|
Bulletin of the Technical Committee on Learning Technology
(ISSN: 2306-0212) |
Authors:Binglin Zhuang1, Mengqi Yu1, and Weiyang Liu*
1
1: Northeast Normal University, Changchun, China
Abstract:
In drama education, script situational understanding is a key educational challenge. When actors or students read script excerpts, they are often constrained by limited text and limited time, making it difficult for them to accurately identify a character’s preceding events, interpersonal relationships, action goals, conflicts and obstacles, subtext, and emotional changes. However, traditional script analysis mainly relies on close reading, teacher explanation, and classroom discussion, which may be time-consuming, fragmented across different dimensions of understanding, and insufficient in providing individualized feedback. To address this issue, a visual support system was developed for script situational understanding and character-state analysis. Taking scenes and characters as the basic units of analysis, this system uses a large language model to generate character-state cards, key-line intention annotations, emotional change curves, and interactive question-answering results. To evaluate the effectiveness and contribution of the system, a 60-minute classroom-based experimental activity was conducted as a case application evaluation with a script situational understanding questionnaire, system acceptance questionnaire, and semi-structured interview. The findings show that students generally gave positive evaluations of the system’s support for script situational understanding. In particular, students recognized the system’s usefulness in supporting action-goal understanding, emotional-change understanding, and contextual understanding. In addition, the interview results further indicated that the character-state cards, line intention annotations, and emotional change curves helped students understand character situations, action logic, and emotional progression more clearly.
Keywords: Drama education; script situational understanding; large language models; character state; visualized analysis.
I. INTRODUCTION
Drama education has important value in contemporary education, and its significance has gone beyond the training of performance skills alone. UNESCO’s Road Map for Arts Education points out that arts education contributes to the development of learners’ creativity, cultural awareness, and expressive ability [1]. Within this field, drama education is often considered to be related to creative thinking, communication, collaboration, and social-emotional competence [2][3]. However, the effects of drama education are influenced by specific teaching methods [4], and one of the core challenges is script situational understanding. Specifically, the key to script situational understanding is not to explain the literal meaning of lines, but to help actors form a performance-ready understanding of the character’s situation [5], action goals [6], and emotional logic [7].
The difficulty of script situational understanding is mainly reflected in the contradiction between fragmented script reading and the construction of a complete character state. This contradiction has become a common situation in the industry. According to the commercial industry platform Backstage, audition sides commonly used in auditions are usually excerpts from a complete script, which may be an entire scene or only part of a scene [8]. At the same time, Spotlight, another important casting platform in the European performance industry, also defines sides as small sections of a script that actors receive before an audition [9]. This means that actors often need to quickly judge a character’s state based on limited text. However, script situational understanding itself relies on an overall grasp of character relationships, event sequences, and dialogic interactions [10]. Therefore, when actors face fragmented texts, they often lack sufficient context to judge the character’s preceding experiences, relationship status, action goals, subtext, and emotional changes.
To address the difficulties of script situational understanding, existing methods mainly include close reading of scripts, scene action analysis, line and subtext annotation, and feedback from teachers, directors, and peers. Close reading helps actors grasp the story background, character relationships, and scene functions [7]， but it depends on complete reading and sufficient discussion. Scene action analysis emphasizes goals, obstacles, and tactics [11], but provides insufficient support for preceding situations and relationship development. In addition, line and subtext annotation focuses on the real intentions and hidden emotions behind specific lines [11], but it can easily remain at the level of local text. Feedback from teachers, directors, and peers can help revise character understanding [12], but it relies on on-site interaction and the experience of the instructor. Overall, although these methods can support script understanding from different perspectives, they still have difficulty resolving the contradiction between fragmented script reading and the construction of a complete character state.
The educational application of large language models (LLMs) provides new possibilities for script situational understanding [13]. Scripts are organized around scenes and interweave character relationships, event progression, action descriptions, and dialogic interactions [14]. Correspondingly, LLMs are trained on large-scale natural language corpora [15] and have capabilities such as question answering, summarization, information extraction, and text generation [16]. These capabilities can be used to support the analysis of characters, events, dialogues, and situational information in scripts [17]. At the same time, script situational understanding is not a single task of plot summarization, but involves multiple dimensions, including given circumstances, character relationships, action goals, obstacles, tactics, subtext, and emotional changes [23]. Therefore, this study embeds LLMs into the task of actors’ script situational understanding, using them to generate structured and visualized character-state information and to support actors in quickly building a performance-ready understanding of characters.
Therefore, this study designs a visual support system for script situational understanding. Taking scenes and characters as the units of analysis, the system uses structured script parsing, character-state cards, key-line intention annotation, emotional change curves, and an actor question-answering module to help actors quickly understand the preceding context, character relationships, action goals, obstacles, subtext, and emotional changes in a single scene. Based on this, this study proposes the following research questions:
- What major difficulties do actors or students face in script situational understanding when reading a single scene within a short period of time?
