Database Reference
In-Depth Information
Feedback Extractors: What Are People Saying?
As we've just seen, to get a sense of the meaning behind people's social media
messages, it's not enough to score positive and negative words. Semantic
understanding of raw text, such as tweets, is complicated and requires careful
analysis. It's for this reason that IBM has built industry-specific extractors for
the sentiment analysis tools in the SDA. These feedback extractors represent
sets of industry-specific rules (for retail, finance, and entertainment) that
apply to the brands, products, or services that you provide. Following are the
kinds of feedback that the SDA rules are looking for.
•
Buzz
The amount of discussion
•
Sentiment
The degree of satisfaction or dissatisfaction. This can get
quite granular. For example, if you were to tweet that you like the
movie
Moneyball
, but dislike the lead actor (it's just an example; we
think Brad Pitt is pretty cool), the SDA can isolate the sentiment for
this movie.
•
Intent to buy
The level of commitment to spend money
•
CustomerOf
Whether someone is an existing customer (for example,
someone said they bought a particular product)
Profile Extractors: Who Are These People?
For the context of what people are saying, it's extremely valuable to get a
sense of who has been expressing sentiment. Profile extractors are typically
applied over large sets of social media messages each time new batches are
added. The first few times that you run these extractors, the profile data that
you acquire will be sparse, but as your collection of social media messages
grows, additional profile elements will be populated for more of the users
that you are tracking. Profile extractors determine the following information
from social media messages, in addition to the poster's online profile infor-
mation (for example their user name, or Twitter profile):
•
Gender
The poster's sex, usually determined from the poster's
online profile information
•
Location
Where the poster lives
•
Parental status
Whether the poster has children
