本論文では,マルコフ確率場を用いた自動作曲を提案する.ある作曲家の曲を,マルコフ確率場の特徴関数と重みで表現することで特徴を学習し,この確率場から確率の高い標本をモンテカルロ法により得ることで自動作曲を行う. / In this paper, we propose an automatic composition method for polyphony music using Markov random fields. Given pieces of music by a composer, we learn the characteristics of the pieces in the form of feature functions and their weights for Markov random fields. Then, a new piece is automatically composed by obtaining a sample with high probability for the random field using a Monte Carlo method.