Download as pdf or txt
Download as pdf or txt
You are on page 1of 2

PROCEDURE

For Query Processing:


1. The process starts with the user either job provider or seeker giving the
query in the form of audio or text, if audio then it is converted into text and
passed to our llm model for query processing.
2. At the time of giving the prompt to the chatbot, the user preferences
mentioned in the chat will be extracted using the llm for user preference
extraction.
3. The user preference will be stored as vector in the user's local machine.
4. Job providers can upload a job notification or new job postings in the form
of pdf or image in the portal’s dbms. We will extract the semi structured
data (tables, text from pdf) from these pdf or images using the help of OCR,
and embed them into vectors and store them in our additional job info
vector store.
5. There will be the existing dbms of the portal for job and job seeker.
6. Now, coming back to our query processing llm, our original prompt will be
supplemented by our knowledge repository i.e. the combination of user
preference vector store, job vector store, dbms for job and job seeker,
making our chatbot architecture a RAG.
7. In response to the user prompt, based on the user preference data, current
available job etc. the final answer will be generated and will be displayed in
audio or/and text.

Recommendation System:
1. Whenever job seeker will visit the chatbot, by default the chatbot will refer
to the user history and preference vector store and all the job vector store
and job dbms, and create a list of suitable and available jobs for the seeker
and recommend the list to the user.
Additional Feature of Form filling using llm:
1. After the main prompt is passed and if it is observed that the prompt is
about filling up the registration form for job seeker or provider or it is
about job providers wanting to register a new job posting through a form fill
up, then our chatbot will help the user filling up that form.
2. After understanding that it is a query about form fill up, the flow of control
is transferred to the llm for form filling from the main llm for query
processing. Now it is the new llm's responsibility to fill out the form.
3. This new llm will chat back and forth with the user until all the data
required to fill out the form is gathered.
4. After all the data is gathered through the chat, the llm then makes an object
of the form data, and the system then saves the object into the form thus
inserting a new record in the DBMS for job seeker or DBMS for job based on
the user type.

You might also like