Event & event actor identification with event-sentiment relation

In any natural language, each sentence contains a subject which triggers an emotion affecting the object in a certain way that can be positive, negative and even neutral at times. The words that imply the emotional states are referred to as Sentiment Words for the rest of this paper. Identifying Human Sentiments from various texts is quite a task itself, leaving alone their extraction. We have proposed as well as implemented a new technique to first identify the actions (which are formally described as Events in this context) and their respective actors (referred as Event Actors) from a certain text and then to extract them. Later on, we have contrived a method for identifying the relation between the Events and their Sentiments. The technique we have used is Sentence Parsing with the help of Natural Language Processing (NLP) libraries. Our system will identify Events, Event Actors & the Sentiment word related to the Events. We have embedded an accuracy checking in our system to verify the correctness in identifying the Events and its Actors. Furthermore, the system is capable of extracting the Sentiment Word in the sentence along with the Human Emotion coupled with it (Human Emotion can hold the value of positive, negative or neutral in our implementation). We have considered the `TimeML´ corpus data as our Gold Data. This system can also be used in any field where Event identification is necessary to track the current occurring of an incident. Evaluation on a collection of TempEval-2 corpus shows the precision and recall values for the basic model as 72.31%, 75.44% respectively. Our System accuracy is approximately 80%-85% for Event identification and for Event Actor identification it is approximately 77%-82%.