̽��ֱ�� of Cambridge - twitter

Slamming political rivals may be the most effective way to go viral

fpjl2 — Tue, 22 Jun 2021 08:02:51 +0000

Study of almost 3 million Facebook and Twitter posts from US media and politicians shows divisive posts dunking on opponents drive engagement on social media.

Celebrity Twitter accounts display ‘bot-like’ behaviour

sc604 — Tue, 01 Aug 2017 23:01:58 +0000

̽��ֱ��researchers, from the ̽��ֱ�� of Cambridge, used data from Twitter to determine whether bots can be accurately detected, how bots behave, and how they impact Twitter activity.

They divided accounts into categories based on total number of followers, and found that accounts with more than 10 million followers tend to retweet at similar rates to bots. In accounts with fewer followers however, bots tend to retweet far more than humans. These celebrity-level accounts also tweet at roughly the same pace as bots with similar follower numbers, whereas in smaller accounts, bots tweet far more than humans. Their results will be presented at the IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM) in Sydney, Australia.

Bots, like people, can be malicious or benign. ̽��ֱ��term ‘bot’ is often associated with spam, offensive content or political infiltration, but many of the most reputable organisations in the world also rely on bots for their social media channels. For example, major news organisations, such as CNN or the BBC, who produce hundreds of pieces of content daily, rely on automation to share the news in the most efficient way. These accounts, while classified as bots, are seen by users as trustworthy sources of information.

“A Twitter user can be a human and still be a spammer, and an account can be operated by a bot and still be benign,” said Zafar Gilani, a PhD student at Cambridge’s Computer Laboratory, who led the research. “We’re interested in seeing how effectively we can detect automated accounts and what effects they have.”

Bots have been on Twitter for the majority of the social network’s existence – it’s been estimated that anywhere between 40 and 60% of all Twitter accounts are bots. Some bots have tens of millions of followers, although the vast majority have less than a thousand – human accounts have a similar distribution.

In order to reliably detect bots, the researchers first used the online tool BotOrNot (since renamed BotOMeter), which is one of the only available online bot detection tools. However, their initial results showed high levels of inaccuracy. BotOrNot showed low precision in detecting bots that had bot-like characteristics in their account name, profile info, content tweeting frequency and especially redirection to external sources. Gilani and his colleagues then decided to take a manual approach to bot detection.

Four undergraduate students were recruited to manually inspect accounts and determine whether they were bots. This was done using a tool that automatically presented Twitter profiles, and allowed the students to classify the profile and make notes. Each account was collectively reviewed before a final decision was reached.

In order to determine whether an account was a bot (or not), the students looked at different characteristics of each account. These included the account creation date, average tweet frequency, content posted, account description, whether the user replies to tweets, likes or favourites received and the follower to friend ratio. A total of 3,535 accounts were analysed: 1,525 were classified as bots and 2010 as humans.

̽��ֱ��students showed very high levels of agreement on whether individual accounts were bots. However, they showed significantly lower levels of agreement with the BotOrNot tool.

̽��ֱ��bot detection algorithm they subsequently developed achieved roughly 86% accuracy in detecting bots on Twitter. ̽��ֱ��algorithm uses a type of classifier known as Random Forests, which uses 21 different features to detect bots, and the classifier itself is trained by the original dataset annotated by the human annotators.

̽��ֱ��researchers found that bot accounts differ from humans in several key ways. Overall, bot accounts generate more tweets than human accounts. They also retweet far more often, and redirect users to external websites far more frequently than human users. ̽��ֱ��only exception to this was in accounts with more than 10 million followers, where bots and humans showed far more similarity in terms of the volume of tweets and retweets.

“We think this is probably because bots aren’t that good at creating original Twitter content, so they rely a lot more on retweets and redirecting followers to external websites,” said Gilani. “While bots are getting more sophisticated all the time, they’re still pretty bad at one-on-one Twitter conversations, for instance – most of the time, a conversation with a bot will be mostly gibberish.”

Despite the sheer volume of Tweets produced by bots, humans still have better quality and more engaging tweets – tweets by human accounts receive on average 19 times more likes and 10 times more retweets than tweets by bot accounts. Bots also spend less time liking other users’ tweets.

“Many people tend to think that bots are nefarious or evil, but that’s not true,” said Gilani. “They can be anything, just like a person. Some of them aren’t exactly legal or moral, but many of them are completely harmless. What I’m doing next is modelling the social cost of these bots – how are they changing the nature and quality of conversations online? What is clear though, is that bots are here to stay.”

