First, effective writing will almost always restate the main idea somehow; you'll want your algorithm to be more sophisticated than "demote content that seems to repeat itself."
Second, your novel algorithm will be expensive to compute, so you'll need some other signals to filter out very low value pages before you pay to analyze others.
Finally, your technique will be valuable as a ranking signal until you (or someone else) begins to use your algorithm to help steer non trivial purchasing decisions. At that point, someone will dress up their low-value (or negative value) content to appear "high value" to your algorithm and your output will be filled with spam. You can delay this effect by keeping your user-base small, so consider selling private access to institutions who want a unique view of the internet for research or intelligence purposes.
Good luck!