- Can character-state cards, line intention annotations, and emotional change curves based on LLMs help actors understand a character’s preceding context, relationships, action goals, obstacles, subtext, and emotional changes more clearly?
- How do actors or drama learners evaluate the usefulness, comprehensibility, and operability of the system?
- In what ways can the system support character understanding, and in what aspects does it still require supplementary judgment from teachers, directors, or actors themselves?
II. LITERATURE REVIEW
A. Definition of Drama Education
Based on Bolton’s view of drama education, this study defines drama education as an educational process that uses theatre arts as a medium to promote learners’ understanding of characters, relationship construction, and meaning expression through role playing, situational construction, physical and verbal expression, performance practice, and reflection [18]. While Bolton emphasizes the educational need for meaning construction, O’Neill also stresses that the core of drama education lies in learners’ entry into fictional situations and their formation of understanding through actions in role [19]. It is also worth noting that the National Core Arts Standards summarize drama education as a process involving creating, performing, responding, and connecting [20]. Specifically, drama education includes not only character building and performance practice in professional actor training but also learning activities in general education that promote expression, collaboration, imagination, and social understanding through dramatic situations [21] [22].
B. Challenges in Script Situational Understanding in Drama Education
Existing research and industry materials indicate that difficulties in script situational understanding do not simply come from textual complexity, but also from the fragmented reading contexts in which actors use scripts. In audition practice, audition sides usually provide only partial materials from a complete script [8], and actors need to make character judgments and prepare performances based on limited text [9]. This practice creates a contradiction with the internal requirements of script understanding. Tian et al. point out that a script is a long narrative text that contains character relationships, temporally ordered events, and dialogic interactions, and that a complete understanding of a script requires the construction of a coherent story world [10]. Correspondingly, actor training theory also emphasizes that actors need to understand a character’s given circumstances, preceding context, interpersonal relationships, action goals, obstacles, and subtext [7] [11]. Therefore, fragmented script reading may weaken actors’ grasp of the overall character state and make it difficult for them to form a performance-ready understanding of the character in a timely manner.
This study summarizes the difficulties of script situational understanding in drama education into six categories: insufficient context, unclear character relationships, unclear action goals, difficulty in identifying conflicts and obstacles, difficulty in recognizing subtext, and lack of an overall grasp of emotional changes. Starting from the concept of given circumstances, Baumrin regards context and preceding circumstances as core issues in character understanding [23]. Bucs and Pye, following Hagen’s perspective, approach character analysis by emphasizing that character relationships, action goals, obstacles, and action strategies constitute key components of script situational understanding [24] [25]. As an important dimension of line analysis, subtext and line intention are also included by Thomas in the scope of script situational understanding [7]. In addition, emotional changes need to be understood in relation to character goals, conflict progression, and scene events [10].
C. Applications and Limitations of Existing Methods for Script Situational Understanding
To address difficulties in script situational understanding, drama education and actor training have developed four main support methods: close reading of scripts, scene action analysis, line and subtext annotation, and feedback from teachers, directors, and peers. Close reading focuses on the whole text and helps actors understand the story background, character relationships, and scene functions [7], but it depends heavily on complete reading and sufficient discussion time. Scene action analysis further transforms script texts into character goals, obstacles, and action strategies [11], but it mainly focuses on how a character acts and provides insufficient immediate support for preceding circumstances and relationship development. Line and subtext annotation focuses on the real intentions and hidden emotions behind specific lines [11], but it can easily remain at the level of local lines and may fail to construct an overall character state. Feedback from teachers, directors, and peers can revise actors’ understanding during rehearsal [12], but its effectiveness depends on on-site interaction and the experience of the instructor, making it difficult to provide structured support that can be saved, reviewed, and personalized. Therefore, although existing methods can support script understanding from different perspectives, including text, action, lines, and interactive feedback, they still have difficulty effectively reducing the contradiction between fragmented script reading and the construction of a complete character state.
D. Potential of Artificial Intelligence in Drama Education
In recent years, the development of LLMs has provided new technical possibilities for script situational understanding. A script is a narrative text organized around scenes and interwoven with character relationships, event progression, action descriptions, and dialogic interactions [14]. LLMs are trained on large-scale natural language corpora and can be adapted to various natural language processing tasks, such as question answering, summarization, information extraction, and text generation [15]. Therefore, they can be used to support the analysis of characters, events, dialogues, and situational information in scripts [17]. More importantly, script situational understanding is not a single task of plot summarization [11], but requires the simultaneous understanding of multiple dimensions, including given circumstances, character relationships, action goals, obstacles, tactics, subtext, and emotional changes [23]. Therefore, this study does not simply use LLMs for plot summarization. Instead, it embeds them into the task of actors’ script situational understanding, using them to generate structured and visualized character-state information, thereby supporting actors in quickly building a performance-ready understanding of characters.
