3 ms·
I am researching the history of my field, cheminformatics. It grew out of documentation management, more specifically, chemical documentation. While that topic
by dalke 11y ago
I am researching the history of my field, cheminformatics. It grew out of documentation management, more specifically, chemical documentation. While that topic goes back to the 1800s, I limit my research to mechanical methods for chemical documentation search and management.
This started with punched cards in the 1940s. James Perry, then at MIT, was a nexus in the field. It's difficult to find people of the era that he wasn't connected to.
One of these is Calvin Mooers, who is a special focus of mine. Mooers developed some of the fundamental concepts that formed the basis for the computerization of chemical information. (To be more precise, he came up with ideas of how to solve problems. These ideas weren't solutions, but were in the ballpark, and ended up influencing the people who actually solved and implemented the problems.) Mooers was a Master's student at MIT, where he met Perry, and was influenced by the sorts of problems that Perry was interested in.
Mooers' Master's thesis was on a method for superimposed coding which he termed Zatocoding. A superimposed code is one where the keys are assigned to 1 or more bits, and where the bits for different keys may overlap. A Bloom filter is a type of superimpose code. In Zatocoding, there is one pass over the keys to get population statistics, then a bit of information theory to determine the number of bits to use for each key and its bit assignments. A common key ends up with 1 bit, while a rare key might have 7 bits.
Mooers is one of the contributors to "Punched Cards". I likely read the book while trying to track down his publications. I may have read it because Perry was the editor. Or I might have read it because I was trying to understand more about edge-notched punched cards.
As for 'seminal', I based that on the number of sources I've read which cite it. For something more outwardly visible, there are 100 citations at https://scholar.google.com/scholar?cites=464718015108686727&as_sdt=2005&sciodt=0,5 https://scholar.google.com/scholar?cites=464718015108686727&... . And of course the book went through two editions.
Interesting. One of those references is to "Automated concordances and word indexes: The fifties". That article says 'Father Busa inaugurated the history of the automatic production of concordances and word indexes' and 'In 1951, using the punched-card equipment available to him, Father Busa produced and published the first automated word index and concordance to four hymns written by Aquinas for the Feast of Corpus Christi, ...'
It's the same Busa who had the chapter in "Punched Cards" about working with the Dead Sea Scrolls; the one that got me to track down the Popular Science sciation. While that chapter doesn't establish the seminal nature, it does give an idea that the book was presenting cutting edge work with punched cards.
- pvg 11y agoThanks for the detailed reply! I mentioned the PopSci article to an acquaintance of mine who is a New Testament scholar and he brought up Busa, although could not recall the name offhand. I took a closer look at your Punched Cards reference and it was quickly clear Busa was the person he had in mind. I'm somewhat surprised this sort of interesting early work on coding and natural language processing using mechanical and electronic devices with very limited online storage (not to mention speed) is so seemingly obscure.
- dalke 11y agoGlad to be of help. I find this era fascinating. I think it's obscure because it solves a problem we don't really have. We take computers for granted, and mostly group solutions into "with a computer" and "manually" (with pencil and paper), because that's what we are used to. We aren't used to thinking about mechanical solutions, and clever indexing schemes for how to organize piles of paper. Before I started on this research, I had never heard of edge-notched cards, much less appreciated coding schemes that minimize the number of manual needle insertions needed to search a deck. I've been trying to understand why superimposed coding systems did not reach the modern world. The principles behind it really helped me understand how some modern techniques work, but those modern techniques are ad hoc. Already by 1974 superimposed coding was considered an 'old technique.' I quote from "The Large Data Base File Structure Dilemma", http://dx.doi.org/10.1021/ci60001a005 http://dx.doi.org/10.1021/ci60001a005 : > The second solution to the dilemma, suggested in Figure 7, is sequential search. This may appear to be a throwback, particularly within the framework of on-line systems, but it should be noted that the speed of the CPU combined with the extremely high data transfer rate from secondary storage of third and fourth generation computers has reached a point where one might profitably reexamine the potential of sequential processing. I interpret this and similar articles to mean that the older techniques weren't a good fit for the then-current computer architectures. After two generations of disuse, institutional memory fades. And reading the old literature doesn't help if you read it through the eyes of people with 1GHz processors - many of the driving problems are trivially brute-forceable. I read "Punched Cards" in part to get a sense of what it was like to think in those old terms. Now I want to acquire edge-notched cards and physically experience it. But no one sells those cards any more, and haven't for decades. There's a lovely article about this 'dead medium' at http://kk.org/thetechnium/one-dead-media/ http://kk.org/thetechnium/one-dead-media/ . Luckily, we also have new solutions. This modern crafting era has machines that can mark and cut paper product with ease. I think it would be possible to make each card "from scratch" on the computer, rather than the old school style of using large machinery to cut and print a template, followed by manually content addition.