| Back to Top | White Papers | Keith & Faye's Home Page | Photo Albums |
Introduction
The FAX Processing and Scoring System (FPSS) provides a low cost, efficient method of capturing multiple choice information for computer processing via standard FAX machines from anywhere in the world. The processing of the information is performed automatically with a flexible expert system. This replaces such current options as: manual keying of FAXed or mailed information, respondent keying, or scanning of original documents. With the exploding need to collect market data from end users and the extensive data collection needs of today's enterprises, this capability introduces the potential benefits of real-time Electronic Data Interchange (EDI) to companies that are unwilling to wait for the delays or unable to afford the costs of such instant electronic access.
Existing page scanners are expensive and require the data to be collected physically, an expensive and time-consuming process. FPSS introduces the ubiquitous office scanner, the FAX machine, into the data collection process. This makes it nearly universally accessible to respondents. Many FAX machines have retry facilities, reducing the need for multiple receiving lines for the data collection process. Using modular table-driven design and a sophisticated expert system, new applications can be implemented in real time, reducing costs and improving flexibility in collecting data.
General Description
Marks were chosen as the input method because OCR has not achieved sufficient universal reliability to use it for input. The design point for FPSS is to eliminate the need for any human intervention. Marked answer sheets are also more likely to be completed thoughtfully if they can be done at leisure away from a work station. Asking users to key information will often cause respondent fatigue or lack of response at all. Sending files from remote offices requires the sender to be relatively sophisticated. It also forces the called number to be equipped with banks of phone lines to assure reasonable response times, and relatively sophisticated receiving equipment to make it easy for the diverse users to use. The use of FAX turns the process into a technology-managed batch process, while leaving the user with their source document automatically. It was chosen after investigating the total business cycle of creating and sending input.
Although it would appear to be a simple matter to capture the presence or absence of a mark at a particular location on a received FAX, this is not the case. Some of the reasons are outlined in the section below titled "Characteristics of FAXes".
YTI has developed a comprehensive set of mathematical corrections and enhancements to raw FAXes in order to make it possible to reliably determine the presence or absence of a mark at a specific location on the original document.
The primary advantage of this technology is to provide a simple cost effective means of capturing a large volume of repetitive data very accurately from geographically-dispersed originators. Because FAX machines have become ubiquitous within the business community it provides a simple path to collect information quickly without the complications involved in establishing more formal electronic communications links, or the extra cost of couriers for timely input.
One of the many applications for this technology involves the processing of the received data to "Score" a test and interrelate the data itself to draw a series of conclusions. YTI has integrated its FAX processing technology with a sophisticated scoring module that permits complex relationships between answers to be drawn. For example; the full FPSS has been deployed to extract the percentile score within a sample population for 129 character traits of the person completing a questionnaire (using up to 1000 multiple choice questions). In this application, the accuracy has been demonstrated to be 99.8%, considerable higher than other inexpensive conversion methods. The system can be set to fail to read rather than misread an answer. Since the image received from the FAX is available to the operator, it is possible for the operator to supply just a missing answer. But in many cases of failure to read, the human operator is also not able to determine the intended answer due to the quality of the original marking on the received FAX.
Characteristics of FAXes
Because of the nature of FAX technology, the clarity of received FAXes is well below the original document, due to the relatively few pixels or dots that make up the received image. In addition "noise" on phone lines will add or remove dots at random and "light" or "dark" images are common. These things in themselves do not make it particularly difficult to determine the presence or absence of a mark as long as the system is looking for a difference between a mark and no mark. The real problem in locating a specific location within an image is a result of the variations between the mechanics of individual FAX machines. Many machines will shrink images to varying degrees while some will even enlarge them. Rarely will a FAX be fed straight and the amount of skew will vary significantly and unevenly between the top and the bottom of the document as the machine guides try to straighten it. FAXes are also frequently sent with their pages out of order or upside down.
These problems could be addressed by covering the page with "registration" marks, but this was rejected for two reasons. First, a page with a hundred or more answer locations would become too cluttered with multiple marks, and second; the failure to read a registration mark would be an additional source of error especially since the marks themselves would need to be quite small. In addition, more black on a page slows down transmission.
Technical Solutions
YTI has addressed the problems of FAX images using mathematical techniques similar to those used for satellite image enhancements. In this way YTI has made it possible to reduce the required registration marks to three. In order to reduce transmission time to the minimum, all that is actually sent from the typical answer sheet is the registration marks, a sheet number code, the answer numbers (to make it possible for a human to deal with failure to read) and the marks themselves. The location for the marks is indicated on the original document in a transparent bright yellow ink that does not get transmitted by the FAX machine.