III. DESIGN AND IMPLEMENTATION OF A VISUAL SUPPORT SYSTEM
A. System Design Principles
In drama education, both students and actors need to identify a character’s preceding events, interpersonal relationships, action goals, obstacles, subtext, and emotional changes from script excerpts to complete script situational understanding. Therefore, this study first analyzes the six major difficulties that hinder script situational understanding and translates them into system functional requirements. Second, taking scenes and characters as the units of analysis, the system enhances its support functions by transforming script texts into structured, visualized, and follow-up question-based character-state information.
First, six categories of difficulties in script situational understanding in drama education are identified. Based on the literature analysis, this study summarizes the difficulties of script situational understanding into six categories: insufficient context, unclear character relationships, unclear action goals, difficulty in identifying conflicts and obstacles, difficulty in recognizing subtext, and lack of an overall grasp of emotional changes. In response to these problems, the system designs corresponding functions. As shown in Table I, the difficulties of script situational understanding correspond to six system functions, including structured script parsing, scene and character selection, character-state cards, key-line intention annotation, emotional change curves, and an actor question-answering assistant.
TABLE I CORRESPONDENCE BETWEEN DIFFICULTIES IN SCRIPT SITUATIONAL UNDERSTANDING AND SYSTEM FUNCTIONS.
| Difficulties | System design requirement | Corresponding function |
|---|---|---|
| Insufficient context | Supplement situational information related to the current scene | Script parsing; scene selection |
| Unclear character relationships | Present the relationship state of the character | Character-state card; question-answering assistant |
| Unclear action goals | Extract the character’s goals and obstacles | Character-state card |
| Difficulty in recognizing subtext | Analyze real intentions | Line intention annotation |
| Unclear emotional changes | Present the progression path of emotions | Emotional change curve |
| Individual questions not covered | Support immediate follow-up questions | Actor question-answering assistant |
Second, system support strategies are proposed to address the six categories of difficulties in script situational understanding. Based on performance training theory and script analysis methods, this study proposes three system support strategies for script understanding. The first is structured character-state generation. This strategy draws on Baumrin’s discussion of given circumstances and avoids keeping the system output at the level of general plot summarization [23]. Instead, it organizes context, character relationships, pre-entry state, action goals, obstacles, and hidden emotions into character-state information.
The second is visualized scene action analysis. This strategy adapts Bucs and Pye’s scene action analysis theory and transforms questions such as “what the character wants,” “why the character is blocked,” and “what action the character takes” into character-state cards, line intention annotations, and emotional change curves. In this way, script understanding becomes closer to the action tasks involved in actors’ rehearsal [24].
The third is follow-up question-based script situational analysis. This strategy draws on Dewey’s theory of reflective thinking and integrates response, reflection, and situational construction into script understanding [26]. The system allows actors to continue asking questions about the current scene and character, thereby supplementing individualized understanding problems that cannot be fully covered by fixed analysis templates. Through these three strategies, the system transforms script situational understanding from static textual interpretation into a structured, visualized, and interactive process of character-state construc
B. System Architecture and Functions
1) System Framework
To address the six categories of difficulties in script situational understanding, this study adapts a publicly available script visualization system [27] and integrates six core system functions with three support strategies. The system is named the Model-Assisted Script Analysis System, namely MASA. Its core function is to leverage large language models to generate character status cards, annotations of dialogue intent, emotional trajectory curves, and Q&A results, thereby helping students understand the dramatic context and character states. Especially, it adopts a lightweight web application architecture, mainly consisting of a user interaction layer, a script parsing layer, a model analysis layer, and a visualization output layer. In this framework, the process of script understanding is divided into four stages including of text parsing, character focusing, model analysis, and visualized presentation. With this design, it not only reduces the scope of model analysis, but also improves the relevance of system outputs to the actor’s current performance task.
From the perspective of system workflow, users first upload or paste the script text. Receiving the data, the system will use natural language processing techniques to complete scene segmentation, character recognition, and line attribution recognition. Based on the parsing results, users select a specific scene and character. The system backend then calls an LLMs or a local heuristic analysis module to generate structured results. Finally, the system presents the character-state card, line intention annotation, emotional change curve, and question-answering results through the front-end interface.
2) Core Functional Modules
Script input and structured parsing module. This module is responsible for transforming the original script text into a data structure that can be processed by the system, including scene titles, character names, line content, and action descriptions. The system uses natural language processing techniques to conduct structured parsing, identify which character each line belongs to, and provide a foundation for subsequent character-state analysis.
