@kensio/pinyinjs 1.9.3 → 1.9.4

package/dist/decode/cai-rule.d.ts

@@ -0,0 +1,25 @@
+/**
+ * The 才能 that is 才 *only then* and 能 *can*, and not the noun *talent*.
+ *
+ * 才能 is a word, *talent* or *ability*, and every source counts it. The same
+ * two characters are far more often the adverb 才 and the modal 能 in front of
+ * a verb: 怎样才能提高爆发力, 必须先完成这件事才能回家, 到北方才能吃到真正的饺子.
+ * The decode read every one of them as the noun.
+ */
+import { type EdgeRule } from "./rules.js";
+/**
+ * 才能 is taken apart where a verb is left for 能 to be the modal of.
+ *
+ * The noun is kept after a 的 or a 有, and before a 的, a 是, a degree word or
+ * the end of the run.
+ * Everywhere else the 才 is the adverb and the 能 the modal. Taking the edge
+ * away leaves the two to read as themselves. The readings are `cái néng`
+ * either way, so this changes the segmentation and the spacing.
+ *
+ * Over the 139,893 distinct Han runs of Tatoeba and the Chinese Boost example
+ * sentences, the decode read 才能 as the word in 184 runs. The rule splits 164,
+ * and one of them is wrong: 他的钢琴才能为他在音乐专家中赢得了美名 is talent before a
+ * 为 that reads like a verb.
+ */
+export declare const MODAL_CAI: EdgeRule;
+//# sourceMappingURL=cai-rule.d.ts.map

package/dist/decode/cai-rule.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"cai-rule.d.ts","sourceRoot":"","sources":["../../src/decode/cai-rule.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AACH,OAAO,EAEL,KAAK,QAAQ,EAId,MAAM,YAAY,CAAC;AAwDpB;;;;;;;;;;;;;GAaG;AACH,eAAO,MAAM,SAAS,EAAE,QAUvB,CAAC"}

package/dist/decode/cai-rule.js

@@ -0,0 +1,78 @@
+/**
+ * The 才能 that is 才 *only then* and 能 *can*, and not the noun *talent*.
+ *
+ * 才能 is a word, *talent* or *ability*, and every source counts it. The same
+ * two characters are far more often the adverb 才 and the modal 能 in front of
+ * a verb: 怎样才能提高爆发力, 必须先完成这件事才能回家, 到北方才能吃到真正的饺子.
+ * The decode read every one of them as the noun.
+ */
+import { tagOf, wordsEndingAt, wordsStartingAt, } from "./rules.js";
+/**
+ * The word, in both scripts.
+ */
+const TALENT = new Set(["才能"]);
+/**
+ * The characters that close a word in front of the noun.
+ *
+ * 的 makes 才能 something owned or described (他的才能, 一定的才能), and 有 is
+ * having it (有才能, 富有才能, 具有才能, 很有才能). 才 and 能 can follow neither.
+ */
+const BEFORE_NOUN = new Set(["的", "有"]);
+/**
+ * What can follow the noun and not the modal.
+ *
+ * 能 needs a verb after it, so a 的 (有才能的人) or the end of the run (你很有才能,
+ * 她具有卓越才能) leaves it nothing to be the modal of. A 是 or a degree word
+ * says something about the talent (音乐才能是可以培养的, 她的唱歌才能非常出众).
+ */
+const AFTER_NOUN = new Set(["的", "是", "很", "非常", "十分"]);
+/**
+ * Whether a longer word ends at the 才, and would take it once 才能 is gone.
+ *
+ * 处方才能买 is 处方 and then 才能, and taking the edge away let 方才 *just
+ * now* take the 才, reading 处方 as `chù fāng`. 刚才能 and 人才能 have the same
+ * shape. Which word the 才 belongs to there is a question this rule cannot
+ * answer, so it leaves the decode as it was. Only a tagged word counts, since
+ * the full tier holds untagged keys such as 时才 that no decode would choose.
+ */
+function isClaimed(context, from) {
+    return wordsEndingAt(context, from + 1).some((word) => word.length > 1 && tagOf(context, word) !== "");
+}
+/**
+ * Whether the context at a 才能 says it is the noun.
+ */
+function isNoun(context, from, to) {
+    if (wordsEndingAt(context, from).some((word) => BEFORE_NOUN.has(word.at(-1) ?? ""))) {
+        return true;
+    }
+    // Every word starting there is asked, since the longest hides the one the
+    // rule is looking for: 音乐才能是否可以培养 has 是否 wrapped around its 是.
+    const next = wordsStartingAt(context, to);
+    return next.length === 0 || next.some((word) => AFTER_NOUN.has(word));
+}
+/**
+ * 才能 is taken apart where a verb is left for 能 to be the modal of.
+ *
+ * The noun is kept after a 的 or a 有, and before a 的, a 是, a degree word or
+ * the end of the run.
+ * Everywhere else the 才 is the adverb and the 能 the modal. Taking the edge
+ * away leaves the two to read as themselves. The readings are `cái néng`
+ * either way, so this changes the segmentation and the spacing.
+ *
+ * Over the 139,893 distinct Han runs of Tatoeba and the Chinese Boost example
+ * sentences, the decode read 才能 as the word in 184 runs. The rule splits 164,
+ * and one of them is wrong: 他的钢琴才能为他在音乐专家中赢得了美名 is talent before a
+ * 为 that reads like a verb.
+ */
+export const MODAL_CAI = {
+    name: "modal-cai",
+    verdictFor: (context) => {
+        const { edge } = context;
+        return TALENT.has(edge.text) &&
+            !isClaimed(context, edge.from) &&
+            !isNoun(context, edge.from, edge.to)
+            ? "forbid"
+            : "keep";
+    },
+};
+//# sourceMappingURL=cai-rule.js.map

