Wanna find typos in your resources? Search for "modem"

Todd Phillips said:

I found a scan error in one of my resources: "modem" instead of "modern".  I figured this was a common error so I searched my library for "modem" so I could report more typos.  Then I got discouraged:

This is a theological library, so the probably is that most of them are errors. Maybe someday I'll get them all reported.

I suggest Logos build a list of these common scan errors (especially words with "rn") and search for them to help auto-proofread.

Good suggestion.

Todd Phillips said:

Then I got discouraged:

Goodness! You are are right--they are all over the place! Wish I could help Logos get them corrected easily.

Eeek! 796 occurrences in my library.  zowie.

Here we are almost two years later and I have 1403 occurrences of modem in my library. I'm actually surprised how many do refer to the computer modem but it seems no other typos have been fixed. e.g. Modein is still misspelled as modem. I wonder what percentage of typo reports are actually dealt with...in a timely manner.

NB.Mick said:

I repeat: the problem is in the process. With a growing number of resources and a growing userbase, the fine people taking care of this can't possible fix the typos in a reaonable amount of time.

Some changes are discussed on the forum from time to time. I would ask them to go to the publisher only in cases that require factual clarification. I would like my Logos resources to be better than the printed copies and there can't be a reason to retain an error just because the publisher made it and not Logos. The publishers should be happy that we make the work of their downsized editing department! I would propose checking the (scan of the) paper copy only in dubious cases.

For me, the process should utilize crowd intelligence and be as much automated as possible. Proposed typo corrections should automatically be converted into a textual change in a sort of add-on layer - such as those possible with a visual filter, marking up corrected text somehow. Then I could choose to see the text with or without correction. The entries should work automatically for single correction proposals and logos should provide mass changes (like you propose) with a script. If they want to empower us even more, they could have us confirm or dispute a proposed correction and not show disputed ones.

[Y] good suggestion

NB.Mick said:

I would ask them to go to the publisher only in cases that require factual clarification.

I suspect that's impossible. It's probably a copyright issue.

NB.Mick said:

The problem is the process

The main problem is that they don't get it right from the start.

WBC aren't produced Logos, they are 3rd party - "Thomas Nelson/Word".. mine is just as bad though I do get false positives.

I went with the overkill approach, and made a VF

I must confess Logos are getting better at fixing typos, but I also agree that you/me would not want a re-download for each typo fix, >200 in my case, so I think frustrating as it is the user experience is better as it is, but if they want to employ a typo fixer, then I may be willing to emigrate LOL

NB.Mick said:

(at least that's what I remember from a Logos employee explainig it).

Highlight of typo mine: "employee explainig it" [:O]   [:D]  I think the process will never be perfect as long as it involves machines or humans.

There is a reason we measure OCR programs by "accuracy rates."  Not one of them is perfect.  I do think the output can be improved by running the scans through multiple and varied OCR programs and compare the differences of the output.

LOGOS could hire typo correction staff and have the them "telecommute." After all, the typos are distributed through the cloud......[st]

I believe I've found another high frequency OCR error word.

My pastor quoted Spurgeon comparing clowns to certain pastors and upon searching for the quote in my library, I noticed many instances of the word "down" interpreted as the word "clown"

A search of my library returns 905 results for the word "clown".  In many of the older  public domain works though highly comical are clearly errors.

Can anyone else confirm this?

With the release of all the new data sets we can see why typos haven't been a priority.  I don't find them a problem but report what I come across.  My only concern would be incorrect hyperlinks being fixed as this effects function.  Typos can on the other hand improve daily smile factor so they are less of a prority

Rosie Perera said:

Andrew Mckenzie said:

With the release of all the new data sets we can see why typos haven't been a priority.  I don't find them a problem but report what I come across.  My only concern would be incorrect hyperlinks being fixed as this effects function.  Typos can on the other hand improve daily smile factor so they are less of a prority

Typos aren't a problem for me when I'm reading, because I know what was intended. But they're a problem when I'm searching. For example, if I were searching for Soren Kierkegaard, it wouldn't find any occurrences that were spelled S0ren Kierkegaard.

I think this would the greatest incentive for Logos to focus on correcting such scanning errors/inconsistencies as they are found.  The percentage of error is incredibly small, yet they still cause innacuracies in searches.

fgh said:

" rel="nofollow">Keep Smiling 4 Jesus said:

After reading about "scannos", wonder about "be" for "he", "arid" for "and", ...