Scene and character selection module. This module reflects the actor-oriented design of the system. Users first select a specific scene and then choose a character within that scene. The system then conducts subsequent analysis based on the selected character. This module avoids making general explanations of the entire script and makes the output more closely aligned with the actor’s current performance task.
Character-state card module. The character-state card is the core output module of the system. It mainly includes “Who am I?”, “given circumstances,” “pre-entry state,” “goal in the scene,” “main obstacle,” “action strategy,” “hidden emotion,” “what cannot be spoken,” and “performance focus.” This module transforms complex script text into character-state information that actors can directly use.
Key-line intention annotation module. This module is used to analyze the action logic behind key lines. The system not only explains the surface meaning of lines, but also analyzes real intentions, subtext, emotions, strategies, and their influence on the opposite character. This helps actors understand how lines promote changes in relationships and conflicts.
Emotional change curve module. The emotional change curve is used to present the emotional progression of a character in the current scene. The system transforms emotional nodes, intensity, triggering reasons, and corresponding events into a visualized curve, helping actors grasp the character’s psychological development path.
Actor question-answering assistant module. The actor question-answering assistant supports individualized follow-up questions. Actors can continue to ask questions about the current scene and character, such as the character’s goals, emotional sources, interpersonal relationships, or the subtext of a specific line. The system answers based on the current text and existing analysis results, thereby supplementing the fixed analysis output.
C. Large Language Model Analysis Mechanism
The purpose of using an LLMs in this system is not to allow the model to freely interpret the script, but to constrain it within the task of actors’ script situational understanding. Through prompt design, the system clearly defines the analysis object, analysis scope, and output structure. Also, the model is required to generate only the character-state card, line intention annotation, and emotional change curve for the selected character, thereby avoiding ordinary plot summaries or general literary comments. Meanwhile, the model output adopts a structured format, which facilitates stable front-end rendering. Finally, this mechanism transforms the LLMs from an open-ended text generation tool into a character-state analysis tool oriented toward actors’ tasks.
D. System Implementation
The front end of the system adopts a web-based format, supporting script input, scene selection, character selection, analysis result display, and actor question answering, as shown in the figure. First, the interface layout follows the actor’s usage process. Users first input the script and select a scene and a character, then view the character-state card, line intention annotation, and emotional change curve, and finally ask further questions through the question-answering assistant. Second, the backend is responsible for receiving front-end requests, organizing analysis prompts, calling the large language model, and returning structured results. To improve system stability, the system also includes a local heuristic analysis mechanism. When the model interface is unavailable, the network is abnormal, or the LLMs API is not configured, the system can still generate a basic character-state card, simplified line intention annotations, and an emotional change curve, ensuring that the system has basic demonstration and usage capability. Third, in terms of implementation, this study only retains technical descriptions related to the research goal. Specific startup commands, interface addresses, file structures, and complete JSON fields can be placed in the appendix or system documentation.
IV. SYSTEM EVALUATION EXPERIMENT DESIGN
A. Experimental Purpose
This study adopts a case application evaluation method to examine the support provided by the developed system for students’ script situational understanding. Since the purpose of the experiment is to validate the effectiveness and contribution of the system prototype in a teaching context, a combination of questionnaire surveys and interviews was used for evaluation. The experiment adopted a classroom case validation approach, in which learners’ feedback on the system’s usefulness, ease of use, and learning support effects was collected through a simulated teaching task.
B. Experimental Activity Design
The experiment was conducted in a 60-minute small-class activity. We recruited nine undergraduate students majoring in film and television-related fields, specifically in Radio and Television Directing. Prior to the experiment, all students had completed relevant courses such as acting, script analysis, or the analysis of film and television works. Participants were recruited through an on-campus call, and all took part voluntarily after being informed of the study’s purpose and procedures. They were divided into three groups, with three students in each group. The experimental script was selected from Thunderstorm, using a core excerpt from this classic play. The selected excerpt has the characteristics of dense dialogue, concentrated conflict, rich subtext, and clear emotional changes.
Before the experimental activity, a five-minute introduction was given. During the activity, students first read the assigned single scene and developed an initial understanding of the target character. Subsequently, they used the system to view the character-state card, key-line intention annotations, and emotional change curve, and further asked questions about the character’s motivation and performance focus through the question-answering assistant. After the activity, students completed the script situational understanding questionnaire, the system acceptance questionnaire, and semi-structured interviews.
C. Evaluation Indicators
First, script situational understanding was evaluated. The questionnaire was designed around six dimensions: contextual understanding, character relationship understanding, action goal understanding, conflict and obstacle understanding, subtext understanding, and emotional change understanding. It was used to examine whether the system helped students form a more complete understanding of character states. The questionnaire contained 12 items in total, with two items for each dimension. Second, system acceptance was evaluated. The system acceptance questionnaire examined students’ overall evaluation of the system from the perspectives of perceived usefulness, ease of use, and continuance intention. The questionnaire consisted of nine items, covering three dimensions: perceived usefulness, ease of use, and continuance intention, with three items for each dimension. Both questionnaires used a five-point Likert scale, where 1 indicated “strongly disagree,” 2 indicated “disagree,” 3 indicated “neutral,” 4 indicated “agree,” and 5 indicated “strongly agree.”