package/dist/decode/cai-rule.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"cai-rule.js","sourceRoot":"","sources":["../../src/decode/cai-rule.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AACH,OAAO,EAGL,KAAK,EACL,aAAa,EACb,eAAe,GAChB,MAAM,YAAY,CAAC;AAEpB;;GAEG;AACH,MAAM,MAAM,GAAG,IAAI,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC;AAE/B;;;;;GAKG;AACH,MAAM,WAAW,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC;AAExC;;;;;;GAMG;AACH,MAAM,UAAU,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,IAAI,EAAE,IAAI,CAAC,CAAC,CAAC;AAExD;;;;;;;;GAQG;AACH,SAAS,SAAS,CAAC,OAAoB,EAAE,IAAY;IACnD,OAAO,aAAa,CAAC,OAAO,EAAE,IAAI,GAAG,CAAC,CAAC,CAAC,IAAI,CAC1C,CAAC,IAAI,EAAE,EAAE,CAAC,IAAI,CAAC,MAAM,GAAG,CAAC,IAAI,KAAK,CAAC,OAAO,EAAE,IAAI,CAAC,KAAK,EAAE,CACzD,CAAC;AACJ,CAAC;AAED;;GAEG;AACH,SAAS,MAAM,CAAC,OAAoB,EAAE,IAAY,EAAE,EAAU;IAC5D,IACE,aAAa,CAAC,OAAO,EAAE,IAAI,CAAC,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,EAAE,CACzC,WAAW,CAAC,GAAG,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,CACnC,EACD,CAAC;QACD,OAAO,IAAI,CAAC;IACd,CAAC;IACD,0EAA0E;IAC1E,+DAA+D;IAC/D,MAAM,IAAI,GAAG,eAAe,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC;IAC1C,OAAO,IAAI,CAAC,MAAM,KAAK,CAAC,IAAI,IAAI,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,UAAU,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC,CAAC;AACxE,CAAC;AAED;;;;;;;;;;;;;GAaG;AACH,MAAM,CAAC,MAAM,SAAS,GAAa;IACjC,IAAI,EAAE,WAAW;IACjB,UAAU,EAAE,CAAC,OAAoB,EAAE,EAAE;QACnC,MAAM,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;QACzB,OAAO,MAAM,CAAC,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC;YAC1B,CAAC,SAAS,CAAC,OAAO,EAAE,IAAI,CAAC,IAAI,CAAC;YAC9B,CAAC,MAAM,CAAC,OAAO,EAAE,IAAI,CAAC,IAAI,EAAE,IAAI,CAAC,EAAE,CAAC;YACpC,CAAC,CAAC,QAAQ;YACV,CAAC,CAAC,MAAM,CAAC;IACb,CAAC;CACF,CAAC"}

package/dist/decode/dou-rule.d.ts

@@ -0,0 +1,23 @@
+/**
+ * The 都 that is the adverb *all*, read into a place name in front of it.
+ *
+ * 花都 is a district of Guangzhou, `Huādū`, and the dictionary holds it as a
+ * proper noun. In running text it is nearly always 花 *flowers* and 都 *all*:
+ * 花都开了, 公园里的花都开了, 所有的花都是纸做的. The decode read the name,
+ * capital and `dū` included.
+ */
+import { type EdgeRule } from "./rules.js";
+/**
+ * 花都 is taken apart where no noun follows it.
+ *
+ * The district names what it holds: 花都区, 花都机场. Where a verb, an
+ * adverb, a 是 or the end of the run follows, the 都 is the adverb and sums up
+ * the flowers. Taking the edge away leaves 花 and 都 to read as themselves,
+ * `huā dōu`.
+ *
+ * Over the 139,893 distinct Han runs of Tatoeba and the Chinese Boost example
+ * sentences, the decode read 花都 as the name in 11 runs, and none of them was
+ * the district.
+ */
+export declare const SUMMING_DOU: EdgeRule;
+//# sourceMappingURL=dou-rule.d.ts.map

package/dist/decode/dou-rule.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"dou-rule.d.ts","sourceRoot":"","sources":["../../src/decode/dou-rule.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AACH,OAAO,EAEL,KAAK,QAAQ,EAGd,MAAM,YAAY,CAAC;AAYpB;;;;;;;;;;;GAWG;AACH,eAAO,MAAM,WAAW,EAAE,QASzB,CAAC"}

package/dist/decode/dou-rule.js

@@ -0,0 +1,40 @@
+/**
+ * The 都 that is the adverb *all*, read into a place name in front of it.
+ *
+ * 花都 is a district of Guangzhou, `Huādū`, and the dictionary holds it as a
+ * proper noun. In running text it is nearly always 花 *flowers* and 都 *all*:
+ * 花都开了, 公园里的花都开了, 所有的花都是纸做的. The decode read the name,
+ * capital and `dū` included.
+ */
+import { tagOf, wordStartingAt, } from "./rules.js";
+/**
+ * The place names whose first character is a common noun that 都 can sum up.
+ */
+const NAMES = new Set(["花都"]);
+/**
+ * The tag prefix jieba gives a noun.
+ */
+const NOUN_TAG = "n";
+/**
+ * 花都 is taken apart where no noun follows it.
+ *
+ * The district names what it holds: 花都区, 花都机场. Where a verb, an
+ * adverb, a 是 or the end of the run follows, the 都 is the adverb and sums up
+ * the flowers. Taking the edge away leaves 花 and 都 to read as themselves,
+ * `huā dōu`.
+ *
+ * Over the 139,893 distinct Han runs of Tatoeba and the Chinese Boost example
+ * sentences, the decode read 花都 as the name in 11 runs, and none of them was
+ * the district.
+ */
+export const SUMMING_DOU = {
+    name: "summing-dou",
+    verdictFor: (context) => {
+        const { edge } = context;
+        return NAMES.has(edge.text) &&
+            !tagOf(context, wordStartingAt(context, edge.to)).startsWith(NOUN_TAG)
+            ? "forbid"
+            : "keep";
+    },
+};
+//# sourceMappingURL=dou-rule.js.map

