PersonaGuess
PersonaGuessDataset loads turn-based games where two personas ask each other questions.
Loader
from persona_data.persona_guess import PersonaGuessDataset
games = PersonaGuessDataset()
small = PersonaGuessDataset(sample_size=10)
The default dataset source is implicit-personalization/persona-guess. The loader reads games.jsonl.
sample_size keeps the leading games rather than sampling randomly.
Records
GameRecord: one match between two personasTurn: one question-answer turn within a game
Game fields
GameRecord includes:
game_idpersona_a_idpersona_b_idturns
Turn includes:
roundaskerquestionanswer
Queries
game = games[0]
turns = games.get_qa(game.game_id)
turns = games.get_qa(game.game_id, player="A")
questions = games.questions(game.game_id, player="B")
get_qa() returns typed Turn records. questions() returns question strings only.
Notes
playercan be"A","B", or omitted for all turns.sample_sizecan limit how many games are kept in memory.sample_sizekeeps a leading slice rather than sampling randomly.- The dataset is small enough to load eagerly into memory.