Social media mining can and should have broad interpretation. It is not the intent of the authors to confine social media mining to sentiments or opinions, but rather we suggest that a sentiment or opinion is a useful tool for many research pursuits.
Until recently, sentiment was understood as a ubiquitous and constant part of the human experience, with variations in sentiments changing only slightly up or down. Klaus Scherer (2000) developed a working definition as follows:
"Emotions (sentiments) are episodes of coordinated changes in several components in response to external and internal events of major significance to the organism."
It is our intent to understand, measure, and interrelate these changes in a sentiment. Scherer's typology of emotions is a useful grounding point for the understanding of sentiments, and as a jumping-off point for a discussion of the difficulty in measuring sentiment-laden text.
Scherer's typology of emotions is briefly explained as follows:
- Emotion: This is a brief, organically synchronized evaluation of a major event, for example, being angry, sad, joyful, ashamed, proud, or elated
- Mood: This is a diffused, non-caused, low-intensity, long-duration change in subjective feeling, for example, being cheerful, gloomy, irritable, listless, depressed, or buoyant
- Interpersonal stance: This is an affective stance towards another person in a specific interaction, for example, being friendly, flirtatious, distant, cold, warm, supportive, or contemptuous
- Attitude: This is enduring, affectively colored beliefs or dispositions towards objects or persons, for example, being liking, loving, hating, valuing, or desiring
- Personality traits: These are stable personality dispositions and typical behavior tendencies, for example, being nervous, anxious, reckless, morose, hostile, or jealous
Generally, when we try to measure a sentiment, we talk about Scherer's emotions; though, in some situations, we might try to capture longer-term phenomena such as moods.
Anchoring the neo-social science approach using Twitter data versus other types of social media data is important as well because not all data is equal. Twitter data differs from data derived from sites such as Yelp and Google Reviews due to the simple fact that Twitter does not have ratings or explicit targets. If we want to know the sentiment of a given source or topic, whether it is the iPhone 5S or something more sensitive such as social policy, we have to discover that signal in a corpus of other signals. However, Yelp and Google Reviews (just two examples of many) have explicitly accounted for the source or topic by design and have ratings designed to measure sentiments.
A tweet is what Twitter users send to each other and to the Twittersphere. A tweet is sometimes a sentence and other times not, but it is restricted to 140 characters or approximately 11 words. Twitter therefore provides sentence-level sentiment analysis as opposed to reviews on Yelp or Google Review, which usually constitute the entire documents.