Dynamically Determining Affect During Scripted Dialogue

Autonomous synthetic actors performing in live stage performances must be capable of generating speech that is not only lexically correct, but emotionally correct; affect can be conveyed through synthesised speech via modification of various parameters, but first the correct emotional state must be determined, for line performances, this must be done dynamically. The words in each line formed the basic tokens; results were improved with each piece of additional information (simple words, punctuation, parts of speech, character name and previous lines) that was generated from the lines of the script, and also by normalisation of the basic tokens (normalisation of case, removal of punctuation, and word stemming). Results are two to three times improved over a simple random choice, when allowing just a single best choice, or, allowing three attempts, four to five times improved. Synthetic Actors Synthetic actors and computer theatre Although the use of computers to create otherwise imposs...

• Type: Conference
• Year: 2004
• Where: PRICAI
• Authors: Tony A. Meyer
• Keywords: Artificial Intelligence, Autonomous Synthetic Actors, PRICAI 2004, Stage Performances, Synthetic Actors