Adversative conjunctions cause neutrality when used with a negator

[ashleywysocki]

I noticed when analyzing a sentence like "But I'm not happy" (sentiment: 0) that it differs from "I'm not happy" (sentiment: -0.4330127) by making it a neutral statement. However, in my understanding of adversative conjunctions, they should amplify the current clause which should make "But I'm not happy" more negative. I tested some other negator words (didn't, isn't, etc) and when those are used before the polarized word, adding an adversative conjunction like "but" always makes the sentiment score become 0. I looked at the underlying code but couldn't figure out what the issue is so while I can neutralize the adversative weight for now, I was hoping there could be a fix to this at some point.

[trinker]

Thanks for trying sentimentr and reporting this issue. Can you please provide a minimal working example with all the code and strings you've discussed here? This makes it easier to understand, reproduce, and solve. Github allows you to make nice code chunks to display this code.

…

On Fri, Mar 30, 2018, 7:56 PM Ashley Wysocki ***@***.***> wrote: I noticed when analyzing a sentence like "But I'm not happy" (sentiment: 0) that it differs from "But I'm not happy" (sentiment: -0.4330127) by making it a neutral statement. However, in my understanding of adversative conjunctions, they should amplify the current clause which should make "But I'm not happy" more negative. I tested some other negator words (didn't, isn't, etc) and when those are used before the polarized word, adding an adversative conjunction like "but" always makes the sentiment score become 0. I looked at the underlying code but couldn't figure out what the issue is so while I can neutralize the adversative weight for now, I was hoping there could be a fix to this at some point. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#80>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABrnzqpzE6eAT1dLrS4316tAuQQNaJDyks5tjrj3gaJpZM4TCOGB> .

[ashleywysocki]

Yes, it's very simple really.

test_sentences <- c("I'm not happy", "But I'm not happy", "I didn't like it", "But I didn't like it") test_sentences_sentiment <- sentiment(test_sentences)

[trinker]

I see what you mean. I'll have a look when I get development time for sentimentr unless someone provides a pull request in the eman time:

library(sentimentr)

test_sentences <- c(
    "I'm not happy", 
    "But I'm not happy", 
    "I didn't like it", 
    "But I didn't like it",
    "I like it", 
    "But I like it",
    "I hate it", 
    "But I hate it"
)


test_sentences_sentiment <- sentiment(test_sentences)

data.frame(text=test_sentences, test_sentences_sentiment)

                  text element_id sentence_id word_count  sentiment
1        I'm not happy          1           1          3 -0.4330127
2    But I'm not happy          2           1          4  0.0000000
3     I didn't like it          3           1          4 -0.2500000
4 But I didn't like it          4           1          5  0.0000000
5            I like it          5           1          3  0.2886751
6        But I like it          6           1          4  0.6200000
7            I hate it          7           1          3 -0.4330127
8        But I hate it          8           1          4 -0.9300000