package/dist/decode/dou-rule.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"dou-rule.js","sourceRoot":"","sources":["../../src/decode/dou-rule.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AACH,OAAO,EAGL,KAAK,EACL,cAAc,GACf,MAAM,YAAY,CAAC;AAEpB;;GAEG;AACH,MAAM,KAAK,GAAG,IAAI,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC;AAE9B;;GAEG;AACH,MAAM,QAAQ,GAAG,GAAG,CAAC;AAErB;;;;;;;;;;;GAWG;AACH,MAAM,CAAC,MAAM,WAAW,GAAa;IACnC,IAAI,EAAE,aAAa;IACnB,UAAU,EAAE,CAAC,OAAoB,EAAE,EAAE;QACnC,MAAM,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;QACzB,OAAO,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC;YACzB,CAAC,KAAK,CAAC,OAAO,EAAE,cAAc,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC,UAAU,CAAC,QAAQ,CAAC;YACtE,CAAC,CAAC,QAAQ;YACV,CAAC,CAAC,MAAM,CAAC;IACb,CAAC;CACF,CAAC"}

package/dist/decode/erhua-rule.d.ts

@@ -0,0 +1,26 @@
+/**
+ * The 儿 that belongs to the word in front of it.
+ */
+import type { EdgeRule } from "./rules.js";
+/**
+ * 儿 does not stand on its own where the dictionary attests the 儿化.
+ *
+ * 儿化 is a per-word dictionary fact taken from CC-CEDICT's explicit `r5`
+ * token, and 2,009 of the 2,067 keys ending in 儿 carry it. What that leaves is
+ * an edge: 这边儿, 上边儿 and 旁边儿 are all listed and 那边儿 is not, so
+ * 我得跑到那边儿去 read the 儿 as a syllable of its own — `nàbian ér` — and the
+ * decode was choosing between 那边 plus 儿 and 那 plus 边儿 on frequency alone.
+ *
+ * This forbids the bare 儿 edge when the character in front of it makes an
+ * attested 儿化 word, so the only paths left are the ones that absorb it. That
+ * asserts nothing the dictionary does not already say — 边儿 is `biānr` because
+ * CC-CEDICT says so — which is what separates it from synthesising 儿化 with a
+ * suffix rule, the thing ORTHOGRAPHY.md rejects and the old project got wrong.
+ *
+ * Measured over the same 88,866 lines it fires 7 times, and all 7 are right:
+ * 照片儿, 出门儿, 热点儿, 辣味儿, 晚会儿, 针管儿, 碍事儿. It is deliberately
+ * silent on the 20 bare 儿 with no attested 儿化 behind them, which are 马儿,
+ * 门铃儿 and a row of Mongol names transliterated with 儿.
+ */
+export declare const ATTESTED_ERHUA: EdgeRule;
+//# sourceMappingURL=erhua-rule.d.ts.map

package/dist/decode/erhua-rule.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"erhua-rule.d.ts","sourceRoot":"","sources":["../../src/decode/erhua-rule.ts"],"names":[],"mappings":"AAAA;;GAEG;AACH,OAAO,KAAK,EAAe,QAAQ,EAAE,MAAM,YAAY,CAAC;AAExD;;;;;;;;;;;;;;;;;;;GAmBG;AACH,eAAO,MAAM,cAAc,EAAE,QAY5B,CAAC"}

package/dist/decode/erhua-rule.js

@@ -0,0 +1,34 @@
+/**
+ * 儿 does not stand on its own where the dictionary attests the 儿化.
+ *
+ * 儿化 is a per-word dictionary fact taken from CC-CEDICT's explicit `r5`
+ * token, and 2,009 of the 2,067 keys ending in 儿 carry it. What that leaves is
+ * an edge: 这边儿, 上边儿 and 旁边儿 are all listed and 那边儿 is not, so
+ * 我得跑到那边儿去 read the 儿 as a syllable of its own — `nàbian ér` — and the
+ * decode was choosing between 那边 plus 儿 and 那 plus 边儿 on frequency alone.
+ *
+ * This forbids the bare 儿 edge when the character in front of it makes an
+ * attested 儿化 word, so the only paths left are the ones that absorb it. That
+ * asserts nothing the dictionary does not already say — 边儿 is `biānr` because
+ * CC-CEDICT says so — which is what separates it from synthesising 儿化 with a
+ * suffix rule, the thing ORTHOGRAPHY.md rejects and the old project got wrong.
+ *
+ * Measured over the same 88,866 lines it fires 7 times, and all 7 are right:
+ * 照片儿, 出门儿, 热点儿, 辣味儿, 晚会儿, 针管儿, 碍事儿. It is deliberately
+ * silent on the 20 bare 儿 with no attested 儿化 behind them, which are 马儿,
+ * 门铃儿 and a row of Mongol names transliterated with 儿.
+ */
+export const ATTESTED_ERHUA = {
+    name: "attested-erhua",
+    verdictFor: (context) => {
+        const { characters, edge } = context;
+        if (edge.text !== "儿" || edge.from === 0) {
+            return "keep";
+        }
+        /* c8 ignore next -- the edge starts inside the run it was built from */
+        const absorbed = `${characters[edge.from - 1] ?? ""}儿`;
+        const reading = context.dictionary.lookup(absorbed)?.reading;
+        return (reading?.at(-1)?.erhua ?? false) ? "forbid" : "keep";
+    },
+};
+//# sourceMappingURL=erhua-rule.js.map

