tlg0086.tlg031.1st1K-grc1.xml - metadata typos and 2 possible systematic metadata errors

[julowe]

Hello!

I was looking through this text (Aristotle's Physics) and found a few typos in the metadata. I intend to submit a pull request soon-ish, but documenting typos here.

Oooof, as I was checking possible errors against the source text, I was led down a rabbit-hole of what I think are two systematic errors. The 'False/Extra Line Number at Chapter Start' should be easy enough to regex fix. The '<lb> usage/numbering' question, if it is an error and not just my poor assumption on vocabulary meanings, would be a bit more difficult to fix.

This metadata is useful for a project I am doing, but I am curious - how critical are the line beginning marks for the Scaife viewer and other interfaces that use this data? Are there other digitized-to-text sources of this Greek text (with line numbers) that I could use in the meantime?

I don't have the time to correct all this right now, but I wanted to document this and I also wanted to get feedback on the two bigger systematic errors/questions I had before sinking time into them.

Source Text

Going off of:

First1KGreek/data/tlg0086/tlg031/__cts__.xml

Line 5 in 6812dab

<ti:description xml:lang="mul">Aristotle. Aristotelis Physica. Ross, W.D., editor. Oxford: Clarendon, 1960.</ti:description>

I am using this Internet Archive edition: https://archive.org/details/aristotlesphysic0000wdro/ (and also https://archive.org/details/aristotelisopera01arisrich as a rough check on text & line numbers)

Ok, I missed this link until later http://digital.slub-dresden.de/id416133894 at the head of the tei file, but that page is painfully slow, and I don't want to update all the links below now. A quick spot-check of a few pages show the same Greek text pages here and on Internet Archive (but no intro etc as in Internet Archive edition)

✅ False/Extra Line Number at Chapter Start

First1KGreek/data/tlg0086/tlg031/tlg0086.tlg031.1st1K-grc1.xml

Lines 1595 to 1597 in 6812dab

	<div type="textpart" subtype="chapter" n="6">
	<p>Ὅτι δʼ εἰ μὴ ἔστιν ἄπειρον ἀπλῶς, πολλὰ ἀδύνατα <lb n="6"/>
	<lb n="10"/> συμβαίνει, δῆλον. τοῦ τε γὰρ χρόνου ἔσται τις ἀρχὴ καὶ

on line 1596: <lb n="6"/> should be removed, it is the chapter number (which is captured in metadata one line above) See: https://archive.org/details/aristotlesphysic0000wdro/page/178/mode/1up

Oh, hmmm, and this seems like it is a systematic error. I was looking for this error above to get the line permalink, and instead I first found this error below and others like it. I'd imagine most chapter number marks have this problem?

TODO: look at other subtype="chapter" lines for spurious lb tags This did affect many chapters, fixed with regex, see commit message in linked PR.

First1KGreek/data/tlg0086/tlg031/tlg0086.tlg031.1st1K-grc1.xml

Lines 418 to 422 in 6812dab

	<div type="textpart" subtype="chapter" n="6">
	<p>Ἐχόμενον δʼ ἂν εἴη λέγειν πότερον δύο ἢ τρεῖς ἢ πλείους <lb n="6"/>
	εἰσίν. μίαν μὲν γὰρ οὐχ οἷόν τε, ὅτι οὐχ ἓν τὰ ἐναντία, ἀπείρους <pb n="13"/> δʼ, ὅτι οὐκ ἐπιστητὸν τὸ ὂν ἔσται, μία τε ἐναντίωσις ἐν
	παντὶ γένε ἑνί, ἡ δʼ οὐσία ἕν τι γένος, καὶ ὅτι ἐνδέχεται ἐκ
	πεπερασμένων, βέλτιον δʼ ἐκ πεπερασμένων, ὥσπερ Ἐμπε· <lb n="15"/>

on line 419: <lb n="6"/> should be removed, it is the chapter number (which is captured in metadata one line above)

✅ Line number typo

First1KGreek/data/tlg0086/tlg031/tlg0086.tlg031.1st1K-grc1.xml

Line 1669 in 6812dab

ἔχειν τινὰ ὁμοιότητα <lb n="10"/> τῷ ὅλῳ. ἔστι γὰρ τὸ ἄπειρον τῆς τοῦ

should be line begin 20, not 10. This <lb n="10"/> appears between a line 15 and then 25, so I assumed it should be 10. But I went to source to make sure it wasn't a typo there or something. And indeed, it should be 20 - see https://archive.org/details/aristotlesphysic0000wdro/page/181/mode/1up and https://archive.org/details/aristotelisopera01arisrich/page/207/mode/1up But those sources led me to another question...

