testrun.txt

Hi
hello si
r

so you have a query coming in like this

Create a repository with name abcd
, right ? okk sir



keep this kind of dictionary, at the base
from github_module import github_create_repo_func 
from bitbucket_module import bitbucket_create_repo_func


FUNC_MAP = {
    "repository": {
        "create": {
            "github": github_create_repo_func,
            "bitbucket": bitbucket_create_repo_func 
        },
        "delete" .. etc.
    }
}

makes sense ?

Then call it like this

arguments will be a dictionary extracted from query,

for this particular instance it would be this:

arguments = {
    "name": "abcd"
}

FUNC_MAP[query_domain][query_type][domain_provider](**arguments)

This much makes sense ?
yes sir surely I m getitng your point
this will make the process really simplified.
But sir, tell me one more thing, how will I extract the arguments?

using NER ?

yes sir
The previous one file I have of ner that I have showed you. Should I use that one only?

That or something similar to that should work. I mean even if you use the same file it should be an issue.

Are you able to extract query_type and domain ?
yes sir for domain i.e. whether the query is create, delete or etc we are getting that

and sir for query_type, I have to figure out something
isnt cosine similarity working for this ?

Just tokenize the sentence and see if a token has > 0.80 similarity to a word .. it should be pretty accurate.
okk sir I will do it that way
Sir I m checking the word2vec and I , thinking to use that for finding query_type, yeah I sent that gensim thing
did u check it ?
It should be pretty straightforward as well.
Yes sir I have used that a little to understand its use
Tonight I will work on that and also I have studied about making a dockerfile
I will show u that also tomorrow
okay sure.

You understood the approach right ?
yes sir I understand that
So sir should I change this file naa
which file name ?

sir??
sir are u here??

I would suggest you to keep a central query.json for NER, because the task doesnt really depend on the domain anymore.

What I mean is,

Create a repository named xyz
Create a task with description abcd efgh
Create a task that has descrption abcd efgh
Create a story that has description ..

Doesnt really matter right whether it comes for JIRA, Github, Confluence or whatever..?

sir then where we will run that query matters naa
say a new project or tasks are created for JIRA
and creating a new repo or opening a new issue or pull reqest is for bitbucket or github

yes sir obviously
sir actually for that we have 4 different pages separate for github, jira, bitbucket and Confluence
so we actually dont need NER for thatw
NER will be used for getting query_type and domain to get the right function to call

Yes but thats something "static" isnt it ?

either it comes directly from settings file or it has a fixed token.

You dont really need to do NER for this.
Does it make sense ?

I would suggest that NER should be used for getting the arguments only.

domain and query_type are very trivial to determine.

Domain = Cosine similarity of token against ("repository", "project", "user" ,... etc.)
Query_Type = Word2Vec similarity of token against ("create", "delete", "update", "get")

Ohh okk okkk now I get the complete point
Yes sir I get what you are telling

Also as you just told that you have one "deterministic" thing that is domain_provider
ie. Confluence, Github, Bitbucket etc. then make the FUNC_MAP like this.



FUNC_MAP = {
    "github": {
        "create": {
            "repository": github_create_repo_func,
        },
        "delete" .. etc.
    },
    "bitbucket": {
        "create": {
            "repository": bitbucket_create_repo_func,
        },
        "delete" .. etc.
    }
}

Reason is that you go from highest confidence to lowest.

"github", "bitbucket" ... etc. is very deterministic. You have that value beforehands with 100% accuracy.
"create", "delete" etc. can also be pretty accurately determied. with a very ..
no problem.

Friend tha koi ? ;D
yes sir
actually satyarth anydesk se connected hai
toh usne glti se type kr diya
sir please continue
I have disconnected him
Sahi hai.

keep the code flow like this.

def get_query_result():
    query_sentence = "whatever"
    query_result = process_query(query_sentence, domain_provider)
    return query_result


def process_query(query_sentence, domain_provider):
    preprocessed_query = preprocess_query(query_sentence)
    query_domain = get_domain(preprocessed_query)
    query_type = get_query_type(preprocessed_query)
    query_args = get_query_args(preprocessed_query)
    return FUNC_MAP[domain_provider][query_type][query_domain](**query_args)

should do the job.

yes sir I understand this thing now
sir yha yhi hoga naa
what ?

yes

query_domain = repository, user, etc.
query_type = create, delete, update, get
domain_providr = confluence, bitbucket etc. 
query_args = {"name": "xyz",} etc.
haan sir I get all this
I will implement it tonight and will show you tomorrow t