Audience Segmentation in Social Media

Understanding the social media audience is becoming increasingly important for social media analysis. This paper presents an approach that detects various audience attributes, including author location, demographics, behavior and interests. It works both for a variety of social media sources and for multiple languages. The approach has been implemented within IBM Watson Analytics for Social MediaTM and creates author profiles for more than 300 different analysis domains every day. 1 Understanding the Social Media Audience – Why Bother? The social media audience shows Who is talking about a company’s products and services in social media. This is increasingly important for various teams within an organization: Marketing: Does our latest social media campaign resonate more with men or with women? What are social media sites where actual users of our products congregate? What other interests do authors have that talk about our products, so we can create co-selling opportunities? Sales: Which people are disgruntled with our products or services and consider churning? Can we find social media authors interested in buying our type of product, so we can engage with them? Product management and product research: Which product features are important specifically for women, or parents? What aspects do actual users of our product highlight—and how does this compare to the competition’s product? Besides commercial scenarios, social media is becoming relevant for the social and political sciences to understand opinions and attitudes towards various topics. Audience insights are key to put these opinions into the right context. 2 Audience Segmentation with IBM Watson Analytics for Social Media IBM Watson Analytics for Social MediaTM (WASM) is a cloud-based social media analysis tool for line-of-business users from public relations, marketing or product management1. The tool is domain-independent: users configure topics of interest to analyze, e.g., products, brands, services or politics. Based on the user’s topics, WASM retrieves all relevant content from a variety of social media sources (Twitter, Facebook, blogs, forums, reviews, video comments and news) across multiple languages (currently Arabic, English, French, German, Italian, Portuguese and Spanish) and applies various natural language processing steps: • Dynamic topic modeling to spot ambiguous user topics, and suggest variants. • Aspect-oriented sentiment analysis. • Detection of spam and advertisements. • Audience segmentation, including author location, demographics, behavior, interests, and account types. While the system demo will show all steps, this paper focuses on audience segmentation. Audience segmentation relies on: (i) The author’s name and nick name. (ii) The author’s biography: a short, optional self-description of the author (see Figure 1), which often includes the author’s location. (iii) The author’s content: social media posts (Tweets, blog posts, forum entries, reviews) that contain topics configured by a WASM user. https://watson.analytics.ibmcloud.com