Consider the following text:
“Target refused to sign Bangladesh Fire and Safety Agreement “
When “Bangladesh” is added as a sentiment-bearing phrase in the sentiment dictionary with a score of -1.5, the engine does not extract “Bangladesh” with a negative sentiment. This is because according to the engine proper nouns are not accepted in the dictionary. (The dictionary considers any entry with a capitalized first letter as a proper noun and therefore, only lower case words should be added to as a sentiment-bearing phrase in the dictionary.)
According to Semantria’s supported pattern, sentiment-bearing phrases should follow these rules:
Adverb OR any type of verb OR any type of adjective OR common noun OR plural common noun OR any symbol OR : OR ; OR - OR open parenthesis OR close parenthesis OR cardinal
This pattern can appear once, twice, or three times in a row, meaning that a sentiment-bearing phrase can consist of one, two or three words – excluding determiners (a, the, an) and prepositions (of, on, in).
If the pattern appears only once it can include a hashtag, such as #sentiment.
“any type of verb” means the following:
Modal verb (can, could, may, must)
Base verb (take)
Future tense, conditional
Past tense (took)
Gerund, present participle (taking)
Past participle (taken)
Present tense (take)
Present 3rd person singular (takes)
“any type of adjective” means the following:
According to these rules, correct sentiment-bearing phrases are:
victim (common noun)
slap (common noun or verb)
well served (adverb and verb)
well-appointed (adverb, - (dash) and adjective)
wide of the mark (adjective and common noun)
:) (colon and close parenthesis)
;-) (semicolon, dash and close parenthesis)
*) (symbol and close parenthesis)
#sentiment (hash tag)
Therefore, proper nouns like Bangladesh can’t be used.
A sentiment-bearing phrase can be excluded from sentiment extraction with a ~ (tilde) prefix. Therefore if a sentiment-bearing phrase looks like, ~victim, the engine will ignore the word 'victim' and not assign it a sentiment score for the phrase and will exclude its score from the overall document sentiment score.