Text this: PASeg: positional-guided segmenter with multimodal semantic alignment for enhancing urban scene 3D semantic segmentation