You are on page 1of 3

Basic Plan of

Intelligent Bengali Social Virtual Assistant (IBSVA)


Abstract: An intelligent virtual assistant is a software-based agent. It performs tasks from verbal
commands. There are a lot of virtual assistants of different purposes and some of them perform with great
efficiency. However, in the modern age, still, the performance of virtual assistants mostly suffers from
different kinds of noise and lack of intelligence on a subject base. Our main goal is to provide an
innovative software Bengali social agent. Which is capable to identify the corresponding speaker and
his/her language (especially Bangla). Moreover, as an additional feature, the proposed system will
analyze the speaker’s sentiment during the conversation and provide a cognitive response, like as human
though. It will have the ability to train up himself day by day according to communication with people.

Introduction: Every human being always wants a friend/partner with whom he can share everything or at
least can talk when he needs. It would be very much better if that particular friend could maintain privacy
and also could do some counseling understanding human mind when it’s necessary. Considering these
particular things, we are going to design/program a virtual assistant.

Virtual assistants have become a common part of modern life. They have made life easier and make
happy. Different technical companies have developed various types of standalone virtual assistants and
also many built-in virtual assistants for smartphones and computers. An intelligent virtual assistant means
this can adapt the behavior of the user and act based on these data.

Google Assistant (Allo and Google Home), Alexa, Siri or Cortana could be an example for an intelligent
assistant, but they have some specific task. Though they have not yet been able to adapt to complete
human nature in question answering, they are trying to improve further.

On the other hand a Social Assistant or Robot capable to communicate with human and robot. Typically,
it works based on cognitive computing model that simulates human thought. The example of a social
robot is Sophia, it can display more than 50 facial expressions. Another one is Pepper, It’s design to real
emotion. Nao is another social robot with can used in Academic and scientific research. It’s hase more
update version, etc

A few things we do not see in intelligent social assistants’ ability to analyze sentiment. Another notable
mark is all of the assistant does not support Bangla Language. That is why our proposed to make a Social
Assistant.

This social assistant will have the capability to detect our native language (Bengali) and keep up a normal
conversation with the users in Bengali like a social robot or a human. It will also be able to make a
decision based on sentiment analysis. Detecting sentiment from the speech during communication with
corresponding speaker.

Technically our virtual assistant will listen to user any time he/she needs to. Even it will talk with him and
ask questions whenever it is necessary. Our assistant will try to know for what that particular user likes to
eat, what color dress he/she wears even it will try to talk with that particular user as his/her best friend. By
continuing conversations, our assistant will turn itself as user’s day to day partner at some point. User can
share all his/her feelings and thoughts with our assistant and for example- if some day user’s mood is not
good or if he goes through tough situation our assistant would provide him counseling by doing
sentimental analysis based on their previous conversation. It would try to provide condolences to the user
suggest him something which will make user feel better eventually. Besides this , our assistant will also
try to remind user if he had any important meeting or schedule if user wants it to do so. In a word, our
assistant will become user’s personal virtual assistant, friend and counselor. We will look forward to
increase the features of virtual assistant in the near future.

Motivation: There are various types of virtual assistants such as Siri, Alexa, and Google Assistant. These
virtual assistants only support multiple popular languages and also can perform various tasks such as
informing about weather, date, set a reminder or alarm, calling, messaging, controlling IoT based devices
and many more. But these featured and not available for the Bengali language or neither they can detect
sentiment or emotion in this language. Even though those virtual assistants can take real-time speech as
short commands but the real-time long term conversation with efficient noise elimination is yet to be
seen.

Objectives: Our objectives are:

 Eliminate background noise.


 Identify speaker so that the system will only listen to his voice.
 Convert uttered speech to text (Automatic Speech Recognition).
 Detect the assistance the user needed and provide a solution as a conventional virtual assistant.
 From converted text and voice tone after the interview, the system will get the emotional state of
the user.
 The decision making capability and continual self-updating knowledge-base.

Proposed Solution:

Figure 1.a is illustrating the proposed stepwise solution for the intelligent social virtual assistant. First of
all, the user will provide voice input to the assistant; then our voice bot will transfer the noisy speech
signal to the noise suppression phase. After the noise suppression phase, the signal will be analyzed into
two groups: (i) Automatic Speech Recognition and (ii) Sentiment Analysis. Then finally from his
knowledge-base the assistant will make a decision based on the user’s behavior.

1. Noise reduction.
2. Speech to text.
3. Convert already existing virtual assists into Bangla.
4. Our assistant File management.
Figure 1.a: Workflow of the proposed system

Conclusion: This proposed system will work as a regular social virtual assistant. The system will mine
the user’s conversation with the system itself as well as will suggest the counter-measures or relevant
counselors to the user.

You might also like