|
AN OVERVIEW
If you need to collect data from some population
such as your membership, there
are many possible approaches, including use of paper forms and a specially designed
Internet site. We are primarily concerned with obtaining data from forms here,
although we are involved with the other approaches.
Manual entry of information from forms is a possibility,
but it is normally so time-consuming that it is impractical. For example,
for 2500 forms and a staff person who can process 1 form every 2 minutes, a couple
weeks of tedious work by that person would be required.
For more than a couple hundred forms, some sort of scanning technology is normally
cost-effective. The process is this. Properly formatted forms are prepared
and distributed to the population who then mark the forms and return them to you.
You send the forms to us and we scan them with an appropriate high-speed scanner.
Further processing of the scanned results might be necessary to extract the desired
information and put it in the desired format. You would receive either the
raw data of marked fields or professional reports showing the statistical characteristics
of the data, such as counts, mean responses to questions, standard deviations of
responses, charts of response distributions, and so forth. Our turnaround
time is typically a few days.
There would be costs for designing and printing the forms, scanning the forms, possibly
performing additional processing on the scanned data, performing statistical analysis,
and generating reports of the results. When compared to the cost for the staff
person to enter the information manually, run the statistics, and generate
the reports, the cost for the scanning-based process will normally be less, accuracy
will be higher, and total turnaround time will be much shorter.
A LITTLE HISTORY --
In 1990, there was a need in the Metropolitan DC area for OMR (Optical Mark Recognition)
scanning and related services. Organizations in the area needed to extract
data from batches of thousands of surveys, questionnaires, tests, continuing education
credit forms, registration forms, and ballots. They also needed statistical
analysis of the data and professional reports for distribution to certifying organizations,
speakers, committees, members, and so forth.
Two PhDs involved in creating, scoring, and analyzing medical and allied health
certification tests recognized the need and founded Infoscan. Over the past
19 years, it has grown and served the scanning needs of the DC/MD/VA area.
With its reputation for quick, quality work, Infoscan has acquired clients throughout
the United States. We work with many clients year after year to analyze and
report on their annual meetings, courses, score their tests, calculate and report
their continuing education credits, tally their ballots, and much more.
TYPES OF SCANNING --
Infoscan initially only worked with "bubble" forms and special OMR scanning equipment
that could sense whether fixed bubbles were marked or not. These forms and
the scanning equipment were produced by NCS (National Computer Systems, later bought
by Pearson, and now owned by Scantron). Over time, Infoscan has cautiously experimented with improvements in
technology to other types of scanning that use image scanners and software to process
the images and recover not only marked bubbles but also typewritten text (OCR, Optical
Character Recognition) and even handwritten text (ICR, Intelligent Character Recognition).
Each type of scanning has advantages and disadvanges; and, although there is overlap
in true OMR and image-based technologies, it is often better to use one approach rather than
the other to satisfy requirements for accuracy, reliability, flexibility, time constraints,
or cost.
WEB FORMS -- In addition
to scanning of paper forms, Infoscan performs many projects using Web forms or a
combination of Web and paper forms. We specialize in selecting and implementing
the approach that is most advantageous to your situation. Finally, Infoscan
is a division of a larger organization involved with Windows software development,
network design and installation, and implementation of turn-key solutions with the
necessary hardware, software, and support. Infoscan has the resources and
experience to make your data collection project a success.
When either mark-sense (optical mark)
or image scanning
is to be used, advance preparation of the entire process is paramount for success.
Mark-sense scanning has historically been used where critical information is collected,
such as SAT tests and ballots. The special scanner employes a series of LED sensors
lined up across the scanned page. As each page passes under these sensors,
they determine whether specified areas are marked or not. The output is a
series of alphanumeric characters indicating how each area was marked.
This scanning is relatively fast, several thousand sheets per hour. Little
if any further processing of the scanned data is required other than statistical
analysis and reporting.
The other type of scanning is image scanning which produces an image file for each
side of a scanned page. The file contains everything on the page encoded as
a binary image. Image scanning is relatively slower than optical mark scanning
and much larger output files are produced. An image file could be sent directly
to a printer to reproduce the page. In some cases, as in document management,
these images are simply collected in archives and indexed for easy access.
Image scanning is the basis of newer, more sophisticated technologies that extract
information from the images, such as determining whether bubbles have been marked,
converting selected text parts of the image to Word text, interpreting hand-written
block characters, or extracting individual images such as signatures. This
technology has always been problematic because of many factors, such as inconsistent
form printing, fonts that are too small or light, hand markings that are not within
the specified bounds (bubbles or boxes), or characters that are not carefully formed.
Image processing accuracy rates of 95% might be achieved with careful form design,
printing, marking, and scanning. But that is not really very good. A
sheet with 1,000 characters interpreted with a 5% error rate would mean that 50
characters are misinterpreted and need to be fixed. There are ways to improve
on this by limiting the possible range of characters with dictionaries; however,
95% to 98% accuracy is about the best that is ever done. Accuracies in the
85% range are more typical. Compare this with 99.999% accuracy for true OMR.
That means that almost all character recognition approaches (software OMR, Optical
Character Recognition, and Intelligent Character Recognition) have to be reviewed
and corrected manually as part of the process. There are tools that facilitate
that, but it is still a time-consuming process.
WHICH APPROACH SHOULD I USE? LET'S MAKE
IT SIMPLE:
TRUE OMR: If you need
the highest accuracy and reliability or need a large number of forms, use true OMR
forms and scanning. One drawback is that a lot of form space is required to
enter alphanumeric data. The cost of these forms is a little higher, but the
cost of scanning is lower and requires no further processing to extract the data.
True OMR is the only practical and safe approach for large meeting session evaluations,
tests, and recording Continuing Education contact hours. Some will say this
is an "outdated" technology, but it remains the only really reliable
technology.
SOFTWARE OMR/ICR/OCR:
If you have smaller evaluations (such as 50 evaluations each for occasional courses)
or forms from which you must capture hand-written block characters, you can use
software OMR/ICR forms with an image scanner. If the forms are near-perfect,
this method can be viewed as the preliminary to manual entry of the data.
That is, your staff will still have to spend time confirming the accuracy of the
data. The OCR version of this is a good technology, quite well developed,
and really the only way to capture text from printed documents other than manually
entering it all. It still requires editing after the scanning.
WEB: If you have a distributed
population that needs to provide information, this can be good. It is not
good, though, if people have to remember a lot of information from a meeting and
enter it later. It is singularly poor for evaluation of many sessions that
have previously been attended. Similarly, it is not good for recording Continuing
Education credits. It is good for ballots where logins can be compared against
a membership database.
KIOSKS AT MEETING: This
seems like a great idea. The technology for electronic entry is almost ready
for prime time, but the implementation with kiosks is usually much less than expected.
Relatively small amounts of data can be collected without risking long lines of
members missing sessions and meetings waiting for their turn to enter evaluations.
Simple paper evaluations are far better and don't suffer from computer and network problems.
The electronic approach will probably have to wait until everyone has a larger-screen
handheld device with Web access. Cellphone devices are getting there, but
you cannot yet count on most of your membership's having the more sophisticated
units such as the iPhone or Blackberry.
MANUAL DATA ENTRY: Sometimes,
just going through each form manually and entering the data is the most efficient
or only solution. If you have fewer than 100 forms, you can spend more time
and money designing scan forms, getting them printed, and scanning them than you
would spend simply doing the job manually. If forms are not scannable, we
are unfortunately left with only this solution, which we occasionally encounter
when an organization brings us forms that have not been properly designed, printed,
or tested before using them in a major meeting.
|