API Interaction Portal Hosted here
Database Interaction Portal Hosted here
Demo Webiste Integration hosted here
Product Idea Video present here
Product Prototype Video Present here
This is a developmental project that acts as a solution to problem 2 of this competition.
The various tools and services used include
- PythonAnywhere
- Google Dialogflow
- Kommunicate
Approximately 100 rows of Product Data was scraped using a python script that made use of Selenium.
This is the csv file that was obtained after scraping the data.
This is the python script that handled data scraping.
These lines of data concentrate on only the following goods that Lowe's Sells on its website
- Appliances
- Refrigerators
- Washing Machines
- Microwaves
- Dishwashers
- Oven / Ranges
- Driers
- Home-Decor
- Area Mats
- Bathroom Mats
- Rugs
NOTE: The values of Prices (Discounted and Current) are dummy values. This had to be done due to inconsistencies.
The motivation to choose SQLite3 is due to the fact that the list of products always increases, and we needed a mechanism to enable that products get added. Along with this, we felt the need for a querying mechanism. This is help us answer more advanced questions from the user.
Additionally, it seemed to match the requirements for an Open Source RDBMS as in the Rules, and had a vast community and good support.
This is the python scipt that enabled creation of the SQLite3 database.
The initial_data_dump_in_db() was the responsible method.
Good database practices were followed while CREATEing and INSERTing values into the database, such as making it SQL-injection proof.
Each product description is parsed and is normalized by using regular expressions.
This is done to ensure that natural language processing happens with utmost ease.
For example:
A product that has a description
"Samsung 24.52-cu ft Side-by-Side Refrigerator with Ice Maker (Stainless Steel)"
will become
22 23 24 25 26 Samsung cubic feet Side by Side Refrigerator with Ice Maker Stainless Steel
This python script handles the string parsing on the product descriptions.
The parse_string() function was the responsible method.
Each string in a product description is stemmed and then fed to a NLP based synset generator.
A synset is a group of words that are similar to a given word. This is not contextualized.
This synset can be used for a hash-table data-structure for fast retrieval of data.
This python script handles the both the stemming and synset generation of individual words in the parsed product descriptions.
The all_syno() method was the responsible method.
For the efficient retrieval and verification of query strings, a hash-table is an ideal data structure.
A garbage value that is guarenteed to have no adverse affects on the hash-table, is initially pushed into it.
Soon after this, the hash-table is pickled (saved), as product_synonym_hash.pkl.
This ensures that it can be used further on.
This python script handles the creation of the hash-table.
The create_hash_map() method was the responsible method.
The pickled hash-table is loaded into the memory.
Synsets of each string in the processed product description, are found and updated in the hash-table.
Soon after this, the hash-table is pickled (saved), as product_synonym_hash.pkl.
This ensures that it can be used further on.
This python script handles the updation of the hash-table.
The update_hash_map() method was the responsible method.
A flask application app.py, was created, and a template inde.html is linked along with the database table inventory.db.
A trigger is set, for updating hash-table with inputs, as soon as inputs are sent into db for insertion.
This python script is the Flask Web Application.
The update_db() method was the responsible method.
This python script contains the code responsible for triggering the updating of the hash table.
The dynamic_data_dump() method was the responsible method.
A REST-API is created around code that accepts a string or customer query and returns JSON for 2 of the most similar products.
The 2 most similar products are found, and their corresponding asset_links, ratings, dummy prices, website links, and descriptions are generated in JSON.
This python script contains the code responsible for creating the API around processing the input and delivers JSON.
The general_search() method was the responsible method.
Step 9 : Ensure that Dialogflow bot has above API method present in Fulfillment, and data is recieved and displayed on iframe, or window.
This package assumes you use Python 3.x.
Expected package dependencies are listed in the "requirements.txt" file for PIP, you need to run the following command to get dependencies:
pip install -r requirements.txt
-
clone this repo
-
install requirements from requirements.txt
-
pip install nltk.downkload('wordnet')
-
Grant permission for SQLite3 Database access
-
To update database, run app.py in the Scripts folder, and insert values into template.
-
Go to localhost (127.0.0.1)
-
To access API go to (127.0.0.1/general_search)
-
To just use chatbot, go to 127.0.0.1/demo
| Name | Github | Phone | ||
|---|---|---|---|---|
| Viswalahiri Swamy Hejeebu | @viswalahiri | @viswalahiri | (+91) 630-152-9655 | |
| Kurapati Abhilash | @abhilash | @abhilash | (+91) 850-072-8944 |
Distributed under the MIT Liscence. See LICENSE for more information.