Knowledge-Based Question Answering System

chevron-icon
Back
project-presentation-img
Mukhtar A.Algezoli
Project Owner

Knowledge-Based Question Answering System

Funding Awarded

$20,000 USD

Expert Review
Star Filled Image Star Filled Image Star Filled Image Star Filled Image Star Filled Image 0
Community
Star Filled Image Star Filled Image Star Filled Image Star Filled Image Star Filled Image 0 (0)

Status

  • Overall Status

    🛠️ In Progress

  • Funding Transfered

    $15,000 USD

  • Max Funding Amount

    $20,000 USD

Funding Schedule

View Milestones
Milestone Release 1
$5,000 USD Transfer Complete TBD
Milestone Release 2
$3,750 USD Transfer Complete TBD
Milestone Release 3
$3,125 USD Transfer Complete TBD
Milestone Release 4
$5,000 USD Pending TBD
Milestone Release 5
$3,125 USD Transfer Complete TBD

Project AI Services

No Service Available

Overview

This service is an extension to an already running service on Enigma AI platform (https://platform.enigma-ai.com/), It is intended to work as a knowledge-based question-answering system. The cost of running this service is still too high to be opened to small and medium businesses but it could be optimized using a new technology called Adapters which will significantly lower operation costs and make it suitable to work as an add-on in other platforms or websites.

Proposal Description

Compnay Name

Enigma AI

Service Details


This service is an extension to an already running service on Enigma AI platform (https://platform.enigma-ai.com/), It is intended to work as a knowledge-based question-answering system, i.e., the customer would ask a question and the service (chatbot) will try matching the question with the most similar question from a database (containing question-answer pairs) then the answer (or flow branch) for this chosen question is returned. We have done this by utilizing the State-Of-The_Art model in Natural Language understanding called BERT (https://arxiv.org/abs/1810.04805 ) which is usually trained in two steps: First pretraining in a big dataset in a specific language and since we are currently operating on a region with mostly Arabic speakers (Sudanese dialect) we collected around 15 million sentences from Twitter, Facebook, Books, and news sites and pretrained BERT on this data until it became proficiently able to understand. The second step is fine-tuning where the model is trained on domain-specific data using a task that we developed called semantic similarity, currently, this step should be done for each new customer (which takes a long time) and a version of the model should always be loaded in memory, because of these two issues this intelligent AI model is only available to premium customers (like Banks, Telecom companies,...etc) and not available to everybody on our Saas platform (https://platform.enigma-ai.com/).

Enigma AI is a startup tech company that provides cutting-edge technology solutions using state-of-the-art Artificial Intelligence. It provides a high quality and interactive alternative to conventional call centers for big enterprises on social media. Moreover, Enigma AI enables small digital business owners to launch their businesses at low costs and connect with their audience. It enables businesses to create chatbots on social media channels (Facebook, Instagram, WhatsApp) for customer service and commerce to improve their customers' satisfaction, increase sales opportunities and optimize their operations while reducing costs. In short, it is like Shopify but using chatbots (Try our limited-time free experience by signing up to https://platform.enigma-ai.com/ and linking your social media page).

Problem Description

This proposal is intended to solve this scalability problem.

Solution Description

To make the AI-powered chatbot available to every user, we have already started researching and come up with very promising results from a new way of training BERT called Adapters, but in order to adapt it to our semantic similarity task and package it as a final product we need to assemble a team of data scientists and MLOps engineers leading us to this grant.