❓ <lb> usage/numbering?

Is <lb> 'line beginning' as here: https://tei-c.org/release/doc/tei-p5-doc/en/html/ref-lb.html ? Because I think the lb tags in this section of text (same as above, just more lines of context here) are misplaced:

First1KGreek/data/tlg0086/tlg031/tlg0086.tlg031.1st1K-grc1.xml

Lines 1661 to 1670 in 6812dab

	ἔξω, τοῦτʼ ἔστι τέλειον καὶ ὅλον· οὕτω γὰρ ὁριζόμεθα τὸ ὅλον, οὗ <lb n="10"/> μηδὲν ἄπεστιν, οἱον ἄνθρωπον ὅλον ἢ κιβώτιον. ὥσπερ δὲ <pb n="60"/> τὸ καθʼ ἕκαστον, οὕτω καὶ τὸ κυρίως, οἷον τὸ ὅλον οὗ μηδέν
	ἐστιν ἔξω· οὗ δʼ ἔστιν ἀπουσία ἔξω, οὐ πᾶν, ὅ τι ἂν ἀπῇ. ὅλον δὲ καὶ
	τέλειον ἢ τὸ αὐτὸ πάμπαν ἢ σύνεγγυς τὴν φύσιν. τέλειον δʼ οὐδὲν μὴ ἔχον
	τέλος· τὸ δὲ τέλος πέρας. διὸ βέλτιον οἰητέον Παρμενίδην Μελίσσου
	εἰρηκέναι· <lb n="15"/> ὁ μὲν γὰρ τὸ ἄπειρον ὅλον φησίν, ὁ δὲ τὸ ὅλον
	πεπεράν· θαι, “μεσσόθεν ἰσοπαλές ”. οὐ γὰρ λίνον λίνῳ συνάπτειν ἐστὶν τῷ
	ἅπαντι καὶ ὅλῳ τὸ ἄπειρον, ἐπεὶ ἐντεῦθέν γε λαμβάνουσι τὴν σεμνότητα
	κατὰ τοῦ ἀπείρου, τὸ πάντα περιέχειν καὶ τὸ πᾶν ἐν ἑαυτῷ ἔχειν, διὰ τὸ
	ἔχειν τινὰ ὁμοιότητα <lb n="10"/> τῷ ὅλῳ. ἔστι γὰρ τὸ ἄπειρον τῆς τοῦ
	μεγέθους τελειότητος ὕλη καὶ τὸ δυνάμει ὅλον, ἔντελεχείᾳ δʼ οὔ,

the lb tag should corrected and moved to instead be like so (code-lines 1668-1669):

 κατὰ τοῦ ἀπείρου, τὸ πάντα περιέχειν <lb n="20"/> καὶ τὸ πᾶν ἐν ἑαυτῷ ἔχειν, διὰ τὸ 
 ἔχειν τινὰ ὁμοιότητα τῷ ὅλῳ. ἔστι γὰρ τὸ ἄπειρον τῆς τοῦ 

I've included more context because this looks like an error on all right-hand side pages (that I checked) of transcribing this source text (in the file, the even numbered pbs)

Important

TODO: see if this end-of-line vs beginning-of-line error occurs on all right-hand pages of this text

A quick scan of this text versus source images shows:

• pb 59: correct, has lb marks at the beginning of the lines. see: https://archive.org/details/aristotlesphysic0000wdro/page/180/mode/1up
• pb 60: incorrect, has lb marks at end of line of text (said another way, all lb numbers are off by one, e.g. lb 15 would be accurate if it was instead lb 16). see: https://archive.org/details/aristotlesphysic0000wdro/page/181/mode/1up
• pb 61: correct, has lb marks at the beginning of the lines. see: https://archive.org/details/aristotlesphysic0000wdro/page/182/mode/1up
• pb 62: incorrect, has lb marks at end of line of text. see: https://archive.org/details/aristotlesphysic0000wdro/page/183/mode/1up

This error occurs with the <note type="marginal"> marks as well on right-hand side pages, depending on how those marks are used/interpreted. I was taking these to be an implied 'beginning of line number 1 of page 207 column a'. But as the tei file is now, it is technically correct - the text of the marginal note is correct, and the marginal note is correctly placed according to the source text. It is only 'wrongly' placed if we assume it to mark the beginning of the line. If all the 'incorrect' lb tags on the even pb pages were instead just <note type="marginal"> types, then they too would be in the correct place.

