Sentiment scoring

We keep enjoying the rubbish that is generated by automated sentiment scoring. Here's some classics from twitrratr.com which rated a search for what people were saying about a particular website, as 9 positive, 39 Neutral, and 4 negative. Looking deeper however, it includes classic mis-scoring such as:

• "fuck you*****. i need to pay my bill but because you've cut me off i can't call you to pay. & your website payment pages don't work. good one" - rated as POSITIVE
• "Oh my God, FUCK YOU ***** WEBSITE. Why can't you just be straight forward! Maybe if you didn't have so much shit going on, you would load! - rated as NEUTRAL
• "soooo pee'd off with *****. Not the ppl on the helpline but the**** website need shooting" - rated as NEUTRAL
• "Why is the **** website always down whenever I need to try access my account? #fail" - rated as NEUTRAL
• so excited for our new website - check it out! www.royalhuntoxford.com" - rated as POSITIVE (but totally irrelevant).
  

As always, garbage in garbage out. Any brand manager relying on this kind of analysis is going to be missing the real picture by a mile. Lovely graphs and great to have metrics, but if the underlying analysis is false it is of little value.

Machine driven semantic analysis can only get you so far, as so much of sentiment depends on context and perspective. Humans still rule at this, and this is why organisations that need accurate data, and reliable information use the high quality premium service that UKNetMonitor offers.