Project Benefits for SNET AI platform

The onboarding of our service into the SingularityNET platform can provide a variety of benefits. One of the platform's strong features is that it connects a number of AI services that can in the future enhance and inspire each other. Our system will be available on our SaaS platform which can be used both by end users or adapted and integrated into other services at the SingularityNET platform. These kinds of collaborations and elaborate use cases are essential for the growth of the services and the platform as a whole. Additionally, Onboarding our service with its existing customer will introduce the platform to new demographics of customers/businesses who are interested in utilizing the capabilities of AI.  This will contribute to the growth of the SingularityNET platform and make beneficial use of AI tools more accessible across the world. In turn, the SingularityNET platform can provide valuable support to our service. The platform's community of expertise can help improve the model technical capabilities. Furthermore, onboarding on the platform can help our service reach a broader spectrum of consumers, both financially and geographically.
In conclusion, the onboarding of our service into the SingularityNET platform has the potential to provide a variety of benefits for both our service and the platform as a whole. The platform's ability to connect a number of AI services and its existing customer base can enhance and inspire the development of our service. Additionally, the platform's community can help improve the model and its reach to a broader spectrum of users.

Service License Info

MIT license

Marketing & Competition

The chatbots market in MEA is largely untapped requiring more effort in leveraging chatbots in a seamless way to get higher conversion rates and customer satisfaction. Although many global companies can provide chatbot services their costs are much higher leading them to only work with big enterprises leaving out small and medium businesses, we have another competitive edge in understanding and dealing with the customers' literacy, especially in Sudan, which has many similarities in the challenges and opportunities in the region, for example, developing and tuning Arabic-engines to boost the quality of conversations, particularly on social media as most of the people access internet from smartphones and spend a lot of time in social media. This helped us to build various use cases for customer service, social commerce, and process automation as well

Previous Progress

We have been validating different use cases in many sectors while working on automating and scaling the service before investing resources in marketing activities, yet we achieved amazing results for businesses in terms of automating and increasing their sales through social channels (https://wa.me/249116558800, https://wa.me/249116409490, https://wa.me/249999990029, https://www.facebook.com/NationalBankofSudan/), and transforming businesses processes and generating qualified leads for telecom sector with MTN Sudan (https://wa.me/249930012340 (Beta testing)), we are also in very advanced stages with two of the biggest banks in the country, while getting ready to launch our first service for the insurance sector by the end of February Our strategy has been product-focused for most of the past periods, but for this quarter we are working on the UI/UX of our platform before preparing tutorials to help us manage the increasing demand for the service as we launch our marketing campaign next Month. We are planning to use story-telling, email campaigns, workshops, forums, and partnerships to present the value of chatbots for business owners in a meaningful and exciting way.

Related Links

https://www.enigma-ai.com, https://platform.enigma-ai.com

Proposal Video

Placeholder for Spotlight Day Pitch-presentations. Video's will be added by the DF team when available.

  • Total Milestones

    5

  • Total Budget

    $20,000 USD

  • Last Updated

    8 Mar 2024

Milestone 1 - Research new training method

Status
😀 Completed
Description

Research how to train BERT model using Adapters in the Semantic Similarity task and produce good results.

Deliverables

Budget

$5,000 USD

Link URL

Milestone 2 - Implement parallel computing feature

Status
😀 Completed
Description

Implement the solution to use the parallel computing feature of Adapters to facilitate deployment of multiple instances (for multiple businesses) using the same model.

Deliverables

Budget

$3,750 USD

Link URL

Milestone 3 - Automated retraining

Status
😀 Completed
Description

Deploy the solution to allow for automated retraining (allow customers to retrain their models themselves)

Deliverables

Budget

$3,125 USD

Link URL

Milestone 4 - API

Status
🧐 In Progress
Description

Create an API for the solution

Deliverables

Budget

$5,000 USD

Link URL

Milestone 5 - Incorporate into Platform

Status
😀 Completed
Description

Incorporate the solution into the Enigma platform (https://platform.enigma-ai.com/) to work with the existing chatbot experience.

Deliverables

Budget

$3,125 USD

Link URL

Join the Discussion (0)

Reviews & Rating

New reviews and ratings are disabled for Awarded Projects

Sort by

0 ratings

Summary

Overall Community

0

from 0 reviews
  • 5
    0
  • 4
    0
  • 3
    0
  • 2
    0
  • 1
    0

Feasibility

0

from 0 reviews

Viability

0

from 0 reviews

Desirabilty

0

from 0 reviews

Usefulness

0

from 0 reviews