The Go Transformer: Natural Language Modeling for Game Play

  • 2020-09-07 19:37:21
  • Matthew Ciolino, David Noever, Josh Kalin
  • 0


This work applies natural language modeling to generate plausible strategicmoves in the ancient game of Go. We train the Generative Pretrained Transformer(GPT-2) to mimic the style of Go champions as archived in Smart Game Format(SGF), which offers a text description of move sequences. The trained modelfurther generates valid but previously unseen strategies for Go. Because GPT-2preserves punctuation and spacing, the raw output of the text generatorprovides inputs to game visualization and creative patterns, such as the Sabakiproject's game engine using auto-replays. Results demonstrate that languagemodeling can capture both the sequencing format of championship Go games andtheir strategic formations. Compared to random game boards, the GPT-2fine-tuning shows efficient opening move sequences favoring corner play overless advantageous center and side play. Game generation as a language modelingtask offers novel approaches to more than 40 other board games where historicaltext annotation provides training data (e.g., Amazons & Connect 4/6).


Quick Read (beta)

loading the full paper ...