So the bigger question is: Is the intent of this file to accurately encode the location of marginal notes according to the source text? Or to accurately mark and number the beginning of lines of text? Both could be done, but not sure how these files are used/interpreted downstream. e.g. on left-hand side pages this tei file could, for marked line numbers, say: <note type="marginal">15</note><lb n="15"/>... and right-hand side pages would say: <lb n="20"/> καὶ τὸ πᾶν ἐν ἑαυτῷ ἔχειν, διὰ τὸ ἔχειν τινὰ ὁμοιότητα <note type="marginal">20</note>

Also ref: https://archive.org/details/aristotelisopera01arisrich/page/206/mode/1up, https://archive.org/details/aristotelisopera01arisrich/page/207/mode/1up

✅ Line Number in incorrect place

First1KGreek/data/tlg0086/tlg031/tlg0086.tlg031.1st1K-grc1.xml

Lines 1648 to 1652 in 6812dab

	καθαίρεσιν ἄπειρον ὑπάρχει (ἡ γὰρ μονὰς ἐλάχι· <lb n="33"/> στον, οὔτε
	<add cause="fix">τὸ</add> ἐπὶ τὴν αὔξην (μέχρι γὰρ δεκάδος ποιεῖ τὸν
	ἀριθμόν). <lg>
	<lb n="33"/>
	<l>συμβαίνει δὲ τοὐναντίον εἶναι ἄπειρον ἢ ὡς λέγουσιν.</l>

should be:

 καθαίρεσιν ἄπειρον ὑπάρχει (ἡ γὰρ μονὰς ἐλάχι· στον, οὔτε 
     <add cause="fix">τὸ</add> ἐπὶ τὴν αὔξην (μέχρι γὰρ δεκάδος ποιεῖ τὸν 
 <lb n="33"/> ἀριθμόν). <lg> 

assuming that: if a word is split over a line ending, then place the lb mark at the beginning of the word (and un-split it in this tei file). If the lb mark should go after the end of the split word, then it should be: ἀριθμόν<lb n="33"/>). <lg> ? see: https://archive.org/details/aristotlesphysic0000wdro/page/180/mode/1up

❓ Confusing Marginal Note

First1KGreek/data/tlg0086/tlg031/tlg0086.tlg031.1st1K-grc1.xml

Line 1615 in 6812dab

<note type="marginal">29a</note>

This is technically a marginal note, but is really a line number where the source text has a deletion marked. Not sure if anything really needs to be done for this, it doesn't seem to mess up the Scaife Viewer of this portion

See: https://archive.org/details/aristotlesphysic0000wdro/page/179/mode/1up and
actual version used: https://digital.slub-dresden.de/werkansicht?tx_dlf%5Bid%5D=109594&tx_dlf%5Bpage%5D=67

[lcerrato]

@julowe
Thanks for this.

The short answer is that <lb/> is largely irrelevant in a prose text and is not used here as any sort of marker or container. This is true for most works.

In the instance here where you have the facing pages representing the breaks differently (start of vs end of line), I agree that, in the absence of manually correcting this, a better representation would have been notes with a place marker in the margin or milestones with an attribution to Bekker. I believe other Aristotle has Bekker milestones. Ideally, these would be moved to their appropriate locations to be most useful.

As Perseus markup typically employs <lb/> for presentation clarity in instances where lines would otherwise run-on, this particular usage is not in keeping with that. I also see stray and unnecessary <lg> throughout.

I agree that any markers that represent chapters should be removed.

Where the <lb/> (or whatever I change this to) goes in terms of before or after a word around 207a:

it does not matter. Some editors move to the start of the word and others move to the end. In this case, we would not break up the word and the subsequent punctuation. Before seems more intuitive here. In any case, there need not be two "33" markers and the <lg> is completely wrong (as I suspect it elsewhere).

