Language-Driven Action-Adventure Game - Design Process
Language-Driven Action-Adventure Game - Design Process
This project is a 2D exploration focused action-adventure game with progression driven by learning a fictional language through play. Players begin unable to understand the world around them and gradually learn to interpret dialogue and environmental text through exploration, NPC interaction, and in-game discoveries.
Progression is driven not by changing the world itself, but by transforming the player’s understanding of it, allowing the world to open up to the player as their knowledge increases.
This project explores using language as the primary mechanism of exploration, progression, and the ability to overcome isolation in an adventure game.
Rather than using traditional gated exploration based on character abilities, progression is tied to comprehension and learning. The player not only unlocks new movement options and tools, but also gains perspective and new meaning from the systems they've already interacted with.
This reframes exploration as a cognitive process while still preserving the expected interactions of traditional action-adventure gameplay.
The player explores interconnected environments while encountering NPC dialogue and environmental text that is written in a unique fictional language. Over time, revisiting NPC dialogue with new context allows the player to recognize patterns, discover meaning, and gradually build language comprehension.
Previously visited areas remain structurally unchanged but gain new significance as understanding improves.
The loop is designed so that learning the language is inseparable from navigating the world.
The language system is built in Python using spaCy and NLTK WordNet, and is integrated into Unity through a state-based translation pipeline.
Each dialogue entry exists in multiple comprehension states. Early in the game, the displayed text is fully rendered in the fictional language. As progression increases, words and grammatical structures are gradually revealed until understanding is achieved.
Under the hood, token parsing and semantic relationships from WordNet ensure consistency across states. Synonyms and hypernyms are used to preserve meaning even as surface-level language changes.
The intent is to maintain the meaning in the world and NPC dialogue while shifting only the player’s interpretation of every encounter.
The game is structured as a connected set of 2D rooms designed around backtracking and revisitation rather than linear progression.
Early areas remain relevant throughout the game experience. As comprehension improves, players naturally return to these familiar spaces and reinterpret previously opaque dialogue, signage, and environmental cues.
For example, a previously unreadable sign instructs the player to dive into the ground at a specific location. The mechanic was always available, but the player simply lacked the contextual understanding to recognize its importance.
Navigation, storytelling, and language learning are intertwined throughout the game so that even world traversal contributes to player understanding.
A key design constraint was avoiding a binary fully translated vs. fully untranslated system.
Instead, language clarity evolves gradually across five progression states. Untranslated text is always present, but becomes increasingly interpretable by the player in controlled stages until the final stage where all NPC dialogue is fully in English.
To make learning and progression feel organic and rewarding to the player, feedback is designed to reinforce discovery rather than correctness. The system emphasizes gradual improvement in understanding. For example, some discoveries can be made through partial understanding alone, allowing players to act before achieving full comprehension.
The project was developed using iterative prototyping with frequent playtesting and feedback.
Early playtesting revealed two major issues:
Players struggled when early language exposure was too opaque
Later progression felt too abrupt when language comprehension unlocked too quickly
Iteration focused on:
Highlighting key vocabulary terms to reinforce recognition and progression pacing
Introducing earlier context for pattern recognition
Reordering NPC placement to encourage discovery-driven understanding rather than prematurely resolving unanswered questions
Improving consistency of NPC language usage to reinforce learning signals
Navigation was also refined, since language ambiguity directly impacts the players' spatial understanding.
This project involved several intentional tradeoffs between clarity, immersion, and systemic depth.
Readability vs. authenticity
A fully realistic language system would introduce depth, but reduces player readability.
I decided to prioritize pattern recognition over realism to ensure players can form early hypotheses (highlighting important untranslated text in bold).
Static vs. dynamic world transformation
The alternative approach would be changing environments based on language mastery.
I decided to keep the world static and shift only interpretation, reinforcing the theme that understanding is what changes.
Explicit translation UI vs. embedded comprehension system
Traditional UI could have shown direct translations or glossaries.
I decided to embed understanding into the dialogue system itself to simplify systems for the sake of demo clarity.
Pacing vs. reward feedback
Slower unlocks would improve depth but decrease readability in a small scale demo.
I decided to speed language comprehension discovery through book pickups to properly get across intention in a small scale demo.
While the current system demonstrates language as a progression mechanic, there are several directions for future improvement.
Allowing players to guess words based on context clues would help improve immersion and keep track of internal player thinking.
Implementing more obvious grammatical clues could ensure future iterations better distinguish between grammatical rules versus vocabulary.
Adding more NPC based progression would allow different characters to teach the player specific words and clues rather than progression coming from simple pickups.
Implementing word reinterpretation and regional meanings would allow NPCs to express language differently based on their personality or the area they are found in, making comprehension feel contextual and adding depth and re-discovery.
Introducing controlled misinterpretation could add depth and immersion to interactions and strengthen NPC-player interactions.
Navigation could be vastly improved by an in-game map and clearer visual distinction between rooms.
The system is implemented in Unity as a 2D exploration framework, with external Python-based language processing handling semantic structure and progression logic.
SpaCy and NLTK WordNet provide semantic relationships that support consistent meaning mapping across translation states. Unity manages runtime state transitions and dialogue rendering based on player comprehension level.
The architecture is modular, separating language processing, progression logic, and gameplay systems to support rapid iteration and tuning.
This project demonstrates a progression model where gameplay advancement is defined not by mechanical expansion, but by increasing interpretive depth.
Exploration and learning become innately linked, allowing the world to gradually open up through player understanding.
This project also included a published research paper documenting the system architecture, design decisions, and iterative development process.
https://scholarcommons.scu.edu/cseng_senior/339/
Some sprites were sourced from publicly available libraries under license. All gameplay systems, 3D models, interaction design, progression logic, and implementation were independently developed as part of this project.
Attribution links are provided below:
[1] Pixel Frog, ”Pixel Adventure,” Itch.io. [Online]. Available: https://pixelfrog-assets.itch.io/pixel-adventure-1.
[2] Szym, ”SpearFishing assets pack,” Itch.io. [Online]. Available: https://nszym.itch.io/spearfishing-assets-pack.
[3] CraftPix, ”Water Enemies: Octopus, Jellyfish, Shark, and Turtle,” CraftPix.net. [Online]. Available: https://craftpix.net/freebies/octopus-jellyfish-shark-and-turtle-free-sprite-pixel-art/.
[4] Flixberry Entertainment, ”Heart Pixel Art,” OpenGameArt.org. [Online]. Available: https://opengameart.org/content/heart-pixel-art.
[5] CreativeKind, ”Flying Obelisk,” Itch.io. [Online]. Available: https://creativekind.itch.io/flying-obelisk.
[6] GX310, ”Sign Posts,” Itch.io. [Online]. Available: https://gx310.itch.io/sign-posts.
[7] malekith97, ”Crab,” OpenGameArt.org. [Online]. Available: https://opengameart.org/content/crab.
[8] serenajeanne, ”Bubbling Potion Bottle,” OpenGameArt.org. [Online]. Available: https://opengameart.org/content/bubbling-potion-bottle.