Skip to main content
The Spell Correction feature corrects misspelled words in the user’s utterance based on the conversation language. Version 10.1.19 introduces Spell Correction Version 2 for English, French, and Dutch with the following advantages:
  • Consistent Spell Correction Experience: Spell correction happens at a central place immediately after language detection, unlike Version 1 where it occurred separately at each NLP engine.
  • Enhanced Accuracy: Words are checked against comprehensive dictionaries using edit distance to find and rank corrections based on their commonness.

​
Configurations to Manage Spell Correction

Two new language-specific settings are available in Advanced NLP Configurations (disabled by default):
  • Spell Correction Version: Select from the drop-down and choose the version.
  • Spell Correction Status: Select from the drop-down and toggle the status to Enabled.
Manage Spell Correction V2 When enabled, spell correction runs immediately after language detection, and the corrected input is used by all engines for further processing. Once V2 is enabled, the following legacy settings are removed:
  • Auto-Correction in Knowledge Graph.
  • ML_Spell_Correction custom flag in Advanced NLP Configurations.

​
Impact on New and Existing Bots

  • New bots and existing bots without spell correction enabled (in configured or published copy for ML and KG engines): Only Spell Correction V2 is supported. Configure and enable the settings as described above.
  • Existing bots with spell correction already enabled (in either configured or published copy in ML or KG engine): Both versions are supported. An upgrade banner appears on the NLU Config screen. You can initiate the upgrade from the banner, and the platform will automatically configure the required settings.
Spell Correction Version Upgrade banner Alternatively, manually select Version 2 as the Spell Correction Version. To downgrade, select Version 1 and the platform will automatically handle the configuration changes.