/* wordChain - Create Markov chain of words. */ #include "common.h" #include "linefile.h" #include "hash.h" #include "localmem.h" #include "options.h" #include "dlist.h" #include "rbTree.h" /* Global vars - all of which can be set by command line options. */ int maxChainSize = 3; int maxNonsenseSize = 10000; int minUse = 1; boolean lower = FALSE; boolean unpunc = FALSE; boolean fullOnly = FALSE; void usage() /* Explain usage and exit. */ { errAbort( "wordChain - Create Markov chain of words\n" "usage:\n" " wordChain in.txt\n" "options:\n" " -size=N - Set max chain size, default %d\n" " -chain=fileName - Write out word chain to file\n" " -nonsense=fileName - Write out predicted nonsense to file\n" " -maxNonsenseSize=N - Keep nonsense output to this many words.\n" " -lower - Lowercase all words\n" " -unpunc - Strip punctuation\n" " -fullOnly - Only output chains of size\n" " -minUse=N - Set minimum use in output chain, default %d\n" , maxChainSize, minUse ); } /* Command line validation table. */ static struct optionSpec options[] = { {"size", OPTION_INT}, {"minUse", OPTION_INT}, {"nonsense", OPTION_STRING}, {"chain", OPTION_STRING}, {"lower", OPTION_BOOLEAN}, {"unpunc", OPTION_BOOLEAN}, {"fullOnly", OPTION_BOOLEAN}, {"maxNonsenseSize", OPTION_INT}, {NULL, 0}, }; /* The wordTree structure below is the central data structure for this program. It is * used to build up a tree that contains all observed N-word-long sequences observed in * the text, where N corresponds to the "size" command line option which defaults to 3, * an option that in turn is stored in the maxChainSize variable. At this chain size the * text * this is the black dog and the black cat * would have the chains * this is the * is the black * the black dog * black dog and * dog and the * and the black * the black cat * and turn into the tree * this * is * the * is * the * black * the * black * dog * cat * black * dog * and * dog * and * the * and * the * black * Note how the tree is able to compress the two chains "the black dog" and "the black cat." * * A node in the tree can have as many children as it needs to at each node. The depth of * the tree is the same as the chain size, by default 3. At each node in the tree you get * a word, and a list of all words that are observed in the text to follow that word. * * There are special cases in the code so that the first and last words in the text get included * as much as possible in the tree. * * Once the program has build up the wordTree, it can output it in a couple of fashions. */ struct wordTree /* A node in a tree of words. The head of the tree is a node with word value the empty string. */ { struct rbTree *following; /* Contains words (as struct wordTree) that follow us. */ char *word; /* The word itself including comma, period etc. */ int useCount; /* Number of times word used. */ }; struct wordTree *wordTreeNew(char *word) /* Create and return new wordTree element. */ { struct wordTree *wt; AllocVar(wt); wt->word = cloneString(word); return wt; } int wordTreeCmpWord(void *va, void *vb) /* Compare two wordTree. */ { struct wordTree *a = va, *b = vb; return strcmp(a->word, b->word); } struct wordTree *wordTreeAddFollowing(struct wordTree *wt, char *word, struct lm *lm, struct rbTreeNode **stack) /* Make word follow wt in tree. If word already exists among followers * return it and bump use count. Otherwise create new one. */ { struct wordTree *w; /* Points to following element if any */ if (wt->following == NULL) { /* Allocate new if you've never seen it before. */ wt->following = rbTreeNewDetailed(wordTreeCmpWord, lm, stack); w = NULL; } else { /* Find word in existing tree */ struct wordTree key; key.word = word; w = rbTreeFind(wt->following, &key); } if (w == NULL) { w = wordTreeNew(word); rbTreeAdd(wt->following, w); } w->useCount += 1; return w; } void addChainToTree(struct wordTree *wt, struct dlList *chain, struct lm *lm, struct rbTreeNode **stack) /* Add chain of words to tree. */ { struct dlNode *node; wt->useCount += 1; for (node = chain->head; !dlEnd(node); node = node->next) { char *word = node->val; verbose(2, " %s\n", word); wt = wordTreeAddFollowing(wt, word, lm, stack); } } void wordTreeDump(int level, struct wordTree *wt, FILE *f) /* Write out wordTree to file. */ { static char *words[64]; struct slRef *list, *ref; int i; assert(level < ArraySize(words)); words[level] = wt->word; if (wt->useCount >= minUse) { if (!fullOnly || level == maxChainSize) { fprintf(f, "%d\t", wt->useCount); for (i=1; i<=level; ++i) fprintf(f, "%s ", words[i]); fprintf(f, "\n"); } } if (wt->following != NULL) { list = rbTreeItems(wt->following); for (ref = list; ref != NULL; ref = ref->next) wordTreeDump(level+1, ref->val, f); slFreeList(&list); } } int totalUses = 0; int curUses = 0; int useThreshold = 0; char *pickedWord; void addUse(void *v) /* Add up to total uses. */ { struct wordTree *wt = v; totalUses += wt->useCount; } void pickIfInThreshold(void *v) /* See if inside threshold, and if so store it in pickedWord. */ { struct wordTree *wt = v; int top = curUses + wt->useCount; if (curUses <= useThreshold && useThreshold < top) pickedWord = wt->word; curUses = top; } char *pickRandomWord(struct rbTree *rbTree) /* Pick word from list randomly, but so that words more * commonly seen are picked