package/dist/decode/erhua-rule.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"erhua-rule.js","sourceRoot":"","sources":["../../src/decode/erhua-rule.ts"],"names":[],"mappings":"AAKA;;;;;;;;;;;;;;;;;;;GAmBG;AACH,MAAM,CAAC,MAAM,cAAc,GAAa;IACtC,IAAI,EAAE,gBAAgB;IACtB,UAAU,EAAE,CAAC,OAAoB,EAAE,EAAE;QACnC,MAAM,EAAE,UAAU,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;QACrC,IAAI,IAAI,CAAC,IAAI,KAAK,GAAG,IAAI,IAAI,CAAC,IAAI,KAAK,CAAC,EAAE,CAAC;YACzC,OAAO,MAAM,CAAC;QAChB,CAAC;QACD,wEAAwE;QACxE,MAAM,QAAQ,GAAG,GAAG,UAAU,CAAC,IAAI,CAAC,IAAI,GAAG,CAAC,CAAC,IAAI,EAAE,GAAG,CAAC;QACvD,MAAM,OAAO,GAAG,OAAO,CAAC,UAAU,CAAC,MAAM,CAAC,QAAQ,CAAC,EAAE,OAAO,CAAC;QAC7D,OAAO,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,EAAE,KAAK,IAAI,KAAK,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,CAAC,MAAM,CAAC;IAC/D,CAAC;CACF,CAAC"}

package/dist/decode/ge-rule.d.ts

@@ -0,0 +1,23 @@
+/**
+ * The 个人 whose 个 is counting one 人, and not the word *individual*.
+ *
+ * {@link import("./measure-rule.js").COUNTED_MEASURE} takes 个人 apart after a
+ * number, so 三个人 is three people. A bare 个 with no number at all stands for
+ * 一个 after a verb that introduces somebody, and the measure rule has nothing
+ * to go on there: 门口有个人, 想找个人聊聊天, 我今天雇了个人.
+ */
+import { type EdgeRule } from "./rules.js";
+/**
+ * 个人 is taken apart where an introducing verb makes its 个 count one 人.
+ *
+ * Taking the edge away leaves 个 and 人 to read as themselves. The readings
+ * are `gè rén` either way, so this changes the segmentation and the spacing.
+ *
+ * Over the 139,893 distinct Han runs of Tatoeba and the Chinese Boost example
+ * sentences, the decode kept 个人 as one word in 61 runs. The rule splits 12,
+ * and every one of them is somebody. A rough number in front (十几个人,
+ * 一百多个人) is left alone, because taking the edge away there moves the 个
+ * onto the 几 and reads 十 几个 人.
+ */
+export declare const COUNTED_GE: EdgeRule;
+//# sourceMappingURL=ge-rule.d.ts.map

package/dist/decode/ge-rule.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"ge-rule.d.ts","sourceRoot":"","sources":["../../src/decode/ge-rule.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AACH,OAAO,EAEL,KAAK,QAAQ,EAGd,MAAM,YAAY,CAAC;AAoDpB;;;;;;;;;;;GAWG;AACH,eAAO,MAAM,UAAU,EAAE,QASxB,CAAC"}

package/dist/decode/ge-rule.js

@@ -0,0 +1,77 @@
+/**
+ * The 个人 whose 个 is counting one 人, and not the word *individual*.
+ *
+ * {@link import("./measure-rule.js").COUNTED_MEASURE} takes 个人 apart after a
+ * number, so 三个人 is three people. A bare 个 with no number at all stands for
+ * 一个 after a verb that introduces somebody, and the measure rule has nothing
+ * to go on there: 门口有个人, 想找个人聊聊天, 我今天雇了个人.
+ */
+import { tagOf, wordStartingAt, } from "./rules.js";
+/**
+ * The word, in both scripts.
+ */
+const INDIVIDUAL = new Set(["个人", "個人"]);
+/**
+ * The verbs that bring an unnamed somebody into a sentence.
+ *
+ * 有个人 is *somebody*, 找个人 and 叫个人 are finding and sending one, 雇了个人
+ * is hiring one and 换了个人 is becoming somebody else.
+ */
+const INTRODUCING = new Set([
+    "有",
+    "找",
+    "叫",
+    "请",
+    "請",
+    "雇",
+    "换",
+    "換",
+    "派",
+]);
+/**
+ * The aspect particle that can sit between the verb and the 个.
+ */
+const PERFECTIVE = "了";
+/**
+ * The tag prefix jieba gives a noun.
+ */
+const NOUN_TAG = "n";
+/**
+ * Whether a verb that introduces somebody stands in front of a 个人, with or
+ * without a 了 after it.
+ *
+ * The word after the 个人 is asked about too. 个人 in front of a noun is the
+ * modifier *personal* (我有个人问题, 有个人信息), and only a verb, a particle
+ * or the end of the run leaves the 人 to be the person.
+ */
+function isIntroduced(context, from, to) {
+    const { characters } = context;
+    const verb = characters[from - 1] === PERFECTIVE ? from - 2 : from - 1;
+    if (!INTRODUCING.has(characters[verb] ?? "")) {
+        return false;
+    }
+    return !tagOf(context, wordStartingAt(context, to)).startsWith(NOUN_TAG);
+}
+/**
+ * 个人 is taken apart where an introducing verb makes its 个 count one 人.
+ *
+ * Taking the edge away leaves 个 and 人 to read as themselves. The readings
+ * are `gè rén` either way, so this changes the segmentation and the spacing.
+ *
+ * Over the 139,893 distinct Han runs of Tatoeba and the Chinese Boost example
+ * sentences, the decode kept 个人 as one word in 61 runs. The rule splits 12,
+ * and every one of them is somebody. A rough number in front (十几个人,
+ * 一百多个人) is left alone, because taking the edge away there moves the 个
+ * onto the 几 and reads 十 几个 人.
+ */
+export const COUNTED_GE = {
+    name: "counted-ge",
+    verdictFor: (context) => {
+        const { edge } = context;
+        return INDIVIDUAL.has(edge.text) &&
+            isIntroduced(context, edge.from, edge.to)
+            ? "forbid"
+            : "keep";
+    },
+};
+//# sourceMappingURL=ge-rule.js.map

