The Scotland Debate #10: The Future Role of the Queen.

by Nigel Filed under: Scotland

This is the tenth in our series on mentions of Scotland and Independence in Social media and online news.  As before, the posts collected have been classified as “pro”, “undec”, “anti”, and “news”, and the chart below tracks these through the week.

The chart shows that both pro-Independence and anti-Independence posts have been reduced this week, with “undec” and “news” taking a larger proportion of the volume.

In the “pro” category, there is a continuation of a series, with Liz Lochhead, the current Scottish Makar, or National Poet, giving her views via the @Celebs4Indy account, which was retweeted a number of times:

RT @Celebs4indy: LIZ LOCHHEAD ‘if there was a referendum tomorrow I’d vote yes-I don’t see why Scotland & England can’t remain friendly co-operative equals.[ http://twitter.com/DanSmatterings/statuses/200254164721344512]

My question is whether her sentiment adds up to a valid motivation for voting “Yes”.

Tweets like

RT @albagubrath_61: The Labour/Tory Govt’s did this to Scotland. #McCrone http://t.co/H8Ak0jjE and this http://t.co/S2C0ar3p Please vote Yes For Independence.[ http://twitter.com/ivanelrick/statuses/199428773232386050]

suggest that problems of Scotland are the fault of the English, but forget that Labour have been the majority party representing Scotland in Westminster for a long time.

In the “undec” category, the comedians and satirical twitter accounts were, once again, well represented, with:

Alex Salmond rang, asking if I’d be Chancellor for Scotland after independence. Trump & Murdoch both suggested me. Said I’ll think about it. [http://twitter.com/OsborneDrunk/statuses/201622605118193664]

And

100 Celtic fans were asked if Scotland gained independence should we change currency?98% said NO we should keep the giro [http://twitter.com/delb1978/statuses/201248758288236544]

Both of these were retweeted quite a bit.

As the title above may suggest, the news category this week was dominated by the news that the Scottish National Party intend to hold a referendum after independence on the future role of the Monarchy in Scotland:

THE Queen may not remain head of state in an independent Scotland after one of the Scottish National Party’s most senior MPs confirmed it is party policy to hold a referendum on the monarchy. Christine Grahame, the convener of the Scottish Parliament’s … [http://www.devonporttimes.com.au/news/world/world/general/queen-could-be-ousted-as-scotlands-head-of-state/2554343.aspx]

As with the quoted example, a large proportion of the posts on this subject were from Australian and New Zealand new outlets.

Another important topic was the future of Scotland’s historic regiments, with the SNP now claiming that the military would be safer under an independent government:

 RT @theSNP: Independence opportunity to save Scotland’s historic regimental tradition http://t.co/wKHPSsJx #IndyRef [http://twitter.com/cusackandrew/statuses/201761417173532673]

 

I would imagine that those people who have actively been posting pro and anti messages on social media in the past are reconsidering following the local elections, or are posting news articles and links rather than direct comment.

From now on we will produce this post on a fortnightly basis, as the referendum is going to be some way off.  We will be producing reports every few months, combining data across the period, and will let you know, here and on twitter, when these are available.

Scotland #9 – The Local Elections.

by Nigel Filed under: Scotland

We have continued our search across social media and online news for “Scotland” AND (“Independence” OR “Referendum”), and here present our analysis of the conversations found in the week from 30th April to 6th May, 2012.

There were local elections across the United Kingdom on Thursday May the 3rd, to elect local councillors, elected mayors, and, in ten cities, referendums on whether to have an elected mayor. As part of the campaign for these elections in Scotland, Alex Salmond and the Scottish National Party (SNP) attempted to claim that the local elections were an important step on the road to independence, and that a vote for the SNP on the 3rd would be a sign of support for independence.  This campaigning explains the large peak in “news” posts on the 2nd of May.  The fact that the party was largely unsuccessful – they did not take Glasgow council, their main target, and the only city they held on to was Dundee – accounts for a lot of the news posts on the 4th of May.

The lack of electoral success for the SNP lead also to the peak for “undec” on the 5th of May, with the comment

So, the SNP were going to control all of Scotland’s cities on the road to independence. They actually control……Dundee.

being retweeted frequently (as a comment frrom a memeber of the public, which does not demonstrate a voting intention, this was classified as “undec” or undecided).

