Results of Various Tests of K-Nearest-Neighbour to Recognise a Paraphrased Statement
#knn #paraphrasing #essayhelper

I wanted an algorithm that could recognise a paraphrased version of a statement.
This was part of my Essay Helper GitHub Repository, which recognises reused statements while questioning the user to write a humanities-style essay.

The point of paragraph 2 is to introduce KNNs... I found them after querying "Prolog machine learning" on YouTube. K-Nearest-Neighbour was the successful candidate to select the best answer using machine learning while not sacrificing rule writing with the result. (They produce a list of results, i.e. 0 for exact match and increase with "greater distance" away from this.)

In the tests, I wanted to test for success in finding that a statement contains a paraphrased statement.

First, Control: this test should return true if the statement is exactly the paraphrased statement:

?- Phrase="I'd like that very much",downcase_and_split(Phrase,Key_Phrase),get_string("",either,one,"",Key_Phrase,String).

|: I'd like that very much
Attempt 1
[found,[i,d,like,that,very,much],[i]]
Success

Good! Now on to the tests.

I tested for paraphrased statements in 6 word statements (surrounded by other words), where the replacement, addition or deletion of word(s) occurred at the start, middle or end of the 6 word paraphrased version of the statements.

Replacing 1 word with 1 word

Find "I'd like that very much." in "It goes unsaid that you'd like that very much my dear."
[found,[i,d,like,that,very,much],[you,d,like,that,very,much]]

Find "I'd like that very much." in "It goes unsaid that I'd like those very much my dear."
[found,[i,d,like,that,very,much],[i]]

Find "I'd like that very much." in "It goes unsaid that I'd like that very happily my dear."
[found,[i,d,like,that,very,much],[i]]

Replacing 1 word with 2 words

Find "I'd like that very much." in "It goes unsaid that you will like that very much my dear."
Failed

Find "I'd like that very much." in "It goes unsaid that I'd like those greatly much my dear."
[found,[i,d,like,that,very,much],[i]]

Find "I'd like that very much." in "It goes unsaid that I'd like that very happily most dear."
[found,[i,d,like,that,very,much],[i]]

Replacing 2 words with 1 word

Find "I'd like that very much." in "It goes unsaid that you like that very much my dear."
Failed

Find "I'd like that very much." in "It goes unsaid that I'd like those much my dear."
[found,[i,d,like,that,very,much],[i]]

Find "I'd like that very much." in "It goes unsaid that I'd like that very happily dear."
[found,[i,d,like,that,very,much],[i]]

Adding 1 word

Find "I'd like that very much." in "It goes unsaid that nevertheless I'd like that very much my dear."
[found,[i,d,like,that,very,much],[i]]

Find "I'd like that very much." in "It goes unsaid that I'd like that one very much my dear."
[found,[i,d,like,that,very,much],[i]]

Find "I'd like that very much." in "It goes unsaid that I'd like that very much love my dear."
[found,[i,d,like,that,very,much],[i]]

Adding 2 words

Find "I'd like that very much." in "It goes unsaid that in time I'd like that very much my dear."
[found,[i,d,like,that,very,much],[i]]

Find "I'd like that very much." in "It goes unsaid that I'd like that one there very much my dear."
[found,[i,d,like,that,very,much],[i]]

Find "I'd like that very much." in "It goes unsaid that I'd like that very much loved one my dear."
[found,[i,d,like,that,very,much],[i]]

Deleting 1 word

Find "I'd like that very much." in "It goes unsaid that I like that very much my dear."
[found,[i,d,like,that,very,much],[i]]

Find "I'd like that very much." in "It goes unsaid that I'd like very much my dear."
[found,[i,d,like,that,very,much],[i]]

Find "I'd like that very much." in "It goes unsaid that I'd like that much my dear."
[found,[i,d,like,that,very,much],[i]]

Deleting 2 words

Find "I'd like that very much." in "It goes unsaid that like that very much my dear."
Failed

Find "I'd like that very much." in "It goes unsaid that I'd like much my dear."
[found,[i,d,like,that,very,much],[i]]

Find "I'd like that very much." in "It goes unsaid that I'd like that my dear."
[found,[i,d,like,that,very,much],[i]]

Combinations

Replacing 1 word with 1 word at start
and adding 2 words in middle

Find "I'd like that very much." in "It goes unsaid that you'd like, indubitably so, that very much my dear." Failed
Replacing 1 word with 1 word in middle
and adding 2 words at start

Find "I'd like that very much." in "It goes unsaid that in time I'd like those very much my dear." [found,[i,d,like,that,very,much],[i]]
Adding 1 word in middle and

Deleting 2 words at start

Find "I'd like that very much." in "It goes unsaid that like that, timely, very much my dear." Failed
Replacing 1 word with 2 words at start and
Deleting 1 word at end

Find "I'd like that very much." in "It goes unsaid that I definitely equally like that very my dear." [found,[i,d,like,that,very,much],[i]]
Replacing 1 word with 2 words at end,
adding 1 word at start and
deleting 2 words in middle

Find "I'd like that very much." in "It goes unsaid that timelily I'd like definitely so my dear." [found,[i,d,like,that,very,much],[i]]
Replacing 1 word with 2 words at start,
adding 1 word at middle and
deleting 2 words in end

Find "I'd like that very much." in "It goes unsaid that I definitely nonequally like that infinitesimally my dear." [found,[i,d,like,that,very,much],[i]]
In summary, the best result came from replacing the first word of the statement with another word. The failures came from doing these at the start:

replacing 1 word with 2 words
replacing 2 words with 1 word
deleting 2 words (but not always, as the combinations show).
See my GitHub Repository.

Lucian Green