Collocations as Cues to Semantic Orientation

  • Faye Baron ,
  • Graeme Hirst

Proceedings of the AAAI Spring Symposium on Exploring Attitude and Affect in Text: Theories and Applications |

Techniques to classify opinion or sentiment in text as either positive or negative employ features such as individual words, bigrams, and part-of-speech patterns. When individual words and bigrams are employed, both the frequency of their presence and their predetermined probable polarity are used to decide the polarity of the text to be classified; neutral words and bigrams are not useful in this task. We refer to words that provide polarity cues as nuance-bearing.