The author of this document has limited its availability to on-campus or logged-in CSUSB users only.

Off-campus CSUSB users: To download restricted items, please log in to our proxy server with your MyCoyote username and password.

Date of Award


Document Type

Restricted Project: Campus only access

Degree Name

Master of Science in Computer Science


School of Computer Science and Engineering

First Reader/Committee Chair

Gomez, Ernesto


A chatbot is software which simulates a human being in a conversation. This project involved developing a text-based general conversation chatbot. The chatbot is based on neural network technology where the neural network is trained using conversations from classic films. A Chatbot refers to software that chats. It is indeed a modern technology that simulates conversation. The interaction with any customer is really all about it.

It answers the questions asked by the user. Modeling verbal exchange is a vital task in natural language processing and synthetic intelligence (AI). A neural network is a type of machine studying algorithm that is stimulated by means of the behavior of organic neurons within the mind. Deep neural networks can be trained to do greater tasks than a mapping of complicated capabilities to complex functions. An example of this mapping is the series to series mapping completed in language translation. This same mapping can be implemented to conversational retailers. They consist of various layers for analyzing and studying statistics. This paper shows the avenues of making computer systems to learn user language in a textual content by developing a chat bot. A chatbot allows a person to in reality ask questions inside the equal way that they might cope with a human.

Resources like those of search engine results and individual agents lack individuality, shown by their technical approval of input and output creation, and practicality shown by a single form of input, such as keywords, being brought in.

In addition to a new paradigm of GCA, this paper introduces a new form of adversarial training for Generative Conversational Agents (GCA). Our method assumes the GCA as a generator which intends to fool a discriminator who marks dialogs as human-generated or device-generated. The discriminator carries out a token-level (word-level) classification in our method like that of it shows if the current token was created by humans or devices. To use it, the discriminator often gets the background statements (the existence of the dialog) and the inadequate answer as feedback up to the current token. This latest statement makes end-to-end learning accessible due to back propagation.

The trained discriminator could be used to select the best answer amongst the responses created by the various trained models to further boost performance. A self-communication process allows for the adversarial preparation to create a collection of generated data with even more diversity. This method increases the efficiency of non-training information issues related. Experimental tests of assessments of humans and adversaries indicate that perhaps the adversarial approach produces substantial efficiency improvements throughout the normal curriculum pressuring learners.