Here's my latest cut at the numberbatch file parsing:
nb =: <'C:\numberbatch-en.txt'
load 'files'
$nbs =. fread nb
NB. nbs is the raw numberbatch character array
NB. parse the raw text & get the words
$ words=. <@({.~ i.&' ');._2 nbs
417195
NB. Get the vectors
$ vecs =. 0 1 }. _&".;._2 nbs
417195 300
ts 'vecs =. 0 1 }. _&".;._2 nbs'
61.1788 2.78677e9
<<<<>>>>
Wow! the vector extraction takes over a minute on my machine.
Is there a more efficient way to extract the vectors?
The good news is that I only have to do this once. After that,
I can use the vecs noun for looking up a words' vector:
NB. Raul's 'get' verb:
get=: 13 :'vecs{~words i. ;:^:(0=L.)y'
get 'bake'
0.0666 0.0191 0.1889 0.1288 _0.0716 _0.1245 0.1423 _0.1784 0.0352 _0.0095
0.0624 _0.0423 _0.0237 _0.1382 _0.1238 0.085 _0.0188 _0.1 0.0225 _0.0264
_0.1443 _0.1018 0.0856 _0.0824 0.082 _0.1749 _0.0436 0.1718 0.0024 0.0018
_0.0029 0.1591 0.0497 0.0199 _0.031...
ConceptNet Numberbatch 17.06
<https://conceptnet.s3.amazonaws.com/downloads/2017/numberbatch/numberbatch-17.06.txt.gz>
is the current recommended download link for the full numberbatch file.
Skip
Skip Cave
Cave Consulting LLC
On Wed, Feb 21, 2018 at 1:28 PM, Raul Miller <[email protected]> wrote:
> I suppose the shape of the variable 'words' depends on how you built
> the value for 'words'?
>
> If you used
> words=: <@({.~ i.&' ');._2 text
>
> the noun will be rank 1, and there's no need to ravel it.
>
> If you built it at rank 2, rather than rank 1... well... I'm not sure
> why you would want to do that - it seems like if you used a process
> that gives you a single column rank 2 noun you probably should ravel
> it before assigning it to the variable?
>
> Thanks,
>
> --
> Raul
>
>
> On Wed, Feb 21, 2018 at 1:46 PM, Don Guinn <[email protected]> wrote:
> > So words should be a list instead of a one column table. So we would have
> > words&i.
> > instead of
> > (,words)&i.
> >
> > Correct? Doesn't the raveling prevent sharing of the contents of words in
> > the new verb?
> >
> > And perhaps get should be
> > get=:13 : 'words&i.boxopen y'
> > instead of
> > get=:13 : 'words i.boxopen y'
> >
> > Does the building of the hash table require that i. be bound to the left
> > argument with & or will it still build the hash table only once in the
> > tacit definition where i. is in the dyadic form where the & is not there?
> >
> > It would probably be safer to put the & in.
> >
> > On Wed, Feb 21, 2018 at 11:00 AM, Henry Rich <[email protected]>
> wrote:
> >
> >> I don't think this prescription is accurate. When m&i. is executed to
> >> create a fast search verb, the value of m is put into the new verb. If
> m
> >> is a name, the value of the name is NOT copied, but instead referred to.
> >> If the name m is subsequently reassigned, the old value is retained,
> >> referred to by the m&i. verb, and the new value is assigned to the name
> m.
> >>
> >> So, deleting words will not actually free any memory. On the other
> hand,
> >> executing words&i. didn't consume any memory either.
> >>
> >> (this is all from memory & I haven't checked it with 7!:2)
> >>
> >> Henry Rich
> >>
> >>
> >> On 2/21/2018 12:08 PM, Don Guinn wrote:
> >>
> >>> Defining a verb get to retrieve the index of the desired word as tacit
> >>> does
> >>> make get pretty much unreadable; however, there is a possible
> performance
> >>> gain as the hash table for i. gets built only once when get is
> defined. If
> >>> you will be running get many times this could result in a significant
> >>> performance gain.
