In Chapter 5, we learned about sentiment analysis using Watson Analytics' new social media feature to automatically analyze and categorize text posted to social media, in an attempt to determine an audience's feeling about a topic.

Key takeaways:

• Sentiment analysis can often be referred to as opinion mining and IBM Watson Analytics for Social Media is a relatively new offering in Watson Analytics and can be used for this purpose.
• The workflow in using Watson Analytics for social media is similar to the workflow that should be followed for all other Watson Analytics project types.
• The social media workflow involves: specifying topics, specifying the date range, specifying the types of source, reviewing the suggestions for your topics, and optionally defining themes.
• Depending on which version (or your subscription level) of IBM Watson Analytics for Social Media you are using, the number and type of social media documents that are supported vary.
• To control the volume of documents that is retrieved, you specify topic keywords, context keywords, and exclude keywords.
• To create a social media process, you click New project, then on the Create a new social media project tab you enter a name for the project, and click on the button labeled Next. At that point a new Watson Analytics Social Media project is then created.
• A topic embodies a portion of social media content that you want to retrieve and analyze. To add topics to a social media project, you type it into the Enter a topic text area and then click Add.
• You can click on any one of the projects topics and create topic keywords, context keywords, and exclusion keywords for that topic.
• As an option, you can also create investigative themes consisting of an attribute or list of attributes on which you want to break down a topic.
• Just like with any data query, it is always a good idea to try and limit the amount of social media data to retrieve and process. In Watson Analytics for Social Media, you can use a date range to accomplish this.
• You can select which language(s) you want to include in your Watson Analytics social media project.
• Watson Analytics for Social Media can collect from a number of media sources and you have the ability to determine which you want to include within your project.
• Once a Watson Analytics project is set up, you create a dataset with the query results by clicking Create data. Watson Analytics for Social Media will then provide an estimate of resources the project will consume and allow you to go back or continue. The results are then written to an Analysis tab where you can click on View Analysis to view the project's results in a variety of ways.
• Watson Analytics for Social Media automatically generates visualizations as part of the project analysis results and you can interact with those visualizations to do things such as filtering.
• Use Conversations Cluster visualizations to understand key terms that appear in social media posts about the topics you added as part of your project. This allows you to identify trends and shows you facets and insights that you may not have been thinking of back when you were defining your project topics and themes.
• When you click on the View Analysis icon, the default view starts with Conversation Clusters already selected for you. At any time, you can click on the selected tab, then click to navigate to any of the project analysis result types grouped under What, Where and Who.
• You can use the Topics visualization to find trends and share of voice for the topics that you defined. To view Topics, you can click on Conversation Clusters and then select Topics (as shown here):
• You can use the Sentiment visualization to understand the tone of the social media content found during the analysis. Sentiment is broken down by the topics that we specified when defining our project. Relative sentiment shows the dispersal of positive, negative, neutral, and ambivalent sentiment. Sentiment terms are words that measure the tone of a mention. Sentiment indicates whether a mention is positive or negative. A mention is categorized as ambivalent when it has the same number of positive and negative sentiment terms. A mention is categorized as neutral when no sentiment terms are detected in it.
• Sentiment terms are words that measure the overall tone of a mention. Sentiment indicates whether a mention is positive or negative. Sentiment terms visualization can be used to refine a sentiment analysis dictionary. This page groups terms into two groups: Top positive and Top negative. A neat feature is the ability for you again to click on a particular term that you may find interesting or most relevant and see its references (shown in the right pane).
• You can use Geography visualization to inspect where social media documents have been posted.
  • Sources and Sites visualization can be used to compare the different sources of topics, then drill-down to see what sites had the most posts.
• The Watson Analytics for Social Media Influential Authors visualization shows influential authors by source type. Influential authors are those in social media who are actively talking about the topics that we are interested in.
• The Watson Analytics for Social Media Author Interests visualization allows you to look a bit closer at the interest of content authors. The visualization starts with the number of authors broken down by topics.
• Watson Analytics for Social Media analyzes text found in social media content in an attempt to determine the behavior of the authors. The Behavior visualization shows the number of mentions by authors from defined behavior categories.
• Behavior categories include: authors that are users of a topic, authors that are prospective users of a topic, and authors that are churners of a topic.
• The Demographics Visualization analyzes project mentions by gender, marital status, and parental status automatically.
• Sentiment specifies whether a mention is positive or negative. IBM Watson Analytics for Social Media provides you with a default dictionary of sentiment terms. A sentiment term can consist of more than one word and is manageable.
• When you use Watson Analytics for Social Media and create a project, a new Social Media dataset is created for you. After the dataset is created, you can find it on the Data page in Watson Analytics.