The Quotable Architecture
MIDS Capstone Project Fall 2020

Quotable: Making Data Come Alive!

Quotable is a data assistant that helps consumers and enterprises query their datasets in natural language. Quotable uses cutting-edge transformers (BERT, Google TaPas for tabular data) and layers our innovation called "AQuA" (Augmentation for Question Answering) to provide simple answers to queries in English.

Generally, you will need to be an expert at Structured Query Language (SQL) to query tables. SQL can be complex and difficult to learn. With Quotable, you can query tables in plain English! It is very simple to use - Just upload your tables, ask your questions, and make your data talk to you. Quotable is perfect for you if you are experiencing "SQL pains" such as below:

  • You have to generate instantaneous insights for your execs as a part of an Enterprise BI Team
  • You are a lay consumer who wants insights into data, but don't have SQL expertise

Quotable for Enterprise BI

As an executive in a large enterprise, you want your BI team to provide quick insights on revenues for your product lines. However, every time you ask a question, the BI team needs another week to get back to you. With Quotable, you don't need to wait, and you don't need to know SQL either! Just ask your question in English and get an instantaneous response.

Quotable for Consumers

You are a basketbal enthusiast, say, and no stats expert. You want to know how well Lebron James did in the 2020 season compared to others. You do a Google search, but there are no straight answers. What do you do? With Quotable, you just type your question in, and we do the rest of the magic and get you a straight answer.

Key Technologies

Quotable uses two main technologies for its product:

  1. Augmentation for Question Answering: AQuA is the patentable innovation which prepares user tables for querying. AQuA has 3 major functions: Splitting larger tables into smaller ones to feed our NLP model, Preprocessing and modification of table rows and colums, Augmenting existing rows and columns with derived rows and columns
  2. Pretrained BERT model for Question Answering:  We used a BERT model which is pretrained with table data to perform natural language based Question Answering. For our first version of Quotable, we are using a Google model called TAPAS.

More Information

Quotable: Making Data Come Alive!
Quotable: Making Data Come Alive!

Quotable is simple to use

  1. Upload your tables: ou can upload .csv or .xls/.xlsx formats. We will support other formats in the future.
  2.  Quotable prepares your tables for Querying: On the fly, Quotable prepares your tabes for Querying. We perform a number of operations such as identifying the table to query, breaking down tables into smaller chunks for NLP processing, adding new rows or colums if needed.
  3. Ask Questions and Get Answers: You can then query your tables in English. Ask questions and get instant answers

Video

QuotableAI Demo

QuotableAI Demo

Last updated:

December 8, 2020