The FPSS contains a number of variables that can be used to "tune" the decoding software to the YTI FPSS Page 3 specific needs of the application. Some applications will require almost perfect accuracy while others need to virtually eliminate all manual intervention. The tuning function can be used to meet these different requirements.
Implementation
The system is currently delivered on a multi-threaded OS/2 or Warp platform to permit the receipt of new FAXes at the same time that existing FAXes are being processed, summary reports are being produced, and other work is going on at the keyboard. The Windows version requires the capabilities of Windows 95 because the limitations in the Windows 3.1 environment do not make it practical to receive FAXes while others are being processed.
The design of a new entry sheet and scoring scheme is simply a matter of creating description files in any word processor. For this reason, it is possible to change the scoring algorithms at any time. It is also possible to modify the results with such things as normalization tables since the scoring module itself is designed for maximum flexibility. In fact, YTI has deployed the FPSS with Artificial Intelligence (AI) components imbedded in scoring schemes. Because YTI specializes in all forms of AI including Neural Networks, very sophisticated schemes could be developed or AI techniques could be used to identify very subtle differences between respondents.
It is possible to use a "standard" answer sheet and develop a new turnaround process for data collection, literally creating the scoring scheme in the time it takes to FAX out the questionnaire itself and be ready to process responses before the person could mark their answers and FAX them back. (assuming a straight-forward scoring scheme, of course)
FPSS Product Summary
The FPSS introduces a new alternative for data collection. It uses standard office components and relies on the existing industry infrastructures. Test scoring, market data collection, and order entry are all prime candidates to benefit from the reduced cycle times, lower costs, and improved turnaround potential.
FPSS Applications
Anonymous Survey
This type is characterized by a relatively short survey of under 50 questions with a simple reporting and scoring structure. Such a survey requires period over period reporting showing change by group and trait.
Tracked Survey
Similar to Anonymous survey with addition of respondent IDs in the database and changes by respondent period by period.
Complex Survey
Similar to tracked survey with addition of complex scoring including normalization tables etc.
Market Qualification
Characterized by a survey sent to potential customers in order to pre qualify them for a product or service. The product or service is likely to by somewhat complex and need both qualification of prospect as well as the need to gain an understanding of prospects requirements. The result would be a focused sales pitch.
This could also be used as a simpler prospecting vehicle. For example; the auto industry could use it to do mass mailings to previous customers to capture some who may be in the market today.
Initial Survey
This would be used by clients such as consultants who today do complex initial studies of their potential clients prior to engaging in a major contract. They frequently charge a significant fee to perform this work ($25K+).
The survey could run to 1000 + questions and have very complex scoring schemes. The output would be a document similar to all scores and a comprehensive report for the customer. The report would be a mix of boiler plate, selected boiler plate and calculated customized data.
Order Entry
The focus of an order entry, especially for a catalogue-based customer, would be on encouraging impulse buying.
The major work in this flavour would be in dealing with the problem of providing a score sheet design that could be simple to fill in while at the same time comprehensive and capable of conveying a great deal of information. Page, item, quantity, special features (colour etc.).
FPSS Survey Data Categories
+----------------+---+-----+-----+-----+-----+ | Data Type |Dif| Cvr | Auto| Rep | Man | +----------------+---+-----+-----+-----+-----+ | Psychological. Tests | 9 | Yes | No | Yes | Yes | +----------------+---+-----+-----+-----+-----+ | Anon. Survey | 2 | No | No | No | No | +----------------+---+-----+-----+-----+-----+ | Tracked Survey | 4 | No | Yes | No | No | +----------------+---+-----+-----+-----+-----+ | Complex Survey | 5 | Yes | Yes | Yes | No | +----------------+---+-----+-----+-----+-----+ | Market Qual. | 6 | Yes | Yes | No | No | +----------------+---+-----+-----+-----+-----+ | Initial Survey | 9 | Yes | Yes | Yes | No | +----------------+---+-----+-----+-----+-----+ | Order Entry | 5 | Yes | Yes | No | No | +----------------+---+-----+-----+-----+-----+
Explanation of table columns
To send us an eMail, just click here and include your CV details
======> Return to YTI table of contents/Reload home page
Copyright the authors and/or Keith Cowan