Anyone know of a reverse parsing tool on the web?

[JuniperPark]

Is there a tool on the web that does a good job of reverse parsing strings into phrases?

Example #1 (easy):
FIREDEPARTMENT => FIRE DEPARTMENT

Example #2 (harder):
NEWYORKFIREDEPARTMENT => NEW YORK FIRE DEPARTMENT

I would be willing to pay for something like this if necessary.

Thanks!

 

[madcamel]

http://www.optimize.net/tools/word-parser.html

 

[Salient]

Nice one MC. Seems to work as advertised.


I was curious if someone would respond to JP's thread.

 

[JuniperPark]

I wasn't able to get it to work for more than a couple dozen domains. In fact most of the tools on that site seem to either simply not work, or have a very limited function. Sometimes they show code debug responses, and sometimes they give no response at all. But this is along the idea of what I was looking for.

 

[Ubiquitous]

Is there a tool on the web that does a good job of reverse parsing strings into phrases?

Example #1 (easy):
FIREDEPARTMENT => FIRE DEPARTMENT

Example #2 (harder):
NEWYORKFIREDEPARTMENT => NEW YORK FIRE DEPARTMENT

I would be willing to pay for something like this if necessary.

Thanks!

I actually know of a few + wrote something similar not too long ago. Are you doing one word at a time or manually putting the phrases in? If you let me know a few more details I can guide you in the right direction ... just let me know.

Regards ~

 

[Dale Hubbard]

To do this effectively and in bulk you need a massive amount of recursive code. And you need a pretty huge dictionary.

Consider this:

NEWYORKFIREDEPARTMENT

NEW
YORK
FIRE
RED
FIRED
PAR
ART
PART
MEN

See the problem?

 

[JuniperPark]

I actually know of a few + wrote something similar not too long ago. Are you doing one word at a time or manually putting the phrases in? If you let me know a few more details I can guide you in the right direction ... just let me know.

Regards ~

Ideally I would like a SQL Server stored procedure... failing that, VbScript code, and failing that a bulk checker somewhat like the one suggested above, except that it works

Thanks!

To do this effectively and in bulk you need a massive amount of recursive code. And you need a pretty huge dictionary.

Consider this:

NEWYORKFIREDEPARTMENT

NEW
YORK
FIRE
RED
FIRED
PAR
ART
PART
MEN

See the problem?

I definately understand the complexity, in fact I've written smaller versions of these myself.. that's why I'm looking outside to see if there is something better. I already have 'large' and 'small' dictionaries on my SQL server for high speed lookups.

 

[Dale Hubbard]

It's an old problem, and unfortunately one in which computers are slower than humans.

It would be quicker to eyeball the list.

 

[JuniperPark]

It's an old problem, and unfortunately one in which computers are slower than humans.

It would be quicker to eyeball the list.

Nah... it can be done. Step 1 would be to try all possible combinations of spacing (well, up to 4 spaces) where every word is in the dictionary. WHere there is more than one possible result, hit Google for a popularity check to determine the best word split. This should be mroe than 0.1 seconds per domain on a modern server, then run the Google side for as long as that takes.

 

[Dale Hubbard]

Well, you obviously seem to know best. You cannot arbitrarily decide on 4-space steps. You don't want to be hitting Google until your parsing is done. So how would your method work on a domain such as CREDITCARDSFORALL.COM? I'm keen to know...

 

[Ubiquitous]

For manual input dictionary.com has the best splitting algorithm ... take a look at the handy work it did for the more difficult example you listed. Give that a try and see if something like that (although in bulk) would ultimately work.

One question I had for you - are the names always all uppercase?

Regards ~

 

[Dale Hubbard]

Do a logical AND &HDF on the list and they'll all be upper case. Or use a Linux awk line which includes a 'toupper' command. Dictionary.com is excellent, but as per my example it can't do this: http://www.reference.com/search?q=CREDITCARDSFORALL

 

[Ubiquitous]

Do a logical AND &HDF on the list and they'll all be upper case. Or use a Linux awk line which includes a 'toupper' command. Dictionary.com is excellent, but as per my example it can't do this: http://www.reference.com/search?q=CREDITCARDSFORALL

That is a definitely a tough one ... I'm thinking we'll need to borrow some supercomputing power to take care of CREDITCARDSFORALL.com - In the meantime I think this might take the prize for the most comprehensive tool for those pesky dot coms!

BTW ~ I would jump on Flaccid Lard Resort.com while it's still available ... :yes:

 

[Dale Hubbard]

Anagrams - useful but of no direct consequence to this debate I fear

Anagrams are far easier to produce - you have a pool of letters to strike out against a dictionary - not so difficult - but amusing nevertheless

 

[DomainingCom]

We have such technology (check DomainScore.com), PM to explain why you intend to do with this, maybe we can partner.

 

[JuniperPark]

OK, spent a little time in this stuff this morning. Here's the MS SQL for the basics, just need to make it recursive for the 3 and 4 word splits. Actually, I think I can make it work for unlimited word counts. This runs in < 1 second per name on my older, stessed server running 13,000 websites.


declare @str varchar(100)
set @str = 'firedepartment'
--set @str = 'CREDITCARDSFORALL'
set @str = lower(@str)

declare @pos1 int, @pos2 int, @done char(1), @word1 varchar(100), @word2 varchar(100), @pos1_stop int
declare @word1_dict int, @word2_dict int

set @done = 'N'
set @pos1 = 0
set @pos2 = 0

set @pos1_stop = len(@str) - 2

while @done='N'
begin

set @word1_dict = 0
set @word2_dict = 0

set @pos1 = @pos1 + 1
set @word1 = substring(@str,1,@pos1)
set @word2 = substring(@str,@pos1+1,100)

if @pos1 = 1
begin
if @word1 = 'a'
set @word1_dict = 1
end
else
begin
select @word1_dict = count(*) from dictionary.dbo.dictionary where @word1 = word
end
if @word1_dict > 0
select @word2_dict = count(*) from dictionary.dbo.dictionary where @word2 = word
-- TODO: recursive check of remainder of string for 2 more words, then 3 more words

-- if @word1_dict > 0 and @word2_dict > 0
-- TODO: Insert into DB for Google pop check if more than one result

-- Test-only Check of results
select @pos1, @word1, @word1_dict, @word2, @word2_dict

if @pos1 = @pos1_stop
set @done = 'Y'

end