Pi memorisation tool using the Major System and the CMU phonetic dictionary.
Using nodejs, express, jade/pug, JQuery, Ajax, etc.
- The /pi page allows user to grab digits of pi in blocks of 10 from the first 10000 digits
- The /scandig page allows the user to scan the CMU phonetic dictionary for words that appear as major digit sequences within a block of digits. Data is grabbed with Ajax and stuffed in a JQuery DataTables table.
- There are other hidden routes for testing purposes!
Memorising long sequences of digits is hard! Unless, of course, the digits follow some pattern so you can easily generate digits on demand from some starting condition - I'm not talking about those numbers! I'm talking about phone numbers or something like pi, e, or root-2 where the digit sequence is seemingly random and follows no recognisable pattern. For these, the human brain requires something less abstract to hook into in order to retain this dry, obscure data. It is much, much easier, as a human, to remember stories, songs, scenarios, feelings - something sensory and tangible. That's the type of information that our brains are wired for! A system is needed to encode the number sequence into something more memorable. Then to recall the digits the decoding just needs to be error free and unambiguous. The Mnemonic Major System is a pretty standard way of encoding numeric digits as consonant sounds. Then words can be chosen to fit the sequence of consonants, sentences built up from the words, and a narrative can be built to make very long sequences memorable.
There are way more sounds (aka phonemes - maybe 40 in English) than there are digits (10) so only "hard" consonants sounds are used and these can be freely interspersed with vowels and soft consonant sounds ("w", "y", etc.) so the mapping is loose in one direction and strict in the other; by that I mean that you may be able to make umpteen different words from a sequence of digits but going from the words back to the digits will always get the correct digits back (as long as your encoding is accurate and complete).
See Wikipedia for detail but in brief, the digit mappings are as follows...
- 1 = t, d, th
- 2 = n, ng
- 3 = m
- 4 = r
- 5 = l
- 6 = j, ch, sh
- 7 = k, g, (hard-c)
- 8 = f, v
- 9 = b, p
- 0 = s, z
In order to become proficient in the Major system, I decided to learn some digits of pi. "A hundred sounds doable!" I thought. This was the start of January 2016, and I set myself a deadline of learning this by World Pi Day on the 14th of March. I reached my goal of 100 digits in about a week so I carried on learning digits, ten at a time, and by pi day I had about 600 digits! I soon realised that what could be viewed as an arduous task, was an enjoyable and rewarding creative writing exercise! Somewhat closer to the restrictive patterns of song writing or creating haiku than free prose storytelling.
I have chosen to memorise in blocks of 10 digits. This restriction helps keep me on track: the words that I choose cannot cross a 10 digit boundary, and the little group of 3 to 5 words makes a little sentence or phrase. The phrase is usually silly or nonsense but I make sure that it is always memorable in the general ongoing narrative within its neighbours.
As a computer nerd, I quickly became interested in what I could possibly create to help me choose interesting words from the rich language we have. I had previously worked with open source speech synthesis and speech recognition software as part of my Cheesoid robot project so I was familiar with the state-of-the-art. In trying to make text into recognisable speech and vice-versa, some amazingly useful resources have emerged, the most important of which is the Carnegie Mellon University (CMU) Pronouncing Dictionary.
The CMU dictionary is in the ARPABET transcription encoding which is a plain ASCII representation of the phoneme set for English. I created a simple mapping of ARPABET to Major so that each word in the CMU dictionary can be translated into its Major Mnemonic representation. When searching for words that fit within a digit sequence we are doing simple string matching of each word's Major phoneme string within the digit string.
For a decent modern web app we want minimal delays and to make full use of proper browser capabilities. The dictionary is big so we don't want to search the whole thing before coming back with results. It makes sense to use a stateless API so that a search of any length can be started from any position. The browser can keep state, i.e. track the progress, and do the appropriately smart things like batching queries and issuing AJAX requests.
When getting results with AJAX we might want to chunk them or just grab single results. We might want to send back a message saying "I'm still searching" if it is taking a long time to find the next item. We want to send back a message saying "finished" when all items have been processed. In general we want live progress feedback for a search. We want to be able to cancel a long search, and this all has to be multi-user and multi-session aware!
No problem! We have NodeJS to do the back-end stuff and JQuery to do the front-end.
- Use Bower for JS packages on the browser side
- DONE - Dynamic use of JQuery DataTables for results paging on the browser
- CSS it up! Bootstrap styling perhaps
- pi route to provide form when no index arg
- scanfull also to default to a form
- General CMU browsing especially of single digit "filler" words ** DONE for special single digit case in /scanfull/(digits) route ** need little GUI table for it (open filler digit tab in new window) ** CMU word search lookup
- word saving when clicking table row - goes to a div - checking routine - DataTables search box interaction ** clickable rows DONE ** rows added to box based on position and clash indicated