Database Reference
In-Depth Information
Figure 9.6
Sentiment140 [41], an online tool for Twitter sentiment analysis
Emoticons make it easy and fast to detect sentiments of millions or billions of
tweets. However, using emoticons as the sole indicator of sentiments sometimes
can be misleading, as emoticons may not necessarily correspond to the sentiments
in the accompanied text. For example, the sample tweet shown in
Figure 9.7
contains the
:)
emoticon, but the text does not express a positive sentiment.
Figure 9.7
Tweet with the :) emoticon does not necessarily correspond to a
positive sentiment
To address this problem, related research usually uses Amazon Mechanical Turk
(MTurk) [44] to collect human-tagged reviews. MTurk is a crowdsourcing Internet
marketplace that enables individuals or businesses to coordinate the use of human
intelligence to perform tasks that are difficult for computers to do. In many cases,
MTurk has been shown to collect human input much faster compared to traditional
channels such as door-to-door surveys. For the example sentiment analysis task,
the Data Science team can publish the tweets collected from Section 9.3 to MTurk
as Human Intelligence Tasks (HITs). The team can then ask human workers to tag
each tweet as positive, neutral, or negative. The result can be used to train one or