The large peak in “undec” posts on the 30th of Aperil was accounted for by photographs of the opening of “Democracy for Scotland: The Referendum Experience” being posted by the Scottish Political Archive to Flickr on that date.   I would not wnat to try and guess which side of the debate the organisers of this exhibition, made up it seems of archive material from past and future referenda, take.

The pie chart below demonstreates the continuing low level of posts that are straighforwardly anti independence, when compare to the undecided and pro categories.

The table below presents the data for the week.

Tech pro undec anti news Total
MicroBlog 100 368 4 133 605
Friendfeed 2 9 0 5 16
Facebook 1 23 0 18 42
Twitter 97 336 4 110 547
Blog 0 6 0 8 14
News 0 10 0 241 251
WWW 5 103 1 288 397
Media 0 42 0 13 55
Total 105 529 5 683 1322
Percent 8% 40% 0% 52% 1

Scotland #8: The Debate continues.

by Nigel Filed under: Scotland,Uncategorized

This post covers social media posts returned by the search term (“Scotland” AND (“Independence” OR “Referendum”)) between 23rd April and 29th April.

The posts found were classified, using our human / machine system, into the following categories:

  • “pro” – pro Independence;
  • “undec” – undecided (mainly questions about outcomes or comments from the public on the process);
  • “anti” – anti independence or pro-Union;
  • “news” – news reports on the process.

 

The chart shows that, once again, the “news” category dominates the posts found, though during the latter part of Wednesday and the early part of Thursday there was a large peak for the “undec” category, largely made up of retweets of

RT @frankieboyle: Independence might cost the average Scot A?A?300. Doesn’t sound much but that’s money we need to get the fuck out of Scotland 2 weeks a year [http://twitter.com/zozozozozozoxox/statuses/196268101048864769]

The news this week was a combination of pessimism over the economic future of the country, with claims that an independent Scotland would be dependent on oil and gas, and statements from energy companies that they will not invest in Scottish renewables :

Renewable energy investors will not commit funds to projects in Scotland unless they have 100% guarantees of continuing subsidy support in the wake of Scottish independence, the Energy and Climate Change Committee (ECCC) was told today. [http://www.edie.net/news/news_story.asp?id=22322]

The viability of the economy of an independent Scotland was again called into question, this time by the ratings agency Fitch:

NEWSFLASH: Scotland won’t keep UK’s AAA rating if it gains independence, says Fitch http://t.co/jmkDLCaB [http://twitter.com/DarrenGershin/statuses/195074037016313856]

Questions were also asked about an independent Scotland’s position regarding Nato and The European Union, and employment and tax policies, though the major story of the week was the interview given by Michael Ignatieff, the ex-Liberal Party Leader in Canada, for BBC Scotland.

 Ignatieff, who left the Liberal leadership post after his party was drubbed in the 2011 election, suggested in an interview broadcast Monday on BBC about Scotland’s independence referendum that Quebec and the rest of Canada have little to say to each … [http://www.thestarphoenix.com/news/Ignatieff%20backtracks%20Quebec%20separation%20comments/6514101/story.html]

Ignatieff initially said that independence for both Scotland and Quebec was the natural and logical end result, but had to backtrack.  As both Quebec and Scotland were mentioned, this interview, and the retraction, got a considerable amount of coverage.

 

The data:

Tech pro undec anti news Total
MicroBlog 64 463 15 438 980
Friendfeed 0 4 3 18 25
Facebook 3 1 0 14 18
Twitter 61 458 12 406 937
Blog 0 2 0 18 20
News 2 13 0 268 283
WWW 1 31 28 383 443
Media 0 0 0 12 12
Total 67 509 43 1119 1738
Percent 4% 29% 2% 64% 100%

 

Churnbar Supermarket Tracker (and maybe others).

by Nigel Filed under: News

I’ve spent a lot of time recently working with social media posts about British supermarkets, and I’ve got a feeling that there’s a wealth of information in there which could be useful to anyone in the retail business, be they store managers, company directors, product marketers, or retail analysts.

So we have decided to go back to the social search and look for everything we can find mentioning Asda, Morrisons, Sainsburys, Tesco, and Waitrose, the big five (here in alphabetical order, rather than any temporary ranking according to market share or sales volume or whatever).  We’ll start this off today, and produce the first report from the data for release on Friday July 6th.  Further reports will be produced every two months from then on.