V. EVALUATION ANALYSIS OF SYSTEM APPLICATION
This study evaluated the effectiveness and contribution of the system in script situational understanding learning through a classroom case application. Quantitative and qualitative analyses were conducted to examine the script situational understanding questionnaire, the system acceptance questionnaire, and the structured interview data. Since this study adopted a case application evaluation method and the sample size was nine students, the questionnaire results were mainly analyzed using descriptive statistics, including mean, standard deviation, and agreement rate. The agreement rate refers to the proportion of students who selected 4 or 5 points. In addition, to further explore the relationship between students’ system acceptance and their perceived support for script situational understanding, Spearman rank correlation analysis was used as an exploratory supplementary analysis.
A. Analysis of the Script Situational Understanding Questionnaire
Overall, the total mean score of the script situational understanding questionnaire was 4.23, and the overall agreement rate was 91.7%, as shown in the table II. The results indicate that students generally believed that the system helped them understand the script situation and form a more complete character state. Among the dimensions, action goal understanding and emotional change understanding received the highest scores, with both mean values reaching 4.50 and agreement rates reaching 100.0%. This result suggests that students believed the system had a positive effect on helping them understand the character’s behavioral purpose and grasp the process of emotional change.
At the same time, the mean score for contextual understanding was 4.28, with an agreement rate of 94.4%. This also indicates that the character-state card generated by the system helped students supplement information about the character’s situation before entering the current scene and understand the causes and consequences of the scene. In addition, the mean score for character relationship understanding was 4.11, with an agreement rate of 88.9%, indicating that the system also provided some support for identifying character relationships.
TABLE II. STATISTICAL RESULTS OF THE DIMENSIONS IN THE SCRIPT SITUATIONAL UNDERSTANDING QUESTIONNAIRE
| Dimension | Mean | Standard deviation | Agreement rate |
|---|---|---|---|
| Contextual understanding | 4.28 | 0.57 | 94.4% |
| Character relationship understanding | 4.11 | 0.58 | 88.9% |
| Action goal understanding | 4.50 | 0.51 | 100.0% |
| Conflict and obstacle understanding | 3.94 | 0.54 | 83.3% |
| Subtext understanding | 4.06 | 0.64 | 83.3% |
| Emotional change understanding | 4.50 | 0.51 | 100.0% |
| Overall | 4.23 | 0.59 | 91.7% |
In comparison, the scores for conflict and obstacle understanding and subtext understanding were relatively lower, with mean values of 3.94 and 4.06, respectively, and agreement rates of 83.3% for both dimensions. Although these two dimensions still remained at a relatively high level, the results suggest that the understanding of complex conflicts and implicit subtext still needs to be further supported by teacher explanation, classroom discussion, and students’ own performance experience.
Therefore, the system’s support for students’ script situational understanding is mainly reflected in three aspects. First, it helps students quickly grasp the character’s situation from local script excerpts. Second, it helps students transform line reading into the analysis of action goals and emotional changes. Third, it provides structured references for students’ subsequent rehearsal and classroom discussion.
B. Analysis of the System Acceptance Questionnaire
Overall, the analysis of system acceptance indicates that the system was not only regarded by students as having high learning support value, but also showed good classroom operability and potential for continued use. As shown in Table III, the total mean score of the system acceptance questionnaire was 4.42, with an overall agreement rate of 96.3%, indicating that students had a high level of overall acceptance of the system. Among the three dimensions, perceived usefulness received the highest score, with a mean value of 4.63 and an agreement rate of 100.0%. This indicates that students generally believed that the system could provide effective support for script analysis, character understanding, and performance preparation. At the same time, students’ high recognition of the system’s usefulness corresponds to the high scores for action goal understanding and emotional change understanding in the script situational understanding questionnaire, suggesting that the core functions of the system effectively responded to students’ actual needs in script situational understanding.
TABLE III. STATISTICAL RESULTS OF THE DIMENSIONS IN THE SYSTEM ACCEPTANCE QUESTIONNAIRE
| Dimension | Mean | Standard deviation | Agreement rate |
|---|---|---|---|
| Perceived usefulness | 4.63 | 0.49 | 100.0% |
| Ease of use | 4.22 | 0.58 | 92.6% |
| Continuance intention | 4.41 | 0.57 | 96.3% |
| Overall | 4.42 | 0.57 | 96.3% |
The mean score for ease of use was 4.22, with an agreement rate of 92.6%, indicating that most students were able to understand the system operation process and smoothly view the character-state card, line intention annotations, emotional change curve, and question-answering results. However, compared with perceived usefulness, the score for ease of use was slightly lower, suggesting that there is still room for further improvement in the system interface, operation guidance, or result presentation. The mean score for continuance intention was 4.41, with an agreement rate of 96.3%, indicating that most students were willing to continue using the MASA system in future script learning, classroom discussion, or performance preparation.
