Saturday, 26 September 2015

Virtual Automated Assistance Technology

Say, you have a product that is used by hundreds and thousands of people, enterprise or public is not a question. You want to keep touch with your users but you also do not want to have an overhead of maintaining a separate team for this. You also want your users to get in touch with you with their questions (which, by experience, one can say are part of FAQs).
You want your user's FAQs answered but at the minimum involvement of a dedicated team.

A simple concept where a computer program responds to human's questions, this will reduce cost and sounds cool too, Right?
It could use any communication medium such as email, facebook, twitter, SMS or any enterprise/customized application used for request-response.


The Application will be capable and can be intelligent enough to do any of the following basic items:
1. Respond to users based on the set of words used by them
2. Respond by using the same communication medium such as Twitter
3. Inform other teams to act further on user's request

My Current Implementation

1. Twitter Account @bbgn123 is mapped as Virtual Automated Assistant
2. When anyone mentions this Twitter handle in his/her Tweet, then the message is read by the Virtual Automated Assistant
3. The Virtual Automated Assistant then parses the messages, and Classification algorithm is executed on the message
4. Once the classification is completed the a Response is generated for the message based on the semantic of the message
5. The Response is then Sent out as Reply to the Tweet via Twitter itself
6. User can see the whole communication as Twitter Conversation in its Tweets and Replies section
7. Analytics part uses classification algorithms (advance steps in progress) and if required, for some implementation, Clustering algorithms will also be explored for unsupervised training
8. List of Default applications to be implemented so that the Virtual Automated Assistant works as Plug and Play (in progress)
9. Sample Requests and Responses from the BabaGyan's implementation of Virtual Automated Assistant @bbgn will be soon published for all to explore
10. Below set of Industrial Applications of the Virtual Automated Assistant shows some potential advantages of the program

Application for Technical Support Teams

1. Automatic Reply to FAQ when User's Tweet on Twitter
2. Automatic Reply on Downtime Schedule when User asks on Twitter
3. Automatic Routing of ticket to Relevant Team based on User's tweet


Application for Transportation Services

1. Automatic Reply with next Bus/Train/Airline Schedule on User's request via Twitter
2. Automatic Reply Cheaper options for Airline on User's request via Twitter
3. Automatic Reply Booking Payment link for Airline on User's request via Twitter


Application for E-Commerce Services

1. Automatic Reply with Brand options for User's Tweet on Twitter
2. Automatic Reply with Product details when User asks on
3. Automatic Arrangement of Cart and Purchase Link based on User's tweets

Application for On Demand information Services

1. Status of server downtime to vendor/employee
2. Status of scheduled meeting to all attendees
3. Contact details of people in organization, etc. on demand

What I am calling here as Virtual Automated Assistance Technology is not a new concept. What I have written is not even a new implementation of this concept. However, just focussing more on specific domain for its implementation separates it from Artificial Intelligence.

The focus area will be a specific domain like ticketing system. Simple set of programs required for my implementation includes communication integration, reading User requests, classification or clustering of User requests and sending appropriate responses.

Implementing these under one application has a capability of radically minimizing the amount of effort spent on support or online communication. It also shows an opportunity to enhance User's experience my many folds.

This can theoretically sound and grow into something like Siri or Google Now or S Voice or Cortana or any other fully implemented Application will be, all I am talking here about is how similar assitance technology can benefit various domains, and may be your business too, and that too, Right Away.

The Future Plans

More communication channels
Current Support is for Twitter but the BabaGyan's implementation of Virtual Automated Assistance Technology will be enhanced to use any other communication channel such as corporate email, gmail, facebook, etc.

More techniques inclusions
BabaGyan's implementation of Virtual Automated Assistance Technology currently uses an analytics engine that works on Text mining algorithms to classify the user's requests.
It currently uses a mix of classification algorithms along with use of Term Document Matrix.
This engine will be enhanced to provide use (with options) of various classification text mining techniques such as Naive Bayes', K-Nearest Neighbors, Support Vector Machine, Artificial Nueral Networks, Decision Tree, Random Forest, etc.
This engine will also be enhanced by implementing Clustering techniques which will handle the unsupervised learning part. These can be Hierarchical, K-Means, K-Medoids, etc.

Inclusion of more advance techniques
BabaGyan's implementation of Virtual Automated Assistance Technology also plans to use Natural Language Processing in future so that the user can freely provide its requests and still get correct responses.
Along with this, speech to text synthesis will be implemented to create a complete Automated Assistance Application.

Related Article on previous Twitter exploration at CLT20 2013 Twitter Analysis of League