References:
Zafar Gilani, Reza Farahbakhsh, Gareth Tyson, Liang Wang, Jon Crowcroft. Of Bots and Humans (on Twitter). Paper presented at 9th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM'17). Sydney, New South Wales, Australia.

Zafar Gilani, Ekaterina Kochmar, Jon Crowcroft. Classification of Twitter Accounts into Automated Agents and Human Users. Paper presented at 9th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM'17). Sydney, New South Wales, Australia.

‘Celebrity’ Twitter accounts – those with more than 10 million followers – display more bot-like behaviour than users with fewer followers, according to new research.

A Twitter user can be a human and still be a spammer, and an account can be operated by a bot and still be benign.

Zafar Gilani

Esther Vargas

Twitter

̽��ֱ��text in this work is licensed under a Creative Commons Attribution 4.0 International License. For image use please see separate credits above.

Yes

Licence type:

Attribution-ShareAlike

Welsh Twitter: capturing language change in real time

lw355 — Wed, 29 May 2013 07:50:07 +0000

Twitter keeps millions of people in touch, whether it’s sharing their politics with followers or updating their mates with the trivia of everyday life. These tweets are in Welsh: ‘loaaaads o gwaith i neud a di’r laptop ’cau gwithio!’, ‘dio cau dod on!! Mar bwtwm di tori.’ Roughly translated, they read: ‘loads of work to do and the laptop won’t work’ and ‘it won’t come on!! ̽��ֱ��button’s broke.’

How do you capture changes as they take place in the language we use in everyday life – from buzz words such as ‘sweet’ to tags such as ‘innit’? One answer is to look at tweets. Because they don’t follow the conventions of written language, tweets provide an authentic snapshot of the spoken language. By analysing the content of the 140-character messages, linguists can get to grips with the dynamics of the language played out in real time.

Welsh is spoken by 562,000 people in Wales; 8% of the country’s children learn it at home as their first language and 22% are educated in Welsh.

Like all living languages, Welsh is constantly changing and new varieties are emerging. When Dr David Willis from Cambridge’s Department of Theoretical and Applied Linguistics set out to research the shifts taking place in Welsh, he used a database of Welsh tweets as a means of identifying aspects of the language that were changing, and then used that information to devise the questionnaires used for oral interviews.

He explained: “When your intention is to capture everyday usage, one of the greatest challenges is to develop questions that don’t lead the respondent towards a particular answer but give you answers that provide the material you need.”

“If I want to find out whether a particular construction is emerging, and where the people who use it come from, I would normally have to conduct a time-consuming pilot study, but with Twitter I can get a rough and ready answer in 30 minutes as people tweet much as they speak,” he said. “My focus is on the syntax of language – the structure or grammar of sentences – and my long-term aim is to produce a syntactic atlas of Welsh dialects that will add to our understanding of current usage of the language and the multi-stranded influences on it. To do this relies on gathering spoken material from different sectors of the Welsh-speaking population to make comparisons across time and space.”

In the late 17th century, the antiquarian Edward Lhuyd conducted an investigation into the dialects of Wales. By the 19th century, Welsh was attracting the attention of European historical linguists such as Johann Kaspar Zeuss. Later, scholars all over Europe, realising that local dialects were receding in the face of industrialisation, sought to record variations in language. Large dialect atlases were undertaken in Germany and France, and speech archives were begun, such as the one that laid the foundations for the National History Museum at St Fagan’s near Cardiff.

In the 1960s the attention moved away from rural areas to the cities where most people by then lived – and researchers started to look at sentence structure, an area of language that presents particular challenges for investigators. Willis’s interest in syntax stemmed from his study of a wide range of minority languages, including Breton, which is, like Welsh, a Celtic language. To create the biggest possible picture of syntactic changes in Welsh as it’s spoken today, he decided to take an inclusive approach and set out to investigate day-to-day speech patterns of a broad range of speakers, aged 18–80.

British Academy funding for a year-long study has enabled Willis and assistant researchers to interview around 160 people across Wales, beginning his analysis with North Wales where the language is thriving and a significant number of children use Welsh as their home language. ̽��ֱ��study included both those who had acquired Welsh at home and at school.

̽��ֱ��spoken questionnaire asked interviewees to repeat in their own words sentences that were presented to them in deliberately ‘odd’ Welsh that mixed different dialects, inviting the interviewee to rephrase the awkwardly phrased sentence to sound more ‘natural’. An example in English might be ‘we’ve not to be there yet, don’t we?’ which a British speaker might be expected to rephrase as ‘we haven’t got to be there yet, have we?’

