#include "sentence.h" #include #include #include sentence::sentence( const verbly::database& database, std::mt19937& rng) : database_(database), rng_(rng) { } std::string sentence::generate() const { // Generate the form that the sentence should take. std::vector actions { {"like", verbly::token(std::set({"infinitive_phrase", "subjectless"}))}, {"have", verbly::token(std::set({"gerund_phrase", "subjectless", "past_participle"}))} }; verbly::token form = actions[ std::uniform_int_distribution(0, actions.size()-1)(rng_)]; // Compile the form. while (!form.isComplete()) { visit(form); } std::string compiled = form.compile(); std::list words = verbly::split>(compiled, " "); verbly::token cased; for (std::string& word : words) { if (std::bernoulli_distribution(1.0/2.0)(rng_)) { cased << verbly::token::capitalize(verbly::token::casing::all_caps, word); } else { cased << word; } } return cased.compile(); } bool sentence::chooseSelrestr(std::set selrestrs, std::set choices) const { int validChoices = 0; for (const std::string& choice : choices) { if (selrestrs.count(choice)) { validChoices++; } } return std::bernoulli_distribution(static_cast(validChoices)/static_cast(selrestrs.size()))(rng_); } verbly::word sentence::generateStandardNoun( std::string role, std::set selrestrs) const { std::geometric_distribution tagdist(0.5); // 0.06 std::vector result; bool trySelection = true; while (result.empty()) { verbly::filter condition = (verbly::notion::partOfSpeech == verbly::part_of_speech::noun) && (verbly::form::proper == false) //&& (verbly::form::complexity == 1) // && (verbly::word::tagCount >= tagdist(rng_)) // Favor more common words && (verbly::word::tagCount >= 1) && !(verbly::word::usageDomains %= (verbly::notion::wnid == 106718862)); // Blacklist ethnic slurs // Only use selection restrictions for a first attempt. if (trySelection) { verbly::filter selection(true); for (const std::string& selrestr : selrestrs) { if (selrestr == "concrete") { selection += (verbly::notion::wnid == 100001930); // physical entity } else if (selrestr == "time") { selection += (verbly::notion::wnid == 100028270); // time } else if (selrestr == "state") { selection += (verbly::notion::wnid == 100024720); // state } else if (selrestr == "abstract") { selection += (verbly::notion::wnid == 100002137); // abstract entity } else if (selrestr == "scalar") { selection += (verbly::notion::wnid == 103835412); // number } else if (selrestr == "currency") { selection += (verbly::notion::wnid == 105050379); // currency } else if (selrestr == "location") { selection += (verbly::notion::wnid == 100027167); // location } else if (selrestr == "organization") { selection += (verbly::notion::wnid == 100237078); // organization } else if (selrestr == "int_control") { selection += (verbly::notion::wnid == 100007347); // causal agent } else if (selrestr == "natural") { selection += (verbly::notion::wnid == 100019128); // natural object } else if (selrestr == "phys_obj") { selection += (verbly::notion::wnid == 100002684); // physical object } else if (selrestr == "solid") { selection += (verbly::notion::wnid == 113860793); // solid } else if (selrestr == "shape") { selection += (verbly::notion::wnid == 100027807); // shape } else if (selrestr == "substance") { selection += (verbly::notion::wnid == 100019613); // substance } else if (selrestr == "idea") { selection += (verbly::notion::wnid == 105803379); // idea } else if (selrestr == "sound") { selection += (verbly::notion::wnid == 107111047); // sound } else if (selrestr == "communication") { selection += (verbly::notion::wnid == 100033020); // communication } else if (selrestr == "region") { selection += (verbly::notion::wnid == 105221895); // region } else if (selrestr == "place") { selection += (verbly::notion::wnid == 100586262); // place } else if (selrestr == "machine") { selection += (verbly::notion::wnid == 102958343); // machine } else if (selrestr == "animate") { selection += (verbly::notion::wnid == 100004258); // animate thing } else if (selrestr == "plant") { selection += (verbly::notion::wnid == 103956922); // plant } else if (selrestr == "comestible") { selection += (verbly::notion::wnid == 100021265); // food } else if (selrestr == "artifact") { selection += (verbly::notion::wnid == 100021939); // artifact } else if (selrestr == "vehicle") { selection += (verbly::notion::wnid == 104524313); // vehicle } else if (selrestr == "human") { selection += (verbly::notion::wnid == 100007846); // person } else if (selrestr == "animal") { selection += (verbly::notion::wnid == 100015388); // animal } else if (selrestr == "body_part") { selection += (verbly::notion::wnid == 105220461); // body part } else if (selrestr == "garment") { selection += (verbly::notion::wnid == 103051540); // clothing } else if (selrestr == "tool") { selection += (verbly::notion::wnid == 104451818); // tool } else if ((selrestr == "concrete_inanimate") || (selrestr == "inanimate")) { selection += (verbly::notion::wnid == 100021939); // artifact selection += (verbly::notion::wnid == 100019128); // natural object } else if (selrestr == "non_region_location") { selection += (verbly::notion::wnid == 102913152); // building } else if (selrestr == "non_solid_food") { selection += (verbly::notion::wnid == 107881800); // beverage } else if (selrestr == "solid_food") { selection += (verbly::notion::wnid == 107555863); // solid food } else if (selrestr == "slinky") { selection += (verbly::notion::wnid == 103670849); // line } } if (selection.compact().getType() != verbly::filter::type::empty) { condition &= (verbly::notion::fullHypernyms %= std::move(selection)); } else if (role == "Attribute") { condition &= (verbly::notion::fullHypernyms %= (verbly::notion::wnid == 100024264)); // attribute } else if (role == "Instrument") { condition &= (verbly::notion::fullHypernyms %= (verbly::notion::wnid == 104451818)); // tool } else if (role == "Agent") { condition &= (verbly::notion::fullHypernyms %= (verbly::notion::wnid == 100007347)); // causal agent } trySelection = false; } else { std::cout << "Selection failed" << std::endl; } result = database_.words(condition).all(); } return result.front(); } verbly::token sentence::generateStandardNounPhrase( const verbly::word& noun, std::string role, bool plural, bool definite) const { verbly::token utter; bool indefiniteArticle = false; /*if (std::bernoulli_distribution(1.0/8.0)(rng_)) { std::geometric_distribution tagdist(0.2); descript = database_.words( (verbly::word::tagCount >= tagdist(rng_)) && (verbly::notion::partOfSpeech == verbly::part_of_speech::adjective)).first(); sounder = descript; }*/ if ((std::bernoulli_distribution(1.0/3.0)(rng_)) && (definite)) { utter << "the"; if (std::bernoulli_distribution(1.0/2.0)(rng_)) { plural = true; } } else { if ((role != "Theme") && (role != "Attribute") && std::bernoulli_distribution(1.0/2.0)(rng_)) { utter << "your"; } else if (!plural) { indefiniteArticle = true; } } if (plural && noun.hasInflection(verbly::inflection::plural)) { utter << verbly::token(noun, verbly::inflection::plural); } else { utter << noun; } if (indefiniteArticle) { return verbly::token::indefiniteArticle(utter); } else { return utter; } } verbly::token sentence::generateClause( const verbly::token& it) const { verbly::token utter; std::geometric_distribution tagdist(0.07); std::vector verbDataset; verbly::filter frameCondition = (verbly::frame::length >= 2) && (verbly::frame::parts(0) %= ( (verbly::part::type == verbly::part_type::noun_phrase) && (verbly::part::role == "Agent")) && (verbly::frame::parts(1) %= (verbly::part::type == verbly::part_type::verb)) && !(verbly::frame::parts() %= ( verbly::part::synrestrs %= "adjp"))); if (it.hasSynrestr("experiencer")) { frameCondition &= (verbly::frame::parts(2) %= (verbly::part::type == verbly::part_type::noun_phrase) && !