Hi everyone! I wanted to know if anybody knows how we code random words as positive (+1), negative (-1) and neutral (0). Like how do we know if a word is positive, negative or neutral. For example, house, happy, hand, hall etc..Do we look at it's valence and if yes, how can we find the valence of a word? It is for a psychology research project. Many thanks :)
The ANEW dataset was built based on the statistical annotation of words. It may be that words have different meaning to different people, but there is a general trend.
All words could be potentially interpreted as one or another, e.g. "love" could be wonderful as a mother thinking of her children, or "love" hurts for someone that is going through a break-up. In generally you could say it is a matter of perception.
But this is why you get the average in a sentence and not for each word individually. For instance:
(1) Love is wonderful, I am so happy with my family
(2) I hate the world, it is an ugly life, I would rather die
The above sentences as scored in ANEW for mean valence:
(1) 8.6 (love) + 8.25 (happy) + 7.5 (family) = 24.35
(2) 2.92 (hate) + 2.5 (ugly) + 7 (life) = 12.42
*not all words were available in the dataset
What we can deduce from the above example, although "life" has a positive value, when taken into the context of sentence (2) it adds to a lower sum of valence for the complete sentence. From scoring, as well as from common sense, it is quite clear which sentence is positive and which negative.
In the overall I would agree that it may be hard to have an objective rating for a single word, and this is why it should be put in the context of a sentence.
Perhaps the easiest method is to have a sample of people rate the words from positive to negative, perhaps on a 7 point scale ranging from extremely positive to extremely negative with neutral in the middle. The mean response could be taken as the value. There are more complex ways to scale the words as explained in JP Guilford, 1954, Psychometric Methods.
Check out ANEW dataset (which is free) here: https://tomlee.wtf/2010/06/16/anew/
And also this: http://liwc.wpengine.com/ but it is commercial, so you have to pay in order to use it. But you can try it online too.
Because words have a range of meanings and because different meanings will be salient to different individuals, there is no objective way to identify the valence of words. Paul Spector's suggestion of collecting data is the best way of getting the information you're looking for.
The ANEW dataset was built based on the statistical annotation of words. It may be that words have different meaning to different people, but there is a general trend.
All words could be potentially interpreted as one or another, e.g. "love" could be wonderful as a mother thinking of her children, or "love" hurts for someone that is going through a break-up. In generally you could say it is a matter of perception.
But this is why you get the average in a sentence and not for each word individually. For instance:
(1) Love is wonderful, I am so happy with my family
(2) I hate the world, it is an ugly life, I would rather die
The above sentences as scored in ANEW for mean valence:
(1) 8.6 (love) + 8.25 (happy) + 7.5 (family) = 24.35
(2) 2.92 (hate) + 2.5 (ugly) + 7 (life) = 12.42
*not all words were available in the dataset
What we can deduce from the above example, although "life" has a positive value, when taken into the context of sentence (2) it adds to a lower sum of valence for the complete sentence. From scoring, as well as from common sense, it is quite clear which sentence is positive and which negative.
In the overall I would agree that it may be hard to have an objective rating for a single word, and this is why it should be put in the context of a sentence.
- Badges
- Science topic
- Similar topics
- Psychology
- Emotional Psychology
- Happiness