In this article, we develop a system with a natural language interface to generate animation of small groups in environments with ambient crowds. Our system takes simple sentences of English as input and these sentences are parsed to obtain information about character attributes, behaviors, and locations for constructing situation nodes. These situation nodes form an animation graph for producing an animation of small groups. The interface assists in interactively managing and editing the animation graph. We demonstrate the effectiveness of the system in some examples. The user study results indicate that the proposed system is user-friendly and flexible in producing animation of small groups in various scenes. K E Y W O R D S crowd animation, interactive interface, natural language processing, small groups 1 INTRODUCTION Crowds are composed of individual characters and small groups (e.g., families, friends) which exhibit rich behaviors, e.g., watching a performance, wandering, and dancing, that are common in various situations, e.g., parks, subway stations, streets, markets, and exhibition halls. Graphical authoring tools and behavior trees have been adopted in producing animations with character models. 1,2 Natural language can be treated as a kind of input medium for constructing virtual objects and animations. 3-5 A story maker tool incorporates natural language processing (NLP) and graphics to generate 2D animated scenes. 3 The components of a scene, including actors, actions, objects, and locations, are determined based on input sentences. A story telling system 5 adopts a module, called subject-predicate-object (SPO), 6 to extract the subject, predicate and object in an input sentence. The aforementioned systems mainly focus on one or a few characters, which do not handle groups, crowd behaviors, and groups of ambient crowds. In this article, we implement a system with a natural language interface and crowd simulation techniques to animate groups (Figure 1). We extract group-related information, such as characters, group behaviors, group targets, and locations from the input sentences. Situation nodes are constructed to form an entire animation. In this study, we take input as simple sentences of English. A user study was conducted to evaluate the performance and utility of our system.

In order to improve the efficiency of voice analysis, the authors make a series of study on the cache technologies of improving VoiceXML voice system performance, and in this paper the authors propose a new method of using sharing DOM tree to improve VoiceXML voice system performance.