(verbly::part::synrestrs %= "genitive") && ((verbly::part::role == "Patient") || (verbly::part::role == "Experiencer"))); } verbly::filter verbCondition = (verbly::notion::partOfSpeech == verbly::part_of_speech::verb) && frameCondition; if (it.hasSynrestr("participle_phrase")) { verbCondition &= (verbly::word::forms(verbly::inflection::ing_form)); } else if (it.hasSynrestr("progressive")) { verbCondition &= (verbly::word::forms(verbly::inflection::s_form)); } else if (it.hasSynrestr("past_participle")) { verbCondition &= (verbly::word::forms(verbly::inflection::past_participle)); } // Because of the tag distribution, it's possible (albeit extremely unlikely) // for the verb query to fail, so we loop until it succeeds. while (verbDataset.empty()) { verbDataset = database_.words( verbCondition && (verbly::word::tagCount >= tagdist(rng_)) ).all(); } verbly::word verb = verbDataset.front(); verbly::frame frame = database_.frames(frameCondition && verb).first(); std::list parts(std::begin(frame.getParts()), std::end(frame.getParts())); if (it.hasSynrestr("experiencer")) { // Ignore the direct object. parts.erase(std::next(parts.begin(), 2)); } if (it.hasSynrestr("subjectless")) { // Ignore the subject. parts.pop_front(); } for (const verbly::part& part : parts) { switch (part.getType()) { case verbly::part_type::noun_phrase: { std::cout << "NP: "; for (auto& s : part.getNounSynrestrs()) { std::cout << s << " "; } std::cout << std::endl; if (chooseSelrestr(part.getNounSelrestrs(), {"currency"})) { int lead = std::uniform_int_distribution(1,9)(rng_); int tail = std::uniform_int_distribution(0,6)(rng_); std::string tailStr(tail, '0'); utter << ("$" + std::to_string(lead) + tailStr); } else if (part.nounHasSynrestr("adjp")) { utter << std::set({"adjective_phrase"}); } else if ((part.nounHasSynrestr("be_sc_ing")) || (part.nounHasSynrestr("ac_ing")) || (part.nounHasSynrestr("sc_ing")) || (part.nounHasSynrestr("np_omit_ing")) || (part.nounHasSynrestr("oc_ing"))) { utter << std::set({"participle_phrase", "subjectless"}); } else if ((part.nounHasSynrestr("poss_ing")) || (part.nounHasSynrestr("possing")) || (part.nounHasSynrestr("pos_ing"))) { utter << "your"; utter << std::set({"participle_phrase", "subjectless"}); } else if (part.nounHasSynrestr("adv_loc")) { if (std::bernoulli_distribution(1.0/2.0)(rng_)) { utter << "here"; } else { utter << "there"; } } else if (part.nounHasSynrestr("refl")) { utter << "yourself"; } else if ((part.nounHasSynrestr("sc_to_inf")) || (part.nounHasSynrestr("ac_to_inf")) || (part.nounHasSynrestr("vc_to_inf")) || (part.nounHasSynrestr("rs_to_inf")) || (part.nounHasSynrestr("oc_to_inf"))) { utter << std::set({"infinitive_phrase", "subjectless"}); } else if (part.nounHasSynrestr("oc_bare_inf")) { utter << std::set({"infinitive_phrase", "bare", "subjectless"}); } else if (part.nounHasSynrestr("wh_comp")) { utter << "whether"; verbly::token sentence(std::set({"progressive"})); utter << generateClause(sentence); } else if (part.nounHasSynrestr("that_comp")) { utter << "that"; utter << "they"; verbly::token sentence(std::set({"subjectless"})); utter << generateClause(sentence); } else if (part.nounHasSynrestr("what_extract")) { utter << "what"; verbly::token sentence(std::set({"progressive", "experiencer"})); utter << generateClause(sentence); } else if (part.nounHasSynrestr("how_extract")) { utter << "how"; verbly::token sentence(std::set({"progressive"})); utter << generateClause(sentence); } else if (part.nounHasSynrestr("wh_inf")) { utter << "how"; verbly::token sentence(std::set({"infinitive_phrase", "subjectless"})); utter << generateClause(sentence); } else if (part.nounHasSynrestr("what_inf")) { utter << "what"; verbly::token sentence(std::set({"infinitive_phrase", "subjectless", "experiencer"})); utter << generateClause(sentence); } else if (part.nounHasSynrestr("wheth_inf")) { utter << "whether"; verbly::token sentence(std::set({"infinitive_phrase", "subjectless"})); utter << generateClause(sentence); } else if (part.nounHasSynrestr("quotation")) { utter << verbly::token::quote("\"", "\"", verbly::token(std::set({"participle_phrase"}))); } else { if (part.nounHasSynrestr("genitive")) { verbly::word noun = generateStandardNoun("Passive", {"animate"}); verbly::token owner = generateStandardNounPhrase(noun, "Passive", false, true); utter << verbly::token::punctuation("'s", owner); } verbly::word noun = generateStandardNoun(part.getNounRole(), part.getNounSelrestrs()); bool plural = part.nounHasSynrestr("plural") || chooseSelrestr(part.getNounSelrestrs(), {"group", "plural"}); utter << generateStandardNounPhrase( noun, part.getNounRole(), plural, part.nounHasSynrestr("definite")); if (part.nounHasSynrestr("acc_ing") || part.nounHasSynrestr("ac_ing")) { utter << std::set({"participle_phrase", "subjectless"}); } } break; } case verbly::part_type::verb: { std::cout << "V: " << verb.getBaseForm().getText() << std::endl; if (it.hasSynrestr("progressive")) { utter << verbly::token(verb, verbly::inflection::s_form); } else if (it.hasSynrestr("past_participle")) { utter << verbly::token(verb, verbly::inflection::past_participle); } else if (it.hasSynrestr("infinitive_phrase")) { if (!it.hasSynrestr("bare")) { utter << "to"; } utter << verb; } else if (it.hasSynrestr("participle_phrase")) { utter << verbly::token(verb, verbly::inflection::ing_form); } else if (it.hasSynrestr("gerund_phrase")) { utter << verbly::token(verb, verbly::inflection::past_participle); } else { utter << verb; } break; } case verbly::part_type::preposition: { std::cout << "PREP" << std::endl; if (part.isPrepositionLiteral()) { int choiceIndex = std::uniform_int_distribution(0, part.getPrepositionChoices().size()-1)(rng_); utter << part.getPrepositionChoices()[choiceIndex]; } else { verbly::filter pgf(true); for (const std::string& choice : part.getPrepositionChoices()) { pgf += (verbly::notion::prepositionGroups == choice); } utter << database_.words(pgf && (verbly::notion::partOfSpeech == verbly::part_of_speech::preposition)).first(); } break; } case verbly::part_type::adjective: { std::cout << "ADJ" << std::endl; utter << std::set({"adjective_phrase"}); break; } case verbly::part_type::adverb: { std::cout << "ADV" << std::endl; utter << std::set({"adverb_phrase"}); break; } case verbly::part_type::literal: { std::cout << "LIT" << std::endl; utter << part.getLiteralValue(); break; } case verbly::part_type::invalid: { // Nope break; } } } if ((parts.size() == 1) && (std::bernoulli_distribution(1.0/4.0)(rng_))) { utter << std::set({"adverb_phrase"}); } return utter; } void sentence::visit(verbly::token& it) const { switch (it.getType()) { case verbly::token::type::utterance: { for (verbly::token& token : it) { if (!token.isComplete()) { visit(token); break; } } break; } case verbly::token::type::fillin: { if (it.hasSynrestr("infinitive_phrase") || it.hasSynrestr("gerund_phrase")) { it = generateClause(it); } else if (it.hasSynrestr("adjective_phrase")) { verbly::token phrase; if (std::bernoulli_distribution(1.0/6.0)(rng_)) { phrase << std::set({"adverb_phrase"}); } if (std::bernoulli_distribution(1.0/4.0)(rng_)) { phrase << std::set({"participle_phrase", "subjectless"}); } else { std::geometric_distribution tagdist(0.2); phrase << database_.words( (verbly::word::tagCount >= tagdist(rng_)) && (verbly::notion::partOfSpeech == verbly::part_of_speech::adjective)).first(); } it = phrase; } else if (it.hasSynrestr("adverb_phrase")) { std::geometric_distribution tagdist(1.0/23.0); it = database_.words( (verbly::notion::partOfSpeech == verbly::part_of_speech::adverb) && (verbly::word::tagCount >= tagdist(rng_)) ).first(); } else if (it.hasSynrestr("participle_phrase")) { if (std::bernoulli_distribution(1.0/2.0)(rng_)) { it = verbly::token( database_.words( (verbly::notion::partOfSpeech == verbly::part_of_speech::verb) && (verbly::word::forms(verbly::inflection::ing_form))).first(), verbly::inflection::ing_form); } else { it = generateClause(it); } } else { it = "*the reality of the situation*"; } break; } case verbly::token::type::transform: { visit(it.getInnerToken()); break; } case verbly::token::type::word: case verbly::token::type::literal: case verbly::token::type::part: { // Nope break; } } }