The Corpus of North American Spoken English (CoNASE) is a 1.29-billion-word corpus of geolocated automatic speech recognition (ASR) YouTube transcripts from the United States and Canada. It was created for the study of lexical, grammatical, and discourse-pragmatic phenomena of spoken language, including their geographical distribution, in North American English. The size of the corpus allows rare phenomena to be considered, and because the annotation includes the video IDs of transcripts, search hits can be manually inspected and video or audio data can be accessed.

Corpus description

The corpus was created from 301,846 ASR transcripts from 2,572 YouTube channels, corresponding to 154,041 hours of video. The size of the corpus is 1,294,885,016 word tokens. The channels sampled in the corpus are associated with local government entities such as town, city, or county boards and councils, school or utility districts, regional authorities such as provincial or territorial governments, or other governmental organizations. The transcripts are primarily of recordings of public meetings, although other genres are also present. Video transcripts have been assigned exact latitude-longitude coordinates using a geocoding script.


Format and metadata

The corpus exists in two versions: one in which the transcripts have not been annotated, and one in which word tokens have been annotated with part-of-speech tags and timing. Data is in tabular form in csv format (with the pipe character "|" as a separator). The columns of the table are 'country', 'state', 'channel_title', 'video_title', 'video_id', 'name', 'type', 'channel_id', 'channel_username', 'video_length', 'location', 'address', 'nr_words', 'text_pos' (or 'text', in the untagged version), and 'latlong'. Each row corresponds to an individual transcript (see excerpt below).

The columns 'name' and 'type' correspond to US Census Bureau designations for the government entity the channel represents (see Coats, in review). Transcripts are in the 'text' column (or 'text_pos', for the tagged version). Tagging was done with spaCy and the Penn Treebank tagset. Tagged tokens have the format "token_POS_10.0", where "token" is the transcribed lexical item, "POS" the tag, and "10.0" the time offset in seconds from the start of the corresponding video.

country state channel_title video_title video_id name type channel_id channel_username video_length location address nr_words text_pos latlong
0 USA Alabama Baldwin County Alabama Loxley School KAFtEkKE_ik COUNTY OF BALDWIN general purpose UCVEowdBqDvlT_TJDjf8YuyA BCCommission 1651.38 Baldwin County, AL, USA Baldwin County, AL, USA 3513 the_DT_25.619 Baldwin_NNP_26.619 County_NNP_27... (30.6010744, -87.77633329999999)
1 USA Alabama Baldwin County Alabama 2019 Baldwin County Sewer Utilities Informatio... 6KvmxXRoMGA COUNTY OF BALDWIN general purpose UCVEowdBqDvlT_TJDjf8YuyA BCCommission 6110.09 Baldwin County, AL, USA Baldwin County, AL, USA 16090 welcome_UH_1.37 everybody_NN_2.37 glad_JJ_3.89... (30.6010744, -87.77633329999999)
2 USA Alabama Baldwin County Alabama Heritage Museum of Baldwin County Alabama bGL_FlSulZ4 COUNTY OF BALDWIN general purpose UCVEowdBqDvlT_TJDjf8YuyA BCCommission 1255.31 Baldwin County, AL, USA Baldwin County, AL, USA 2318 located_VBN_22.039 in_IN_23.039 the_DT_23.13 t... (30.6010744, -87.77633329999999)
3 USA Alabama Baldwin County Alabama AL State Veterans Memorial Cemetery at Spanish... BX18sxhdL5Y COUNTY OF BALDWIN general purpose UCVEowdBqDvlT_TJDjf8YuyA BCCommission 4376.31 Baldwin County, AL, USA Baldwin County, AL, USA 7991 good_JJ_15.94 morning_NN_16.94 everyone_NN_17.... (30.6010744, -87.77633329999999)
4 USA Alabama Baldwin County Alabama USO Canteen Dance Re-enactment VqCvFSku9Es COUNTY OF BALDWIN general purpose UCVEowdBqDvlT_TJDjf8YuyA BCCommission 3893.26 Baldwin County, AL, USA Baldwin County, AL, USA 6846 [Music]_XX_5.1 [Music]_XX_14.86 [Music]_XX_104... (30.6010744, -87.77633329999999)

Locations

Channels have been sampled in all of the 50 United States, the District of Columbia, and all 13 Canadian Provinces or Territories. The map below shows the locations, names, and sizes of the sampled channels.


Sample script

Corpus hits can be presented as a data table with links to the videos at the time of utterance. This Python script will retrieve all instances of "is there any way" in the PoS-tagged text in the table conase_df:


import re
hits = []
for i,x in conase_df.iterrows():
pat1 = re.compile("is_\\w+_\\S+ there_\\w+_\\S+ any_\\w+_\\S+ way_\\w+_\\S+",re.IGNORECASE)
if pat1.search(x["text_pos"]):
	finds1 = pat1.findall(x["text_pos"])[0]
	seq = " ".join([x.split("_")[0] for x in finds1.split()])
	time = finds1.split()[0].split("_")[-1] 
	hits.append((x["state"],x["channel_title"],seq,"https://youtu.be/"+x["video_id"]+"?t="+str(round(float(time)-3))))
pd.DataFrame(hits)
       (...)

Access

The corpus is available free of charge for the purposes of research, education, and scholarship. Users may not redistribute the corpus or use the corpus for commercial purposes. The corpus is available through the Harvard Dataverse here. Please also see the Corpus of British Isles Spoken English, created using similar methods, here.


Citation and contact

To cite the corpus, please use