- Understand your audience
- Research methods
- Overview of Research Methods
- Social media research
Social media research
What to do with the data?
There are three main approaches to analyse social media data: trend analysis, content analysis and social networks analysis.
A wide number of social media data collection tools – both open source and commercial – is now available to gain a first general impression of particular social media channels. These tools usually provide quantitative data, usually displayed in a graphical or tabular way, focused on the keywords used in the search criteria and their associations. Most tools sort the quantitative data by location and channel (e.g., Twitter or blog), providing the opportunity for comparisons. For instance, it is possible to compare the prevalence of a certain topic between different countries and channels (see case study). Some tools apply text mining and semantic technology to the data, providing additional insights. Popular and common examples of this are sentiment analysis and word clusters analysis.
Sentiment analysis refers to a popular text-mining technique that takes written text and translates it into evaluative information, that is, determines if a specific conversation is positive, neutral or negative. Sentiment analysis can be useful to track and evaluate the overall impact of communications, campaigns or news about a specific event or topic, and/or to determine the themes and topics that are driving both good and bad conversations. However, the use of sentiment analysis as a stand-alone method to evaluate conversations is incomplete, because the tools analytic accuracy is still under development. Thus, they do not provide a context to help understand the deeper meaning of the evaluative judgment (e.g., does it reflect the users emotion, attitude or behaviour, does it refer to the topic of interest or to a related topic).
Word cluster analysis
Word cluster analysis refers to a set of new visualization techniques based on text-mining, such as word cloud or tag cloud, that help translate and summarize vast quantities of text into recurrent topics and themes. This type of analysis can be useful to have a general overview of the different topics and themes that emerge associated with a specific event or issue.
When to use trend analysis?
Trend analysis seems particularly suited when there is a need to gain a quick and general perspective about the relative popularity of the event and issue under study in different locations and/or channels, what type of associated themes and topics emerge in conversations, and the users’ feelings
Although reporting analysis performed by social media tools can be very valuable by providing a first general impression of what is happening in social media regarding a specific event or topic, it is not suitable to gain a deeper insight into the needs, concerns, opinions, behaviours or trends of consumers. This is mainly because these tools do not provide any context for the summarized data being presented, making it difficult the correct interpretation of data.
Content analysis is a systematic and objective technique that provides a quantitative description of observations, in this case, of conversations and messages in social media. This approach encompasses rigorous procedures that emphasize the reliability and replicability of observations and subsequent interpretations. By using content analysis it is possible to gain a deeper insight into what is being communicated in social media, given that the unit of analysis is usually at the individual level (i.e., message and user) and takes into consideration all the information included in the message. Given this thorough process of analysis, this approach can be time-consuming. However, computer-assisted approaches to content analysis can be used to reduce the time and cost required by traditional techniques.
Content analysis can include psychosocial analysis. This refers to an event-centered quantitative and qualitative analysis of written text, which takes into consideration social and psychological dimensions. These can be in the form of perceptions, beliefs, expressions of coping, and spontaneous reactions vs. deliberation. This analysis can complement sentiment analysis and go beyond its limitations, by providing assess to meaning at the individual and social level.
Generally, the structuring process of content analysis of social media data follows three main steps:
Identification of the categories of analysis
This first step involves deciding which themes are recurrent and give meaning to what is being communicated in social media data. This decision can follow two approaches: data-driven and/or a theory-driven approach.
In a data-driven approach, the themes or categories are selected based on a detailed analysis of all data. An example of this approach would be looking at all tweets referring a health risk (e.g., EHEC) and identify the recurrent themes that summarize the meaning conveyed by them. This approach is particularly suited when there is little knowledge about the topics and themes that may come up in the messages or when the goal is to make an in-depth exploration of the data.
In a theory-driven approach, the themes or categories selected are predetermined by an existing theory. Thus, in this approach it is not strictly necessary to go through all data in order to select the categories, making it less time-consuming than the data-driven approach. The theory-driven approach is particularly suited when there is already knowledge and a conceptual organization of the themes that should be analysed in the messages or when the goal is to test a theory.
In both approaches, the end product of this step should be a checklist or coding system for the organization of the data.
Organizing the content into categories of analysis
This step involves organizing the data in a way that ensures reliability and meaningfulness. The previously defined themes are used to classify the social media messages content into meaningful and explicative categories. Thus, this step requires the execution of an explicit set of recording instructions about the rules for coding the data into categories. Recording should involve more than one judge so that the coding of each message can be examined for reliability, and sources of disagreement can be identified and corrected. Reliability of the coding system can also be evaluated through computation of coefficients of agreement between two or more different judges/coders (e.g., Kappa and Pi).
Analysis and interpretation
Once all data is organized, qualitative (content) and quantitative analysis (e.g., prevalence) can be performed and followed by an interpretation of the results.
When to use content analysis?
Content analysis seems particularly suited when it is important to gain a deeper insight of what is being communicated in social media and what it reveals about the users attitudes, opinions, concerns, needs and behaviour.
Social networks analysis
Social network analysis uses the communication of social media users on particular topics to determine connections between them, providing a graphical representation of the relationships between the different users/profiles. This representation of the users social networks can be very complex, but advanced networking techniques coupled with the use of web analytics (e.g., centrality index) can be used. This can help for example to identify the users that act as leaders or influencers and the users that act as followers, as well as determining the relative strength of particular topic leader(s). In social networks, leaders are those users that are at the center of most of the activity for a topic, i.e., promoting discussion (e.g., comments to their messages).
When to use social network analysis?
Social network analysis seems particularly suited to determine how information flows between social media users and to identify which users are the most influential and act as sources of information and discussion on a specific topic.