Sentiment analysis

During extraction, i2 TextChart provides sentiment analysis at the entity and the document level to provide a detailed representation of the sentiment of the text.

This analysis leverages psycholinguistic research across multiple sentiment vectors to provide precise information about an author's language, and pinpoint documents that do not follow predicted patterns. Multi-vector sentiment analysis uses four metrics to help analysts identify outliers in their data, recognize heightened emotional language, sort data by media type, and subset large data sets into only the documents that require further assessment.

<lex>
  <word>arson</word>
  <aspect>-2</aspect>
  <intensity>3</intensity>
  <mood>-3</mood>
  <polarity>-3</polarity>
  <sv><CRIME/></sv>
</lex>

Aspect

What level of control is a reader likely to feel?

Aspect measures how controlled or in control the language in a text makes a reader feel.

Aspect is measured on a floating point scale ranging from −3 (controlled) to +3 (in control). Very positive or negative aspect values generally indicate that the author is part of the narrative, while neutral aspect values generally indicate that the author is not part of the narrative. Very negative values for aspect tend to indicate that the author is attempting to bully or dominate the reader, while very positive values tend to indicate that the author is attempting to persuade the reader.

For example, words like abandonment and abduction have a very negative aspect because these words tend to make readers feel a loss of control. By contrast, words like accomplish and motivate have a very positive aspect because readers tend to feel in control when reading these words.

Intensity

How motivated is a reader likely to be to react?

Intensity is a measurement of the level of activation used to describe a particular entity; in other words, whether an entity is activated/aroused or deactivated/calm. It is measured on a floating point scale ranging from 0 (no activation) to +3 (high activation).

Unlike polarity, mood, and aspect, intensity does not have a negative range.

Words like catastrophic or prestigious evoke a high level of activation with a value of +3, while words like thermometer and syllabus evoke a low level of activation with a value of 0.

Mood

How happy or sad is your text?

Mood is a measurement of the level of emotion of the language used to describe a particular entity in a document.

This measure roughly indicates the degree of happiness or sadness associated with the language.

While polarity measures the valence of the events themselves, mood measures an emotional response typically elicited in readers or listeners upon hearing the words. This metric is also computed at both the entity and document level, and represented as a floating point value on a scale from −3 (negative mood - "sad") to +3 (positive mood - "happy").

For example, words like funny and affection have a positive mood, while words like cancer and racist have a negative mood.

Polarity

How positive or negative is your text?

Polarity is a measurement of how positive, negative, or neutral the language is about a particular entity or the document as a whole, taking into consideration the salience of various words and phrases.

This metric is calculated at both the entity and document level, and is measured on a floating point scale from −3 (negative) to +3 (positive), where 0 represents neutral polarity.

For example, words like elegant and strongest have positive polarity, while words like arson and lousy have negative polarity.

Salience

How important is your text?

Salience refers to how important a word or phrase (or entity) is within a document or set of documents.