C. Exploratory Correlation Analysis Between the Two Questionnaires
To further explore the relationship between students’ system acceptance and their perceived support for script situational understanding, this study calculated Spearman rank correlations between the total score of system acceptance and each dimension of script situational understanding. Spearman rank correlation does not require the data to follow a normal distribution and is suitable for small samples and ordinal data. However, because this study included only nine participants, the correlation results are used only as exploratory evidence and are not used for causal inference. The analysis results are shown in Table IV. Overall, the exploratory correlation analysis indicates a consistent trend between students’ acceptance of the system and their perceived support for script situational understanding.
TABLE IV. SPEARMAN CORRELATION ANALYSIS BETWEEN SYSTEM ACCEPTANCE AND DIMENSIONS OF SCRIPT SITUATIONAL UNDERSTANDING
| Dimension of script situational understanding | Spearman ρ | p-value | Result description |
|---|---|---|---|
| Contextual understanding | 0.37 | 0.332 | Weak correlation; not significant |
| Character relationship understanding | 0.88 | 0.002 | Strong positive correlation |
| Action goal understanding | 0.88 | 0.002 | Strong positive correlation |
| Conflict and obstacle understanding | 0.76 | 0.017 | Relatively strong positive correlation |
| Subtext understanding | 0.75 | 0.021 | Relatively strong positive correlation |
| Emotional change understanding | 0.64 | 0.065 | Moderately strong correlation; approaching significance |
| Overall script situational understanding | 0.88 | 0.002 | Strong positive correlation |
As shown in Table IV, there was a strong positive correlation between the total score of system acceptance and the total score of script situational understanding, with a Spearman ρ of 0.88 and a p-value of 0.002. This indicates that students who gave higher evaluations of the system’s overall usefulness, ease of use, and continuance intention were also more likely to believe that the system helped them form a more complete understanding of the script situation.
From the perspective of specific dimensions, system acceptance showed relatively strong positive correlations with character relationship understanding, action goal understanding, conflict and obstacle understanding, and subtext understanding. Among these, the correlation coefficients for character relationship understanding and action goal understanding were both 0.88. This suggests that the more students recognized the system, the more likely they were to perceive its support in judging character relationships and analyzing characters’ action goals. Meanwhile, the correlation coefficients for conflict and obstacle understanding and subtext understanding were 0.76 and 0.75, respectively, indicating that students with higher system acceptance were more likely to believe that the system helped them understand why a character’s action was blocked and what real intention lay behind the lines.
However, the correlation between system acceptance and contextual understanding was relatively weak, with a Spearman ρ of 0.37 and a p-value of 0.332. A possible reason is that the script excerpt selected for this experiment had a relatively clear plot background. Therefore, some students could form a basic contextual understanding through text reading even without relying on the system, which may have made the perceived gain provided by the system in this dimension relatively limited.
D. Interview Data Analysis
To further explain the questionnaire results, this study conducted semi-structured interviews with nine students. The interview questions mainly focused on students’ difficulties in understanding the script before using the system, the support provided by the character-state card, the role of line intention annotation, the value of the emotional change curve, and the positioning and limitations of the system in classroom teaching. The interview data were organized using thematic analysis, and four main themes were identified: character-state cards helped establish an overall understanding, line intention annotations promoted subtext understanding, emotional change curves helped students grasp emotional progression, and the system still required supplementation through teacher guidance and classroom discussion.
TABLE V. RESULTS OF THE THEMATIC ANALYSIS OF INTERVIEWS
| Theme | Number of supporting participants | Main meaning |
|---|---|---|
| Character-state cards helped establish an overall understanding | 8/9 | Students believed that the system helped them grasp the character’s situation, interpersonal relationships, action goals, and hidden emotions |
| Line intention annotations promoted subtext understanding | 7/9 | Students believed that the system helped them understand the real intentions and action strategies behind the lines |
| Emotional change curves helped students grasp emotional progression | 6/9 | Students believed that the visualized curve helped them understand the process of emotional change |
| The system still required supplementation through teacher guidance and classroom discussion | 5/9 | Students believed that the system results had reference value, but could not replace the judgment of teachers, directors, or actors |
The interview results of table Ⅴ showed that most students believed the most obvious function of the system was to transform scattered script information into structured character states. Eight students mentioned that the character-state card helped them understand the character’s preceding circumstances, interpersonal relationships, action goals, and hidden emotions. For example, S1 stated that the system helped them understand “to what extent the character’s emotions had accumulated before entering this scene.” Meanwhile, S5 believed that the “goal in the scene” and the “main obstacle” could be directly transformed into rehearsal tasks. This is consistent with the questionnaire result that action goal understanding received the highest score, indicating that the character-state card helped students shift from plot reading to character action analysis.