The reports will be available for purchase and immediate download (one of my tasks over the next two months will be to identify the best method of doing this), though we have not finalised pricing yet.

If the Supermarket tracker is successful, we will launch trackers along similar lines for other vertical sectors; if there is a sector that you are particularly interested in, please let me know. We are currently considering banking and financial services as interesting areas.

We will be setting up a landing page for the Supermarket Tracker in the next couple of days with more information and a sign-up form to register your interest, please watch out for it.

 

NEW: Topic Modeling in Reports.

by Nigel Filed under: News

To use the now-fashionable technique, the definition of topic models on Wikipedia is:

a type of statistical model for discovering the abstract “topics” that occur in a collection of documents.

These techniques have been developed over the last twenty years or so, and, in simple terms, calculate the probabilities that words will be found together in the same document, looking across a collection of documents.  These probabilities are then used to define the abstract or latent topics – latent because they are hidden from view until you tackle the documents with your statistical tools.

Topic models are commonly found in search applications; how can they relate to social media analytics? From our recent report on posts mentioning Asda, you will see that each topic is defined by a set of words, and you can then infer what the conversations are that contribute to that topic.  So for example topic #0 for LSI contains the words

“I”, “asda”, “my”, “you”, “@”, “is”, “was”, “get”, “on”,  

and we can speculate from this that this topic is about location, about being in the supermarket. From this interpretation, we can summarise what the documents (in this case, social media posts matching your query) are about, and thus get a deeper level of understanding and insight into how people feel about your brand.

From running a number of comparisons, it appears to me that, with the data sets we are working with at the moment, Latent Semantic Indexing will provide greater insight than Latent Dirichlet Allocation, though my maths is not good enough to explain how they work or why there is the difference.

As of this week, we are offering topic modelling as an add-on to our standard reports.  If you are interested in getting a report of social media mentions about your brand or company with Topic Modeling included, talk to one of our resellers or contact us direct.

Scottish Independence Referendum – #7.

by Nigel Filed under: Scotland

Continuing our weekly series looking at the social media and online news around the Scottish Independence debate, in the last week there were 1,139 posts, which we classified as pro, undecided, anti, and news, as before.

Thus News is by far the majority category (see table below).  This includes posts about the Economist’s front cover from April 13th, posts regarding an Independent Scotland’s membership of NATO:

The SNP’s NATO u-turn: In its drive to sell independence to middle-Scotland, the nationalist leadership is negle… http://t.co/H6BQtowe

This lead to a number of retweets, commenting on the direction the SNP are taking, so on the one hand there were posts like

RT @OpenUnionism: “In its drive to sell independence to middle-Scotland, the nationalist leadership is neglecting the Scottish left”: http://t.co/GpDxsowk

While on the other there were posts such as:

SNP will ensure an independent Scotland stays in NATO. The correct decision if you want the oil fields protected! http://t.co/8qcBFvZj

The news category also included considerable comment regarding the local elections on May 3rd, and whether these would be seen as a dry run for the referendum, though there was also a fair amount of comment along the lines that people who supported independence did not necessarily support the policies of the SNP:

@PeatWorrier ie “Future policy for Scotland is for future Scotland to decide at that point, not the SNP, pre-referendum”

The pro camp retained it’s lead over the anti’s, though there were reports of an opinion survey which I found interesting:

THE majority of Scots believe the Queen contributes nothing to their sense of national pride, according to a new survey of national identity, calling into question support for the monarchy if Scotland becomes independent. (http://www.scotsman.com/scots-take-more-pride-in-billy-connolly-than-the-queen-says-survey-1-2250061 – and the URL says it all.)

The data:

 

Tech pro undec anti news Total
MicroBlog 191 60 1 273 525
Friendfeed 1 3 0 14 18
Facebook 2 1 0 8 11
Twitter 188 56 1 251 496
Blog 2 0 0 15 17
News 4 2 0 208 214
WWW 7 3 8 357 375
Media 0 0 0 8 8
Total 204 65 9 861 1139
Percent 18% 6% 1% 76% 100%

Is There Value in Sentiment Analysis?

by Nigel Filed under: Tech

Sentiment is one of the key analyses in social media monitoring and in many other fields that use text analytics. But does it have any value?

