Data Repository

Welcome to the Parlance Data Repository. Below is an outline of all the datasets collected on the project. All datasets are for the restaurant domain for San Francisco.

To gain access to these datasets please click here

Dataset

System

Num total Utterances

(System and User)

Attributes (# values)

Hand-annotation

Notes

SFCore

1

~4,800

food(59), area(155), pricerange(3)

Word transcription, task success, Dialogue Acts/ Semantics,

Translated for Mandarin and Spanish

End-to-End Eval

SFCore1.5

1.5

~10,000

food(59), area(155), pricerange(3)

None

End-to-End Eval

SFExt

SF1Ext

2

~8,000

SFCore+ near(39)

Dialogue Acts/ Semantics

Data used for algorithm dev of a number of modules

SF2Ext

2

~6,500

SF1Ext + allowedforkids (2)

Dialogue Acts/ Semantics

SF3Ext

2

~6,300

SF2Ext + goodformeal(4)

Dialogue Acts/ Semantics

SFExtEval

2

~15,000

as SFExt

Transcription

End-to-End Eval

SFExt_generic

2

~9,000

as SFExt

None

Data used for algorithm dev for Task 2.2

 

SFExt_generic1

2

~8,000

as SFExt

None

SFExt_generic2

2

~8,200

as SFExt

None

SFExt_scratch

2

~9,200

as SFExt

None

SFExt_scratch1

2

~9,500

as SFExt

None

SFExt_scratch2

2

~10,000

as SFExt

None

Comments