package/dist/decode/ge-rule.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"ge-rule.js","sourceRoot":"","sources":["../../src/decode/ge-rule.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AACH,OAAO,EAGL,KAAK,EACL,cAAc,GACf,MAAM,YAAY,CAAC;AAEpB;;GAEG;AACH,MAAM,UAAU,GAAG,IAAI,GAAG,CAAC,CAAC,IAAI,EAAE,IAAI,CAAC,CAAC,CAAC;AAEzC;;;;;GAKG;AACH,MAAM,WAAW,GAAG,IAAI,GAAG,CAAC;IAC1B,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;CACJ,CAAC,CAAC;AAEH;;GAEG;AACH,MAAM,UAAU,GAAG,GAAG,CAAC;AAEvB;;GAEG;AACH,MAAM,QAAQ,GAAG,GAAG,CAAC;AAErB;;;;;;;GAOG;AACH,SAAS,YAAY,CAAC,OAAoB,EAAE,IAAY,EAAE,EAAU;IAClE,MAAM,EAAE,UAAU,EAAE,GAAG,OAAO,CAAC;IAC/B,MAAM,IAAI,GAAG,UAAU,CAAC,IAAI,GAAG,CAAC,CAAC,KAAK,UAAU,CAAC,CAAC,CAAC,IAAI,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,GAAG,CAAC,CAAC;IACvE,IAAI,CAAC,WAAW,CAAC,GAAG,CAAC,UAAU,CAAC,IAAI,CAAC,IAAI,EAAE,CAAC,EAAE,CAAC;QAC7C,OAAO,KAAK,CAAC;IACf,CAAC;IACD,OAAO,CAAC,KAAK,CAAC,OAAO,EAAE,cAAc,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC,CAAC,UAAU,CAAC,QAAQ,CAAC,CAAC;AAC3E,CAAC;AAED;;;;;;;;;;;GAWG;AACH,MAAM,CAAC,MAAM,UAAU,GAAa;IAClC,IAAI,EAAE,YAAY;IAClB,UAAU,EAAE,CAAC,OAAoB,EAAE,EAAE;QACnC,MAAM,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;QACzB,OAAO,UAAU,CAAC,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC;YAC9B,YAAY,CAAC,OAAO,EAAE,IAAI,CAAC,IAAI,EAAE,IAAI,CAAC,EAAE,CAAC;YACzC,CAAC,CAAC,QAAQ;YACV,CAAC,CAAC,MAAM,CAAC;IACb,CAAC;CACF,CAAC"}

package/dist/decode/mei-rule.d.ts

@@ -0,0 +1,22 @@
+/**
+ * The 美的 that is an adjective and its 的, and not the brand.
+ *
+ * 美的 is a word, the name of the appliance maker read `Měidí`, and the
+ * dictionary holds it as a proper noun. The same two characters are far more
+ * often 美 *beautiful* and the particle after it: 最美的姑娘, 很美的动物, 多美的日出.
+ * The decode took the name every time, so a learner's sentence about a
+ * sunrise came out with a capitalised `Měidí` in it.
+ */
+import { type EdgeRule } from "./rules.js";
+/**
+ * 美的 is taken apart where a degree word or a 是 grades it.
+ *
+ * Taking the edge away leaves 美 and 的 to read as themselves, `měi de`, which
+ * is the reading as well as the spacing: the name's 的 is `dí`.
+ *
+ * Over the 139,893 distinct Han runs of Tatoeba and the Chinese Boost example
+ * sentences, the decode read 美的 as the name in 28 runs and none of them was
+ * the company.
+ */
+export declare const ADJECTIVAL_MEI: EdgeRule;
+//# sourceMappingURL=mei-rule.d.ts.map

package/dist/decode/mei-rule.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"mei-rule.d.ts","sourceRoot":"","sources":["../../src/decode/mei-rule.ts"],"names":[],"mappings":"AAAA;;;;;;;;GAQG;AACH,OAAO,EAEL,KAAK,QAAQ,EAId,MAAM,YAAY,CAAC;AA0EpB;;;;;;;;;GASG;AACH,eAAO,MAAM,cAAc,EAAE,QAQ5B,CAAC"}

