Calculating similarity between tags/keywords (NLP)

Hello! Thank you for having interest in this topic.

I am trying to have a tool calculating distances/simialrities between keywords. 

in fact, Tried it for multiple times by myself using Gensim/Spacy/Fasttext. usually very well known libraries. 

but everytime I tried this I’ve only acquired very horrible results. non-usable. not-practical. no-value or efficiency.

though I believe it was more like because my inexperience and my faults. 

So, here I am wanting to ask, if you have much more broaden experience and view then me. I wanna ask for help about it. (but in a way you making this practical functioning library/function in computer sripts)

condition 1: 

nodejs or Python or AHK format. 

condition 2: 

or else, please suggest me which language format to use (which language your function is scripted) so I could check. 

condition 3:

please enable ‘categorizing function’/’bundling function’ inside of your script. 


for example) 

If you have made a variable called ‘standard_point’, 

I could put some amount of number 0.1 ~ 0.99 in that variable to scale the similarity, 

and see the calculated result which has been categorized and bundled. based on that variable number. 

easily overviewing your pakage’s performance. 

rough example will be something like this. 

target_keyword = ‘flower’

standard_point = ‘0.5’

results(outcome) =


tulip (0.01)

rose (0.01)

scent (0.2)

girl (0.2)

valentine (0.3)

wedding (0.4)

children (0.5)


condition 4:

please show me how it does work by picking 10 different  cases of examples to show me your tool’s performance

condition 5:

after you showing me as condition 4, I can cross check bringing examples from my backgrounds. 

condition 6:

when cross checking in condition 5 is done, I will release the milestone. 

condition 6:

This is just an option (not essential), but 

if you can get it worked even in global scale, (for example standardform = english, additional form = russian, hindi, japanese, mandarin) 

or suggesting any legitimate idea for this one, 

I will make sure to tip fair amount of money for this. 

condition 7: 

other practical recommandation or suggestion would be welcomed. 

condition 8 (important): 

I know that there are lots of companies who need pretty much the same function, 

So I assume maybe you might have already built one. no problem with this.

So please show me any usage examples of your tools/functions/scripts. 

until then, I can not be assured you have made this one. so I can not release any milestone. even few percentage. 

I can not release milestone before seeing the results from your tools. Please be notified this. 

Thank you for reading! Have a good day and bid me if you think you can build one. Anytime ! 

And I am willing to respond very quickly to explain more what I exactly need. 

also, if you can really make it out well this time in this project, I will promise to carry out further projects with you. Thanks alot.