The torn notice at the crossing that taught a computer to read better
I stood at the crossing, staring at a torn notice taped to a pole. The middle was gone, but the top line and the bottom line were still there. I guessed the missing words by using both ends, like reading with both eyes open.
Older computer readers were more one-way. They’d follow the line from the start and guess before they’d “seen” what came later. Some tried a backwards pass too, but it felt like taping two half-reads together and hoping they matched.
Then someone taught a new reader by messing up sentences on purpose. A few words were covered, some were swapped for the wrong ones, and some were left looking normal but still treated like a test. Like my notice with bits taped over, it had to use the words around the gap, not just copy what was there.
Here’s what was new: it learned with both sides of the sentence available at the same time, all the way through its reading. At the crossing, the bottom line helped me decide if the missing word was “closed” or “crowded”. Takeaway: looking both ways cuts down daft guesses.
It also got practice at spotting whether two bits of writing truly belonged together. Imagine a second page under my notice: sometimes it’s the real next part, sometimes it’s a random page someone slapped on. The reader learned to tell when the pages actually fit.
After all that practice, the same reader could be reused for lots of jobs with only a small extra piece added at the end. It’s like keeping my notice-reading habit, then changing the question: fill the gap, pick out a person’s name, or find the line that answers a question.
When the lights changed, I wasn’t inventing a brand-new trick for every notice. I just kept using both ends of the message and checking if the next page belonged. That’s why this kind of reader mattered: it learned a general, two-sided way of reading first, then carried it into many everyday text tools.