trinker
commented
2 years ago Hello. Thanks for the comments.
Some sentiment dictionaries offer categorisation for a given text or calculated scores. Is there any categorisation of scores or text in sentimentR. Can we say like 0 is neutral, below 0 negative and above 0 positive? This may not be three class categorisation, it can consist of only negative or positive categories too.
I added a section to ?sentiment describing:
sentiment - Sentiment/polarity score (note: sentiments less than zero is negative, 0 is neutral, and greater than zero positive polarity)
Turning numeric into catgories is pretty standard so I have not provided a function to do this explicitly. Instead, I added this code chunk to the examles in sentimentr describing how to make categories (you could use sign and or cut to do this as well):
## Categorize the polarity (tidyverse vs. data.table):
library(dplyr)
sentiment(mytext) %>%
as_tibble() %>%
mutate(category = case_when(
sentiment < 0 ~ 'Negative',
sentiment == 0 ~ 'Neutral',
sentiment > 0 ~ 'Positive'
) %>%
factor(levels = c('Negative', 'Neutral', 'Positive'))
)
## element_id sentence_id word_count sentiment category
## <int> <int> <int> <dbl> <fct>
## 1 1 1 4 0.25 Positive
## 2 1 2 6 -1.87 Negative
## 3 2 1 5 0.581 Positive
## 4 3 1 5 0.402 Positive
## 5 3 2 4 0 Neutral
## 6 4 1 4 0 Neutral
library(data.table)
dt <- sentiment(mytext)[, category := factor(fcase(
sentiment < 0, 'Negative',
sentiment == 0, 'Neutral',
sentiment > 0, 'Positive'
), levels = c('Negative', 'Neutral', 'Positive'))][]
dt
## element_id sentence_id word_count sentiment category
## 1: 1 1 4 0.2500000 Positive
## 2: 1 2 6 -1.8677359 Negative
## 3: 2 1 5 0.5813777 Positive
## 4: 3 1 5 0.4024922 Positive
## 5: 3 2 4 0.0000000 Neutral
## 6: 4 1 4 0.0000000 NeutralAnother thing is not an issue but just a curiosity. Have you ever used VADER for sentiment analysis? It is also developed for valence aware sentiment analysis. It has also ported for R I saw your performace review of other sentiment libraries, if you used it, what do you think about VADER?
No I have not tried it. I am aware of it. At the time ran the tests I don't think it had been ported to R , or at least, I was unaware of it. If the community ran testing on VADAR vs. other approaches it would be interesting to see the results.
Some sentiment dictionaries offer categorisation for a given text or calculated scores. Is there any categorisation of scores or text in sentimentR. Can we say like 0 is neutral, below 0 negative and above 0 positive? This may not be three class categorisation, it can consist of only negative or positive categories too.
Another thing is not an issue but just a curiosity. Have you ever used VADER for sentiment analysis? It is also developed for valence aware sentiment analysis. It has also ported for R I saw your performace review of other sentiment libraries, if you used it, what do you think about VADER?