> >>>
> >>> Of course, once read in words must not be modified without rebuilding
> get.
> >>> But if it turns out that you don't need words for anything else than in
> >>> get
> >>> then you could erase words after get is defined so storage used by a
> big
> >>> verb is offset by not having words around any more.
> >>>
> >>> On Wed, Feb 21, 2018 at 9:31 AM, R.E. Boss <[email protected]>
> wrote:
> >>>
> >>> vec {~ (<'adults') i.~ words
> >>>> is perhaps what you are looking for
> >>>>
> >>>>
> >>>> R.E. Boss
> >>>>
> >>>>
> >>>> -----Original Message-----
> >>>>> From: Programming [mailto:[email protected]]
> >>>>> On Behalf Of Skip Cave
> >>>>> Sent: woensdag 21 februari 2018 17:09
> >>>>> To: [email protected]
> >>>>> Subject: Re: [Jprogramming] File Cleanup
> >>>>>
> >>>>> Thanks to Raul and Mike for the suggestions.
> >>>>>
> >>>>> I read in the data:
> >>>>>
> >>>>>
> >>>>> nb =: <'C:\numberbatch-en.txt'
> >>>>>
> >>>>> nbs =. fread nb
> >>>>>
> >>>>>
> >>>>> Then I tried to clean it up:
> >>>>>
> >>>>>
> >>>>> Mike's method ran out of memory:
> >>>>>
> >>>>> nbs4 =. ( i.&' ' ({.;0 ". }.)] ) every nbs
> >>>>>
> >>>>> |out of memory
> >>>>>
> >>>>> When I tried to run it on a smaller set:
> >>>>>
> >>>>> nbs4=: (i.&' '({.;0".}.)])every 100000{. nbs
> >>>>>
> >>>>> nbs4
> >>>>>
> >>>>> ...
> >>>>>
> >>>>> │0││
> >>>>>
> >>>>> ├─┼┤
> >>>>>
> >>>>> │0││
> >>>>>
> >>>>> ├─┼┤
> >>>>>
> >>>>> │3││
> >>>>>
> >>>>> ├─┼┤
> >>>>>
> >>>>> │5││
> >>>>>
> >>>>> ├─┼┤
> >>>>>
> >>>>> │ ││
> >>>>>
> >>>>> ├─┼┤
> >>>>>
> >>>>> │0││
> >>>>>
> >>>>> ├─┼┤
> >>>>>
> >>>>> │.││
> >>>>>
> >>>>> ├─┼┤
> >>>>>
> >>>>> │0││
> >>>>>
> >>>>> ├─┼┤
> >>>>>
> >>>>> │7││
> >>>>>
> >>>>> ├─┼┤
> >>>>>
> >>>>> │8││
> >>>>>
> >>>>> ├─┼┤
> >>>>>
> >>>>> │2││
> >>>>>
> >>>>> ├─┼┤
> >>>>>
> >>>>> So that wasn't working for me.
> >>>>>
> >>>>> I tried Raul's suggestion:
> >>>>>
> >>>>> words=. <@({.~ i.&' ');._2 nbs
> >>>>>
> >>>>> vec =. 0 1 }. _&".;._2 nbs
> >>>>>
> >>>>>
> >>>>> $words
> >>>>>
> >>>>> 417195
> >>>>>
> >>>>>
> >>>>> Looking good....
