Skip to content

[BUG] Overlapping slots in KVRET dataset #207

Description

@ytxmobile98

It came to my attention that in the KVRET dataset (permalink), the file dialogues.json has one sample that contains overlapping slots:

{
  "speaker": "user",
  "utterance": "I need the reminder at 7pm for the yoga class on December 12th.",
  "utt_idx": 2,
  "dialogue_acts": {
    "binary": [],
    "categorical": [],
    "non-categorical": [
      {
        "intent": "inform",
        "domain": "schedule",
        "slot": "date",
        "value": "yoga class",
        "start": 35,
        "end": 45
      },
      {
        "intent": "inform",
        "domain": "schedule",
        "slot": "event",
        "value": "yoga class",
        "start": 35,
        "end": 45
      }
    ]
  },
  "state": {
    "schedule": {
      "date": "Yoga class",
      "event": "Yoga Class",
      "agenda": "reminder"
    }
  }
}

Here the term "Yoga class" is both classified as both a date and an event. Shouldn't the date be December 12th instead?

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't working

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions