reversed words in english leading new english words

[robleh373]

I have a simple vba script that runs in excel which reverses english words and checks if they create new english words. However, it is not reliable since it simply doesn't really know the language for obvious reasons. This script is related to a curiosity of mine to know two things:

• how many words can be reversed to form new words in english (a number to be determined)

what are the reversible words.

• I suspect this is possible in any language not just english. In english the shorter the word the greater the probability it forms another I believe.

Here is the code which gets it word source from dictionary.com:
Sub CheckWordsOnDictionaryCom()
Dim IE As Object
Dim URL As String
Dim word As String

' Set the URL to the Dictionary.com search page
URL = "https://www.dictionary.com/browse/"

' Reference to Internet Explorer
Set IE = CreateObject("InternetExplorer.Application")
IE.Visible = False

' Loop through each cell in a column (e.g., column A)
For Each cell In ThisWorkbook.Sheets("EnglishWordlist").Range("b26:b" & ThisWorkbook.Sheets("EnglishWordlist").Cells(Rows.Count, "B").End(xlUp).Row)
word = Trim(cell.Value) ' Get the word from the cell
If word <> "" Then
' Navigate to the Dictionary.com page for the word
IE.navigate URL & word
Do
DoEvents
Loop Until IE.readyState = 4 ' Wait for the page to load

' Check if the word exists on Dictionary.com
If InStr(IE.document.body.innerText, "No results found for") = 0 Then
' Word found, do something with it
cell.Offset(0, 1).Value = "Found"
Else
' Word not found
cell.Offset(0, 1).Value = "Not Found"
End If
End If
Next cell

' Clean up
IE.Quit
Set IE = Nothing
End Sub

I get many meaningless words reversed. They are rare, from 29884 I only get 454 words.

I need to help making this routine more sophisticated. Perhaps splitting it into subsets.

 

[Chriss Miller]

What do you think of when you think of subsets? I imagine you can forget to find words ending with uq, for example, i.e. there are letter combinations only making sense in one way, not reversed. But then there is bouquet, for example, which has uq in it. In short, I think of that as a problem you could solve offline far simpler as long as you find an exhaustive word list, reverse all of them and then do an SQL inner join of words with reversed words of same length. Indexing the data should be sufficient to make it work far better than a brute force comparison of all words with all reversed words.

 

[Deniall]

In english the shorter the word the greater the probability it forms another I believe.

I have no comment to make about your main quixotic quest, nor about the way you are approaching it.

However your statement I quote above is almost certainly correct, and (if so) will apply to all non-logographic languages, not just English. I further suspect that this would be provable using Statistics alone (but not by me). As trivial pointers to this I offer two comments.
(1) For one-letter words the rate of meaningful reversal is 100%.
(2) For squillion-letter words there is probably only one word of that length in the entire language, so unless that word is palindromic its reversibility rate is 0%.

 

[vb5prgrmr]

While my awareness of such words is on the level of mom, dad, and toot, there are already people who asked this question and have compiled some answers already, and there are some words that describe this strange thing...

Words Spelled Backwards: Discover Semordnilaps | Savvy-Writer

Explore the fascinating world of semordnilaps—words spelled backwards. Learn their history and discover how they can boost your creativity and problem-solving skills.

savvy-writer.com

https://en.wikipedia.org/wiki/Anadrome 'I think you like this one and where it points too

derived from a simple google search

Google Search

Good Luck & Have Fun

 

[Chriss Miller]

The gateman checked the nametag against the guest list.

That's my favorite so far from my own list: gateman/nametag.

I got a list of ~1000 from ~466k words from https://github.com/dwyl/english-words/blob/master/words.txt
I removed single letters, words with a period and an apostrophe and only kept words where the first letter is smaller than the last letter or the same, so you mainly have only one word of the two possible except for palindromes and pairs like spoons/snoops.

Still, many of the semordnilaps I have are not well known words.

Problem with the root 466k word list is that it's containing words you don't find on dictionary.com. It has words like
kinnikinic
kinnikinick
kinnikinnic
kinnikinnick
kinnikinnik

Of which the latter obviously is a palindrome and so also a semordnilap. The longest found, if you count it from alternative forms of https://en.wiktionary.org/wiki/kinnikinnick#English

As said many words of words.txt are not on dictionary.com. Some of those not found on dictionary.com are in Merriam Webster, though, or in the Wiktionary. What's the reference source of definitive English vocabulary? The Oxford English Dictionary, perhaps?

If you want a list of well known words (both normal and reversed) that's harder, but could be done by looking up both results in a smaller dictionary. More ideally a dictionary of well known words or a dictionary having the frequency of the word usage where you limit it to a minimum frequency of usage.