> >>>>>
> >>>>>
> >>>>> ,.20{. 6000}. words
> >>>>>
> >>>>> ┌────────────┐
> >>>>>
> >>>>> │adultly │
> >>>>>
> >>>>> ├────────────┤
> >>>>>
> >>>>> │adultness │
> >>>>>
> >>>>> ├────────────┤
> >>>>>
> >>>>> │adultoid │
> >>>>>
> >>>>> ├────────────┤
> >>>>>
> >>>>> │adultress │
> >>>>>
> >>>>> ├────────────┤
> >>>>>
> >>>>> │adults │
> >>>>>
> >>>>> ├────────────┤
> >>>>>
> >>>>> │adultship │
> >>>>>
> >>>>> ├────────────┤
> >>>>>
> >>>>> │adulty │
> >>>>>
> >>>>> ├────────────┤
> >>>>>
> >>>>> │adumbral │
> >>>>>
> >>>>> ├────────────┤
> >>>>>
> >>>>> │adumbrant │
> >>>>>
> >>>>> ├────────────┤
> >>>>>
> >>>>> │adumbrate │
> >>>>>
> >>>>> ├────────────┤
> >>>>>
> >>>>> │adumbrated │
> >>>>>
> >>>>> ├────────────┤
> >>>>>
> >>>>> │adumbrates │
> >>>>>
> >>>>> ├────────────┤
> >>>>>
> >>>>> │adumbrating │
> >>>>>
> >>>>> ├────────────┤
> >>>>>
> >>>>> │adumbration │
> >>>>>
> >>>>> ├────────────┤
> >>>>>
> >>>>> │adumbrations│
> >>>>>
> >>>>> ├────────────┤
> >>>>>
> >>>>> │adumbrative │
> >>>>>
> >>>>> ├────────────┤
> >>>>>
> >>>>> │adunation │
> >>>>>
> >>>>> ├────────────┤
> >>>>>
> >>>>> │adunc │
> >>>>>
> >>>>> ├────────────┤
> >>>>>
> >>>>> │aduncate │
> >>>>>
> >>>>> ├────────────┤
> >>>>>
> >>>>> │aduncity │
> >>>>>
> >>>>> └────────────┘
> >>>>>
> >>>>> $vec
> >>>>>
> >>>>> 417195 300
> >>>>>
> >>>>> 3 {. }.vec
> >>>>>
> >>>>> _0.0264 0.0468 _0.0099 _0.0242 _0.0762 0.0562 0.0863 0.0115 _0.0471
> >>>>>
> >>>> 0.0442
> >>>>
> >>>>> _0.0875 0.0376 _0.0404 _0.0086 0.0161 _0.1689 0.1485 _0.0201 0.1021
> >>>>>
> >>>> _0.0635
> >>>>
> >>>>> _0.0317 0.0142 0.0588 _0.1299 _0.0905 0.0389 _0.0452 0.1352 0.0731
> >>>>> 0.0648
> >>>>> 0.1309 0.0493 0.0785 0.015...
> >>>>>
> >>>>> _0.0096 0.0318 _0.0095 _0.042 _0.0831 0.1103 0.075 0.024 _0.0237
> 0.0398
> >>>>> _0.1274 _0.0299 _0.0209 _0.0195 _0.0043 _0.1033 0.1378 _0.0499 0.0517
> >>>>> _0.0958 _0.0651 0.0214 0.0096 _0.0855 _0.1049 0.036 _0.0562 0.043
> 0.0616
> >>>>> 0.1124 0.152 0.0418 0.0628 _0.018...
> >>>>>
> >>>>> _0.0364 0.0254 _0.0448 _0.0327 _0.0712 0.1548 0.1004 0.0033 _0.039
> >>>>> 0.0635
> >>>>> _0.1179 _0.0703 _0.0359 0.0296 _0.0594 _0.0954 0.1904 _0.0301 0.0078
> >>>>> _0.0607 _0.0344 0.034 _0.0059 _0.1453 _0.0429 _0.0061 _0.05 0.0377
> >>>>> 0.0959
> >>>>> 0.1313 0.1238 0.0302 0.0043 _0.038...
> >>>>>
> >>>>>
> >>>>> So this looks good!
> >>>>>
> >>>>>
> >>>>> Now I need a verb that will let me specify a word, and it will return
> >>>>> the
> >>>>> associated vector.
> >>>>>
> >>>>> Here's how it should work:
> >>>>>
> >>>>>
> >>>>> tst =. get 'adults'
> >>>>>
> >>>>>
> >>>>> tst
> >>>>>
> >>>>> 0.1144 0.0444 0.0574 0.0387 0.082 _0.0271 0.209 _0.006 _0.1896 0.1038
> >>>>> _0.0257 0.0646 0.0488 _0.0065 0.0486 0.0422 0.0239 _0.1006 _0.0541
> >>>>> 0.0511
> >>>>> _0.0254 _0.0121 0.0216 0.0324 _0.1349 0.0237 0.0049 0.0061 0.0349
> >>>>> _0.0264
> >>>>> 0.0086 0.0919 _0.0174 0.0645 ...