Straight off, I’d say a qualified yes, and a qualified no.  Why?

A variety of methods have been developed for measuring sentiment, the majority of them based on the presence or absence of key words (bag of words techniques). There are a number of resources that help with this, such as Sentiwordnet, in which words are given a classification of positive, neutral, or negative. So using this, you can say that if a piece of text contains a positive word, it is likely to be positive, and if it contains a negative word it is likely to be negative.

This method will be relatively simple to work out – find the scores of all the words in your piece of text, then take the average, and you have the sentiment for the text.  But what does it mean?

Even simple texts like tweets, which only contain 140 characters, can discuss more than one product – so is the positive word referring to product (a) or product (b)?  If you take the tweet,

News: Lidl’s Vitafit tomato juice is better than Coop’s own brand. I so wanted Coop’s to be better, but it was too sweet.

It is obviously positive about Vitafit and negative about Coop own brand tomato juice; but would a bag of words method for sentiment analysis be able to show this?

A more disturbing – in my view – result comes when vendors include an emotion score.  Using such a system, I was looking at mentions of cancer, and the system showed a high score for “sexual content”.  On investigation, this was due to campaigns for breast, testicular, and prostate cancer research and awareness – definitely not a sexual matter.

So this shows that a greater degree of human intervention is required. At Churnbar, we use a trained classifier to produce sentiment scores, with a simple scale of negative, neutral and positive – having a larger scale with measures of “somewhat negative” and “somewhat positive” will be prone to add error to the system, as there will be overlap between the terms included in the variations on positive and negative.

A person builds the training set, on a project by project basis, by classifying a random set (usually around 200 – 250).  The person knows which way to classify things – so with the tweet quoted above, they would know to classify it as negative if the client was Coop, but positive if the client was Lidl or Vitafit.

This training set is then used to create an automatic classifier, which is used to assign sentiment scores to the rest of the posts collected for the study – meaning that the person does not have to read through and score every single post.  In recent projects, such as our comparison of three supermarkets, we have used this method, with around an 80% accuracy rate (this is about as good as you would get if two people classified the same set of texts and compared the results).

This shows that a human trained, classification based, method of sentiment analysis can be more discerning and thus more accurate than a sentiwordnet-based key word method.  But – and here’s the real question – does sentiment actually give you anything valuable?

Again, the answer is yes and no, and depends on what you are using social media monitoring for. If you’re aim is to use monitoring as a form of research, to track opinions and attitudes to your brand, to define results of PR or marketing campaigns, and to provide an early warning of problems ahead, then sentiment is a valuable thing to measure. If, however, your primary aim is to use social media as a direct channel of communication with individual clients or prospects, for customer support or sales, then sentiment is not a useful measure, as it is not actionable.  Not every negative mention needs to be followed up.  Case in point – I said a while ago on Twitter that I hated Starbucks, and they immediately sent me a voucher.  But I hate Starbucks, I won’t go in there –so they wasted whatever little effort they put into sending me the voucher.

An additional advantage of using a trained classifier is that it can be trained to do anything you want. For a recent study for a bank, we classified the posts found as “Consumer Banking”, “Investment Banking”, “Business Banking”, “Sponsorship”, “News”, or “Wrong” (wrong being a catch-all class for all posts not in English and not about the bank). To do this did not require any additional software development – it just needed a different mind-set from the person creating the training set.

If you would like to see a report on social media mentions of your company or brand, please contact one of the resellers listed on our website. For further information about our methods, contact us.

 

The Scotland Debate #6: The Economist weighs in

by Nigel Filed under: Scotland

The debate this week contained within the results of our search (“Scotland” AND (“Referendum” OR “Independence”) largely revolved around news items.  It seems that activists on both sides have grown quieter in the time since we started this series of posts.