I am happy to make corrections to the file that better reflect a consistent representation of the marginal numbering, but I suspect that a large text like this will take some time to review and update. There is other missing information here that it would be good to add as well.

Unfortunately, the lack of print page numbers in the SLUB text makes it hard for me to quickly cross check some of the pull request changes, but I can delve into it further over the weekend.

[julowe]

Ah, about the SLUB text numbers. That reminded me of another issue...

I found the SLUB text near the end of my editing (sorry again for introducing yet another link/source to try and deal with! though the bekker edition on Internet Archive did help me understand Ross's more complicated edits and place those line numbers appropriately), and I noticed that the the pb numbers were 8 off of the SLUB pages.

So <pb n="2"/> ὥσπερ οἱ φυσικοί,

First1KGreek/data/tlg0086/tlg031/tlg0086.tlg031.1st1K-grc1.xml

Line 128 in 6812dab

ἤτοι ἀκίνητον, ὥς φησι Παρμενίδης καὶ Μέλισσος, ἢ κινουμένην, <pb n="2"/> ὥσπερ οἱ φυσικοί, οἱ μὲν ἀέρα φάσκοντες εἶναι οἱ δʼ ὕδωρ τὴν πρώτην

corresponds to their permalink doc_id/page_num: http://digital.slub-dresden.de/id416133894/10 (well, with second half of the split-word pushed before the pb tag...)

<pb n="3"/> συμβαίνει λέγειν αὐτοῖς,

First1KGreek/data/tlg0086/tlg031/tlg0086.tlg031.1st1K-grc1.xml

Line 153 in 6812dab

φύσεως μὲν οὔ, φυσικὰς δὲ ἀπορίας <pb n="3"/> συμβαίνει λέγειν αὐτοῖς,

corresponds to http://digital.slub-dresden.de/id416133894/11

and so on.

✅ pb numbers need to be corrected to be consecutive/correspond to SLUB pages

...Except the pb numbers were not always 8 off. Which I thought was odd, but forgot to verify against source text. Oof.

simple sanity check:
grep "<pb" tlg0086.tlg031.1st1K-grc1.xml | tail -1 returns:

                            <pb n="207"/> πάλιν καὶ πάλιν, τῷ ἐφεξῆς εἶναι συνεχῶς· ἢ γὰρ αὐτὸ δεῖ

but grep "<pb" tlg0086.tlg031.1st1K-grc1.xml | wc -l returns: 206...

digging into the grep output and the file, it looks like there are duplicate pb 52, it skips 77, and skips 94.

Fixed in latest commits of PR. So now, if you use my latest commit from the PR, you can use the permalink to SLUB doc/pageNum and just add 8 to every pb (or download their PDF, I just found that as the 2nd option under the download icon...) - hopefully all this helps make it easier to check my PR!

Also, the perl code I used to correct the numbers is in the commit message, just for reference.

While I was thinking of this, I did a quick check with grep and it looks like all the bekker references that are marginal notes are in order (or, in the case of one convoluted edit that Ross made, are transcribed from the source text correctly).

[julowe]

Ah, thanks for explaining about the lg tags because I was curious about that usage - ha I just didn't want to tack on yet one more question!

It seemed like the transcriber of the source text was trying to capture when Ross split one original Bekker line into two, but what is the better way to do that?

ref page 206, column b, line 33 in the Internet Archive version: https://archive.org/details/aristotelisopera01arisrich/page/206/mode/1up

oh, but image attached here to save you jumping through the hoops to borrow the book:

but if they were trying to capture the idea that "yes these are two lines of text in the Ross edition, but actually they are only one line of text in the Bekker edition", then I (with just about zero familiarity with TEI) would have captured both lbs like so (after correcting the transcribing error where the first lb 33 was place one line out of place):

                            <lg>
                                <lb n="33"/><l>ἀριθμόν). </l>
                                <lb n="33"/>
                                <l>συμβαίνει δὲ τοὐναντίον εἶναι ἄπειρον ἢ ὡς λέγουσιν.</l>
                            </lg>

but even that seems a bit odd to me.

ok, I also forgot that this was not the most simple example. In latest commit of this PR I've corrected the first lb 33 to be on the correct line (one line lower) and before the split word as you suggested. So that, for now, this lb tag corresponds to the source text. Oh, and added a missing ) and corrected the split word that instead of a -, had a · and that lb tag that was placed one line early splitting it.

