A (very) experimental look at recent articles in FT.com,
at first, exploring the visual impact of aligning the search results in novel ways,
but then, looking for syllables and rhymes and poetry.

The code for all this sits at github.com/railsagainstignorance/alignment.
It makes use of the enormous and enormously useful phoneme dataset, CMU Pronouncing Dictionary.
The haiku-detection is written up in a FT Labs blog post.

align on matching phrase

The search text used to identify the articles (via keyword search) is used to align the results.

source: any
title-only

align an article's sentences matching meter

You identify an article (with a UUID from the url, by hand, sorry) and specify a meter.
If any of the sentences within that article body contain text conforming to that meter, they are aligned accordingly.
Meter, you ask? See the internet for some examples.
You specify the meter as a sequence of 0s and 1s, where 1 is the emphasised beat.
You can anchor the meter to the ^start or end$ of the text, or leave it free to match anywhere.


uuid 
meter 

find author's recent articles' phrases matching the meter

You identify an author and specify a meter.
If any of the sentences within recent articles by that author contain text conforming to that meter, they are aligned accordingly.
Meter, you ask? See the internet for some examples.
You specify the meter as a sequence of 0s and 1s, where 1 is the emphasised beat.
You can anchor the meter to the ^start or end$ of the text, or leave it free to match anywhere.


author 
meter 
max 
(NB: there will be a bit of a delay)

find author's recent articles' phrases matching the haiku meter, 5-7-5

You identify an author.
If any of the sentences within recent articles by that author contain text conforming to the haiku meter, 5-7-5, they are displayed.
Meter, you ask? See the internet for some examples.
You specify the meter as a sequence of 0s and 1s, where 1 is the emphasised beat.
You can anchor the meter to the ^start or end$ of the text, or leave it free to match anywhere.


author 
max 
(NB: there will be a bit of a delay)

find ontology's recent articles' phrases matching the meter

You identify an ontology and value, and a meter.
If any of the sentences within the relevant recent articles contain text conforming to that meter, they are aligned accordingly.
Meter, you ask? See the internet for some examples.
You specify the meter as a sequence of 0s and 1s, where 1 is the emphasised beat.
You can anchor the meter to the ^start or end$ of the text, or leave it free to match anywhere.

ontology 
value 
meter 
max 
(NB: there will be a bit of a delay)

find ontology's recent articles' phrases matching the haiku meter, 5-7-5

You specify an ontology and value.
If any of the sentences within the relevant recent articles contain text conforming to the haiku meter, 5-7-5, they are displayed.
Meter, you ask? See the internet for some examples.
You specify the meter as a sequence of 0s and 1s, where 1 is the emphasised beat.
You can anchor the meter to the ^start or end$ of the text, or leave it free to match anywhere.

ontology 
value 
max 
(NB: there will be a bit of a delay)

find page's current articles' phrases matching the haiku meter, 5-7-5

You specify a sitepage.
If any of the sentences within the relevant articles contain text conforming to the haiku meter, 5-7-5, they are displayed.
Meter, you ask? See the internet for some examples.
You specify the meter as a sequence of 0s and 1s, where 1 is the emphasised beat.
You can anchor the meter to the ^start or end$ of the text, or leave it free to match anywhere.

page 
(NB: there will be a bit of a delay)

for articles before the date, find phrases matching the haiku meter, 5-7-5

You identify an date.
If any of the sentences within recent articles before that date contain text conforming to the haiku meter, 5-7-5, they are displayed.
Meter, you ask? See the internet for some examples.
You specify the meter as a sequence of 0s and 1s, where 1 is the emphasised beat.
You can anchor the meter to the ^start or end$ of the text, or leave it free to match anywhere.


before 
max 
(NB: there will be a bit of a delay)

explore phrases matching meter

You specify a phrase and a meter.
Details extracted from the phrase and the meter are displayed, mostly for debugging purposes,
but it is interesting to see how the words are fragmented into phonemes and syllables.
If any parts of the phrase contain text conforming to that meter, they are aligned accordingly.
Meter, you ask? See the internet for some examples.
You specify the meter as a sequence of 0s and 1s, where 1 is the emphasised beat.
You can anchor the meter to the ^start or end$ of the text, or leave it free to match anywhere.

Perhaps you want to see why a particular phrase does not get matched by the meter? Most likely first, consider

  1. Its a bug - in which case, feel free to let me know
  2. The phonemes are based on American pronounciation, and they do speak funny. So, sorry, but, shrug.
  3. The phonemes' emphasis points are perhaps too rigid to allow sufficient slack for the word to be spoken in a poetic context. Again, sorry, shrug.

phrase 
meter 

Caveats

  • It has very little error handling
  • It can not work for an excitingly wide variety of reasons.
  • Perhaps the main reason for flakiness is that this is MyFirstGoProgram.
... so choose wisely.


extras


Simply interesting

assorted imagery