Comparative and Score are coming out the same about sentiment HOT 4 CLOSED

@sseetal Can you provide a little more info?

• What versions of node are you running on the server and your computer?
• What are the input strings that are producing these results?
• Can you share the output from both instances? It is difficult to understand what you are expecting.

from sentiment.

Thanks for getting back to me.

1. We're using nvm everywhere. node 4.1.1 locally (MAC), v6.9.4 on Server (Ubuntu)
2. The input streams are in the results I posted above (happy, sad, ok) to try and get the three desired results (positive, negative and neutral results)
3. I tried getting the local results (the server results are above). I can now further understand the problem. It's because they are one word.

INPUT sentiment

For input "happy" (both locally and on server)

{ score: 3,
comparative: 3,
tokens: [ 'happy' ],
words: [ 'happy' ],
positive: [ 'happy' ],
negative: [] }

For input "happy cat"

{ score: 3,
comparative: 1.5,
tokens: [ 'happy', 'cat' ],
words: [ 'happy' ],
positive: [ 'happy' ],
negative: [] }

For Input "I am happy today"

{ score: 3,
comparative: 0.75,
tokens: [ 'i', 'am', 'happy', 'today' ],
words: [ 'happy' ],
positive: [ 'happy' ],
negative: [] }

It was hard to find in the docs but all i really need is the range of values (Min and Max) for score and comparative. I thought comparative was between -1 and 1, but now I can't tell.

from sentiment.

Just to add more context. We're using this to analyze comments left by people in a survey. The whole point is to have a score and an overall score. Ideally a score out of 100. But can't seem to figure out how to calculate a comparable score.

from sentiment.

This issue includes a thread that covers some of the areas you are asking about including how "comparative score" works and how to scale / offset it.

TL;DR: Tokens are scored between -5 (highly negative) and 5 (highly positive) with 0 being "neutral". Comparative scores are simply SUM(tokens) / number of tokens. Details can be found in the AFINN whitepaper:
http://www2.imm.dtu.dk/pubdb/views/publication_details.php?id=6010

Sorry that the docs don't make this more clear. I'll file an issue now to make sure I don't forget to improve them. 😄

from sentiment.