package/dist/decode/mei-rule.js

@@ -0,0 +1,95 @@
+/**
+ * The 美的 that is an adjective and its 的, and not the brand.
+ *
+ * 美的 is a word, the name of the appliance maker read `Měidí`, and the
+ * dictionary holds it as a proper noun. The same two characters are far more
+ * often 美 *beautiful* and the particle after it: 最美的姑娘, 很美的动物, 多美的日出.
+ * The decode took the name every time, so a learner's sentence about a
+ * sunrise came out with a capitalised `Měidí` in it.
+ */
+import { tagOf, wordStartingAt, wordsEndingAt, } from "./rules.js";
+/**
+ * The word, in both scripts.
+ */
+const BRAND = new Set(["美的"]);
+/**
+ * The words that grade or assert an adjective, in both scripts.
+ *
+ * Each one says something about how beautiful a thing is, and a company name
+ * takes none of them: 最美的, 非常美的, 这么美的. 多 is in for 多美的日出啊, the
+ * exclamation.
+ */
+const DEGREE = new Set([
+    "最",
+    "很",
+    "更",
+    "太",
+    "真",
+    "挺",
+    "多",
+    "多么",
+    "多麼",
+    "非常",
+    "十分",
+    "特别",
+    "特別",
+    "极其",
+    "極其",
+    "比较",
+    "比較",
+    "相当",
+    "相當",
+    "如此",
+    "这么",
+    "這麼",
+    "那么",
+    "那麼",
+    "这样",
+    "這樣",
+    "那样",
+    "那樣",
+]);
+/**
+ * The copula, which asserts 美 of its subject (都是美的) and also names what a
+ * thing is (这是美的冰箱, a Midea fridge).
+ */
+const COPULA = "是";
+/**
+ * The tag prefix jieba gives a noun.
+ */
+const NOUN_TAG = "n";
+/**
+ * Whether the 美的 starting at a position is graded or asserted.
+ *
+ * A degree word in front always grades it. A 是 asserts it only where nothing
+ * follows for 美的 to name: 都是美的 is beautiful, and 这是美的冰箱 and
+ * 这是美的牌冰箱 are the brand.
+ */
+function isGraded(context, from, to) {
+    const before = wordsEndingAt(context, from, 2);
+    if (before.some((word) => DEGREE.has(word))) {
+        return true;
+    }
+    return (before.includes(COPULA) &&
+        !tagOf(context, wordStartingAt(context, to)).startsWith(NOUN_TAG));
+}
+/**
+ * 美的 is taken apart where a degree word or a 是 grades it.
+ *
+ * Taking the edge away leaves 美 and 的 to read as themselves, `měi de`, which
+ * is the reading as well as the spacing: the name's 的 is `dí`.
+ *
+ * Over the 139,893 distinct Han runs of Tatoeba and the Chinese Boost example
+ * sentences, the decode read 美的 as the name in 28 runs and none of them was
+ * the company.
+ */
+export const ADJECTIVAL_MEI = {
+    name: "adjectival-mei",
+    verdictFor: (context) => {
+        const { edge } = context;
+        return BRAND.has(edge.text) && isGraded(context, edge.from, edge.to)
+            ? "forbid"
+            : "keep";
+    },
+};
+//# sourceMappingURL=mei-rule.js.map

package/dist/decode/mei-rule.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"mei-rule.js","sourceRoot":"","sources":["../../src/decode/mei-rule.ts"],"names":[],"mappings":"AAAA;;;;;;;;GAQG;AACH,OAAO,EAGL,KAAK,EACL,cAAc,EACd,aAAa,GACd,MAAM,YAAY,CAAC;AAEpB;;GAEG;AACH,MAAM,KAAK,GAAG,IAAI,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC;AAE9B;;;;;;GAMG;AACH,MAAM,MAAM,GAAG,IAAI,GAAG,CAAC;IACrB,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;CACL,CAAC,CAAC;AAEH;;;GAGG;AACH,MAAM,MAAM,GAAG,GAAG,CAAC;AAEnB;;GAEG;AACH,MAAM,QAAQ,GAAG,GAAG,CAAC;AAErB;;;;;;GAMG;AACH,SAAS,QAAQ,CAAC,OAAoB,EAAE,IAAY,EAAE,EAAU;IAC9D,MAAM,MAAM,GAAG,aAAa,CAAC,OAAO,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC;IAC/C,IAAI,MAAM,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,MAAM,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC,EAAE,CAAC;QAC5C,OAAO,IAAI,CAAC;IACd,CAAC;IACD,OAAO,CACL,MAAM,CAAC,QAAQ,CAAC,MAAM,CAAC;QACvB,CAAC,KAAK,CAAC,OAAO,EAAE,cAAc,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC,CAAC,UAAU,CAAC,QAAQ,CAAC,CAClE,CAAC;AACJ,CAAC;AAED;;;;;;;;;GASG;AACH,MAAM,CAAC,MAAM,cAAc,GAAa;IACtC,IAAI,EAAE,gBAAgB;IACtB,UAAU,EAAE,CAAC,OAAoB,EAAE,EAAE;QACnC,MAAM,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;QACzB,OAAO,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC,IAAI,QAAQ,CAAC,OAAO,EAAE,IAAI,CAAC,IAAI,EAAE,IAAI,CAAC,EAAE,CAAC;YAClE,CAAC,CAAC,QAAQ;YACV,CAAC,CAAC,MAAM,CAAC;IACb,CAAC;CACF,CAAC"}

