/
Members: Tom  Amon Aaron Members: Tom  Amon Aaron

Members: Tom Amon Aaron - PowerPoint Presentation

luanne-stotts
luanne-stotts . @luanne-stotts
Follow
346 views
Uploaded On 2018-11-09

Members: Tom Amon Aaron - PPT Presentation

Campbell Daniel Fitzick Shishir Kharel Hoa Vu Nanxi Zhang Stakeholder Marcello Balduccini 52317 CCI Class of 2017 Senior Design Project Drexel Chatbot 1 Introduction amp Features ID: 724551

design drexel senior project drexel design project senior 2017 class chatbot 17cci cci amp generic answer questions question database

Share:

Link:

Embed:

Download Presentation from below link

Download Presentation The PPT/PDF document "Members: Tom Amon Aaron" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.


Presentation Transcript

Slide1

Members:

Tom

Amon

Aaron

Campbell

Daniel FitzickShishir KharelHoa VuNanxi Zhang

Stakeholder:Marcello Balduccini

5/23/17

CCI Class of 2017 Senior Design Project - Drexel Chatbot

1Slide2

Introduction

& Features

Overview

of Algorithm

System Architecture & DesignManagementDemo

Limitations & Technical ChallengesDrexel chatbot5/23/17CCI Class of 2017 Senior Design Project - Drexel Chatbot

2Slide3

Introduction & Features

5/23/17CCI Class of 2017 Senior Design Project - Drexel Chatbot

3Slide4

What is

drexel

chatbot

Question answering system for DrexelRESTful APIDemo applications (Web, Mobile and SMS)

5/23/17CCI Class of 2017 Senior Design Project - Drexel Chatbot4Slide5

Why drexel chatbot

Compared to Google:

More efficientMore user friendlyGets results fasterCompared to Google Assistant/Siri:Optimized for Drexel

More information about Drexel5/23/17CCI Class of 2017 Senior Design Project - Drexel Chatbot

5Slide6

Why drexel chatbot

Why API

Expandable to other applicationsFriendly to other developersResearch projectFocus on the process

5/23/17CCI Class of 2017 Senior Design Project - Drexel Chatbot

6Slide7

features

Input and output in natural language

APIInput by URL parametersOutput formatted in JSONConcurrent usersDemo applications

Simple interfacesEnter and send questions; get responses5/23/17

CCI Class of 2017 Senior Design Project - Drexel Chatbot7Slide8

Supported questions

Faculty

EmailWebsiteTitleOfficePhone number

PictureDepartmentEducationPublicationsResearch interests5/23/17

CCI Class of 2017 Senior Design Project - Drexel Chatbot

8

BuildingsAddress SchedulePictureWebsiteSlide9

Overview of algorithm

5/23/17CCI Class of 2017 Senior Design Project - Drexel Chatbot

9Slide10

inspiration

Google’s “A neural conversational model” (

2015)I am seeing an error related to VPNWhat

is the error message that you are getting when connecting to VPN using NetworkConnect ?

Connection refused May I know the version of NetworkConnect

you are using?Wheres Marcello Balduccini office?Marcello Balduccini office is Rush 233C.

5/23/17CCI Class of 2017 Senior Design Project - Drexel Chatbot10Slide11

5/23/17

CCI Class of 2017 Senior Design Project - Drexel Chatbot

11Slide12

5/23/17

CCI Class of 2017 Senior Design Project - Drexel Chatbot

12Slide13

5/23/17

CCI Class of 2017 Senior Design Project - Drexel Chatbot

13Slide14

5/23/17

CCI Class of 2017 Senior Design Project - Drexel Chatbot

14Slide15

Algorithm design rationale

Less training data

ModularAdaptable5/23/17

CCI Class of 2017 Senior Design Project - Drexel Chatbot15Slide16

System architecture & design

5/23/17CCI Class of 2017 Senior Design Project - Drexel Chatbot

16Slide17

5/23/17

CCI Class of 2017 Senior Design Project - Drexel Chatbot17Slide18

Main languages

Frontend

Java

Java Script

HTML &

CSSAPI: JavaBackend:

Python5/23/17CCI Class of 2017 Senior Design Project - Drexel Chatbot18Slide19

API

Basic Model-View-Controller

Calls backend on each GET requestEach requests gets separate process

5/23/17CCI Class of 2017 Senior Design Project - Drexel Chatbot

19Slide20

Generic question construction

Generalize” stepNatural Language Processing

Identify keywordsPart of speech taggingDatabase lookupReplace keywords with generic form

5/23/17

CCI Class of 2017 Senior Design Project - Drexel Chatbot20Slide21

Generic answer construction

“Classify” step

Neural network5/23/17

CCI Class of 2017 Senior Design Project - Drexel Chatbot21Slide22

Generic answer construction

18

answer categories~200 training questions~7000 training question after adding noise (misspelling)

5/23/17

CCI Class of 2017 Senior Design Project - Drexel Chatbot

22Slide23

Generic answer population

Simplest component

Query the databaseReplace generic form5/23/17

CCI Class of 2017 Senior Design Project - Drexel Chatbot23Slide24

database

RDF Database

No tables or keys

ExtensibilityOne-to-many

5/23/17CCI Class of 2017 Senior Design Project - Drexel Chatbot

24Slide25

Information extraction

Populates the database

20 scriptsCreates TTL(turtle) files<#Denise-Agosto>

   cb:name "Denise Agosto" ;    

cb:room "Rush 214B" ;    cb:email

"dea22@drexel.edu" ; 5/23/17CCI Class of 2017 Senior Design Project - Drexel Chatbot25Slide26

Error handler

Allows failing more gracefully

2 types of errors“I don’t understand the question”“Sorry I don't have an answer for that. But I found this website: http:...”

5/23/17CCI Class of 2017 Senior Design Project - Drexel Chatbot

26Slide27

Management

5/23/17CCI Class of 2017 Senior Design Project - Drexel Chatbot

27Slide28

5/23/17

CCI Class of 2017 Senior Design Project - Drexel Chatbot

28

CommunicationSlide29

Planning

5/23/17

CCI Class of 2017 Senior Design Project - Drexel Chatbot

29Slide30

5/23/17

CCI Class of 2017 Senior Design Project - Drexel Chatbot

30

Repository & bug trackingSlide31

testing

Unit

testingSystem testingUser testing

5/23/17

CCI Class of 2017 Senior Design Project - Drexel Chatbot31Slide32

Demo

5/23/17CCI Class of 2017 Senior Design Project - Drexel Chatbot

32Slide33

Limitations & technical challenges

5/23/17CCI Class of 2017 Senior Design Project - Drexel Chatbot

33Slide34

limitations

Input question

No compound

Context-unaware

Keyword-dependentNLTKM

isidentification5/23/17CCI Class of 2017 Senior Design Project - Drexel Chatbot34Slide35

Technical challenges

Database speed

Neural network speedLack of data

5/23/17CCI Class of 2017 Senior Design Project - Drexel Chatbot

35Slide36

conclusion

5/23/17CCI Class of 2017 Senior Design Project - Drexel Chatbot

36Slide37

Questions

5/23/17CCI Class of 2017 Senior Design Project - Drexel Chatbot

37Slide38

Thank you

Aaron, Daniel,

Hoa, Nanxi,

Shishir, Tom, and Bun5/23/17

CCI Class of 2017 Senior Design Project - Drexel Chatbot38