1. Duplicate the "bike" layer.
2. Create a mask on the top layer. The mask should outline the girl on bike. You only need to mask the section of the girl that the words will go behind. Don't worry about being to precise... if your mask is a little off, you probably won't notice it when it's played in real time.
3. Make a keyframe for "mask shape" at the first frame of the comp.
4. Move forward 5 or 10 frames and readjust your mask like you did in step 2. Depending on how much the girl moves around on the screen, keyframes might need to be made more often or less often than this.
5. When you're finished keyframing the "mask shape" across the entire comp, put your text layer between the 2 bike layers.
6. Keyframe the text position to make it move across the screen.
Hope this helps.