̽��ֱ��data from these interviews are a treasure trove of information in terms of the light their content can shine on how and why the structure of language shifts over time – and give the researcher a valuable database not just for the present study but also for future research.

Changes identified so far include use of pronouns and multiple negatives. An analysis of usage of the Welsh words for ‘anyone’, ‘someone’ and ‘no-one’ reveals that there are differences between those who learnt Welsh in the home (who are more likely to say the equivalent of ‘did someone come to the meeting?’ and ‘I didn’t see no-one’) and those who learnt it at school (who are more likely to say ‘did anyone come to the meeting?’ and ‘I didn’t see anyone’).

One example of multiple negatives reveals a shift in meaning of the Welsh word for refuse, ‘cau’. “We knew that people in the north used the word ‘cau’ to mean ‘won’t’, saying the equivalent of ‘the door refuses to open’ for ‘the door won’t open’. Negative concord – such as saying ‘I haven’t not seen no-one’ for ‘I haven’t seen anyone’ – is a strong feature of Welsh. We’ve now identified two groups in the north: one that still says ‘the door refuses to open’ and the other that have begun to say ‘the door doesn’t refuse to open’. ̽��ֱ��next step is to work out when and how this change occurred.”

In tracking shifts in the language, GIS mapping is used to plot where interviewees were brought up and enables researchers to look at the geographical spread of particular aspects of syntax, making comparisons between age groups, gender and mode of acquisition.

̽��ֱ��research has revealed that, while Welsh does not vary much by social class, there are interesting differences between the variety of Welsh spoken by those who learn it as their first language in the home and that spoken by those who are first exposed to it in nursery or primary school.

“Those who acquire Welsh once they reach school are more likely to use English sentence constructions, which are perfectly good Welsh but differ significantly from the constructions used by those who acquired Welsh at home. For example, they tend to prefer standard focus particles – words that correspond to a strong stress in English sentences like ‘I know YOU’ll be on time’ – over the ones from their local dialect,” said Willis.

With around 22% of the Welsh population educated in Welsh at school, and all children learning it as a second language, data on this aspect of language acquisition may prove valuable in developing Welsh teaching policy – for example, in determining which forms to teach second-language learners or in promoting both dialect and standard written Welsh in schools.

Inset image credit: Howard Beaumont

A database of Welsh tweets is being used to identify the characteristics of an evolving language.

If I want to find out whether a particular construction is emerging, I would normally have to conduct a time-consuming pilot study, but with Twitter I can get a rough and ready answer in 30 minutes

David Willis

̽��ֱ��District

Welsh Twitter

This work is licensed under a Creative Commons Licence. If you use this content on your site please link back to this page.

Yes

Cambridge ranked ‘most visible’ on social networks

fpjl2 — Thu, 23 Feb 2012 15:20:28 +0000

New research by digital publisher Econsultancy into the visibility of universities digital content on some of the major social media sites, including Facebook and Twitter, has placed the ̽��ֱ�� of Cambridge top amongst the Russell Group.

̽��ֱ��ranking is based on the amount of direct links to ̽��ֱ�� websites that are shared, liked and tweeted by social media users over an eight week period across the end of last year and start of this one.

Econsultancy estimates that taken altogether there are roughly 207,900 links every week linking to content on Russell Group university websites, with Facebook being the major player for higher education representation on social sites - accounting for around 80% of all links.

Cambridge came top with a visibility score of 462, 823 - narrowly beating age-old rivals Oxford who came in second with 442, 758. In third was the London School of Economics with just over half Oxford’s score, with Newcastle ̽��ֱ�� in fourth.

Data from twenty Russell Group universities was analysed to compile the ranking. ̽��ֱ��main type of content shared were news stories about research or institutional initiatives, with much of the content targeted at the current university population - although the study suggests that this content provides interesting insights into life at a particular institution.

“ ̽��ֱ�� ̽��ֱ�� is making a concerted effort to use social media effectively. We do this by creating engaging web content through articles, video and imagery. We tailor some content to target a range of channels – Flickr, Twitter, YouTube and Vimeo, iTunesU and Facebook”, says Barney Brown, Digital Communications Manager for Cambridge.

