By: Larry Walker, May 2023; Context: Sperry Knowledge Systems Center.
In the early 1990s, the St. Paul (MN) School District wanted to pursue education research. To do this, they established The Saturn School of the Future for students in the fourth, fifth, and sixth grades. Saturn’s goal was to have a tailored curriculum for each student. Our company, PEAKSolutions, was selected to provide the software that would enable effective creation of individualized curriculum.
Saturn created a set of learning objects which identified all the resources they had available for use. Each learning object was then given a set of attributes which characterized what areas that specific learning object related to. Teachers would then ask their students what things interested them most, input their response into the software, and the software would ‘match’ the words in that query to the learning object resources.
For example, if a student’s interest was horses, the software would retrieve the set of resources that related to horses. The teacher would then select the resource that supported the subject area they were teaching. If they wanted to teach reading, they would select reading materials that involved horses. If they wanted to teach arithmetic, they would find math resources that related to horses, (e.g. if the horse weighed 220 pounds, how much feed should they receive in each feeding?).
This software teamed with committed teachers worked wonderfully. A couple of their success stories:
- One sixth grade girl wrote a story that was so good, her teacher was furiously looking to find a way to get it nationally published. When the girl was told this, she responded, “That would be wonderful. It took me two years to write that story.” How many 6th grade students commit two years of effort to anything!?
- William Norris, one of the founders of the computer industry, was visiting the school when his aide came running over. “Bill, you have to come and see this.” Bill followed the aide to where a fifth grade student was demonstrating how he had programmed the computer to run a small electric car connected to the computer. Using a joy stick, the boy could turn the car left, right, forward, or backward.
When asked to explain how this process worked, the boy responded, “I programmed it using a high level language called Logo. The Logo instructions get converted to a set of machine instructions which tell the computer hardware what to do. This sends a signal to the car attached by the wires and which directs the car to move in the direction I indicated using the joy stick.” A fifth grade student explaining how computer systems work to one of the industry founders!
In 1994, President H. W. Bush visited the Saturn School to present them their award as one of the top four innovative schools in the country.
Sometime after the success of the Saturn School, Mike Raimondi, PEAKS’ top sales person came up with a brilliant idea. One day all excited, he asked, “If our software can match interests to learning objects, why couldn’t it provide answers in response to questions – the very general case?”
Once Mike said it, the answer was obvious – of course, it could. So PEAKS developed a Just-In-Time Answer software solution. It worked well, and we sold it to multiple clients.
When the JIT Answer system was demonstrated to one of our strongest supporters, Bill Parkhurst, a very successful executive level person, responded, “In my experience, when looking into some new area of interest, the challenge is that I do not know what questions to ask, so what good is a JIT Answer system.
His insight was on the money, so what to do about it? Fortunately, we quickly realized that the right questions could be included – they were simply another set of knowledge items. Once we incorporated this thinking we realized that we greatly expanded the universe of questions that our tool could deal with: curriculum, outlines, books, etc. etc. The sky was the limit.
Unfortunately, PEAKS succumbed to the AI Winter in the early 90s. We started a new company, Knowledge Management, Inc. (KMI) which focused on commercializing our JIT Answer + Questions software solution. The original had been done on an Apple computer, and this just was not the primary hardware in use at that time. We also needed a PC-based version of this software.
KMI was a small start-up with no programmers. When KMI sold a PC-based version of this solution, something had to be done. Larry met with Nigel Dolby, a good friend and an excellent programmer. The two of them met for one day and by the end of that day, they had the overall design in place for Knowledge Engine which Nigel would implement using the Lisp language.
The design was simple: there would be knowledge objects. After much discussion it was determined that the ‘paragraph’ was a good definition of a knowledge object, since a paragraph was intended to deliver a unit of thought. The second part of the design was that attributes which dealt with, ‘why, what, where, when, who, how’ would anyone every be interested in that object would be appended to each object. Given the power of Lisp, this list of attributes had no limit. It also could be updated whenever experience determined that a new attribute was needed.
Working with a food industry executive with 40 years of experience, our first serious application of this Knowledge Management Environment (KME), was to input the FDA Food Code. This Food Code was a 450 word document which contained 70 years of experience on what it took to deliver ‘safe food’ from ‘the farm to the table’.
- Each paragraph of the Food Code was enter (manually) with a robust set of attributes.
- The user interface was natural language.
- The response to a query would return a list of all the paragraphs that related to one or more of the words in the query.
- Filter links would identify how many of the selected paragraphs related to each of the attributes, so the user could filter down to get a more useful subset of paragraphs.
