ContentsThe library

Finding Things by Meaning

The Two Methods Fail on Opposite Questions

Last timeWhat Actually Gets Compared

Matching on meaning finds paraphrases and loses part numbers. Matching on words does the reverse. Running both and merging the lists is close to free and fixes most of each.

A customer writes in: part MX-4471B is making a noise. The manual has a page

about MX-4471B. Meaning-based retrieval frequently fails to find it.

This is not a bug. The piece was reduced to a few hundred numbers chosen to place

it near things about the same subject. Nothing in that process was asked to

preserve one eight character string, and in a collection with four hundred part

numbers of the same shape, it generally does not.

What the older method does

Matching on words is the method that predates all of this, and it is extremely

good at exactly the thing meaning matching is bad at.

FIG 1What one question word contributes to one piece
rarity: a word in few pieces of the collection counts for a lot, a word in most of them counts for almost nothing
frequency, flattened, with a correction b for how long the piece is
The rarity term is why this method finds part numbers. A string appearing in one piece out of a million gets an enormous weight, so the single piece containing it wins outright. The same term is why the words how and much contribute nothing at all.

The three ingredients are worth stating plainly. How often the word appears, so a

page about the part beats a page mentioning it once. How rare the word is, so

rare words dominate. How long the piece is, so a long piece does not win just by

containing more words in general.

None of these three are learned. They are counted from the collection you

already have, which has two consequences worth noticing. There is no training

step, no model to choose and no domain mismatch to worry about, so this method

works equally well on case law and chip specifications as on web prose. And it

has no notion whatsoever of meaning, so a question answered in different words is

invisible to it no matter how much text you give it.

The lesson stops here

4 more paragraphs to go

You have read the opening. The rest of the argument, the problems that check whether it landed, and the lines worth keeping at the end all come with a plan.

The first lesson of every course in the library reads the whole way through, free, so you can see exactly what the rest of them are.

See the planThe contents

This is the reading half

Starting the course gives you your own copy of it. Every idea on every page has problems standing under it, marked with a reason rather than a tick, and any sentence you do not believe can be opened and argued with. None of that can happen on a page nobody owns.

The contents

The rest of this course

  1. 01The Step That Decides Everything After It
  2. 02The Decision Made Before Anything Is Searchedopening only
  3. 03Neither Side of the Comparison Is Textopening only
  4. 04The Two Methods Fail on Opposite Questionsyou are here
  5. 05You Cannot Compare the Question Against Everythingopening only
  6. 06One Knob, and What It Is Really Selling Youopening only
  7. 07The Stage That Can Afford to Be Slowopening only
  8. 08A Hundred Questions You Wrote Yourselfopening only

Read alongside