“Our ‘Under the Microscope’ videos, bite-sized snapshots of Cambridge research, have recently produced our most watched film to date on YouTube – over a quarter of a million views in the first week. Our Flickr collaboration with the photographer ‘Sir Cam’ has resulted in well over a million views for the ̽��ֱ��’s photostream.”

“We use social media to open Cambridge up to the world - communicating its landscape and culture, education and world-class research. It is thrilling to see people engaging with one of the world’s oldest universities through some of the latest social technologies – debating and sharing Cambridge research and culture.”

“We believe that this approach, combined with the global reputation and history of an institution such as Cambridge, has led to the ̽��ֱ��’s significant social media visibility among UK universities, as this research shows.”

̽��ֱ��social media visibility study can be viewed here.

A ranking produced by Econsultancy has listed the ̽��ֱ��’s website as the most visible Russell Group website on some of the biggest social media networks.

Our ‘Under the Microscope’ videos, bite-sized snapshots of Cambridge research, have recently produced our most watched film to date on YouTube

Barney Brown

WebTreats ETC

108 High Resolution Dark Denim Social Media Icons

This work is licensed under a Creative Commons Licence. If you use this content on your site please link back to this page.

Yes

̽��ֱ��meaning of emoticons

ns480 — Thu, 13 Oct 2011 14:00:25 +0000

Alex Davies, a Gates scholar at the ̽��ֱ�� of Cambridge, has created a visual map of the words associated with Twitter emoticons. It's not just the usual smiley and sad face emoticons whose meaning is fairly obvious and associated with words such as birthday, weekend and Friday and hospital, cold, stomach and pain respectively.

Other emoticons include:

^_^ This is associated with more immediate pleasures such as food and holidays. Words linked to it include shopping, lunch, dinner and chocolate. It is more associated with Asian Tweeters, but has begun to be used more in the West.

<3 This emoticon, which looks like a heart on its side, is associated with words such as love, music, amazing, proud, beautiful, thankful, Jesus and Justin.

:/ This generally denotes a half-awake, slightly annoyed state, associated with words such as shift, sleeping, busy, class and Monday.

-_- This emoticon is indicative of frustration and is associated with words such as dumb, lame, dick and bitch.

Davies says all the emoticons he has analysed have been around for at least as long as the Internet has been in existence, but some are less prevalent than others. Some were used exclusively in Asia, but have now spread to the West, and being adopted by particular groups.

Davies, who is studying for a PhD in Engineering with a focus on statistical modelling, says: “ ̽��ֱ��creation of new emoticons has essentially stopped, but the context and usage of existing ones is constantly evolving. Take for example Asian style emoticons, such as ^_^ (happy) and -_- (sad).

“Initially these were used almost exclusively by Asian online communities, but have slowly been adopted by different Western sub-cultures and have taken on subtly different meanings in these contexts. One way to visualise this usage is to visualise the words that are strongly associated with these emoticons. What is interesting is that two emoticons that essentially represent the same sentiment, such as :) and ^_^, actually differ substantially in how they are used, and we can see this in the images of the words.”

Davies has also published a list of the happiness/sadness of 7,500 common words on Twitter after he was approached at an international conference about his previous work on creating a Twitter map of happiness. He was asked if he could release a list of words so people could easily create systems that use sentiment analysis of Twitter.

Davies says: “Twitter contains a wealth of sentiment information which researchers and businesses are very interested to explore so they can assess the changing global mood on different issues in real time and make predictions based on this.”

̽��ֱ��emoticons used on Twitter are a language in themselves and are taking on new and often surprising meanings of their own, according to new research.

Initially these were used almost exclusively by Asian online communities, but have slowly been adopted by different Western sub-cultures and have taken on subtly different meanings in these contexts.

Alex Davies

Candie_N from Flickr

smiley

This work is licensed under a Creative Commons Licence. If you use this content on your site please link back to this page.

Yes

Tweeting disasters

ns480 — Wed, 12 Oct 2011 09:52:45 +0000

Professor John Preston, who is based at the ̽��ֱ�� of East London’s Cass School of Education, will tell the ‘Violent Nature’ Research Councils UK debate that Twitter and Facebook have been credited with being able to pick up advance signals of disasters. However, it is only in retrospect that the significance of the signals can be ascertained.

̽��ֱ��debate focuses on whether governments, scientists and aid agencies can manage the risks of living in potentially lethal locations. Other speakers include Professor James Jackson from the ̽��ֱ�� of Cambridge, Daniel Walden, policy adviser for disaster risk reduction at Save the Children UK and Dr Andrew Collins, reader and director of the Disaster and Development Centre at Northumbria ̽��ֱ��. It will be chaired by James Randerson, the Guardian's science and environment editor.

