By Panayiota Skouroumouni, MSc in Management and Organisational Innovation, Queen Mary University of London

For non-profit organisations, Twitter can be a powerful tool for network building and online engagement with key audiences. Since most non-profits operate with minimum resources, it is important that they have the necessary capacity to monitor engagement and make the most our of their social media presence.

As part of an MSc dissertation in Management and Organisational Innovation at Queen Mary University of London, Chorus tools were used to capture Twitter data from non-profits that focus on humanitarian campaigns related to children. With the help of Tweetcatcher, one years worth of data from July 2013 to July 2014 were captured from a selection of organisational accounts – the total number of tweets collected was 15,549.

Tweetcatcher was used for capturing tweets by children-related non-profits

Tweetcatcher was used for capturing tweets posted by children-related non-profit organisations

Tweetvis was then used to understand the patterns of users’ activity through examining structural characteristics of the tweets (e.g. hashtags, retweets, mentions), as well as keyword frequency and sentiment around specific events. Identifying the most frequent words was important as an overview of the most important topics discussed. Hashtags revealed how children-related campaigns were linked to other news and worldwide events.

The study’s main findings how non-profit organisations are trying to build their network through online relationships in several ways, for example:

  • Creating and heavily promoting tweets for causes that tend to receive a lot of attention (mentions, retweets).
  • Mentioning celebrities in their tweets, who receive a lot of publicity and can lead others to engage with their causes.
  • Mentioning/retweeting followers and supporters to involving them more actively in the organisations’ campaigns and possibly expand their influence through further networking effects. This can possibly enhance feeling of belongingness and connectedness among the audience, hence leading to more permanent connections and future action. A common reason of mentioning followers was acknowledgement of support.
  • Creating and promoting hashtags – Each hashtag usually represents an event, which forms a community of users with the same interests, involved in the same conversation.
  • Joining world-wide conversations about major events that attract the attention of Twitter users – During the time of this study, these included: the 3rd anniversary of conflict in Syria, the escalating war in Gaza, the abduction of schoolgirls in Nigeria, the typhoon in Philippines and the World Cup of June 2014 taking place in Brazil. A more advanced tactic used to attract attention to was “newsjacking”, where campaign-related content is inserted into contemporary media coverage like popular or breaking news. For instance, this infographic was created by the charity Save the Children during the London tube strike and received a lot of attention as well as controversy. 

Tweetvis 1.8.5 release

November 6th, 2014 | Posted by ch_admin in Uncategorized - (0 Comments)

This new release contains some small but useful updates, mainly to do with the Cluster Explorer, including:

  • A new function for node scaling creates a more balanced, legible view.
  • Edges between nodes are now coloured grey to create a less cluttered/dense image without loss of information.
  • The word map now offers finer control over the label filter and also the facility to manually increase/decrease the size of nodes.
  • The menu bar has been reorganised such that Main Timeline functions come first, followed by the Term Timeline.
  • The user now has has a choice of visual encodings by which to represent the term timeline (accessed from the menu: Term Time Line -> Cell Style.
  • Improved semantic clustering of nodes in the interval and tweet cluster views

 

TCDE 1.2.5 (beta) release

October 22nd, 2014 | Posted by ch_admin in Uncategorized - (0 Comments)

After a long silence, we are proud to announce a new update to TCDE (Tweetcatcher). You can download the new executable file from here. This latest release contains numerous bug fixes and new features:

  • Fixed the problem where Search no longer works after previous retrieval terminated at historical limit.
  • Time to next quota reset (shown in status bar) is now shown in local (correct) time.
  • Re-tweets can now be automatically removed via the Actions menu.
  • The user list in User Timelines can now be populated automatically with followers/friends of a specific user (see User List menu). These functions output a more comprehensive user table (including friend/follower/status counts, location, language etc.) to the clipboard which can be processed in Excel to produce filtered lists that can be pasted back into TCD for timeline download.
  • Failed requests, which can occur on high volume sessions and used to cause the download to end prematurely, can now be seamlessly re-sent. This means more data can be collected on trending topics and gaps in the timeline should now be a thing of the past.
  • There is now an option to set Search language (defaults to English – en) in the Settings menu. You need to specify your preferred language using the ISO-639-1 code (e.g. en for English or fr for French).
  • Cancelling a download is no longer handled by the button (which wasted space on the UI), but via the Actions menu.

This is currently a beta release and so may still contain minor bugs, but all the main functionality has been tested successfully. Please let me know if you find a problem.