- Food Code did the Government thing of referring to other Government documents which owned authority over certain regulations and/or entity definitions.
- These were embedded in our solution, so users did not have to scurry about finding other information sources.
- Query Example: “What are the duties of the waiter in a restaurant to deliver safe food?”
- Response: Every paragraph related to waiters which defined what actions they had to know and executer.
- This set of paragraphs could be printed as a dynamic white paper which could include the information cross-referenced from other sources as well.
- In many cases these specific white papers may have never been produced before.
- We the price at $100/disc which contained the entire Food Code Software solution.
- We lined up 4 System providers who delivered food to thousands of organizations involved with food.
- Larry visited a ‘food show’ in Florida where 30,000 visitors were expected.
- All four System providers had said they would demo the Food Code solution at this show.
- Larry visited all four of their demo stands, and they did indeed feature the software.
- After watching one of the VPs’ make a demo, Larry asked him, “What do you think of that software?”
- Response: “It is fantastic. We have never seen anything like it.”
- Follow-Up Question: “So have you not sold any?”
- Response: “We don’t know. It is powerful. It is not expensive. Our only thought is – maybe it is too ‘space age’, and people don’t believe it.”
Working with a Scottish-based document scanning company, KMI had the opportunity to demo the knowledge engine to the British Parliament. Parliament captured every word spoken in their sessions starting in 1803. This information resided in the Hansord volumes, several hundred of them, and millions of words. The Hansord volumes were done by professional librarians.
- Leather bound.
- Elegant print fonts
- Beautiful paper
- And – the librarians captured key words on both the top and bottom of each page.
Larry’s son, Tadd. embedded code within Scottish company’s scanning software which captured all the information from each physical page and embedded it in our knowledge engine. Of special importance were the key words from each page. Also, the Scottish company salesman alerted us to importance of the ‘look’ of each of the Hansord document pages. His point:these Hansord librarian ‘love’ their pages. So these physical images were saved by the Scottish scanning mechanisms and were the pages displayed in response to every query.
Hansord gave KMI 3 volumes from the year 1876. The scanning company had to unbound these beautiful books in order to scan them, and then they ran through their scanning machines with our code embedded in their software. The result was every page from 3 volumes were then live for the demonstration.
Sample Query: What did William Gladstone have to say about India? The retrieval would bring back every page from every document that included statements by Gladstone about India. These were the beautiful Hansord pages – not boring computer print output.
Sample Query: What is happening in Bosnia Herzegovina between the Muslims and the Christians? Retrieval would bring back every page that related.
Sample Query: What are we doing to do about the Irish drinking problem? Ditto to above.
KMI was invited to demo this solution at the Parliament – for a full day. We set up and Members of Parliament and their staff passed through all day long. The Parliament Project Manager for this told KMI, “From the reactions he was getting, this was THE best software demo ever given in Parliament. We were on Cloud Nine!
But – Parliament never funded the project.
Working with Potlatch Paper Company technical experts, KMI was asked to deliver operator training. Potlatch had committed to have their 1950s paper machine updated with Digital Equipment mini-computers. The problem was that their workforce was unionized, and Potlatch believed:
- About ⅓ of their operator workforce would be incapable of dealing with this complexity.
- About ⅓ were capable but could care less, so could not be relied on.
- About ⅓ were capable and would do a good job.
In general, Potlatch was dubious that teaching the needed skills in a classroom setting would work. In response, KMI suggested we deliver a Just-in-Time Learning solution.
KMI delivered the Just-in-Time-Learning solution within a month. The new hardware cutover took place shortly after, and the operators received no training ahead of time. This solution was simple in concept:
- When a paper machine was detected by the Digital computers, they would report a numerical number for that specific issue to the software. The software would alert the operators of the error and provide what steps the operator needed to perform.
- With NO operator training, this solutions worked seamlessly.
Success with a wide variety of knowledge solutions led us to wonder if we had discovered the DNA of knowledge? That is to say, “Is it enough to have knowledge objects (content) characterized by a set of attributes that identify all the reasons someone would be interested in seeing that object?”
KMI was very confident that we could deliver useful knowledge solutions for any specific domain of knowledge. We could also work across Domains if the client had multiple solutions in place. This made us realize the challenge of working across domains that were not closely related. The problem centers around the vocabulary associated with each Domain. Many Domains have vocabulary which is unique to that Domain – has NO meaning in unrelated Domains.
We also learned that even words that may appear in two Domains could have a different meaning in each one.
Questions welcome. KME has nearly 100 features for uses and developers – a very complete, rich tool. This article barely taps what can be done. These were the result of meeting with prospects and clients which made us apparent of various user needs.