> >>>>>
> >>>>>
> >>>>> To build the 'get' verb we need to try to find the location of the
> word
> >>>>>
> >>>> 'adults'
> >>>>
> >>>>> in the boxed words array:
> >>>>>
> >>>>> 'adults' = each words
> >>>>>
> >>>>> |length error
> >>>>>
> >>>>> | 'adults' =each words
> >>>>>
> >>>>>
> >>>>> Nope, that didn't work... Do I need to box the word?
> >>>>>
> >>>>>
> >>>>> (<'adults')=each words
> >>>>>
> >>>>> |length error
> >>>>>
> >>>>> | (<'adults') =each words
> >>>>>
> >>>>>
> >>>>> Nope! How do I find a specific word in the boxed word array?
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>> Skip Cave
> >>>>> Cave Consulting LLC
> >>>>>
> >>>>> On Wed, Feb 21, 2018 at 2:36 AM, Skip Cave <[email protected]>
> >>>>> wrote:
> >>>>>
> >>>>> I read in a text file of word vectors using fread. The format looks
> >>>>>> like
> >>>>>> this:
> >>>>>>
> >>>>>> bell 0.0264 -0.2927 -0.0254 -0.1034 0.1672 -0.0440 -0.0019 0.1210
> ...
> >>>>>>
> >>>>>> bell_tower -0.1252 -0.1233 0.1351 0.1897 0.0242 0.0014 0.1942
> -0.0237
> >>>>>>
> >>>>> ...
> >>>>
> >>>>> belt 0.1332 0.0142 -0.1208 -0.0574 0.1451 -0.0731 -0.1293 0.0855 ...
> >>>>>>
> >>>>>> belfast 0.1190 -0.0440 -0.0254 -0.2090 0.2144 0.0348 -0.1467 0.1256
> ...
> >>>>>>
> >>>>>> Everything is literal text.
> >>>>>>
> >>>>>> The basic layout for each line is:
> >>>>>>
> >>>>>> word(s) (could contain multiple words separated by underscores)
> space
> >>>>>> number (positive or negative) in text format space number (positive
> or
> >>>>>> negative) in text format space
> >>>>>> ...... repeat for 300 numbers (in text)
> >>>>>>
> >>>>>> the last number is followed by a line feed for the next line
> >>>>>>
> >>>>>> I need to:
> >>>>>> 1. Convert all the the high minuses (-) to J's low minus (_) 2.
> >>>>>> Extract the word(s) up to the first space into a separate array
> >>>>>> (words) 3. Convert the text numbers into a 2D array of ? x 300
> >>>>>> floating point numbers
> >>>>>>
> >>>>>> I know how to do #1 (string replace), and #3 (".) once I get rid of
> >>>>>> the words, but I don't know how to strip out the initial word on
> each
> >>>>>> line and put them in a separate array. Any help is appreciated.
> >>>>>>
> >>>>>> Skip
> >>>>>>
> >>>>>> ------------------------------------------------------------
> ----------
> >>>>> For information about J forums see http://www.jsoftware.com/
> forums.htm
> >>>>>
> >>>> ------------------------------------------------------------
> ----------
> >>>> For information about J forums see http://www.jsoftware.com/
> forums.htm
> >>>>
> >>>> ------------------------------------------------------------
> ----------
> >>> For information about J forums see http://www.jsoftware.com/forums.htm
> >>>
> >>
> >>
> >> ---
> >> This email has been checked for viruses by AVG.
> >> http://www.avg.com
> >>
> >>
> >> ----------------------------------------------------------------------
> >> For information about J forums see http://www.jsoftware.com/forums.htm
> >>
> > ----------------------------------------------------------------------
> > For information about J forums see http://www.jsoftware.com/forums.htm
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
