Skip to content

Voice UI Design Reference

When developing voice products, users need to pay special attention to the design of wake-up words and command words required for voice interaction to achieve good recognition results. This document describes in detail the design method of voice wake-up words and command words recommended by our company, which is recommended for users.

Chinese

Design of wake-up words

  1. It is generally 4-6 characters, 4 characters is the best, too short for error tolerance, and too long for users to call and remember;
  2. The syllables of the words in the command word are longer and the greater the discrimination between the initial consonants of adjacent Chinese characters, the better;
  3. Repetitious or similar pronunciation words should be avoided in wake-up words. Adjacent syllables should avoid consonants, that is, avoid that the previous vowel is the same or similar to the following vowel, such as “语音”;
  4. The words should be pronounced clearly, with high energy loudness and large opening. It is better to have a plosive tone. The fourth tone is recommended, such as “灌溉”;
  5. Try to select uncommon words as wake-up words, which can effectively reduce or even avoid false wake-up;
  6. Avoid using reduplicated words, such as “你好你好” and “宝宝开机”;
  7. To improve the pronunciation energy, you can select syllables containing the initial of the plosive sound, such as p, b, t, d, k, g, etc;
  8. The vowel should be big mouthed and open accented, such as a, ai, an, ian, ao, avoiding e, i, en, in, etc;
  9. The Chinese command word can only be composed of Chinese characters, and no space, comma and other characters are allowed;
  10. The platform provides wake-up evaluation function. If the fitness of wake-up words is “excellent”, it is a better wake-up word;
  11. A model only allows one wake-up word. If multiple wake-up words are required, please optimize them separately and use the multi network switching function.

Design of Command Words

  1. It is generally 4-6 characters, 4 characters is the best, too short for high error recognition, too long for users to call and remember;
  2. The greater the discrimination between the consonants of adjacent Chinese characters in the command word, the better;
  3. Conform to the user’s language habits, try to use common expressions, and the content is specific and direct;
  4. Avoid using everyday expressions, such as “Have a meal”;
  5. Unusual characters and zero initial characters should be avoided as far as possible, for example, “speech” in “speech recognition” are all zero initial characters;
  6. It is better not to have modal particles in the command words, such as “ah”, “ni”, etc;
  7. Avoid using reduplicated words, such as “你好你好”;
  8. The Chinese command word can only be composed of Chinese characters, and no spaces, commas and other characters are allowed;
  9. The number in the command word shall be expressed in Chinese characters, such as “调高一度”;
  10. If you have not determined the command word, it is recommended that you select it from the “Recommended Command Words” in the document center.

English

Design of wake-up words

  1. It is suggested that the wake-up words consist of 2-4 words (4-6 syllables). If they are too short, high error wake-up is likely to occur, and if they are too long, it is inconvenient for users to call and remember;
  2. The syllable of the word in the wake-up call is longer and the greater the discrimination between adjacent sounds is, the better;
  3. Repetitious words or words with similar pronunciation should be avoided in wake-up words, and adjacent syllables should avoid linking;
  4. The words should be pronounced clearly, with high energy loudness and large opening;
  5. Try to select words that are not commonly used as wake-up words to avoid using everyday words, which can effectively reduce or even avoid false wake-up;
  6. Avoid using reduplicated words, such as “HELLO HELLO”;
  7. A language model only allows one wake-up word. If multiple wake-up words are required, please optimize and use the multiple network switching function separately. For specific methods, please contact our technical support personnel for more information.

Design of Command Words

  1. It is suggested to be composed of 2-4 words (4-6 syllables). If it is too short, it is easy to have high false recognition, and if it is too long, it is inconvenient for users to remember;
  2. The greater the syllable discrimination between command words, the better;
  3. Conform to the user’s language habits, try to use common expressions, and the content is specific and direct;
  4. Avoid using everyday words, such as “HI, HELLO”;
  5. Avoid using similar syllables. The pronunciation of words should be clear and loud. For example, avoid using TURN-ON and TURN-OFF at the same time;
  6. Avoid using reduplicated words, such as “HELLO HELLO”.

Japanese

Design of wake-up words

  1. It is suggested that the wake-up words consist of 4-6 syllables of Japanese. If they are too short, high error wake-up is likely to occur, and if they are too long, it is inconvenient for users to call and remember;
  2. The syllable of the word in the wake-up call is longer and the greater the discrimination between adjacent sounds is, the better;
  3. Repetitious words or words with similar pronunciation should be avoided in arousal words, and adjacent syllables should avoid linking;
  4. Words should be pronounced clearly, with high energy loudness and large opening;
  5. Try to select words that are not commonly used as wake-up words to avoid using everyday words, which can effectively reduce or even avoid false wake-up;
  6. Avoid using reduplicated words, such as “ラ ボ - ラ ボ”;
  7. A language model only allows one wake-up word. If multiple wake-up words are required, please optimize and use the multiple network switching function separately. For specific methods, please contact our technical support personnel for more information.

Design of Command Words

  1. A Japanese command word is suggested to be composed of 4-6 syllables of Japanese. If it is too short, it is easy to have high error recognition, and if it is too long, it is inconvenient for users to call and remember;
  2. The greater the syllable discrimination between command words, the better;
  3. Conform to the user’s language habits, try to use common expressions, and the content is specific and direct;
  4. Avoid using everyday words, such as “は い, お は よ”;
  5. Avoid using similar syllables, and pronounce words clearly and loudly, such as; Lower げ げ (sa ge te), upper げ げ (a ge te);
  6. Avoid using reduplicated words, such as “ラ ボ - ラ ボ”.