Index/ Legal/ Splitting a Contract the Right Way
Legal AI OS — NLP preprocessing

Splitting a Contract the Right Way

The same contract, cut two ways — one that breaks sentences apart, one that keeps them whole.

01

Cut at Every Period

Before — the old way

Even periods that aren’t the end of a sentence

A Contract Read the Words Cut at Every Period Broken Pieces “Acme Corp.” / “§4.2” / “$1,250,000.00” Make It Searchable Searchable Library Wrong Piece Retrieved AI Answer

Cut at every period, period. In a contract, a period is as likely to sit inside “Inc.”, “§4.2”, or “$1,250,000.00” as at the end of a sentence. So the text gets cut in the middle of a clause — and that broken piece is what the search library stores.

02

Clean It, Then Cut at Real Breaks

After — the fix

Two small changes keep whole clauses intact

A Contract Read the Words Clean the Text drop page numbers, repeated headers, stray spaces Knows These Aren’t Ends Inc. / §4.2 / decimals / U.S. Cut at Real Sentence Ends Whole Clauses nothing cut in half Make It Searchable Searchable Library Right Clause Retrieved AI Answer

Same steps, two small changes. First clean the text so it isn’t carrying page numbers and junk. Then cut only at real sentence ends — the kind that know an abbreviation or a section number isn’t the end of a thought.

03

The two changes

First — clean it
Strip the noise
Drop page numbers, repeated headers, and stray spaces before anything gets searched. Clean text searches better, and the cleanup is free and instant.
Second — cut at real breaks
Respect the sentence
Know the difference between a real sentence end and an abbreviation, a section number, or a dollar amount. So pieces stay whole instead of getting cut in half.
04

What changes

Three wins, all before the AI ever reads the text

Better search
results
Whole clauses instead of half a section number. The AI finds the actual provision, not a fragment.
A trail you
can trust
The same text gives the same result, every time. You can show your work.
Less AI cost
Splitting text is free and instant. You only pay the AI for judgment, not for deciding where a sentence ends.