Masked Point-Entity Contrast for Open-Vocabulary 3D Scene Understanding | IEEE Conference Publication | IEEE Xplore