In-browser library classification

This is a web application that can classify any arbitrary text into Library of Congress subclasses or some other library classification schemes. As you wait, your browser will download a neural network trained using Google's TensorFlow; you can then paste any text into the box below, and click "classify" to see what subject areas the classifier thinks it belongs to. Your computer will locally process the text into an SRP projection that the neural network can read, and then infer a number of classes. The numbers are expressed as certainties; there will be a guess for any text, regardless of whether it resembles a library book or not.

Click one of these buttons to load a different model.

goin' to be more'n two hours befo' this wrack breaks up and washes off down the river.  See? He'll be drownded, and won't have nobody to blame for it but his own self.  I reckon that's a considerble sight better 'n killin' of him.  I'm unfavorable to killin' a man as long as you can git aroun' it; it ain't good sense, it ain't good morals.  Ain't I right?"
"Yes, I reck'n you are.  But s'pose she don't break up and wash off?"
"Well, we can wait the two hours anyway and see, can't we?"
"All right, then; come along."
So they started, and I lit out, all in a cold sweat, and scrambled forward. It was dark as pitch there; but I said, in a kind of a coarse whisper, "Jim!" and he answered up, right at my elbow, with a sort of a moan, and I says:
"Quick, Jim, it ain't no time for fooling around and moaning; there's a gang of murderers in yonder, and if we don't hunt up their boat and set her drifting down the river so these fellows can't get away from the wreck there's one of 'em going to be in a bad fix.  But if we find their boat we can put all of 'em in a bad fix—for the sheriff 'll get 'em. Quick—hurry!  I'll hunt the labboard side, you hunt the stabboard. You start at the raft, and—"
"Oh, my lordy, lordy!  raf'?  Dey ain' no raf' no mo'; she done broke loose en gone I—en here we is!"
CHAPTER XIII.
WELL, I catched my breath and most fainted.  Shut up on a wreck with such a gang as that!  But it warn't no time to be sentimentering.  We'd got to find that boat now—had to have it for ourselves.  So we went a-quaking and shaking down the stabboard side, and slow work it was, too—seemed a week before we got to the stern.  No sign of a boat.  Jim said he didn't believe he could go any further—so scared he hadn't hardly any strength left, he said.  But I said, come on, if we get left on this wreck we are in a fix, sure.  So on we prowled again.  We struck for the stern of the texas, and found it, and then scrabbled along forwards on the skylight, hanging on from shutter to shutter, for the edge of the skylight was in the water.  When we got pretty close to the cross-hall door there was the skiff, sure enough!  I could just barely see her.  I felt ever so thankful.  In another second I would a been aboard of her, but just then the door opened.  One of the men stuck his head out only about a couple of foot from me, and I thought I was gone; but he jerked it in again, and says:
"Heave that blame lantern out o' sight, Bill!"
He flung a bag of something into the boat, and then got in himself and set down.  It was Packard.  Then Bill he come out and got in.  Packard says, in a low voice:
"All ready—shove off!"
I couldn't hardly hang on to the shutters, I was so weak.  But Bill says:
"Hold on—'d you go through him?"
"No.  Didn't you?"
"No.  So he's got his share o' the cash yet."
"Well, then, come along; no use to take truck and leave money."
"Say, won't he suspicion what we're up to?"
"Maybe he won't.  But we got to have it anyway. Come along."
So they got out and went in.
The door slammed to because it was on the careened side; and in a half second I was in the boat, and Jim come tumbling after me.  I out with my knife and cut the rope, and away we went!
We didn't touch an oar, and we didn't speak nor whisper, nor hardly even breathe.  We went gliding swift along, dead silent, past the tip of the paddle-box, and past the stern; then in a second or two more we was a hundred yards below the wreck, and the darkness soaked her up, every last sign of her, and we was safe, and knowed it.
When we was three or four hundred yards down-stream we see the lantern show like a little spark at the texas door for a second, and we knowed by that that the rascals had missed their boat, and was beginning to understand that they was in just as much trouble now as Jim Turner was.
Then Jim manned the oars, and we took out after our raft.  Now was the first time that I begun to worry about the men—I reckon I hadn't had time to before.  I begun to think how dreadful it was, even for murderers, to be in such a fix.  I says to myself, there ain't no telling but I might come to be a murderer myself yet, and then how would I like it?  So says I to Jim:
"The first light we see we'll land a hundred yards below it or above it, in a place where it's a good hiding-place for you and the skiff, and then I'll go and fix up some kind of a yarn, and get somebody to go for that gang and get them out of their scrape, so they can be hung when their time comes."
But that idea was a failure; for pretty soon it begun to storm again, and this time worse than ever.  The rain poured down, and never a light showed; everybody in bed, I reckon.  We boomed along down the river, watching for lights and watching for our raft.  After a long time the rain let up, but the clouds stayed, and the lightning kept whimpering, and by and by a flash showed us a black thing ahead, floating, and we made for it.
It was the raft, and mighty glad was we to get aboard of it again.  We seen a light now away down to the right, on shore.  So I said I would go for it. The skiff was half full of plunder which that gang had stole there on the wreck.  We hustled it on to the raft in a pile, and I told Jim to float along down, and show a light when he judged he had gone about two mile, and keep it burning till I come; then I manned my oars and shoved for the light.  As I got down towards it three or four more showed—up on a hillside.  It was a village.  I closed in above the shore light, and laid on my oars and floated.  As I went by I see it was a lantern hanging on the jackstaff of a double-hull ferryboat.  I skimmed around for the watchman, a-wondering whereabouts he slept; and by and by I found him roosting on the bitts forward, with his head down between his knees.  I gave his shoulder two or three little shoves, and begun to cry.
He stirred up in a kind of a startlish way; but when he see it was only me he took a good gap and stretch, and then he says:
"Hello, what's up?  Don't cry, bub.  What's the trouble?"

For instance, the default text is a segment from Huckleberry Finn, which the browser doesn't recognize as American Literature; instead it gives the books original classification PZ, "Fiction and Juvenile belles lettres." Neuralnetworks have a reputation for model uninterpretability, which is only partially fair; although the *model* as a whole is uninterpretable, individual decisions can be explained. If you click "Introspect Model," you can see an explanation of why the network made the decision that it did; the first pair of columns explains why it thought the top answer was plausible at all (or what the strongest negative scores were), while the second two give show it chose the answer it did relative to the other top two or three contenders. (In a multiclass problem like this, both are important: a word like "Paris" might increase the probability of something both American History compared to, say, aluminum processing; but decrease the chance of it being American history compared to French history.) These numbers are inferred by re-running the network many times, each time with a single word dropped out, to see how the output weights differ. They give some sense of what vocabulary the neural network has "learned," although some results may simply be random noise.