DataHub v1 is the original DataHub catalog of 11,000+ open datasets, now migrated to PortalJS Arc.
Dataset

FoodDB

Purpose

ASR reference transcriptions
FoodDB
FoodDB_large
About

Purpose

The database presented here was used in research for a voice search application.

Data Contents

2 food types databases (FoodDB and FoodDB_large) are presented.

The databases each have 2 database fields:

  1. A general specifier: describes a food ingredient or a dish name

  2. A mode specifier: describes a preparation mode of the ingredient or dish

Furthermore, speech queries with reference transcriptions and a statistical language model are also provided.

Sources

FoodDB is based on ‘The Book of Household Management’ by Isabella Beeton [1]. The database contains food types from the index of the book.
The speech queries were mined from the audio book version available on LibriVox [2]. For scalability tests, databases with food types
from Freebase [3] were used to expand the original database.

[1]:http://www.gutenberg.org/ebooks/10136, October, 2015.

[2]:https://librivox.org/the-book-of-household-management-by-isabellabeeton/, November, 2009.

[3]:Google, Freebase Data Dumps, https://developers.google.com/freebase/data, June 30, 2015.

Download & API
ASR reference transcriptionscsv
A list of reference transcriptions per speech query
Download
FoodDBcsv
The food types database of the original size (3349 entries)
Download
FoodDB_largecsv
The food types database of the scaled-up size (9173 entries)
Download
Speech querieszip
179 speech queries from 12 speakers, formatted as ssscuuu.wav with sss=speaker-id and ccc=utterance-id
Download
Statistical language model for the speech queriesgz
The statistical language model used in the speech recognition in '.arpabo'-format (N-grams with backoff)
Download
Sources & license
License
  • cc-by-sa