Scribd
Upload
70 views1 page

02 - 16 - Chitty Chat Bot Paper Summary

This research summary discusses methodologies for developing conversational AI/chatbots, including retrieval based and generation based response systems, as well as systems that combine both approaches. It also addresses how to create context-based responses using previous utterances/conversations, generate diverse responses, and enable proactive conversations. However, the paper does not present any new methods or findings, but rather surveys existing techniques discussed in other papers. It aims to provide an overview of technologies used for conversational agents.

Uploaded by

Nayeem Uzzaman
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
70 views1 page

02 - 16 - Chitty Chat Bot Paper Summary

This research summary discusses methodologies for developing conversational AI/chatbots, including retrieval based and generation based response systems, as well as systems that combine both approaches. It also addresses how to create context-based responses using previous utterances/conversations, generate diverse responses, and enable proactive conversations. However, the paper does not present any new methods or findings, but rather surveys existing techniques discussed in other papers. It aims to provide an overview of technologies used for conversational agents.

Uploaded by

Nayeem Uzzaman
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

Research Summary: “Chitty-Chitty-Chat Bot”: Deep

Learning for Conversational AI


Nayeemuzzaman Farhan Mahmud
Computer Science and Engineering Computer Science and Engineering
University of Dhaka University of Dhaka
Dhaka, Bangladesh Dhaka, Bangladesh
[email protected] [email protected]

I. A BSTRACT 2) Multi-turn Conversation: Context is important to multi-


turn conversations. In creating a response model, adding pre-
This paper highlights data formulation and collection as well
vious utterance weights into it has been stated in the paper.If
as methodologies of response creation of Conversational AI.
the conversation is too long, context segmentation is suggested
II. I NTRODUCTION to be used to keep the relevance.

This paper divides the response methodologies for chatbots C. Diverse Responses
into 3 parts. They are retrieve based conversational sys- Normal models tend to give similar responses to a high portion
tems,generation based conversational systems and system en- of the query domain.To solve this problem it was suggested
semble which is a combination of both. Author also discusses that we add some diverse factors in training our model so that
how these methods can be used in creating response with the candidate response generated from similar queries can be
context,creating diverse and proactive responses diverse as much as possible.

III. OVERVIEW D. Proactive Converstaions


Proactive conversation leads to a much better experience
A. Response Methodologies
conversing with CAs. It can continue the conversation when
1) Retrieval Based Conversational System: One of the ways it comes to a halt. One way is to add proactive suggestions to
of creating responses to a user query is using retrieve based response retrieval.
systems. Here the system searches for candidate responses
by matching metrics. The retrieval system learns the metrics IV. M ETHODOLOGY
using deep learning techniques. A repository is used to store Mainly data from other research papers were used to present
’responses’ for different situations which introduces additional a general survey of methodologies for chatbots. The method
overhead. for giving training data mainly involves deep learning and
2) Generation Based Conversational System: Another representation learning. Millions of conversations found on
method of creating a response from user utterance is using social media and forums are used for training data. The paper
generate based systems. Here a model built using deep learning doesn’t any new methods of its own.
techniques and trained using some dataset to generate a V. R ESULTS
response from a user query. As we are not using any data
repository, the response is not guaranteed to be meaningful. This paper presents a summary of methodologies for conver-
But it saves from the overhead of searching for a response sational AI. Reading this paper gives an idea of what type
from a repository. of technologies are prevalent in this field. Minimal numerical
data were used. Also, it doesn’t present any new findings.
3) System Ensemble: Both methods. Candidate responses
are merged into a list from both retrieve based system and VI. D ISCUSSION
generate based system and then the best outcome is selected This paper can give the reader a good idea of what technolo-
as the response for the user.. gies and methodologies are used for conversational chatbots.
But it doesn’t present with any new ideas of its own or does
B. Context-Based Response
not give any new findings. As such, it lacks originality.
1) Single-turn Conversation: Single turn conversations use
only the current query to utter a response. But most real R EFERENCES
life conversations are contextual and the appropriate response Yan, R. (2018, July). ” Chitty-Chitty-Chat Bot”: Deep Learn-
depends also on the previous occurrences. So, single-turn con- ing for Conversational AI. In IJCAI (Vol. 18, pp. 5520-5526).
versation is not good enough to model real life conversations.

You might also like