BookmarkSubscribeRSS Feed
RuneNielsen
SAS Employee

RuneNielsen_0-1606980002261.jpeg

Photo by David Beale on Unsplash 

 

With the new Nordic sentiment models and the holiday season, we decided to run some text and sentiment analysis on one of the oldest Christmas songs, Silent Night, in Danish, Norwegian, and Swedish. The song is originally written in German and the three translated songs do therefore have the same starting point. Differences between the languages should occur as the translators had to get creative to get the rhymes and rhythm of the songs to work. Let us investigate who will be the happiest when singing the song on Christmas. 

In order to extract sentiment from the text and understand which language had the most positive version of the classic Christmas song “Silent Night”, we started with a text analytics pipeline in SAS Viya Visual Text Analytics 

 

RuneNielsen_1-1606979374592.png

 

Since we were interested in making use of our new Nordic sentiment models in SAS Viya 3.5, we selected the new appropriate language specific model. To get access to the models for the Scandinavian languages, contact Ina or Rune as described in the end of this post.  

In Viya 4 these models will be readily available without having to specify them in the node themselves 

 

RuneNielsen_2-1606979374594.png

 

Below are the results of our sentiment analysis of “Silent Night”.  

 

Verse 

Danish 

Norwegian 

Swedish 

1 

0.60 

0.60 

0.77 

2 

0.60 

0.77 

0.69 

3 

0.88 

0.92 

0.60 

Average 

0.69 

0.76 

0.69 

The sentiment is here reported as the likelihood that the verse is positive, defined as a number between 0 and 1 with 0,5 being neutral. The closer to the more likely it is that the document has a positive sentiment. 

 

All three verses of Silent Night were positive for Danish, Norwegian, and Swedish, however, there were differences in the degree of positive sentiment between the languages. So, the creative process of the translators does change the sentiment of the versesAnd the Norwegian translator was in general able to make the most positively phrased song, peaking in the third verse. 

 

We are excited for the Nordic sentiment models and for the functionality there! If you want to learn more about text analytics or the Nordic sentiment models, feel free to reach out to Rune Hjorth Nielsen (RuneHjorth.Nielsen@sas.com) or Ina Conrado (Ina.Conrado@sas.com). We are happy to chat about it some more 😊 

Happy holidays!  

1 REPLY 1
ErikStrömgren
SAS Employee
Great and very interesting!
Could you maybe also post the different language versions here...? Would be interesting to compare...
Also, I can’t wait to sing the Norwegian verse 3 with the family this year, to really boost happiness! 😄