FAQs

1. We have over 100 document publishers. What is the setup process and fee?

There is never a set up fee with GLYNT. Use our Getting Started Guide to get started with your first publishers. (hot link)

2. We see lots of variation within a utility’s document group. The fields are always moving around the document.

No worries! GLYNT is designed for semi-structured data, including variations in the location of data on documents. GLYNT will find the field no matter which page it moved to.

3. What happens when there is a long list of “possible” fields, but some of them are not on every document?

To handle the variation in fields printed on a document, start with the documents at hand. As long as 5 – 7 documents have a field, GLYNT can be trained to find it. When new fields come in on fresh documents, GLYNT is easily retrained to handle.

4. We want the automated data extraction of machine learning, but get just 5 documents per month from one of our publishers.

GLYNT is ready for this challenge. Upload the small sets each month, and manually process with GLYNT’s point and click web console. Once there are 20 documents in our system, proceed with setup for the utility. Voila! Automated extraction on small data sets.

5. Errors are a huge part of our cost. When they are hard to find, we spend hours fixing problems.

From our work in document processing we know this problem well, so we designed GLYNT for easy error trapping. When GLYNT is not confident about a data item, it returns nothing. This creates a clean trail for the automated and human reviews, the Verification Layer. Every data set extracted by our machine learning models goes through the Verification Layer.

6. Our invoices come in around the first of the month. This surge in volume chokes our current systems and slows everything to a crawl.

Enjoy the power of automation! GLYNT is a fully elastic machine learning system and can scale as needed. Your throughput can be as fast or slow as needed to keep your work flow steady.

7. How does GLYNT handle software bugs and exceptions?

Our experience is that almost all questions about potentially inaccurate data are very specific to data items on a document for a utility. And often these are resolved quickly. So the GLYNT customer success team will respond directly to most questions in a rapid manner, and escalate issues when they are truly bugs.

8. What scan or image quality is needed?

GLYNT integrates multiple OCR engines to reduce OCR errors. GLYNT gets good results at 300 DPI and higher. GLYNT often gets good results at 200 DPI depending on other features of the document. If you are concerned about document readability, send a zip file with sample documents of any type. We’ll let you know if the document quality is sufficient for processing.

9. What if my documents have handwriting?

GLYNT does not read handwriting at this time. Additional handwritten notes on a document cause an OCR error only if the handwriting is near a desired field. Often those emphatic circles around Amount Due are not a problem, there is sufficient white space between the handwriting and the text.

10. I have a document set, but not every field shows up on every document. What documents should we send for training and testing?

Create a document set that includes 7 examples of each field for training. It might take 30 documents to get the training set. That’s ok. Add more documents to the file to get a validation report on the accuracy of the data extraction.

11. We’ve tried other data extraction vendors and they bring back lots of OCR errors. Can we do a quick “readability test” with GLYNT to make sure your system can read our documents, e.g. do a “read validation” before doing an extraction exercise?

Sure! Send us one or two sample documents that you have concerns about in a zip file. We’ll quickly test them in GLYNT to verify that they can be read. This is an efficient use of your time and ours. Ok to bundle a bunch of publishers or sources you have questions about into a single document group for a quick “readability test.”

12. What is a Data Item?

GLYNT extracts a group of text in a single step, this is known as a Data Item. The text group may include several fields. In this example healthcare insurance card, every item highlighted is a Data Item. The extraction of “Karen Carter” is two Data Items.
Similarly a three-line street address is treated as three Data Items:

AddressLine1: 2583 Laguardia Ave.
AddressLine2: Apt. #14
AddressLine3: Smokey, CA 94229

13. What items does GLYNT return?

For each Data Item, the following items are returned:
• Filename: The document filename
• Field Label: As printed on the document
• Data Value: Exactly as printed on the document
• Preferred Field Name: Selected by customer (optional)
• Preferred Format: Determined by customer (optional)

14. When data is returned in a CSV format, what is the layout?

One CSV file is returned per extraction. Each row is a document in the batch, each column is a field label.

15. What is the different between “Missing” and “Not Present”?

Missing and Not Present originate from two sources and are ways that data is not returned by GLYNT. Not Present notes that the field requested is not present on the document. Missing notes that the field requested is present, but the machine learning system is not sufficiently confident that it will extract the correct data item, so a data item is not returned. The user can add this data through the GLYNT web console.