Quantcast
Channel: Valuescope » WEB ANALYZER
Viewing all articles
Browse latest Browse all 5

Social Sentiment Analysis cannot be that hard…

0
0

(Social) Sentiment Analysis is and will continue to be a much discussed topic because it is an ambiguous analysis no matter if conducted automated or manually.
Everyone has their own preferences how to express their opinion and how to interpret the statements of other people.
It is a great challenge to filter the emotion from those statements.
At the same time it is more insightful than surveys because it captures more honest as well as momentary emotions and can thus be more comprehensive.

What is needed is the right technology!

On bernetblog.ch you can find a post related to the topic “What is Social Sentiment Analysis”.
Valuescope as a specialist for automated text analyses in terms of content and emotion appreciates every post that treats the topic on a deeper level.
It is said often that automated Social Sentiment Analysis is not possible. As if to prove this statement, many Monitoring providers offer a Sentiment Analysis for the German language that almost equals the flipping of a coin.

What makes the difference is the technology!

Basically there are 3 possibilities to analyze sentences for emotion:

Keyword method:

 


A list of positive and negative words is created. If a sentence contains one or several of these words it is evaluated as positive or negative.
Disadvantage: This method is problematic for evaluating German sentences because German speakers often times use negations: „Das Essen war nicht schlecht“ (=The food was not bad), „Das Wetter heute ist nicht schön“ (=Today’s weather is not good). Using the keyword method, both sentences are assigned wrong values.
Advantage: relatively fast implementation

Algorithms:

 


In this method sentences are collected that are evaluated as positive or negative. New sentences are compared to those sentences and according to their concordance evaluated as x% positive or negative. Each technology provider decides with which minimum percentage a sentence is displayed as positive, negative, or neutral.
This method can be improved by collecting more and more positive and negative sentences and hence enlarging the basis for comparison.
Disadvantage: Nevertheless it is a method with limited possibilites because it cannot recognize irony or complex structures: “I prefer the t-shirt from H&M to the one from C&A”, or “Yay, what a great day again….”
Advantage:an automatically learning system

Rule or grammar based method:

 


For the grammar based approach the machine has to “learn” the grammar of the language in question – e.g. what are verbs, substantives, articles, adjectives,… and the respective conjugations and declinations. This analysis refers always to a search term. The machine has to recognize if a statement refers to the search term or if it is simply mentioned: “The weather is really bad today, I think I will go to the cinema” is a neutral statement for the search term “cinema” but a negative one for the search term “weather”.
In a next step follow definitions of positive, negative, and neutral statements. For this purpose statements are researched and the machine is taught their positive, negative, or neutral value on the basis of the grammatical structure.
Using this technology it is possible to recognize subtleties or different notions in a text, e.g.

This was not really interesting
This was really not interesting

if it is relevant for the emotion of a statement.
Further examples and explanations .
Disadvantage: Few monitoring providers choose this option because it is very elaborate and has to be created for every language individually.
Advantage:It offers a quality comparable to a manual Sentiment Analysis!

The post Social Sentiment Analysis cannot be that hard… appeared first on Valuescope.


Viewing all articles
Browse latest Browse all 5

Latest Images

Trending Articles





Latest Images