In terms of subtext understanding, seven students mentioned that key-line intention annotation was helpful for understanding the character’s real intentions. S3 believed that the system reminded them to notice that some lines were not simply statements of facts, but might be attempts to test the other character. S9 also pointed out that the system helped her understand why the character did not directly express their true thoughts. These responses indicate that line intention annotation can help students move from the literal meaning of lines to the understanding of subtext and action strategies.
In terms of emotional change understanding, six students believed that the emotional change curve had intuitive value. S1 mentioned that the emotional curve allowed them to see that emotions did not suddenly erupt, but gradually developed along with the progression of scene conflicts. S7 believed that the emotional curve could help actors control emotional layers during rehearsal and avoid pushing the emotion to the highest point at the very beginning. This finding corresponds to the high score for emotional change understanding in the questionnaire, indicating that visualization can help students grasp the continuous development of a character’s emotions.
VI. DISCUSSION AND CONCLUSION
This study focused on the difficulties of script situational understanding in drama education and designed and applied a visual support system for script situational understanding and character-state analysis. Taking “scene and character” as the units of analysis, the system helps students transform local script excerpts into structured, visualized, and follow-up question-based character-state information through character-state cards, key-line intention annotations, emotional change curves, and a question-answering assistant.
The classroom application results showed that students gave generally positive evaluations of the system. The questionnaire results indicated that the system provided relatively clear support for action goal understanding, emotional change understanding, and contextual understanding. Meanwhile, the system acceptance results also showed that students generally recognized the system’s usefulness, ease of use, and value for continued use. The interview results further indicated that the character-state card helped students grasp the character’s situation and action goals, the line intention annotation helped them understand subtext, and the emotional change curve helped them understand the process through which the character’s emotions developed along with the progression of conflict.
At the same time, the results also show that the system is more suitable as an auxiliary tool for script understanding, rather than as a final interpretive tool that replaces the judgment of performance teachers or actors. For scripts such as Thunderstorm, which involve complex character relationships and rich subtext, complex conflicts and hidden emotions still need to be further supplemented through teacher explanation, classroom discussion, and students’ own performance experience. Therefore, the main value of the system lies in providing students with an analytical scaffold for entering the script situation, helping them form an initial understanding of the character more quickly, and providing references for subsequent discussion and rehearsal.
The contributions of this study are reflected in two aspects. On the one hand, this study summarizes the difficulties of script situational understanding into six aspects: insufficient context, unclear character relationships, unclear action goals, difficulty in identifying conflicts and obstacles, difficulty in recognizing subtext, and lack of an overall grasp of emotional changes. On the other hand, this study proposes and implements three system design strategies: structured character-state generation, visualized scene action analysis, and follow-up question-based script situational analysis. These strategies provide a practical example of applying LLMs to drama education.
This study has certain limitations. Most notably, the sample size was small which comprising only nine students, meaning the findings should be viewed as a preliminary case evaluation within a specific classroom setting rather than results that can be directly generalized to other grade levels, disciplines, or institutions. Furthermore, the limited sample size may compromise the stability of the statistical results, rendering metrics such as means, correlation coefficients, and significance levels susceptible to the influence of individual participants’ responses. Consequently, the quantitative findings of this study should be regarded as exploratory evidence. Future research should prioritize expanding the sample size and including participants from diverse educational backgrounds. Second, this study did not include a strict control group or a pre-test/post-test comparison, and therefore cannot directly prove that the system significantly improved students’ script situational understanding ability. Future research can expand the sample size and combine teacher evaluation, the quality of students’ script analysis assignments, and performance presentation outcomes as objective data to further verify the teaching effectiveness of the system.
Reference
[1] UNESCO. (2006). Road map for arts education: Building creative capacities for the 21st century [Conference meeting document]. World Conference on Arts Education, Lisbon, Portugal.
[2] Toivanen, T., Halkilahti, L., & Ruismaki, H. (2013). Creative pedagogy-Supporting children’s creativity through drama. The European Journal of Social & Behavioural Sciences.
[3] Celume, M.-P., Goldstein, T., Besançon, M., & Zenasni, F. (2020). Developing Children’s Socio-Emotional Competencies Through Drama Pedagogy Training: An Experimental Study on Theory of Mind and Collaborative Behavior. Europe’s Journal of Psychology, 16(4), Article e2054.
