Interactive Narrative Personalization with Deep Reinforcement Learning

Authors: Pengcheng Wang, Jonathan Rowe, Wookhee Min, Bradford Mott, James Lester

IJCAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We investigate the deep RL framework s performance with an educational interactive narrative, CRYSTAL ISLAND. Results suggest that the deep RL-based narrative generation framework yields effective personalized interactive narratives. By evaluating the Q-network based interactive narrative planner in a widely used educational interactive narrative, CRYSTAL ISLAND, we demonstrate that deep RL yields more effective policies for interactive narrative personalization than commonly utilized linear RL techniques.
Researcher Affiliation Academia Pengcheng Wang, Jonathan Rowe, Wookhee Min, Bradford Mott, James Lester Department of Computer Science, North Carolina State University, Raleigh, NC 27695, USA {pwang8, jprowe, wmin, bwmott, lester}@ncsu.edu
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any concrete access information for source code.
Open Datasets No The paper mentions using a "human player interaction corpus for CRYSTAL ISLAND" generated from human subject studies but does not provide any access information (link, DOI, repository, or citation to a public source) for this dataset.
Dataset Splits No The paper states that data is divided into training (80%) and test (20%) sets but does not explicitly mention a separate validation set.
Hardware Specification No The paper does not provide specific hardware details (e.g., exact GPU/CPU models, memory amounts) used for running experiments, only general mentions like "without GPU acceleration" and "on 4 threads".
Software Dependencies No The paper mentions the use of 'Adam' optimizer and 'LSTMs' but does not provide specific version numbers for any software dependencies.
Experiment Setup Yes For all of these structures, Hidden Layer 1 has 64 neurons, and Hidden Layer 2 (either fully connected layer or LSTM) has 32 neurons. The advantage stream and state value stream both contain 32 neurons in dueling structures. Discount factor γ is set to 1.