a human behaviour, psychology and mind to create human-like conversation. It
is also important, because no one wants to continue a conversation with a dead-
end robot[6]. Even if we talk about one of the first realistic-bot, SIRI which is
developed by Apple has a vast range of features helping users to manage calendar,
make calls, schedule meetings and what not. However, if you have ever tried to talk
more interactively, it fails to understand and most frequently replies “I am sorry
I couldn’ t understand. Could you try again?”. The most significant drawback of
a chatbot is, it lacks to understand the specific context of the sentences as well as
the right sentiment that needs to be conveyed in response. Usually, it identifies the
most similar keywords and responds to the closely matched text from its database.
Eventually, most of the replies become inconsistent and meaningless. Furthermore,
using a bot in the psychological field needs to be even more human-like giving the
vibe of human-connection and empathy[6]. All these are only possible if the virtual
assistant is able to get the right sentiment, emotion and intent of the person and
clearly have an overall understanding of the user’s situation. Once this problem
can be solved, it can be used in every sector for personalised conversations and
this information can be very helpful to the organizations whether to help a mental
health patient or to improve various services and products sold. Even universities
can use it to know the exact problem that a student is going through in a particular
course that he failed. Therefore, through various research, we are trying to develop
a model that will analyse a text and understand it just like a human so that making
responses can be ever easier and accurate.
1.3 Aim to study
Analysing the various features that are required by a chatbot to gain a human-like
understanding of text is the aim of our research. We are trying to develop a hybrid
model that will understand the sentiment, emotion, named-entity and intent of the
user’s text and also will be able to identify the correct entities mentioned. We chose
particularly deep learning because DL models help to extract the meaning of mixed
contrary complex sentences given by a user. The layers in DL models creates a net-
work that helps bots to learn accurately on their own which normal ML bots fail to
do[8]. It is easy to generate a dialog without having a sense of it, even we humans do
the same. At times humans speak out without knowing the reason behind the for-
mer person’s opinion or statement. These kinds of conversation in real life itself are
not healthy and should not be encouraged. Hence, to replicate an idealistic human
nature, we need to build the chatbot with knowledge, consistency and empathy [9]
so that users can engage in interactive conversations and find interest to share their
views, emotions or motives. Our goal is to implement understanding capabilities in
a chatbot which can be used for developing a human-like bot.
3