Errors in the catalogue’s Chinese point names

Every point in AcuiQ’s catalogue carries a code, an English name, its Chinese name in characters, and its name in pinyin, the standard spelling of Mandarin in the Latin alphabet, with a tone mark over each syllable’s vowel. The pinyin is printed on each point’s page. Its unaccented form, such as Jianyu, goes into the page title, and the search box matches against it. On 9 October 2026 we checked the characters and pinyin of all 773 points against a character dictionary and against China’s national standard for the channel points, and corrected 154 rows: the pinyin of 100, the characters of 59 and the English name of 3. The errors are of several kinds, and each kind takes a different check to catch.

The two references

The first check took each pinyin syllable and asked whether it is a reading of the character it belongs to, using the character dictionary in the open-source pypinyin library (version 0.55). The second compared the 361 channel points, the points on the fourteen channels, with GB/T 12346-2021, which prints every channel point as characters, tone-marked pinyin and code, for instance 肩髃 Jiānyú (LI15). The catalogue’s other 412 points are 214 extra points, 184 ear points and 14 scalp lines, and GB/T 12346-2021 does not cover them, so for those the dictionary was the only check.

Readings the character does not have

In 21 rows the pinyin paired a character with a syllable that is not one of its readings. 髎 (liáo), a bone cleft, was stored as hé in all 14 point names that contain it, from BL31 上髎 to TB14 肩髎. 臑 (nào), the upper arm, was stored as bù in LI14, SI10 and TB13; 髃 (yú) as dòu in LI15; 膕 (guó), the back of the knee, as jiǔ in two ear points; and BL45 譩譆, which GB/T 12346-2021 reads Yìxǐ, as mián àn. Until the correction, LI15’s page was titled LI15 Jiandou, and a search for Jianyu did not find it. We have no record of how these readings were produced. pypinyin, converting the same names, reads every one of these characters correctly, so they are not the output of a converter like it.

The everyday reading in place of the point’s

Many characters have more than one reading, and a point name often uses the one that is rarer in ordinary Mandarin. 少 is shǎo, few, in everyday speech and shào, young, in 少陰 shàoyīn, the lesser yin, and in the points named for it, such as LU11 少商 Shàoshāng. 膀 is bǎng in 肩膀, the shoulder, and páng in 膀胱, the bladder. The catalogue held the everyday reading for 14 such characters, in 77 rows.

The last column is what pypinyin returns for the character on its own, which is what any tool produces when it converts a name one character at a time. For 9 of the 14 characters it is the reading the catalogue held, which is the pattern such a conversion leaves. It is not the whole explanation, since the other five do not match.

The dictionary check cannot find these errors, because every stored reading is a real reading of its character. It also disagrees with the standard on two names: it gives 俞 only as yú or shù, where GB/T 12346-2021 prints shū in every point name that contains it, and it reads 譩譆 as yīxī. Our own first correction of 俞 used shù, the reading the dictionary offered, and one catalogue row, BL29 中膂俞, already had it. The comparison with the standard moved all 27 rows to shū.

Character forms

The catalogue writes names in traditional characters, the forms used in Taiwan and Hong Kong and in the classical texts, rather than the simplified forms used in mainland China since the 1950s. 48 names carried simplified forms or variants regardless, such as 库房 for ST14 庫房 and 阴市 for ST33 陰市, and 37 of those 48 were extra points. Two names used Japanese forms: 霊台 for GV10 靈台 and 偏歴 for LI06 偏歷.

The forms also decide which Chinese-language abstracts reach the extractor. AcuiQ finds points named in Chinese-language abstracts by converting each catalogue name to simplified characters and searching the text for it, because mainland journals print simplified forms. The converter, OpenCC, maps Chinese traditional forms to simplified ones and leaves the Japanese forms as they are, so 霊台 stayed 霊台, and an abstract naming 灵台 or 偏历 would have matched no point. None of the abstracts fetched so far names either point in Chinese, so nothing was missed.

Conversion in the other direction is not one-to-one, because simplification merged characters that traditional script keeps apart. 冲 stands for both 衝, to rush, the character of the thoroughfare vessel 衝脈, and 沖, to rinse. The catalogue used 衝 for three points, among them PC09 中衝, and 沖 for six, among them LV03 太衝; all nine now read 衝, and BL60 昆侖 is 崑崙. 发 stands for both 發, to issue, and 髮, hair. Our own first pass at this correction converted one character at a time and wrote 發際 for 发际, the hairline, which is 髮際; converting the whole name with OpenCC’s phrase table gave the right form.

Wrong characters and a truncated name

Two channel points had a character that GB/T 12346-2021 does not use. TB18 was 契脈, read qì mài, where the standard prints 瘈脈 Chìmài, and BL58 was 飛陽 where the standard prints 飛揚 Fēiyáng; 陽 and 揚 are both read yáng. The pinyin of TB18 followed its wrong character, so neither check could flag it until the characters were compared with the standard. Separately, the English names of the three lowest thoracic Huatuo jiaji points, T10 to T12, each read “Hua Tuo’s Para-Vertebral Point - T1”.

What the catalogue still differs on

After the corrections every channel point matches GB/T 12346-2021 in characters and pinyin except three, where the catalogue keeps an older name: BL43 膏肓俞, which the standard shortens to 膏肓 Gāohuāng; GV26 人中, which the standard names 水溝 Shuǐgōu; and TB11 清冷淵, which the 2021 edition writes 清泠淵 Qīnglíngyuān. The extra points and the ear points have their own national standards, GB/T 40997-2021 and GB/T 13734-2008, and the catalogue has not yet been compared with either, so their characters and pinyin rest on the dictionary check alone.

What the characters of each name mean, and how far the English renderings of one name differ, is set out in Point names in translation. The codes have a separate problem, since a published code names no scheme and can resolve to the wrong point; TF4 is not a point measures how often that happens.