As you will see from the chart, there was a big peak in news coverage on the 13th of April, last Friday.  This was caused by reports and discussion of the Economist’s satirical cover image, a map of Scotland renamed as Skintland, Edinburgh as Edinborrow, and so on, with an accompanying article saying that the economics of Independence would not work out:

The front page of its latest edition shows a map of Scotland with puns instead of place names such as “Glasgone”, “Edinborrow” and the “Highinterestlands”. It is followed by an article concluding that independence would come at a high price and could … (http://www.granthamjournal.co.uk/news/sturgeon-challenges-skintland-map-1-3736576 )

Other news stories that boosted the proportion in this classification included the suggestion by a Conservative MP that

RT @ScotonSunday: Scottish independence: Scotland “needs UK to set up trade missions”  http://t.co/6vOT3Mjk (http://friendfeed.com/kilsally/415cd1a2/rt-scotonsunday-scottish-independence)

It seems that the trend towards more analytical reporting on the subject from politicians, pundits, and journalists is continuing, with some discussion around funding of renewable subsidies after independence:

Scottish households could be left with higher power bills after independence to fund subsidies for wind and wave farms, the country’s major energy companies and the Government have … (http://shamwag.com/2012/04/12/energy-companies-deliver-independent-scotland-power-bill-warning/)

This was posted after an energy company made a presentation to a select committee in Parliament in Westminster, and could call into doubt the proposed energy and economic policies of an independent Scotland – leading back to the Economists article.

At the same time, the pro-Independence politicians and activists claimed that the articles questioning the economic basis for Independence showed ignorance about the realities of Scotland, and anti-Independence activists posted updates along the lines of:

RT @xAwkwardTurtlex: Scotland will be absolutely fucked if it gets independence. (http://twitter.com/hosty96xD/statuses/189310697463758848)

While I am inclined to believe economists over politicians, I would not want to make a judgement.  All I can say is that, among those who care, the debate will continue with new issues coming up each week.

Tech anti undec pro news Total
MicroBlog 41 162 80 566 849
Friendfeed 2 8 0 15 25
Facebook 0 2 0 14 16
Twitter 39 152 80 537 808
Blog 0 4 0 12 16
News 0 1 2 130 133
WWW 4 17 1 311 333
Media 0 0 0 1 1
Total 45 184 83 1020 1332
Percent 3% 14% 6% 77% 100%

From the data, note:

  1. The proportion of the posts classified as “news” has increased.  A large number of tweets and other posts were found reposting news content.
  2. Since last week, we have updated our software, and so the “Microblog” category is currently broken down into posts from Facebook, Friendfeed, and Twitter.  I note that both the Yes and No campaign facebook pages have more posts on them than are shown in the table.  We will be looking into this in coming weeks.

The Scotland Debate: Independence, the Future of the Union, and Social Media #5

by Nigel Filed under: Scotland

I would class this as “History week” in terms of the pro-independence campaigns statements and posts on social media.  The first event to be celebrated was the anniversary of the 1320 signing of the Declaration of Arbroath (see Wikipedia for details):

RT @StatueLibrtyNPS 4/6/1320– Scotland declares independence with the signing of the Declaration of Arbroath.

Taken Up by many more news outlets (including the New York Times) is the link between the timing of the referendum in 1214 and the celebrations of the 700th anniversary of the Battle of Bannockburn in the same year:

Scottish Leader Pins Hopes for Independence Vote on a 700-Year-Old Fervor: http://t.co/iFV2TOmy #history #scotland

This is quite triumphant, and has given the supporters of independence something to talk about, though at the same time doubts have been raised about the viability of an Independent Scotland by defence experts:

Professor Malcolm Chalmers, policy director at the Royal United Services Institute, today warns that “in a world in which the security of states is increasingly interdependent, it is hard to imagine why the prospect of having independent armed forces could, in itself, be a good reason to support….. (full text)

And

Alex Salmond is under growing pressure to set out his plans for Scotland’s defence forces under independence after a leading defence expert raised questions over the the SNP’s flagship policy to quit Nato.(full text)

Pressure has also come in the area of economic policy, for example,

By Dean Herbert ALEX Salmond will launch an extraordinary bid to snatch almost £70billion-worth of assets from Britain as part of an independence settlement for Scotland, it was revealed yesterday. But the move has been slammed by critics, … (full text)

I assume that all of these news reports will make at least some of the electorate reconsider independence, though at the moment, judging purely from social media posts in the last week, the “pro” lobby seems to be making the running:

Tech type pro undec anti news Percent
MicroBlog

307

282

51

224

864

Blog

0

0

0

14

14

News

0

2

0

36

38

WWW

24

16

1

337

378

Media

0

0

0

31

31

Total

331

300

52

642

1325

Percent

25%

23%

4%

48%

100%