This post is aimed at social sciences researchers who want to capture and analyze social media data, such as from Twitter, and it provides a useful collection of resources related to methods and practical tools which can be used for this purpose.
A lot has changed since I published my 2015 edition of this post on the LSE Impact of Social Science blog, with even more software applications with the function of retrieving and analyzing social media data having been released. Additionally, a number of social listening tools have continued to gain popularity among digital marketers looking to gain insight from consumers.
There remains a number of different methods of analyzing social media data. Take text analytics, for example, which can include using sentiment analysis to place bulk social media posts into categories of a particular topic, such as positive, negative, or neutral. Or machine learning, which can automatically assign social media posts to a number of different topics.
There are other methods such as social network analysis, which examines online communities and the relationships between them. A number of qualitative methodologies also exist, such as content analysis and thematic analysis, which can be used to manually label social media posts.
In industry, there has been much focus on gaining insight into users’ personalities, through services such as IBM Watson’s Personality Insights service, for instance. This uses linguistic analytics to derive intrinsic personality insights, such as emotions like anxiety, self-consciousness, and depression. This information can then be used by marketers to target certain products; for example, anti-anxiety medication to users who fit the personality characteristic of being anxious. A list of personality models can be seen here.
Computational methods can often save time for researchers dealing with large datasets or looking to combine efforts; i.e. humans and machines working together to tackle and analyse data. I would highly recommend reading the following paper, “Social media analytics: a survey of techniques, tools and platforms” (Batrinca and Treleaven, 2015), which provides an overview of some of the methods that can be used to analyse social media data.
In both academia and industry there has been a shift towards research projects and research questions which require more than the simple retrieval of data. More complex questions are being asked which require access to more metadata. So a number of tools have started to increase their data retrieval functionalities for the number of data points that can be retrieved.
For my PhD work I reviewed many methods and opted to use a number of computational techniques to locate and eliminate duplicate and near-duplicate tweets to reduce the volume of data I was working with. I used DiscoverText to do this. I then applied the methodology of thematic analysis, which involved reading through thousands of tweets in order to generate nodes and themes from them. Read more on my approach here.
Popularity of Twitter
The popularity of using Twitter for social media research, both in academia and in industry, remains high; no other platform has attracted as much attention from academics. However, Twitter is not the most popular platform in terms of monthly active users, being ranked at eighth in the overall list (see Figure 1). Facebook and WhatsApp are the top two. However, many of the platforms with the highest number of monthly active users do not make their data available on a similar scale to Twitter.
It can be argued that there is no other social media platform with an infrastructure like Twitter. Twitter is unique in the sense that it has an infrastructure which allows any user to be able to follow another user, and it provides almost 100% of its data through its APIs. With such a large number of monthly active users, Twitter is likely to remain popular for social media and industry research.
Developments on ethics and training
When I originally published my 2015 post, I received a number of queries and concerns related to the ethical implications of using social media data. In 2016, the Academy of Social Sciences and the NSMNSS network held an event solely focused on social media research ethics (you can read more about it here). You can also read a follow-up post of mine related to social media research ethics, and view a recorded conference presentation in which I discuss the ethical challenges of my project.
After the previous post was published, issues and concerns were raised over social scientists potentially lacking the skills to analyse social media data. It has been nice to see a number of training events have been held in order to upskill social scientists. For example, the Social Research Associationand the NSMNSS network held an event which provided an introduction to social media tools (more about that here).
So, what are some of the tools available to social scientists looking to retrieve and analyse social media data? In the table below I provide an overview of some the tools that require no prior technical and/or programming skills, updated and expanded for 2017:
An overview of tools
*It is always best to check with the developers of tools as there may be additional platforms that they can access. Moreover, some tools provide users with the ability of importing data into the applications from external sources.
A number of the tools provided in the table have been tested and used by me over a number of years, and the vast majority of these chiefly handle data from Twitter. It would be nice to have academic and social listening tools to retrieve data from other social media platforms, such as Facebook, Instagram, and Amazon, and also dark social media platforms such as WhatsApp. However, this may not be possible because these applications are not likely to provide all of their data to developers as Twitter does. Moreover, there may be ethical implications of accessing data from dark social media platforms.
Other applications are available but these require programming knowledge and/or were not tested as part of this post. These include:
Moreover, there are a number of advanced data analysis and statistical applications which can be used to analyse social media data, such as:
These packages should be researched when deciding which application is to be used for a project. I’d also like to mention The Digital Methods Initiatives list of tools, and Ryerson University’s list of tools from its Social Media Lab.
In future, we should begin to ask questions regarding the types of research made possible by using tools that do not require end users to hold technical knowledge. Moreover, we should seek to better understand the types of questions more technical tools can address. Consequently, developers of tools should seek to liaise with social scientists at the development phase, to allow for the possibility of new features based on social sciences research questions.
Phillip Brooker, research associate at the University of Bath, has noted that it is important to understand how software packages work in order for researchers to better inform their research practices. I highly recommend reading Phillip’s entry on the NSMNSS blog, about the Programming as Social Science (PaSS) network he has helped to establish.