I changed the header in my original comment for this issue to use a check mark, as I'll consider this closed and punt on the bigger issue of lg usage 😹

[lcerrato]

For this part of the question:

It seemed like the transcriber of the source text was trying to capture when Ross split one original Bekker line into two, but what is the better way to do that?

In a prose file, one would indicate an indentation or new paragraph. A <l> tag is for drama or poetry and would not otherwise appear in prose.

If they endeavored to indicate a new line break, this is where you want attribution for the <lb/>. The question is did they intend to include only the Bekker line breaks, in which case these should be indicated as resp="Bekker" or do they wish to also capture the Ross formatting? If the latter, is this a line break or paragraph break (<p resp="editor (or Ross)" rend="align(indent)">? Further, is this the only place they try to capture the Ross formatting? If so, I'd ignore it. Otherwise, I am left to try and reinsert paragraphs where needed.

In no instance would we want to mix and match the line breaks in a work where we need the Bekker numbering to be preserved (if imperfectly) without disambiguating these somehow.

Does that help?

[lcerrato]

I forgot to add that I agree re SLUB. I use a download if I need to check more than a page or two. Their interface is a challenge.

[julowe]

That does help - thank you!

All the <lb> tags I saw were references to Bekker line numbers, and I'd be surprised if any were creations from transcribing the images or creations by Ross to other sources. But I haven't done any kind of systematic check on that. Perhaps one easier fix would be those lines where Ross puts a paragraph break in his book (where there was no break in Bekker) and has those double Bekker line numbers - those should be easy to spot and maybe even easy to regex for? Heh maybe not...

Tackling the left vs right side Ross pages and the alternating beginning vs end of line <lb> marks is further complicated because the original line breaks in Ross are not preserved in the tei file. It looks like the text content was wrapped at 100 characters (but not always). So just using the tei file, you can't tell where the beginning of the line would be and thus can not just easily move the mark from the end to the beginning...

Is there any suggested or set line length requirement - either by TEI or by your org? Is there a preference towards uniform length or towards mirroring the source text's line breaks?

I did manage to get a decent OCR of the SLUB files (using Surya OCR which reliably retains marginal notes, unlike most) so this could be used as a template for formatting only. Something like: a fuzzy match per line of this new OCR text against the TEI file, and then shift around the line breaks in the TEI file's textparts (not changing any characters at all), and then making <lb> placement uniform. I can send along the ocr file if that is of interest. But I didn't think making those placements uniform is a high priority, and the time to code up and edit for this step seemed large, so I didn't sink time into that.

[lcerrato]

I think it's useful to clarify between a CTS container and the other markup. Lines are a CTS container for Homer for example, where the structure is book/line. They are also used in drama: generally there the line is the only container. (Act and scene are preserved in the text but not utilized for CTS purposes). Prose will have books/chapter/section or some combination of those. some have alternatives like paragraphs.

There is no set (line or container) length for Perseus/OGL, but others have brought to my attention texts where the containers are too large for post-processing. I have manually imposed breaks to alleviate this in some works, but our practice is to defer to traditional citation schemes as the basis for containers where possible.

Lines are not a prose container and would not be applicable here.

Aristotle is a special case given that the traditional citation scheme is not an ideal CTS container because the Bekker pages/columns break irregularly. It is preferable to split containers on hard breaks, and barring that, at least something like a comma.

If one chooses Bekker pages as the default CTS container for Aristotle, however, you have breaks in the middle of sentences (and words).

This is one of the works that has an alternative to Bekker for containers, but for those that do not, we have imposed an arbitrary break roughly corresponding to Bekker and preserved the correct break as a milestone. This may not be the best approach as we move forward, though. There is additional work being done on Aristotle and our current markup may need further refinement.

I do not know if the original OCR for these files was preserved but in using Lace, we can put anything where we want it when doing the capture, so the line numbers could have been moved in that format. (Early Perseus data entry was double-keyed so this type of thing was done at the time of capture).

In the past, we have just done the Bekker cross-referencing manually. It's a good student project.

The short version is that I need to take a deeper look at the file and do some cleanup before I delve any further into refinement but much of the OGL work was focused on basic text presentation rather than any uniform markup. There are significant differences between this collection and the Perseus texts in that regard.