LSVA.com

[Mark Lieb]

Be interested to hear people's thoughts on the various autocoding packages. From Syngence and ALCoder to Valora, the options are emerging.

I think all of us can envision a day when all bib coding is done with software. It is inevitable.

What do you think of the technology? Have you used it?

[RowandK]

I think all of us can envision a day when all bib coding is done with software.

Thanks for the vote of confidence, but the ALCoder model actually expects that not all documents are suited to autocoding. Handwritten notes (especially mine) and documents with little structure will probably require human examination for quite some time to come. In the case of these documents, ALCoder flags the records and generates no charge.

The idea with ALCoder is to process a document collection and assign confidence scores to the output. Depending on the end user's available time, budget and comfort with the output, any documents below a certain score would then be manually coded. If the end user decides that only "B-level" are acceptable, then they would manually review the "F's," "D's," and "C's". ALCoder pricing is such that even under these demanding standards, money is still saved.

(Is there a way I can post a spreadsheet showing how this would work?)

[RowandK]

What do you think of the technology? Have you used it?

If anyone is interested in trying ALCoder (on your own PC), please contact me and I'll send you a control key with 1000 clicks.

[Mark Lieb]

The question was how to offer the spreadsheet. Well, if you have it posted on your server, link to it using the URL button when you write your message. You can also add it to the web links directory.

If you cannot host the spreadsheet, let me know and I will see about hosting the info here.

Mark

[dswider]

Be interested to hear people's thoughts on the various autocoding packages. From Syngence and ALCoder to Valora, the options are emerging.

I think all of us can envision a day when all bib coding is done with software. It is inevitable.

What do you think of the technology? Have you used it?

As a fairly interested party in autocoding, I think there are a variety of different service types, which seem to get lumped together under the name "autocoding".

I think we're starting to see the beginnings of an industry that may revolutionize document production the way that imaging did.

Wait until machine generated subjective coding gets here. Trust me, it's coming.

[Pamela]

Dave,

I hate to disagree with such a worthy opponent -BUT, you said that subject auto coding is coming. While that may be true, (just like the "paperless office" is coming some day), I would not hold my breath that it will be soon or that the technology will be accepted.

Subjective coding done by someone other than a Legal Professional is NOT something most law firms are comfortable with, whether it is done by machine or human. Making a determination about what is relevant or responsive is the attorney's job, not the vendors.

Respectfully,

Pamela Towan

[dswider]

Check these guys out.

http://www.h5technologies.com/

They're seeing twice as much recall vs. human review. They work in conjunction with the legal staff to determine what constitutes relevancy and what doesn't. It's pretty amazing stuff. They deliver a Concordance load that indicates which claim the documents are relevant to, along with the document's text that constitutes relevancy.

It's pretty amazing.

Dave

[RowandK]

If anyone is interested in trying ALCoder (on your own PC), please contact me and I'll send you a control key with 1000 clicks.

Please note that the ALCoder Starter Kit is available via PayPal for $100 at www.alcoder.com . Your subsequent purchases of ALCoder will be discounted by this price.

Sorry we've gone to this model -it's meant to weed out less than serious inquiries. Please contact me for an interactive demonstration.

[jfuex]

Outside of forms processing my experience with various autocoding packages is that you would be better off just searching the OCR text with the notable exception of forms processing situations, which I don't see much of in lit support).

Autocoding is still a pipe dream, IMHO.

[Resident Devil's Advocate]

[RowandK]

When I saw this post, I got in touch w jfuex off-line. His opinion here and some of what he said off-line represent the biggest hurdles ALCoder faces – perception and perspective. (For the record, J hadn’t tried ALCoder himself, but someone else in his company has and still expresses interest.)

First, so much of autocoding has been over-promised or dreams have not met reality. ALCoder’s position is that our product is another tool for the arsenal. ALCoder doesn’t expect to replace manual coding, but ALCoder plus manual coding (hybrid coding) sure can reduce expenses and project time.

Second, there are some things ALCoder does really, really well. If you need to identify document dates or names mentioned, ALCoder does as fantastic job DEPENDING ON point #3.

Point #3, Garbage-In, Garbage-Out. If the documents are in poor shape, or the OCR engine sub-standard, then the text ALCoder interprets will be lacking. Which leads to

Point #4. For many document collections, there is no way to tell how well ALCoder will perform until you run a representative sample through. If anyone ever claims they’ve got an automated system that will accomplish XX% accuracy without trying a sample, they’re blowing smoke. Estimates that use a hybrid approach depend on volume – the bigger the collection the more likely the vendor can achieve some metric. Which leads to

Point #5, Click-based autocoding software is a risky proposition. I’ve just claimed that you can’t really tell how well ALCoder will perform without trying a sample. Wouldn’t it suck to burn clicks to find out that your collection is no good for autocoding? For that reason, A.L.Coder LLC has moved away from click based process.

Point #5a is that we have not totally abandoned clicks. If you’ve got 10,000 tiffed emails, hey, $1,000 in clicks will do the job. When I first came out with the no-click license, I called a customer and said hey, for another $3k, I’ll bump you to the no-click license. She replied, no-thanks, we’ve only got another 10k documents, so we’ll just take another $1,000 worth.

I asked J, how much is the improvement worth? If you’ve got 50,000+ documents that need to be coded, would $6,000 justify a doubling (or more) of you manual coding throughput? ALCoder can do that.