well as use the emojis in the sentiment dictionary.