simoneduca
commented
7 years ago I confirm there are 15 duplicates of the following transcription in the raw CSV file
{
"subject_id": 1275389,
"text": "ffor an Ague\nA toy Seigneur s'entend ta Creature\nTake two quartes of good Ale make a possett and take\nof the Crudd then take a good handfull of Ribbe and\nEt en son tempe tu luy donner pasture\nboyle itt a good while in the possett drinke and putt\nOuurant ta main par ta faueur tres grande\nin a little pepp<ex>er</ex> and drinke itt in the morninge\nand fast an hower after and att night when you\nA toy Seigneur s'entend ta Creature\ngoe to bedd for foure or five dayes\nEt en son tempe tu luy donner pasture\ntoe comende wele\nThe herbes are to boyle till the vertue\nOuurant ta main par ta faueur tres grande\nbe boyled out of them the said quantitie\nA toy Seigneur s'entend ta Creature\nare to be droncke at fyue draughts\nEt en son tempe tu luy donner pasture\nblude warme./\nOuurant ta maint",
"variants": "herbes, sentend, luy, boyled, donner, ffor, faueur, ouurant, fyue, draughts",
"user_name": "hwolfe",
"Author": "",
"Call Number": "V.a.140",
"Filename": "123203.jpg",
"Genre": "Receipt Books",
"Hamnet URL": "http://hamnet.folger.edu/cgi-bin/Pwebrecon.cgi?BBID=231384",
"Luna URL": "http://luna.folger.edu/luna/servlet/detail/FOLGERCM1~6~6~1201046~197000",
"Origin": "compiled ca. 1600",
"Page Number": "folio 18 verso || folio 19 recto",
"Page Sort": 24,
"Priority": 28,
"Title": "Receipt book [manuscript]."
}
VVH
vrooje
camallen
eatyourgreens
Recently, @astopy put together a file of transcriptions by a few select users from ShaxWorld whose transcriptions we trust. This is because we've been having trouble with the aggregation code. It is a short term measure. Paul at the Folger noticed that there are some subjects in the csv file with multiple identical transcriptions submitted by the same user. For example, a transcription by hwolfe for a page in V.a.140 (file 123203.jpg, subject id 1275389) appears 15 times. This has implications for aggregation which uses a majority rules mechanism to determine agreement for a line, and to retire lines by putting the grew dots around them. If 15 transcriptions by the same user are thrown into the pot, these could overwhelm submissions by other users. I've talked to @astopy and @simoneduca about this and we think the possibilities for where this error is originating are as follows:
1) somewhere on the front end in the way transcriptions are captured and submitted to the database. If this is the origin, it's something @rogerhutchings would need to look at because it could originate with AnnoTate.
2) somewhere on the backend, but none of us could think how/where/why, so that's possibly a question for @camallen and @marten once we've ruled out front end origins.
3) Somewhere in the pipeline that gets transcriptions into the database and then out again...but I don't really understand how that would work either.
A volunteer raised a very detailed series of questions over on Talk, which might be related. Carl began by saying that he was seeing subjects several times over. He has also said that when he transcribes a page his classification counter sometimes changes by 2 or 3, rather than 1. This might indicate that multiple transcriptions are being submitted per session. See more here.
Have we seen any behaviour like this elsewhere? @simensta, @hughdickinson, @eatyourgreens have you noticed these problems on text projects before? I doubt it's text specific, but I figured best to open this investigation wide to start, and try to narrow down from there. Thanks in advance.