more often. */ { pickedWord = NULL; curUses = 0; totalUses = 0; rbTreeTraverse(rbTree, addUse); useThreshold = rand() % totalUses; rbTreeTraverse(rbTree, pickIfInThreshold); assert(pickedWord != NULL); return pickedWord; } char *predictNext(struct wordTree *wt, struct dlList *recent) /* Predict next word given list of recent words and wordTree. */ { struct dlNode *node; for (node = recent->head; !dlEnd(node); node = node->next) { char *word = node->val; struct wordTree key; key.word = word; wt = rbTreeFind(wt->following, &key); if (wt == NULL) errAbort("%s isn't a follower of %s\n", word, wt->word); } char *result = NULL; if (wt->following != NULL) result = pickRandomWord(wt->following); return result; } static void wordTreeMakeNonsense(struct wordTree *wt, int maxSize, char *firstWord, int maxOutputWords, FILE *f) /* Go spew out a bunch of words according to probabilities in tree. */ { struct dlList *ll = dlListNew(); int listSize = 0; int outputWords = 0; for (;;) { if (++outputWords > maxOutputWords) break; struct dlNode *node; char *word; /* Get next predicted word. */ if (listSize == 0) { AllocVar(node); ++listSize; word = firstWord; } else if (listSize >= maxSize) { node = dlPopHead(ll); word = predictNext(wt, ll); } else { word = predictNext(wt, ll); AllocVar(node); ++listSize; } node->val = word; dlAddTail(ll, node); if (word == NULL) break; /* Output last word in list. */ { node = ll->tail; word = node->val; fprintf(f, "%s", word); if (word[strlen(word)-1] == '.') fprintf(f, "\n"); else fprintf(f, " "); } } dlListFree(&ll); } struct wordTree *wordTreeForChainsInFile(char *fileName, int chainSize, struct lm *lm) /* Return a wordTree of all chains-of-words of length chainSize seen in file. * Allocate the structure in local memory pool lm. */ { /* Stuff for processing file a line at a time. */ struct lineFile *lf = lineFileOpen(fileName, TRUE); char *line, *word; /* We'll keep a chain of three or so words in a doubly linked list. */ struct dlNode *node; struct dlList *chain = dlListNew(); int curSize = 0; /* We'll build up the tree starting with an empty root node. */ struct wordTree *wt = wordTreeNew(""); int wordCount = 0; /* Save time/space by sharing stack between all "following" rbTrees. */ struct rbTreeNode **stack; lmAllocArray(lm, stack, 256); /* Loop through each line of input file, lowercasing the whole line, and then * looping through each word of line, stripping out special chars, and finally * processing each word. */ while (lineFileNext(lf, &line, NULL)) { if (lower) tolowers(line); while ((word = nextWord(&line)) != NULL) { if (unpunc) { stripChar(word, ','); stripChar(word, '.'); stripChar(word, ';'); stripChar(word, '-'); stripChar(word, '"'); stripChar(word, '?'); stripChar(word, '!'); stripChar(word, '('); stripChar(word, ')'); if (word[0] == 0) continue; } verbose(2, "%s\n", word); /* We come to this point in the code for each word in the file. * Here we want to maintain a chain of sequential words up to * chainSize long. We do this with a doubly-linked list structure. * For the first few words in the file we'll just build up the list, * only adding it to the tree when we finally do get to the desired * chain size. Once past the initial section of the file we'll be * getting rid of the first link in the chain as well as adding a new * last link in the chain with each new word we see. */ if (curSize < chainSize) { dlAddValTail(chain, cloneString(word)); ++curSize; if (curSize == chainSize) addChainToTree(wt, chain, lm, stack); } else { /* Reuse doubly-linked-list node, but give it a new value, as we move * it from head to tail of list. */ node = dlPopHead(chain); freeMem(node->val); node->val = cloneString(word); dlAddTail(chain, node); addChainToTree(wt, chain, lm, stack); } ++wordCount; } } /* Handle last few words in file, where can't make a chain of full size. Need * a special case for file that has fewer than chain size words too. */ if (curSize < chainSize) addChainToTree(wt, chain, lm, stack); while ((node = dlPopHead(chain)) != NULL) { addChainToTree(wt, chain, lm, stack); freeMem(node->val); freeMem(node); } dlListFree(&chain); lineFileClose(&lf); return wt; } void wordChain(char *inFile, int maxSize) /* wordChain - Create Markov chain of words and optionally output chain in two formats. */ { struct lm *lm = lmInit(0); struct wordTree *wt = wordTreeForChainsInFile(inFile, maxSize, lm); if (optionExists("chain")) { char *fileName = optionVal("chain", NULL); FILE *f = mustOpen(fileName, "w"); wordTreeDump(0, wt, f); carefulClose(&f); } if (optionExists("nonsense")) { char *fileName = optionVal("nonsense", NULL); FILE *f = mustOpen(fileName, "w"); int maxSize = min(wt->useCount, maxNonsenseSize); wordTreeMakeNonsense(wt, maxChainSize, pickRandomWord(wt->following), maxSize, f); carefulClose(&f); } lmCleanup(&lm); // Not really needed since we're just going to exit. } int main(int argc, char *argv[]) /* Process command line. */ { optionInit(&argc, argv, options); if (argc != 2) usage(); maxChainSize = optionInt("size", maxChainSize); minUse = optionInt("minUse", minUse); maxNonsenseSize = optionInt("maxNonsenseSize", maxNonsenseSize); lower = optionExists("lower"); unpunc = optionExists("unpunc"); fullOnly = optionExists("fullOnly"); wordChain(argv[1], maxChainSize); return 0; }