This may possibly be the last significant version 1 release, although there may be another minor update at some point. We are currently planning a new version of Chorus that will provide better integration between data collection and analytics, many new and powerful features including unsupervised/scheduled data retrieval and demographic profiling, sampling and analysis of users and status content. Please feel free to send us your ideas: team@chorusanalytics.co.uk.

 

A collaborative and evaluative study conducted by members of the Chorus Team at Brunel University and social media representatives from a leading bio-pharmaceutical company who distribute blood glucose monitors (BGM) as part of their treatment portfolio for people living with diabetes. This study takes data from Twitter to attempt to answer questions of genuine interest to this sector of the medical device market, conducting a small-scale (one day) analysis in order to ascertain any valuable insight may quickly and easily be drawn from the data collected. Data was collected, using the Tweetcatcher, over a 10 week period during 2013 and all analyses were conducted using Tweetvis.

Research Interests

  1. The company had a concern with breaking into a BGM market dominated by established companies and devices, and are interested in exploring various questions around this. One question was “Is it possible to locate opinions, on Twitter, around aspects of existing devices which users think are positive/negative?”. Following this, they were also interested in how device design influenced daily usage: “Is there any discussion of the possible impacts that devices themselves might have on users’ adherence to testing schedules, especially where this information can be tied into specific brands of device?”
  2. The company had a secondary concern with how devices are represented by clinicians and other staff in clinical contexts, in terms of whether any particular brands are favoured, and what work might be done to increase brand uptake in specific ways. One key question was “Is there any discussion, on Twitter, of barriers to the uptake of certain devices on a clinical level, for e.g. advice and recommendations from GPs or nurses, etc?”. Related to this, another question was “Do devices users follow clinical advice in this regard?”

Brief Summary of Findings and Recommendations

Twitter proved to be a surprisingly rich source of discourse around the topic of blood glucose monitoring and management. Whilst constructing effective keyword queries proved difficult – acceptably precise queries tended to provide low recall and were dominated by spam – it was still possible to draw out some useful insights relevant to the questions.

Keyword search

Some interesting themes that were quickly identified within query keyword datasets (i.e. “strips AND glucose”, “glucose AND tests” and “glucose AND (device OR monitor OR meter)”) included:

  • Clear expressions of interest in blood glucose monitors that could integrate with existing commonly-used high-uptake technologies, particularly smartphones.
Dario-Video

Frequent mentions of smartphones were found in the “glucose AND (device OR monitor OR meter)” dataset

The perceived benefits were that this would remove the burden of carrying an extra device whilst providing the facility to readily aggregate and monitor glucose levels over time. There was also much excitement over the rumours that Apple were going to release an iWatch, complete with sensors to monitor glucose continuously without the need for needles

iWatch

Discourse within the “glucose AND (device OR monitor OR meter)” dataset was dominated by references to an article about the iWatch

Also:

  • Engaging with the blood glucose device market with a digitally-savvy approach (i.e. demonstrative YouTube tutorials on device operation, answering user questions via a company Twitter account, etc) could provide a cheap and easy way to increase brand/device awareness.
  • Reliability of devices and accuracy of readings (for both strips and devices) is the key concern of device users – in this regard, users desire the automation and hands-off nature of continuous blood glucose monitoring (and as such it may be profitable to develop devices to reflect continuous monitoring techniques).
  • Losing/forgetting devices is a key issue for device users, who report instances when they need to check their blood glucose levels but cannot do so due to having lost the device or left it at home etc. Hence, the ‘loseability’ of devices is a key aspect of peoples’ insecurities about relying on them, and finding ways to make devices less ‘loseable’ or mitigate the consequences of losing a device may improve how those devices are thought of by users.

 

User Following

A more successful strategy was to retrieve the recent timelines of users who followed a popular information service, in this case an account run by a diabetes safety campaign – @StripSafely. We call this strategy “User Following”. User following avoids the need to specify keywords. This is particularly useful for anthropological research – why would a diabetes sufferer mention their condition or related activities specifically every time they tweeted about it? The trade for improved recall is precision, but with sufficient numbers of users in the sample, the key themes tend to bubble to the surface and can be readily identified in the visualisations and statistics provided by Chorus, enabling the analyst to filter their dataset to pursue a more in-depth analysis. This approach revealed the following insights:

