- An intent becomes more valuable to you, and to everyone, after you populate it with training samples.
- An intent is not optimized for NLP intent-matching until it contains at least 10 training samples.
- Intent-matching accuracy improves with each new training sample that you add.
- When you enter and save new training samples, improvements to intent-matching accuracy might not take effect for 20 minutes.
- End-users never see your training samples. They merely benefit from your well-trained Helpshift system.
- can't access
- log in fails
- need new account
- account trouble
- wrong password
- someone broke in
- password change
- how to login?
- my name
- sign in
- where's my account?
- password no good
- log in
- password erased
- my settings
- locked out
- name and password
- let me in
- need my password
- I want back in
Individually, each training sample is limited to a maximum of 1,000 characters. Including spaces as characters, this threshold allows statistically for roughly 165 English-length words per training sample.
If you cannot easily visualize the length of a 1000-character copy block, the following dummy text broadly illustrates the maximum character count per training sample.
Languages supported in Smart Intents
Smart Intent AI supports following languages and hence samples can be added in only these languages:
- Chinese (Simplified)
- Chinese (Traditional)
If an app/device language is supported and text templates based translation for an intent menu exists in that language, intent menu will be shown:
- E.g. Intent display name and samples are set in English and the device/app language is English then in this case intent menu will show up in English. If a user then types in any language (e.g. English or Hindi or Liki), Helpshift’s AI model will still attempt to provide suggestions in English. However, if the intent display name is translated to Hindi using text templates, then only intent menu will be shown and no AI suggestions will be shown.
If any device/app language is not supported by Helpshift (e.g. Liki) or intent menu does not have text templates supported in that specific language then no intent menu will be shown to the user
There is no fall-back to a supported-language menu for non-supported languages or where intents are not translated for requested language. Rather we show pure messaging experience in this case.