Archive for the ‘TwiNL’ Category

Post mentioned on Kennislink

In TwiNL on 24/04/2014 by admin

20140424 In the article Twitter als basis voor taalkundig onderzoek (Using Twitter for Linguistic Research) on the Dutch science blog Kennislink, Mathilde Jansen discusses the pros and cons of using Twitter for linguistic research. is featured prominently in the article.


The Happiest Dutch Municipality (Dec 2013)

In TwiNL on 27/01/2014 by admin

20140127 In December 2013, the island Ameland appeared at the top of our list of happiest Dutch municipalities on Twitter. De Marne was second while Ubbergen and Rucphen shared third place. The rankings is based on automatic sentiment analysis of tweets of Twitter user with public location information. Like in November 2013, each municipality obtained a positive average sentiment score. You van take a look at the associated map (red color: below the median score of 17) or examine the complete municipality list (the higher the score, the more positive).

The map for the happiest provinces of December appeared in an earlier post. Friesland finished on top in that overview, with Drenthe and Overijssel as close shared seconds.

Post in De Volkskrant

In TwiNL on 20/01/2014 by admin

20140120On Monday 20 January 2014, our website was mentioned in the Dutch national newspaper De Volkskrant. The full-page article Ook op Blue Monday zijn we een vrolijk twittervolk (Even on Blue Monday we are happy on Twitter; page 6) focusses on the sentiment of Dutch tweets.

The article spawned an interview on Dutch national radio (Radio 1; from about minute 20:00). Two additional interviews about the topic were broadcasted on regional radio stations: Radio Noord Holland (16:00 frame; start at 52:46) and RTV Emmen (17:15). The results of the article were also mentioned on the prime Dutch Twitter blog Twittermania and in the tv programme RTL Late Night (with host Humberto Tan) in the item of Luuk Ikink (starts at minute 15:46).

On Tuesday 21 January, the results of the sentiment analysis were mentioned by the regional broadcasting corporation of the province Limburg: L1. Belgian national newspaper De Morgen reported about the sentiment in Flanders.

The Volkskrant article contains a map which is based on a municipality map you can find on This map is created from an offical list of municipality borders from 1-1-2012 which means that the municipality fusions of the last two years have not been incorporated. also creates sentiment summaries for provinces related to searches which are restricted to tweets which contain a gps location. An example for this is the search for all geo tweets (twinl-geo) from December 2013 (also on the right). Apart from the 12 Dutch provinces, this map also includes the five Flemish provinces. To our surprise, the Flemish provinces obtain lower sentiment scores than the Dutch provinces. A good explanation for this is not easy to find but the difference could be caused by the dictionary used for assessing sentiment. If Flemish tweets use a different vocabulary for expressing sentiment than the Dutch tweets, it would be harder for Flemish regions to deviate from the neutral sentiment score. This would explain why their averages are lower than the Dutch average scores. crashed at 15:06 today. The last registered processor load was 524. Tuesday is a maintenance day at SURFsara, so the website is going to be unavailable until at least Wednesday 22 January.



In TwiNL on 18/01/2014 by admin

clin24 The 24th meeting of Computational Linguistics in The Netherlands (CLIN 24) was hosted at the University of Leiden on Friday 17 January 2014. Two talks mentioned using the website

Nander Speerstra and Hans van Halteren presented a method for automatic gender detection from tweet texts and compared this with the approach used on By using hundreds of tweets per user they were able to identify the gender of 95% of them correctly, compared to the 87% accuracy achieved by and other related work.

Florian Kunneman and Antal van den Bosch used tweets collected from to indentify future events and predict the day on which the event will happen. For Dutch football games, they were able to automatically predict the correct day of the event about a week in advance, based on tweet data.

A full list of talks and abstracts of the event can be found on the CLIN website.


Twiqs research paper

In TwiNL on 20/12/2013 by admin

Today, our research paper about appeared in the CLIN Journal:

Erik Tjong Kim Sang & Antal van den Bosch, Dealing with Big Data: the Case of Twitter. In: Computational Linguistics in the Netherlands Journal, volume 3, ISSN: 2211-4009, pages 121-134, 2013. [PDF] [bibtex]

The paper describes the website and discusses the options it presents for studying Dutch tweets by examining three case studies. Enjoy!