  • Followers of @StripSafely were found to participate in a regular weekly diabetes discussion  organised around the hashtag #dsma. This is an event which occurs every Thursday at 9PM EST, and this provides a focal point for any social media engagement that the company may wish to pursue (i.e. there are more people on Twitter discussing various issues in and around diabetes at this time, so the potential audience for blood glucose monitoring devices is maximal at this time).
dsma_on_timeline

A regular peak in activity around Thursdays is evident in the Tweetvis Timeline view

The @StripSafely dataset (n = 182809) was filtered down within Excel to only include tweets containing the #dsma tag. This reduced the set size to n = 11880. From this reduced set we found:

  • People engaging in the #dsma debate regularly link specific device issues with specific device brands, and very clearly state a wide array of opinions on what they do and do not like about existing devices. In this sense, it is possible to use social media and user-following data to capture many different positive experiences of devices (and possibly capitalise on them in future device development) as well as negative ones (and work to rectify those in devices endorsed by the company).
  • User-following datasets also allow for explorations of device recommendations and advice coming from a clinical context, and also display their adherence to that device post-consultation (i.e. their clinicians may recommend a specific device, but they may report preferring or even having bought a different one). These insights are locatable through usages of general occupational terms such as “nurse”, “doc”, “doctor”, “endo” etc, which are difficult to handle in a systematic way as part of a query keyword search.

This post is only an abridged version of the method and findings of the above study. If interested, please direct any enquiries to team@chorusanalytics.co.uk

The power and promise of social media as a resource and tool for doing social research is widely recognised and much vaunted. Social media data is becoming an increasingly attractive resource for social scientists, but the question remains as to what exactly we might want to do with data like this. The present study describes a small-scale interdisciplinary project in medical sociology which instigated the development of an innovative method for making practical use of ‘big data’ drawn from Twitter. What results is a depiction of how a collaboration between software developers, requirements engineers and social scientists demonstrated a need for a new method of data capture, a description of the method by which that need was addressed, and a discussion of the value of the insights that can be drawn through using that method.

Brooker, P., Barnett, J., Cribbin, T., Lang, A., & Martin, J. (2013). User-Driven Data Capture: Locating and Analysing Twitter Conversation about Cystic Fibrosis without Keywords. In SAGE Research Methods Cases. London, United Kingdom: SAGE Publications, Ltd. doi: http://dx.doi.org/10.4135/978144627305014526813

Twitter is known for the abundance of racialized messages posted on its platform. Eruptions of racist abuse occur within a contested array of Twitter discourses, e.g. racial banter, ambivalent humour, hate comments and (anti-)racist sentiments. Surprisingly, only limited studies of everyday ‘race-talk’ on Twitter have been undertaken, and little is known about its ambient stream of racialized expression.

This project explored how racialised messages unfold in Twitter by focusing on the hashtag #notracist. Twitter users can include this (and other related) hashtags in messages to label seemingly ‘racist’ statements (and images/videos) as ‘not racist’. The practice of hashtagging as a strategy of denying racist expression or propagating the ambiguities of race talk enables an understanding of the contested racialized digital ecology of Twitter.

The project is funded by British Academy/Leverhulme Small Research Grant (Apr 2013 – May 2014).

 

Multi-Hashtagging

A significant tweeting practice we note in the #notracist dataset is the usage of more than one hashtag in a tweet – a phenomena we call Multi-Hashtagging.1) For the #notracist dataset, aside from the original#notracist term, there are a further 7717 hashtags, which are used in a variety of ways, for example:2)

  • helen_louise_: I literally cant stop eating watermelon. & Im not even black. #NotRacist #JustSaying
  • PaneKilla: How to say the alphabet in vietnamese #funny #notracist #accent #alphabet #vietnamese#peace #lol http://instagram.com/p/

 

Analytic Work

We used Chorus to plot a ‘cluster map’ of key multi-hashtags (fig. 1). The Chorus software suite includes a Cluster Explorer feature, which enables the co-occurrence of any terms within a Tweet to be identified.

Fig. 1: Hashtag cluster map

In this map, each node (i.e. point in the map) is a hashtag, and the position and connectivity of nodes is determined by how frequently those hashtags are used together (co-occurrence), such that ‘similar’ hashtags cluster together (i.e. ones which are used in comparable ways to express similar sentiments). The model reveals how hashtags relate to each other as ‘semantic’ entities. The two radials overlaid on the image facilitate the process of reading the visualisation.

Source: http://www.darkmatter101.org/wiki/notracist_twitter