[4] Winner, E., Goldstein, T., & Vincent, S. (2013). Art for Art’s Sake? The Impact of Arts Education (Educational Research and Innovation). OECD Publishing, Paris.
http://dx.doi.org/10.1787/9789264180789-en
[5] Carnicke, S. M. (2020). Stanislavsky in focus: An acting master for the twenty-first century. Routledge.
[6] Hagen, U. (2023). Respect for Acting: Expanded Version. John Wiley & Sons.
[7] Thomas, J. (2013). Script analysis for actors, directors, and designers. Routledge.
[8] Cain, A. (2022, April 26). What are “sides” in acting? Backstage. https://www.backstage.com/magazine/article/what-are-sides-in-acting-72452/
[9] Marsden, S. (2022, July 28). Audition sides and how to prepare them. Spotlight. https://www.spotlight.com/news-and-advice/the-essentials/audition-sides-and-how-to-prepare-them/
[10] Tian, Q., Li, Y., Chen, F., Liu, Z., Kong, Y., Guo, F., … & Zhang, X. (2026). STAGE: A Benchmark for Knowledge Graph Construction, Question Answering, and In-Script Role-Playing over Movie Screenplays. arXiv preprint arXiv:2601.08510.
[11] Bruder, M., Cohn, L. M., Olnek, M., Pollack, N., Previto, R., & Zigler, S. (1986). A practical handbook for the actor. Vintage.
[12] Schneider, J. J., Crumpler, T. P., & Rogers, T. (2006). Process drama and multiple literacies. Portsmouth, NH.
[13] Bai, Y., & Li, X. (2025). Development and Application of an Automatic Recognition Model for Film and Television Education of Movie Shot Category. Bull. Tech. Comm. Learn. Technol, 25, 26-34.
[14] Baruah, S., & Narayanan, S. (2023, July). Character coreference resolution in movie screenplays. In Findings of the Association for Computational Linguistics: ACL 2023 (pp. 10300-10313).
[15] Bommasani, R., Hudson, D. A., Adeli, E., Altman, R., Arora, S., von Arx, S., … & Liang, P. (2021). On the opportunities and risks of foundation models. arXiv preprint arXiv:2108.07258.
[16] Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J. D., Dhariwal, P., … & Amodei, D. (2020). Language models are few-shot learners. Advances in neural information processing systems, 33, 1877-1901.
[17] Qin, L., Chen, Q., Feng, X., Wu, Y., Zhang, Y., Li, Y., … & Yu, P. S. (2026). Large language models meet nlp: A survey. Frontiers of Computer Science, 20(11), 2011361.
[18] Bolton, G. (1985). Changes in thinking about drama in education. Theory into practice, 24(3), 151-157.
[19] O’Neill, C. (1995). Drama worlds: A framework for process drama. Heinemann.
[20] National Coalition for Core Arts Standards. (2013). Visual/aural composition of design concept: Theatre model cornerstone assessment (HS Accomplished II). State Education Agency Directors of Arts Education. https://deartsstandards.org/sites/default/files/Theatre%20MCA/Theatre%20MCA/Theatre%20HS%20Accomplished%205-15-14.pdf
[21] Davis, D. (2014). Imagining the Real: Towards a New Theory of Drama in Education. Trentham Books. Available from: IOE Press, Institute of Education, 20 Bedford Way, London, WC1H 0AL, UK.
[22] Welksek, G. (2022). Drama Education. Routledge.
[23] Baumrin, S. (2008). An Actor’s Work: A Student’s Diary. Theatre Journal, 60(4), 689-690.
[24] Bucs, H. H., & Pye, V. C. (Eds.). (2019). Objectives, Obstacles, and Tactics in Practice: Perspectives on Activating the Actor. Routledge.
[25] Hagen, U. (1991). Challenge for the Actor. Simon and Schuster.
[26] Dewey, J. (1933). How we think: A restatement of the relation of reflective thinking to the educative process. D.C. Heath and Co.
[27] Zhao, F. (2026). Actor-role-assistant [Computer software]. GitHub. https://github.com/zhaofz635/actor-role-assistant
Authors
Binglin Zhuang
He was born in Changchun, Jilin Province, China, on June 19, 2002. He is currently pursuing the master’s degree in Arts Studies at the College of Media Science (School of Journalism), Northeast Normal University, Changchun, China. His research interests include drama and film education.
Mengqi Yu
She was born in Dalian, Liaoning Province, China, on August 6, 2005. She is currently pursuing the bachelor’s degree in Broadcasting and Hosting Art at the College of Media Science / School of Journalism, Northeast Normal University, Changchun, China. Her research interests include drama, film and television performing arts.
Weiyang Liu
She was born in Shaoyang, Hunan Province, China, on August 20, 1998. She is currently pursuing the Ph.D. degree in Literature and Art Theory at the School of Chinese Language and Literature, Northeast Normal University, Changchun, China. Her research interests include digital media technology and large language models.