Missing morph tagging in new Loeb volumes

The title of the Heracleitus volume is misspelled in Greek, the final sigma has been ocr'd incorrectly as a capital epsilon.

ΠΕΡΙ ΤΟΥ ΠΑΝΤΟΕ

should be:

ΠΕΡΙ ΤΟΥ ΠΑΝΤΟΣ

Thanks for the update Kyle, I appreciate it.

You can add the Celsus De Medicina volumes to the list of resources to be updated to add morphological tagging. The Latin missing tagging too.

And please do consider adding a new data type for Heraclitus. He's a major Greek Pre-Socratic philosopher, quoted by Luther and particularly important for the early church fathers (my library shows him being quoted by Justin Martyr, Tatien, Clement of Alexandria, etc.) Then, of course, there are all the references in secular literature: Dante, Lucretius, Marx, Francis Bacon, Marcus Aurelius, Plutarch, Hegel, etc.

Butters said:

The Loeb volumes are nearly useless; there are simply digital texts that one can get anywhere - but with jumbled/incorrect/missing morphological information, etc.    

I understand your frustration, Butters, and I do get a little tired of doing quality control for Logos for free, but I think your characterization of "nearly useless" might be a bit of an overstatement.

While the automatic morphological tagging is not perfect, I've found it to be good enough for 95% of the words I need to look up. (I even use the lack of tagging as a first indicator that the word might be OCR'd incorrectly: in which case I go to one of the original scans and check before reporting a typo.) I don't expect perfect tagging, because that would require a human being (or more likely, a team of people) to go through and check and tag the texts manually--an unrealistic and incredibly costly scenario, especially for the number of volumes being produced. Automatic tagging is about as good as one can reasonably expect for the price.

Also, the Logos/Noet team has been pretty good about updating resources with problems. See this thread for instance. Others, however, are still waiting.

In the end, while I think the Logos Loeb volumes are fabulously overpriced for what they are, I'm very happy to have purchased all of them progressively through Community Pricing. The note-taking abilities, the cross-device synching, the linking of English and Greek/Latin texts, as well instant access to dictionaries and encyclopedias are all rather wonderful features, especially compared to working with scanned PDF files, even if they are freely available. (Even Perseus is less user-friendly than Logos, in my opinion).

Ah, I see! I'm afraid the problem you're having is to be blamed on Latin, not on Logos.

As you probably already know, Latin morphology is rather polysemantic, meaning a given form can have many different meanings and interpretations. An automatic morphology program is going to give you ALL possible meanings, rather than just the right one.

Take your nostrum example: from a purely morphological standpoint, it can either represent a form of noster, nostra, nostrum (the possessive pronoun our), OR it can be the plural form of ego (a personal pronoun). And then, within those two forms (either an adjective or a pronoun), it can either be singular, neuter, nominative OR singular, neuter, vocative OR singular, masculine, accusative, and so on and so forth.

So what you're seeing in the Logos popup you posted is just a reflection of the nature of Latin: any automatic tagging will list all possible forms, not just the "right" one. Getting the exactly right form for a given sentence would require someone with a good knowledge of Latin go in and choose from among all those possibilities. And, as I said in my previous post, that would cost a fortune. So, instead, Logos lists all possible interpretations for a given morpheme. You happened to pick one with lots of possibilities, but that's the richness (and difficulty) of classical languages.

I hope that makes it clearer!

Is this the Iliad text you used? That one happens to be the only non-Bible-related Greek text I'm aware of that has complete morphological tagging done by a human being, which explains why you were finding precise morphological information and not just a list of different possibilities as per Augustine's Confessions.

I do agree with you that Logos could/should make it clearer that automatic tagging does not mean 100%-accurate morphological information for each and every word.

One suggestion, if you're finding the popup to be unhelpful is to use the Information panel, it offers a somewhat "cleaner" view of all the morphological possibilities for a a given term.

Kyle G. Anderson said:

I have a case in for Meraclitus for work on a data type. It'll probably be a couple of months before I'm able to get to it.

I noticed that a data type has been added for Heraclitus' fragments. Now books can point to individual fragments (Fragmentum 15, for example), rather than just the book as a whole. A big improvement, thanks Kyle!

However, I noticed that the data type is misspelled as "Heraclides", one of the legendary descendants of the Greek hero Heracles, and not "Heraclitus" as the philosopher's name is generally spelled (though it is "Heracleitus" in the Loeb edition).

Sorry to nitpick, but it's pretty important...

Michel Pauw said:

How do you know that that one is done by a human? Do you infer that from the quality of the tagging?

I seem to remember the project description talking about how this text was made, but it's true that now it's not clear that this text was edited "by hand" by someone.

Below is from the introduction by John J. Jackson:

Preface

My purpose has been to produce an interlinear that is both inductive and deductive in approach, allowing the student to quickly read large amounts of Greek or Latin, and provide a platform for comprehensive study. I find the Libronix Library System to be perfectly suited for this task and am proud to make an addition to the library.
John J. Jackson
[...]

The text of Monro and Allen’s 1920 edition of the Iliad contains a total of 111,862 words, generated from 21,418 unique wordforms. These, when lemmatized, yield 18,430 unique dictionary headwords.  
Upon further analysis we find 12766 unambiguous forms (those with only one morphological parse). Forms of this type were parsed by the use of an automatic parser thus reducing the corpus to 8,652 ambiguous wordforms. Additionally we find that wordform to lemma assignment exhibit fewer than 200 ambiguous relationships.
In addition to an automatic parser, lexicon driven knowledge based systems were created to resolve areas not handled by the parser.

So it would appear that automatic parsing was also used on this text, albeit with more human attention..