b00548078 - sentiment analysis of social media content

Upload: heath-jackson

Post on 01-Jun-2018

222 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/9/2019 B00548078 - Sentiment Analysis of Social Media Content

    1/16

    Sentiment Analysis of

    Social Media ContentStuart Blair

    BEng Hons Software Engineering

    Prof. Maurice MulvennaGroup 10 – Quality Control

    0!1"!"01#

    1

  • 8/9/2019 B00548078 - Sentiment Analysis of Social Media Content

    2/16

    Project Introduction

    • Pro$ect will create an application toclassify an% &a'e use of t(e senti&entpolarity of social &e%ia content

    • E)a&ple uses* – Co&panies can get opinions on pro%ucts

     – Politicians can see if t(ey are +eing

    viewe% favoura+ly• ,elivera+les*

     – Senti&ent -nalysis -pplication

    "

  • 8/9/2019 B00548078 - Sentiment Analysis of Social Media Content

    3/16

    Why Twitter?

    • Glo+al own S/uare

    • 00000000 2sers

    134 of users in 25 • "67 &a'e up 804 of 25 users

    • 804 of users actively post.

    9000000 weets per %ay :831 persecon%;

    • Pu+licly availa+le -P<

    #

  • 8/9/2019 B00548078 - Sentiment Analysis of Social Media Content

    4/16

    Problems with Twitter

    • Paro%y!Hu&our -ccounts

    • "0000000 accounts are fa'e.

    "94 of retweets are %ue to =Please>?

    • ,etection of (u&our sarcas& orirony in weets

    7

  • 8/9/2019 B00548078 - Sentiment Analysis of Social Media Content

    5/16

    Aim

    • -i& – My ai& is to create an application t(at

    can pull a speci@e% nu&+er of tweets

    fro& twitter using a 'eywor% +ase%searc( an% t(en classify t(eir senti&entas positive negative or neutral

  • 8/9/2019 B00548078 - Sentiment Analysis of Social Media Content

    6/16

    Objectives

    • Perfor& a literature review of aca%e&ic $ournalsan% %ocu&ents relate% to senti&ent analysisan% opinion &ining

     ry e)isting senti&ent analysers to get an i%eaof w(at 'in% of perfor&ance an% accuracy t(eyac(ieve

    • Perfor& a co&parison an% analysis of currently

    %eploye% classi@cation algorit(&s use% insenti&ent analysis

    • Create a ro+ust an% co&plete re/uire&entsspeci@cation

    8

  • 8/9/2019 B00548078 - Sentiment Analysis of Social Media Content

    7/16

    iterature !eview

    • >eview $ournals fro& S Paris6Su% (ave pu+lis(e% &anypapers a+out t(e su+$ect area

    • -lso review relevant +oo's*

     –Synt(esis ectures on Hu&ananguage ec(nologies :Bing iu"01";

     – witter -Punning :5evinMa'ice "00; 3

  • 8/9/2019 B00548078 - Sentiment Analysis of Social Media Content

    8/16

    System "evelo#ment ifecycle

    • Propose to use -gile %evelop&ent

    • Engage sta'e(ol%ers constantly

    Pre%icta+le %elivery• -llows for c(ange easily

  • 8/9/2019 B00548078 - Sentiment Analysis of Social Media Content

    9/16

    !e$uirements S#eci%cation

    • Dill +e create% using -gile 2ser Stories

    •  (e syste& &ust +e a+le to strea& realti&e weets +ase% on a 'eywor%

    •  (e syste& s(oul% +e a+le to searc((istorical weets +ase% on a 'eywor%

    •  (e syste& &ust +e a+le to classify t(e

    senti&ent polarity of a corpus•  (e syste& s(oul% allow users to save

    t(eir results

  • 8/9/2019 B00548078 - Sentiment Analysis of Social Media Content

    10/16

    Im#lementation

    • Pro$ect will +e i&ple&ente% in avausing witter7

    • Grap(s will +e provi%e% +y FreeC(art

    • G2< will +e %esigne% using Swing an%-D

    10

  • 8/9/2019 B00548078 - Sentiment Analysis of Social Media Content

    11/16

    System &unctionality

    11

  • 8/9/2019 B00548078 - Sentiment Analysis of Social Media Content

    12/16

    '(I)Prototy#e "esi*n

    1"

  • 8/9/2019 B00548078 - Sentiment Analysis of Social Media Content

    13/16

    Twitter e*al !e$uirements

    • Co&pulsory* – in' to sign up for witter

     – in' to creators account

     – ,ata storage

     – weet %isplay

    • >eco&&en%e%* – witter colours an% fonts

     – 2se of logo

    1#

  • 8/9/2019 B00548078 - Sentiment Analysis of Social Media Content

    14/16

    !is+s

    17

    !is+ Im#orta

    nce

    Penalt

    y

    !is+

    evel

    "escri#tion

     (e senti&entclassi@cationalgorit(& is notaccurate enoug(.

    Me%iu& Me%iu&

    Me%iu& -s long asaccuracyisnt too +a%it will +e @ne.

     (e witter -P<(as cras(e% orgoes un%er&aintenance.

    Hig( Hig( Hig(

  • 8/9/2019 B00548078 - Sentiment Analysis of Social Media Content

    15/16

    Pro*ress to "ate

    • iterature review un%erway

    • >e/uire&ents speci@cation al&ostco&plete

    • Basic prototype create%

    1

  • 8/9/2019 B00548078 - Sentiment Analysis of Social Media Content

    16/16

    ,uestions?

    18