 

[Chriss Miller]

For reference I wrote a repro to anyone using VFP.
I now also know what I forgot in the data cleaning: Only keeping all lower case words, which gets rid of any abbreviations and acronyms.

In this run I kept words where the first letter is larger than the last letter, so all pairs and palindromes are in the list, too.
That gives me 1220 words, not ~2000, but also words like tat-tat-tat.

Anyway, I think it's my best shot and it only takes about 15 seconds to run, not counting the download of the words.txt I use in this code:

tstart = Seconds()
Close Databases All
Set Safety Off

Create Cursor Import (Word v(45))
Append From words.txt Type Csv

Select Word, Cast(Len(Word) As Integer) As Len From Import Into Cursor dict Readwrite
Delete For Len=1
Delete For "'" $ Word
Delete For "." $ Word
Delete For Not Lower(Word)==Word
Copy To dict.Dbf For Not Deleted()
Use dict Exclusive
Index On Len Tag Len
Index On Word Tag Word

Select Cast(Reverse(Alltrim(Word)) As v(45)) As drow, Len As nel From words Into Cursor tcid
Copy To tcid.Dbf
Use tcid.Dbf Exclusive
Index On nel Tag nel
Index On drow Tag drow


Create Cursor semordnilaps (emordnilap v(11),Length Integer)
Select dict
Set Order To Word

Set Exact On
Select tcid
Set Relation To drow Into dict
Scan For Found("dict")
   Insert Into semordnilaps Values (tcid.drow, tcid.nel)
Endscan
Select semordnilaps
Copy To semordnilaps.Dbf
Use semordnilaps.Dbf Exclusive
Index On Length Tag Length
Index On emordnilap Tag emordnilap
? Seconds()-tstart
Browse

Function Reverse(tcWord)
   If Len(tcWord)=1
      Return tcWord
   Else
      Return Right(tcWord,1)+Reverse(Left(tcWord,Len(tcWord)-1))
   Endfunc

Notes of minor interest: I used drow and nel as reverse of word an len, emordnilap instead of semordnilap because of the 10 character limit of index tag names and the reverse function could be done better than recursive, it still doesn't take very long.

The list generated with that remains to be cleaned up against a dictionary, of course. Well, and even better the original words.txt could be improved. I guess you could also find longer and yet less redigated lists.

The more you can rely on words.txt to be an accurate list of words by whatever detailed definition of words you want to limit yourself on, the less it would take to run and you could then do without verification as aftermath. On the other hand this raw list already is much shorter than a full dictionary list and is easier to verify against dictionary.com or merriam webster or whatever dictionary you would like to use for that verification step.

Find the list attached...

 

[Chriss Miller]

And here's a list with word and semordnilap in the same row, but also palindromes. It reduces the list to 675 rows.

That computes with the 1220 word list as it has 545 pairs of words differing with their reversed counterpart and 130 palindromes, so 545*2+130=1220.

It makes it easier to lookup whether both word and semordnilap are valid words and the pair can be counted in. The word in this data is always smaller or equal to the semordnilap and the inverse pair is not in the list. The third column is the length of the word and the last column in the data is stating, whether the word is a palindrome and is T for true and F for false. You could also see that from word=semordnilap, of course.

 

[Chriss Miller]

One last thing to say. Since I only look for the reversed word in the words I already have, instead of looking up the reversed word on a website, I first keep it self referential. Doing it the way you do, the reversed word may not be in your original list, but found on dictionary.com anyway, so you extend your chances of getting a semordnilap. I wonder if it's worth to do, though, as it takes so much longer to lookup all the reversed words online instead of first working self referential within the list you already have.

 

[robleh373]

What do you think of when you think of subsets? I imagine you can forget to find words ending with uq, for example, i.e. there are letter combinations only making sense in one way, not reversed. But then there is bouquet, for example, which has uq in it. In short, I think of that as a problem you could solve offline far simpler as long as you find an exhaustive word list, reverse all of them and then do an SQL inner join of words with reversed words of same length. Indexing the data should be sufficient to make it work far better than a brute force comparison of all words with all reversed words.

 

[robleh373]

this is a good idea I hadn't thought about. I could import the huge wordlist excel file to sql, I have 29K word already reversed. I can create a pattern query with various join types....yeah this a great idea. Thx....in fact my code was going to go thru 300k words and I stopped it. But, potentially I can get a comprehensive list of english words and then bring that in SSMS. Might run to some size, memory issues in the importation, but I can fix that with SSIS.

 

[Chriss Miller]

I have 29K word already reversed.

Even VBA has StrReverse(), so whatever word list you have, reversing that is simple and fast, T-SQL also has a function you could use after importing the word list:

REVERSE (Transact-SQL) - SQL Server

REVERSE (Transact-SQL)

learn.microsoft.com