4,713
edits
Dataset Errata: Difference between revisions
m
→Errors noted in various files
| Line 2: | Line 2: | ||
::Gold label NY01075763_lg.txt has Pyrenidium actinellurn, should be Pyrenidium actinellum. Gold Parsed copies the error verbatim (as it should) and needs to be corrected if the .txt file is corrected. | ::Gold label NY01075763_lg.txt has Pyrenidium actinellurn, should be Pyrenidium actinellum. Gold Parsed copies the error verbatim (as it should) and needs to be corrected if the .txt file is corrected. | ||
::::/home/aocr/datasets/lichens/gold/outputs/human/NY01075763_lg.txt fixed | ::::/home/aocr/datasets/lichens/gold/outputs/human/NY01075763_lg.txt fixed --[[User:Dpaul|Dpaul]] 17:28, 26 February 2013 (EST) | ||
::::/home/aocr/datasets/lichens/gold/parsed/human/NY01075763_lg.csv fixed | ::::/home/aocr/datasets/lichens/gold/parsed/human/NY01075763_lg.csv fixed --[[User:Dpaul|Dpaul]] 17:28, 26 February 2013 (EST) | ||
::Inconsistency in Gold Parsed decimalLatitude and decimalLongitude in many labels. All omitted from NYBG lichens and Tennesee lichens. Gold Parsed WIS-L-0011728_lg.csv has decimalLatitude & decimalLongitude rounded to 3 decimal digits (e.g. 60.467). WIS-L-0011729_lg.csv has decimalLatitude rounded to 2 decimal digits (60.15), decimalLongitude rounded to 1 decimal digit (-152.6). Typical of variations found throughout the files. It's possible that trailing zeros were just stripped off, but this inconsistency makes it impossible to match all the labels with a parsing program. | ::Inconsistency in Gold Parsed decimalLatitude and decimalLongitude in many labels. All omitted from NYBG lichens and Tennesee lichens. Gold Parsed WIS-L-0011728_lg.csv has decimalLatitude & decimalLongitude rounded to 3 decimal digits (e.g. 60.467). WIS-L-0011729_lg.csv has decimalLatitude rounded to 2 decimal digits (60.15), decimalLongitude rounded to 1 decimal digit (-152.6). Typical of variations found throughout the files. It's possible that trailing zeros were just stripped off, but this inconsistency makes it impossible to match all the labels with a parsing program. | ||
::Inconsistency in capitalization of verbatim fields in many Gold Parsed lichens. Example: NY01075763_lg.csv. In the label and OCR text the county is capitalized as ST. FRANCOIS, but in NY01075763_lg.csv it is title case: St. Francois. The state MISSOURI is capitalized in both the .txt and the .csv file. The scoring program is case sensitive, so any difference between the gold .csv and the program generated .csv will be marked wrong. | ::Inconsistency in capitalization of verbatim fields in many Gold Parsed lichens. Example: NY01075763_lg.csv. In the label and OCR text the county is capitalized as ST. FRANCOIS, but in NY01075763_lg.csv it is title case: St. Francois. The state MISSOURI is capitalized in both the .txt and the .csv file. The scoring program is case sensitive, so any difference between the gold .csv and the program generated .csv will be marked wrong. | ||
::::Alex will change the metrics to be case-insensitive. --[[User:Dpaul|Dpaul]] 17:28, 26 February 2013 (EST) | |||
::Gold Parsed NY01075759_lg.csv: verbatimEventDate is 1998-04-19, should be 19 April 1998. | ::Gold Parsed NY01075759_lg.csv: verbatimEventDate is 1998-04-19, should be 19 April 1998. | ||