| Resumen |
This paper presents a method for creation of dictionaries marked with specific values (for example, emotions, polarity) for use in various tasks of automatic natural language processing. In the created dictionary, the selected words are tagged with six basic emotions. For this, they are first analyzed (annotated) manually by mul-tiple annotators and automatically weighted on the basis of these evaluations. The method was applied to the Spanish language. The para-digm chosen for tagging the words that form the dictionary corresponds to basic emotional categories: joy, anger, fear, sadness, surprise and disgust. Unlike other dictionaries, our dictionary contains weightings that correspond to percen-tages of probability of being used with the sen-se related to emotion. Each word was evaluated by multiple annotators, and, subsequently, the agreement between them was analyzed with the method of weighted kappa adapted for multiple entries. On the basis of these results, we propose a new measure that estimates the probability of the affective use: probability factor of affective use (PFA), which serves to provide potentially emotional words with the weight. PFA can be used as data in automatic systems for emotional analysis of texts. PFA refers to the use tendency of each word, which is useful for automatic sys-tems. |