package/dist/decode/name-count-rule.d.ts

@@ -0,0 +1,15 @@
+import { type EdgeRule } from "./rules.js";
+/**
+ * A proper noun is taken apart where its 量词 is counting with a number.
+ *
+ * Two shapes are caught. The name starts with the number and the 量词 follows
+ * (三本书, 五本书) and no title follows, or the name starts with the 量词 and the
+ * number is just in front of it (一堵墙). Taking the edge away leaves each
+ * character to read as itself, and the spacing follows the count.
+ *
+ * Over the 139,893 distinct Han runs of Tatoeba and the Chinese Boost example
+ * sentences, the rule changes 13 runs, 9 of them 三本 and 4 堵墙, and each one
+ * is a count.
+ */
+export declare const COUNTED_NAME: EdgeRule;
+//# sourceMappingURL=name-count-rule.d.ts.map

package/dist/decode/name-count-rule.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"name-count-rule.d.ts","sourceRoot":"","sources":["../../src/decode/name-count-rule.ts"],"names":[],"mappings":"AAWA,OAAO,EAAoB,KAAK,QAAQ,EAAkB,MAAM,YAAY,CAAC;AA+C7E;;;;;;;;;;;GAWG;AACH,eAAO,MAAM,YAAY,EAAE,QAkB1B,CAAC"}

package/dist/decode/name-count-rule.js

@@ -0,0 +1,81 @@
+/**
+ * The proper noun that is really a number and the 量词 counting with it.
+ *
+ * 三本 is a Japanese surname and 堵墙 a place, and jieba tags both as names
+ * (`ns` and `nr`). In running text they are 三本书 (three books) and 一堵墙 (a
+ * wall), and the decode read them as one word: `sānběn shū`, `yì dǔqiáng`.
+ * {@link import("./measure-rule.js").COUNTED_MEASURE} keeps a counted 量词
+ * out of the next word, and it finds the 量词 by its tag. 本 is tagged `r` and
+ * 堵 `v`, so it never sees these two.
+ */
+import { QUANTITY_CHARACTERS } from "../numerals/characters.js";
+import { wordStartingAt } from "./rules.js";
+/**
+ * The tags jieba gives a person, a place and another proper name.
+ *
+ * Asked as well as the proper-noun flag, because a word only jieba tags as a
+ * name keeps the tag and loses the flag.
+ */
+const NAME_TAG = /^n[rsz]/u;
+/**
+ * The 量词 the dictionary does not tag `q` and a proper noun has taken in.
+ *
+ * Kept to the two the corpus shows. The untagged 量词 are many (条, 件, 只 and
+ * 块 among them), and a wider set would reach names such as 四川, whose 川 is
+ * no 量词 at all.
+ */
+const UNTAGGED_MEASURES = new Set(["本", "堵"]);
+/**
+ * The titles that follow a surname, in both scripts.
+ *
+ * 三本先生 is Mr Sanbon, and a title after the name says the name is meant.
+ */
+const TITLES = new Set([
+    "先生",
+    "小姐",
+    "女士",
+    "太太",
+    "老师",
+    "老師",
+    "同学",
+    "同學",
+]);
+/**
+ * Whether a character is a numeral, or 几 asking how many.
+ */
+function isNumeral(character) {
+    return (character !== undefined &&
+        (QUANTITY_CHARACTERS.has(character) ||
+            character === "几" ||
+            character === "幾"));
+}
+/**
+ * A proper noun is taken apart where its 量词 is counting with a number.
+ *
+ * Two shapes are caught. The name starts with the number and the 量词 follows
+ * (三本书, 五本书) and no title follows, or the name starts with the 量词 and the
+ * number is just in front of it (一堵墙). Taking the edge away leaves each
+ * character to read as itself, and the spacing follows the count.
+ *
+ * Over the 139,893 distinct Han runs of Tatoeba and the Chinese Boost example
+ * sentences, the rule changes 13 runs, 9 of them 三本 and 4 堵墙, and each one
+ * is a count.
+ */
+export const COUNTED_NAME = {
+    name: "counted-name",
+    verdictFor: (context) => {
+        const { characters, edge } = context;
+        if (!edge.isProperNoun && !NAME_TAG.test(edge.partOfSpeech)) {
+            return "keep";
+        }
+        const first = characters[edge.from];
+        const second = characters[edge.from + 1];
+        const startsWithCount = isNumeral(first) &&
+            UNTAGGED_MEASURES.has(second ?? "") &&
+            !TITLES.has(wordStartingAt(context, edge.to) ?? "");
+        const followsCount = UNTAGGED_MEASURES.has(first ?? "") &&
+            isNumeral(characters[edge.from - 1]);
+        return startsWithCount || followsCount ? "forbid" : "keep";
+    },
+};
+//# sourceMappingURL=name-count-rule.js.map

