With no explanation, label text_A→text_B with either "strengthener" or "weakener".
text_A: A man in an orange jacket and white hat looks off-camera. A man smiling at the camera
text_B: The man is crying.
weakener.