Professor Preston is the leader of a two-year cross-disciplinary research programme, supported by the the Research Councils UK Global Uncertainties Programme, which began in 2010 and seeks to uncover how the likes of Twitter and Facebook could save lives in the event of a national crisis such as a terrorist attack or natural disaster.

He says that where Twitter in particular works well is in correcting information and countering false rumours. “There is an inherent self-correcting bias in Twitter which is like the scientific process. When someone posts it sifts the evidence for and against and the more current information countervails anything that came before,” says Professor Preston, whose book Disaster Education is out early next year.

He adds: “Social networks can be used for malicious reasons to spread rumours by targeting false information at a few super-connected people. Information spread this way would take longer to correct.”

Professor Preston says: “Part of the reason authorities are put off using social media to spread information during disasters is that it can appear quite uncontrollable since information sharing after disasters tends to be followed by a period of emotional reflection on what it means. Emotion is very important in social media. It's not just about information. People use it quite creatively which can make it a little bit uncontrollable.”

̽��ֱ��research programme is looking at how to prepare the UK better for disasters, through, for instance, cell broadcasting and community education, and is looking at lessons that can be learnt from the past.

̽��ֱ��Violent Nature debate will take place at the McCrum Lecture Theatre, Corpus Christi College, Cambridge from 8-9.30pm on 25^th October.

Social networks like Twitter cannot help prevent disasters, but can quickly correct misinformation resulting from false rumours preventing possible further loss of lives, a leading researcher will tell a public debate on 25th October at the Cambridge Festival of Ideas.

Emotion is very important in social media. It's not just about information. People use it quite creatively which can make it a little bit uncontrollable.

Professor John Preston

DVIDSHUB from Flickr

Search and rescue

This work is licensed under a Creative Commons Licence. If you use this content on your site please link back to this page.

Yes

Come here often?

sjr81 — Thu, 04 Aug 2011 09:30:58 +0000

To date, most social networking sites like Facebook and LinkedIn have relied upon the ‘friend-of-a-friend’ approach to try and determine which people may have connections with one another.

However, Salvatore Scellato, Anastasios Noulas and Cecilia Mascolo, of Cambridge’s Computer Laboratory, have devised a new approach that not only looks at friends of friends, but also the places people visit – with incremental weightings given to different places such as airports and gymnasiums.

Scellato said: “Essentially this is a way in which we can predict how people will make new friends. We know that we are likely to become friends with ‘friends of friends’, but what we find is there are specific places which foster the creation of new friendships and that they have specific characteristics.”

Historically, the problem facing social networks has been the sheer volume of users. While millions of users may represent good news from a business perspective, it means the task of recommending friends can become an exponentially difficult one, if, as in the case of Facebook, you have 750 million active users.

̽��ֱ��standard two-hop approach – sharing at least a common friend – has, to date, ignored the possibilities of recommending new friends based on the places where users ‘check-in’.

̽��ֱ��trio’s research is an extension of long-standing sociological theory that people who tend to frequent the same places may be similarly-minded individuals likely to form a connection with one another – but applied to social networking sites.

Added Scellato: “For our research we analysed the location-based social network Gowalla to see how users created social connections over a period of four months. We discovered that about 30 per cent of all new social links appear among users that check-in to the same places. Thus, these ‘place friends’ represent disconnected users becoming direct connections.

“By combining place friends with friends-of-friends, we can make the prediction space about 15 times smaller and yet, cover 66 per cent of new social ties.

“It turns out that the properties of the places we interact can determine how likely we are to develop social ties. Offices, gyms and schools are more likely to aid development rather than other places such as football stadiums or airports. In those places, it’s highly unlikely people will develop a social connection.

“Our results show it’s possible to improve the performance of link prediction systems on location-based services that can be employed to keep the users of social networks interested and engaged with that particular website.”

A new way of predicting which people may become friends on social networks - based on the type of places they visit - has been formulated by ̽��ֱ�� of Cambridge researchers.

It turns out that the properties of the places we interact can determine how likely we are to develop social ties.

Salvatore Scellato

MIDTN.com AKA Brent via Flickr Creative Commons

Ernest Tubb Record Shop sign

This work is licensed under a Creative Commons Licence. If you use this content on your site please link back to this page.

Yes