There are plenty of words that can easily become other words. The thing with modem and clown is that in Logos they're almost certain to be errors, so it shouldn't be that hard to get rid of them. And yet there's been modem's even in resources released this autumn. 

Be vs he, on the other hand, requires careful reading by a real person.

 

Andrew Mckenzie said:

I don't find them a problem (...) they are less of a prority

They should be a very high priority. Logos sells their products with the claim that the software will find "every" occurrence of the words we search for. As long as resources are full of typos, that's not true. And missing the most relevant search result because of a typo isn't fun.

 

Paul Newsome said:

The percentage of error is incredibly small,

Unfortunately, in some resources it's incredibly large.

 

Not suggesting they shouldn't be a priority just if I had to make a choice I'd much rather see hyperlinks fixed, but do take others view in terms of searching being a significant problem seriously.  For the way I use Logos this comes in a very close second.

I tried to find some of these errors, but what is it about Catholic theologians and clowns?

I have almost 200 PBs, and about 20 of them had the modem for modern error. I corrected those!  Check your own PBs...

I have 543 clowns, but looking them over, most of them seem to be real clowns.

Todd Phillips said:

Rosie Perera said:

Todd Phillips said:

After much deliberation, I have concluded that the word "modem" should be thrown out of the English language.  Before long, analog carrier signals will become obsolete. Un-naming the "modem" will hasten the process, which is in everyone's best interest---both for increased network bandwidth and for OCR accuracy.

Throwing the word modem out of the dictionary won't solve the OCR accuracy problem. I encounter tons of OCR errors in creating my PBs where the "words" are not anything recognized in any dictionary. Mixtures of letters and numbers, etc.

I didn't say it would solve all OCR problems. It would just solve the Modem problem...every little bit of accuracy counts.

(You do realize I was being tongue-in-cheek?)

I know you were being tongue-in-cheek, but I like being pedantic about imprecision even with tongue-in-cheek posts. You see, it wouldn't even solve the Modem problem. My point was that OCR doesn't care if something is in the dictionary or not (it seems to have no problem with S0ren Kierkegaard for example; the o in Soren being a zero). If Modem isn't in the dictionary, it could still show up instead of Modern in an OCR'ed text. So there!

elnwood said:

I've ran some searches on the most common misspellings according to the Oxford English Corpus, and almost all of them show up in my library.

http://oxforddictionaries.com/words/common-misspellings

Here are some results from my library:

"occurrence" appears 579 times, 456 of them in the NASB Update.

"posession" appears 261 times, 172 times in The Old Testament Greek According to the Septuagint, and 64 times in the Lexham Greek-English Interlinear Septuagint.

"foward" shows up 249 times, 228 of them in the Dictionary of Latin Forms.

"millenium" shows up 193 times in 92 resources.

"pharoah" shows up 164 times, including 18 times in Theological Wordbook of the Old Testament, and 13 times in the Anchor Yale Bible Dictionary, which I recall was recently revised.

"persistant" appears 147 times, 145 times in the Dictionary of Latin Forms.

"seige" appears 112 times in 57 resources.

"occuring" occurs 81 times, 32 times in "Life and Epistles of St. Paul."

"ecstacy" shows up 67 times, "familar" 40 times, "calender" 38 times, 

"publically" appears 58 times, 15 of which are in the Faithlife Study Bible.

"threshhold" appears 43 times in 40 resources.

"seperate" appears 42 times, 21 in the Lexham Hebrew-English Interlinear Bible.

Also, it seems like ISBE: Revised Edition had the most typos overall, containing some of several different typos.

It suffices to say that I did not report typo" for all of these typos!

Someone should report all of these, or email Logos to let them know about all of them.

Seems like a check against this list of common misspellings should be built into the process of releasing all new resources. In fact, I'm really surprised they didn't run a basic spell check on everything they've released thus far, as it would have found all of these. None of them are actual words. Spell check would of course miss things that are misspelled but are real words nonetheless (their/there, etc.)

EDIT: Incidentally, part of the problem is that Logos usually prefers not to fix typos when they were in the original they are working from -- i.e., it was the print publisher's error. So if they're working with OCRed texts that came from some print publisher who reprinted an old work, they won't fix them. They might notify the print publisher about the problem, but they don't see it as their right to fix it. And if it's a contemporary work that has a misspelling in the print edition, they won't fix that either.

I just found two typos in a row in Religious and Theological Abstracts, which I'm guessing were in the original: "resiliancy and spontananeity." Reported them dutifully but don't expect anything to be done about it.