As promised, here’s minion:
I’m so satisfied with the title of this post, I don’t think a post is necessary at all and we should all spend a few minutes reading the title over and over to properly admire its beauty and elegance. But we still need a post about DGT, so I’ll contain my excitement. I’ll just pat myself on the back and say it loudly an alarming number of times.
DGT is an acronym for Data Gathering Tool. It’s the tool we made for Dr.Hisham, the client, the data analyst, money bringer, and godfather as well as father and stepfather of the project, whom we just love with all our hearts and pockets. Our relationship has evolved from just an employer-employee one into a full fledged familial one, where we ask him for more spending money and he tells us not until you do your homework. He is an MBCHb at Shatby Maternity Hospital and is seeking a doctorate in obstetrics and gynaecology, which roughly means he fixes baby and mommy problems. He masterfully hides his true age, but he has grey hair, so, medically speaking, he is old. We call him The Doctor, which makes me think of a Heisenberg-type drug dealer whom no one ever sees and lives to tell about it, but stories fly here and there about his army of wolves with human heads that he has made of the corpses of his enemies and his biomechanically augmented henchmen who surprisingly die just as easily as human henchmen making you question the whole point of augmentation and this paragraph, but you shouldn’t be so cynical.
This is not just a shameless blatant attempt at flattery, it’s actually necessary. The Doctor is responsible for collecting the data, inputting it, as well as being the main user of the final app and the DGT. Therefore, his input, regarding everything, from the way the tools are set, to the many functions and features included and the way data is represented, that input is of some vitality here. Also of some vitality, is how he is totally not a drug dealer although it would be pretty awesome.
Here’s where this post’s title starts to make sense. Us, the two lead developers, The Doctor and Hamdy, the wealthy but cheap project manager who doesn’t pay us anything and doesn’t get paid himself, are constantly coming across new information about the data we are handling. That sometimes makes us implement significant changes to the DGT. A recent example where we had to redo a big chunk of our data representation, is the problem of relations between anomalies in a single syndrome. Previously, a syndrome was to reach 100% certainty if all the anomalies in the syndrome were found. For example, if syndrome S had anomalies [a1, a2, a3], and the sonographer found and entered into the program anomalies [a1, a2, a3], the top result with a score of 100% should be syndrome S.
After some amount of data gathering, which is a difficult and time-consuming process no doubt, The Doctor informed us of a new piece of information. A syndrome could have two anomalies where only one of them is possible to be present(for example a missing eye and a swollen eye can’t exist for the same eye) and therefore only one is needed to represent the two in a given syndrome. To illustrate with the same example above, syndrome S might need to be represented as follows: [a1, (a2 or a3)]. Which means if the sonographer found a1 and a2 only, the program should still output S as the top choice with 100% score. The same for inputs a1 and a3 only, because only one of a2 and a3 is needed.
This required a change in the way the data is represented in our database, modifications to our classifier, as well as a means of specifying these relationships in the DGT. Already, a big change, but there’s more. First, extend the previous example to be a relation between any number of anomalies not just two, and, second, extend the whole DGT to accept relationships between those relationships. For example, syndrome S2 could be on the form: [((a1 or a2 or a3) or (a2 or a4 or a5) and (a6 or a7))].
These changes also affect how we handle data in our classifier. We will get to that in time.
Our prolonged discussions about these modifications resulted in two pieces of art that we are in the process of making into T-Shirts, so if you want one, just email us you crazy man.
See you soon,