Valence Processor

Computes text valence using two separate stages:

• Comparing each word in text to existing valence word list.
• Obtaining per sentence probability by using naive bayes algorithm on a model pre-trained with emotional text corpus.

This processor determines valence in the following discrete scale:

Name	Definition
NEUTRAL	Contains no emotional information
POSITIVE	Has positive meaning
NEGATIVE	Has negative meaning
MIXED	Contains more than one valence from above

Word List Lookup

Each word is looked up from a predefined emotional dictionary, obtaining the integer value:

Value	Definition	Classification result
0	Word is neutral or does not exist in dictionary	NEUTRAL
1	Word is positive	POSITIVE
-1	Word is negative	NEGATIVE
-8	Word is extremely negative	EXTREME

Aggregation

All tokens not explicitly stated otherwise are aggregated together by group-summarizing results of the sub-tokens. For example, having child tokens classified as:

	Child1 = NEUTRAL(Z1)
	Child2 = POSITIVE(P1)
	Child3 = NEUTRAL(Z2)
	Child4 = NEGATIVE(N1)
	Child5 = NEUTRAL(Z3)
	Child6 = POSITIVE(P2)

result of the current token would be calculated like:

	Token = { NEUTRAL(Z1 + Z2 + Z3), POSITIVE(P1 + P2), NEGATIVE(N1) }

where arguments Zx, Px and Nx of each valence function is a total number of words classified as being in that valence.

Paragraph

As an exception from the above, paragraph level results are aggregated by summarizing all words in all sentences of the paragraph using a following rule set:

Classified word count	Final classification of all words
extreme > 0	Negative
positive > negative	Positive
negative > positive	Negative
positive = negative	Mixed
positive = 0, negative = 0, extreme = 0	Neutral

Above means that if any word in a paragraph was classified as extremely negative, all words in that paragraph are counted as negative etc.

Corpus

Corpus results would first be aggregated from paragraph results using the method described above. After that, further post-processing will be done using following rules.

Valence for a corpus consist of two elements: valence result and a quantifier. Quantifier can be one of ONLY or MOSTLY.

• If only one emotion is found, the corpus would be classified as ONLY that emotion.
• If two emotions were found and one contains more than a half words than the other, the corpus would be classified as MOSTLY that emotion.
• If none of the above rules apply, corpus will be classified as MOSTLY mixed emotion.