Mistakes in Thesaurus

[matthewkumar]

If you notice any mistakes in the thesaurus that need to be manually overridden, please list them below.

[matthewkumar]

animal.csv
line 124
'calandra lark' should normalize to 'lark'

[ronikaufman]

Some files are missing the prefLabel_en column: measurement.csv, material.csv, sensory.csv.
And definition.csv is empty.

[matthewkumar]

All fixed except definition

[danachaillard]

In animal.csv there is a 'smalls female lizards' that goes to smalls female lizard instead of lizard

[matthewkumar]

I imagine 'smalls' is being read as a noun rather than an adjective, which throws off the semantic head identification. I wonder why 'smalls' got added to the english version of the manuscript.

[ps2270]

See fol. 133r - it's an author deletion in the ms. that was carried over as non-deleted into the vocab - is that a common problem?

…

On Sat, May 16, 2020 at 11:58 AM matthewkumar ***@***.***> wrote: I imagine 'smalls' is being read as a noun rather than an adjective, which throws off the semantic head identification. I wonder why 'smalls' got added to the english version of the manuscript. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1616 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AF75UYH6LF7CQNOKBXP33WTRR2Z25ANCNFSM4MQEQFXA> .

-- Pamela H. Smith Seth Low Professor of History <http://history.columbia.edu/faculty/smith-pamela-h/> The Making and Knowing Project Director, Center for Science and Society <http://scienceandsociety.columbia.edu/> Chair, Presidential Scholars in Society and Neuroscience <https://presidentialscholars.columbia.edu/> Columbia University 605 Fayerweather Hall, MC 2516 1180 Amsterdam Avenue New York, N.Y. 10027 (212) 854-7662-Phone (212) 851-5963-Fax [email protected] The Making and Knowing Project has just released *Secrets of Craft and Nature in Renaissance France: A Digital Critical Edition and English Translation of BnF Ms. Fr. 640: *http://edition640.makingandknowing.org *Entangled Itineraries: Materials, Practices, and Knowledges across Eurasia <https://www.upress.pitt.edu/books/9780822965770/> *is now available. *The Body of the Artisan <http://press.uchicago.edu/ucp/books/book/chicago/B/bo3618964.html> *has been reissued in paperback and electronic editions. *The Business of Alchemy <https://press.princeton.edu/titles/10868.html> *is available in a new paperback.

[danachaillard]

From my observations in the vocabulary that is the only case of the problem. So it's okay I was just wondering what hapenned.

[danachaillard]

Few other things I found while making the word clouds, I'm sorry I don't have the lines in the csv though.

In animals : Those with a name containing two nouns do not default (ex: "goat kid"). That's the weakness of the way we do this. I don't see a way to fix this apart from retaining all exceptions. Also there is a "viper color" tagged as an animal, I don't know what to do with that.

In body parts : Do we want "joint of the little finger" to point to "finger". Or will that be part of the abstracting part.

In materials : many things go wrong. Some I don't understand "large steel dice" for example is not changed. Some are just too long or too complicated semantically "latten mad brittle by the calamine. I don't know what to think of "tin is base".

In weapons : We have the double noun problem again with "arquebus powder". "arquebus a croc" doesn't work either but it's probably because it's in french so it doesn't find the name ?

[matthewkumar]

Re: smalls little lizards
It's good that this is the only instance, and we could fix it by adding it to our list of manual corrections. However, it brings up a point Naomi, Terry, and I have discussed on how we should represent del tags throughout the repository. I'll make sure to bring this case up next time we speak.

@danachaillard
It's probably beyond the scope of this project to correctly normalize every term that appears in the manuscript. Our best hope is to maximize the amount we can normalize while minimizing the amount of errors. In general, I'd say its better to avoid generating errors than generate something for everything. I agree that terms like 'large steel dice' would ideally be normalized, but I'm not sure what protocol we would follow to make that happen. What have you been thinking?

[danachaillard]

It depends on how the method to find the 'noun' works, because in every instance where one of he adjectives is also a noun it seems to fail. What does the method return when it doesn't find ? Could it give us all the nouns it finds.