package/dist/decode/name-count-rule.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"name-count-rule.js","sourceRoot":"","sources":["../../src/decode/name-count-rule.ts"],"names":[],"mappings":"AAAA;;;;;;;;;GASG;AACH,OAAO,EAAE,mBAAmB,EAAE,MAAM,2BAA2B,CAAC;AAChE,OAAO,EAAmC,cAAc,EAAE,MAAM,YAAY,CAAC;AAE7E;;;;;GAKG;AACH,MAAM,QAAQ,GAAG,UAAU,CAAC;AAE5B;;;;;;GAMG;AACH,MAAM,iBAAiB,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC;AAE9C;;;;GAIG;AACH,MAAM,MAAM,GAAG,IAAI,GAAG,CAAC;IACrB,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;CACL,CAAC,CAAC;AAEH;;GAEG;AACH,SAAS,SAAS,CAAC,SAA6B;IAC9C,OAAO,CACL,SAAS,KAAK,SAAS;QACvB,CAAC,mBAAmB,CAAC,GAAG,CAAC,SAAS,CAAC;YACjC,SAAS,KAAK,GAAG;YACjB,SAAS,KAAK,GAAG,CAAC,CACrB,CAAC;AACJ,CAAC;AAED;;;;;;;;;;;GAWG;AACH,MAAM,CAAC,MAAM,YAAY,GAAa;IACpC,IAAI,EAAE,cAAc;IACpB,UAAU,EAAE,CAAC,OAAoB,EAAE,EAAE;QACnC,MAAM,EAAE,UAAU,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;QACrC,IAAI,CAAC,IAAI,CAAC,YAAY,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC,IAAI,CAAC,YAAY,CAAC,EAAE,CAAC;YAC5D,OAAO,MAAM,CAAC;QAChB,CAAC;QACD,MAAM,KAAK,GAAG,UAAU,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QACpC,MAAM,MAAM,GAAG,UAAU,CAAC,IAAI,CAAC,IAAI,GAAG,CAAC,CAAC,CAAC;QACzC,MAAM,eAAe,GACnB,SAAS,CAAC,KAAK,CAAC;YAChB,iBAAiB,CAAC,GAAG,CAAC,MAAM,IAAI,EAAE,CAAC;YACnC,CAAC,MAAM,CAAC,GAAG,CAAC,cAAc,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,CAAC,IAAI,EAAE,CAAC,CAAC;QACtD,MAAM,YAAY,GAChB,iBAAiB,CAAC,GAAG,CAAC,KAAK,IAAI,EAAE,CAAC;YAClC,SAAS,CAAC,UAAU,CAAC,IAAI,CAAC,IAAI,GAAG,CAAC,CAAC,CAAC,CAAC;QACvC,OAAO,eAAe,IAAI,YAAY,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,CAAC,MAAM,CAAC;IAC7D,CAAC;CACF,CAAC"}

package/dist/decode/polyphone-rules.d.ts

@@ -3,10 +3,12 @@
  *
  * 弹 is `tán` or `dàn`, and which one a text means is a question about the
  * neighbours rather than about the character. 长 is the same question and has
- * a file of its own, and 好学 is the same question asked of a word. The other rules in `reading-rules.ts` are about particles
+ * a file of its own, and 好学, 美的 and 花都 are the same question asked of a word. The other rules in `reading-rules.ts` are about particles
  * and 儿化, which is a different kind of evidence.
  */
 export { ADJECTIVAL_CHANG } from "./chang-rule.js";
 export { PLAYING_TAN } from "./tan-rule.js";
 export { EASY_HAO } from "./hao-rule.js";
+export { ADJECTIVAL_MEI } from "./mei-rule.js";
+export { SUMMING_DOU } from "./dou-rule.js";
 //# sourceMappingURL=polyphone-rules.d.ts.map

package/dist/decode/polyphone-rules.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"polyphone-rules.d.ts","sourceRoot":"","sources":["../../src/decode/polyphone-rules.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AACH,OAAO,EAAE,gBAAgB,EAAE,MAAM,iBAAiB,CAAC;AACnD,OAAO,EAAE,WAAW,EAAE,MAAM,eAAe,CAAC;AAC5C,OAAO,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC"}
+{"version":3,"file":"polyphone-rules.d.ts","sourceRoot":"","sources":["../../src/decode/polyphone-rules.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AACH,OAAO,EAAE,gBAAgB,EAAE,MAAM,iBAAiB,CAAC;AACnD,OAAO,EAAE,WAAW,EAAE,MAAM,eAAe,CAAC;AAC5C,OAAO,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AACzC,OAAO,EAAE,cAAc,EAAE,MAAM,eAAe,CAAC;AAC/C,OAAO,EAAE,WAAW,EAAE,MAAM,eAAe,CAAC"}

package/dist/decode/polyphone-rules.js

@@ -3,10 +3,12 @@
  *
  * 弹 is `tán` or `dàn`, and which one a text means is a question about the
  * neighbours rather than about the character. 长 is the same question and has
- * a file of its own, and 好学 is the same question asked of a word. The other rules in `reading-rules.ts` are about particles
+ * a file of its own, and 好学, 美的 and 花都 are the same question asked of a word. The other rules in `reading-rules.ts` are about particles
  * and 儿化, which is a different kind of evidence.
  */
 export { ADJECTIVAL_CHANG } from "./chang-rule.js";
 export { PLAYING_TAN } from "./tan-rule.js";
 export { EASY_HAO } from "./hao-rule.js";
+export { ADJECTIVAL_MEI } from "./mei-rule.js";
+export { SUMMING_DOU } from "./dou-rule.js";
 //# sourceMappingURL=polyphone-rules.js.map

package/dist/decode/polyphone-rules.js.map

@@ -1 +1 @@
-{"version":3,"file":"polyphone-rules.js","sourceRoot":"","sources":["../../src/decode/polyphone-rules.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AACH,OAAO,EAAE,gBAAgB,EAAE,MAAM,iBAAiB,CAAC;AACnD,OAAO,EAAE,WAAW,EAAE,MAAM,eAAe,CAAC;AAC5C,OAAO,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC"}
+{"version":3,"file":"polyphone-rules.js","sourceRoot":"","sources":["../../src/decode/polyphone-rules.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AACH,OAAO,EAAE,gBAAgB,EAAE,MAAM,iBAAiB,CAAC;AACnD,OAAO,EAAE,WAAW,EAAE,MAAM,eAAe,CAAC;AAC5C,OAAO,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AACzC,OAAO,EAAE,cAAc,EAAE,MAAM,eAAe,CAAC;AAC/C,OAAO,EAAE,WAAW,EAAE,MAAM,eAAe,CAAC"}