Analisa unjuk kerja sistem penilai esai otomatis berbasis algoritma GLSA (Generalized Latent Sematic Analysis) dan perbandingannya dengan algortima LSA = Performaci analysis of GLSA (Generalized Latent Semantic Analysis) based automated essay garding system compared to LSA based automated essay garnding system
| Main Author: | Henry Artajaya, author |
|---|---|
| Format: | Bachelors |
| Terbitan: |
Fakultas Teknik Universitas Indonesia
, 2012
|
| Subjects: | |
| Online Access: |
http://lib.ui.ac.id/file?file=digital/20307789-S42481-Analisa untuk.pdf |
| ctrlnum |
20307789 |
|---|---|
| fullrecord |
<?xml version="1.0"?>
<dc schemaLocation="http://www.openarchives.org/OAI/2.0/oai_dc/ http://www.openarchives.org/OAI/2.0/oai_dc.xsd"><title>Analisa unjuk kerja sistem penilai esai otomatis berbasis algoritma GLSA (Generalized Latent Sematic Analysis) dan perbandingannya dengan algortima LSA = Performaci analysis of GLSA (Generalized Latent Semantic Analysis) based automated essay garding system compared to LSA based automated essay garnding system</title><creator>Henry Artajaya, author</creator><type>Thesis:Bachelors</type><place/><publisher>Fakultas Teknik Universitas Indonesia</publisher><date>2012</date><description>[<b>ASBTRAK</b><br>
Representasi dokumen sebagai vektor GLSA pada beberapa percobaan seperti uji
sinonim, klasifikasi dokumen, dan clustering terbukti mampu menghasilkan
tingkat akurasi yang lebih baik daripada sistem sejenis yang berbasis algoritma
LSA akan tetapi GLSA belum pernah diujikan pada sistem penilai esay otomatis.
Percobaan ini meneliti pengaruh implementasi GLSA pada sistem penilai esay
otomatis dan perbandingan unjuk kerjanya dengan sistem penilai esay otomatis
berbasis LSA. Unjuk kerja sistem penilai esai otomatis berbasis GLSA lebih
unggul daripada sistem berbasis LSA. Dari 60 kali pengujian, GLSA
menghasilkan nilai yang lebih akurat pada 47 kali pengujian atau 78,3% total
pengujian sedangkan LSA hanya unggul pada 9 kali pengujian atau 15% total
pengujian dan sisanya 4 kali pengujian atau 6,7% total pengujian menghasilkan
nilai dengan tingkat akurasi yang sama. Nilai Pearson Product Moment
Correlation pada percobaan menggunakan sistem LSA 0.57775-0.85868
sedangkan pada GLSA sebesar 0.73335-0.76971. Hal ini mengindikasikan bahwa
sistem berbasis LSA dan GLSA yang diujikan layak pakai karena memiliki
performa yang sama baiknya dengan performa yang dilakukan oleh manusia.
Ditinjau dari waktu proses yang dibutuhkan, LSA unggul pada soal 1 dan 2
dengan rataan 0,07466 detik dan 0,2935 detik sedangkan pada GLSA rataan
waktu proses soal 1 dan 2 sebesar 1,32329 detik dan 17,3641 detik. Waktu proses
yang dibutuhkan sistem penilai esai otomatis berbasis GLSA lebih lama
dibandingkan dengan LSA. Akan tetapi karena GLSA menunjukkan kinerja yang
amat baik, amat dipercaya bahwa manfaatnya lebih besar daripada biaya
komputasi.
<hr>
<b>ABSTRACT</b><br>
Document representation as GLSA vectors were shown to improve performance
on different tasks such as synonymy test, document classification, and clustering
compared to LSA based systems, however GLSA performance has never been
tested on automated essay grading system. This experiment examines the effect of
GLSA implementation on automated essay grading system and evaluates its
performance compared to LSA based system. GLSA performance was shown to
outperform LSA based automated essay grading system. From 60 samples, GLSA
outperform LSA 47 times (78,3%), LSA outperform GLSA 9 times (15%), and 4
times (6,7%) resulted the same score accuracy. Pearson Product Moment
Correlation Value resulted from the experiment using LSA based system is
0.57775-0.85868 and 0.73335-0.76971 for GLSA based system. This result
incidates LSA and GLSA based system used on this experiment are ready to be
used as human rater replacement because both of the system deliver similar
performance with human rater. Processing time of LSA based system is faster
with average processing time consecutively 0,07466 second and 0,2935 second
compared to GLSA consecutively 1,32329 second and 17,3641 second. GLSA
requires more processing time than LSA based system because GLSA based
system has more calculation steps than LSA. However GLSA showed better
performance, therefore it?s believed that its benefits outweigh the computational
cost., Document representation as GLSA vectors were shown to improve performance
on different tasks such as synonymy test, document classification, and clustering
compared to LSA based systems, however GLSA performance has never been
tested on automated essay grading system. This experiment examines the effect of
GLSA implementation on automated essay grading system and evaluates its
performance compared to LSA based system. GLSA performance was shown to
outperform LSA based automated essay grading system. From 60 samples, GLSA
outperform LSA 47 times (78,3%), LSA outperform GLSA 9 times (15%), and 4
times (6,7%) resulted the same score accuracy. Pearson Product Moment
Correlation Value resulted from the experiment using LSA based system is
0.57775-0.85868 and 0.73335-0.76971 for GLSA based system. This result
incidates LSA and GLSA based system used on this experiment are ready to be
used as human rater replacement because both of the system deliver similar
performance with human rater. Processing time of LSA based system is faster
with average processing time consecutively 0,07466 second and 0,2935 second
compared to GLSA consecutively 1,32329 second and 17,3641 second. GLSA
requires more processing time than LSA based system because GLSA based
system has more calculation steps than LSA. However GLSA showed better
performance, therefore it?s believed that its benefits outweigh the computational
cost.]</description><subject>Programming languages (Electronic computers) --Semantics</subject><identifier>20307789</identifier><source>http://lib.ui.ac.id/file?file=digital/20307789-S42481-Analisa untuk.pdf</source><recordID>20307789</recordID></dc>
|
| format |
Thesis:Bachelors Thesis |
| author |
Henry Artajaya, author |
| title |
Analisa unjuk kerja sistem penilai esai otomatis berbasis algoritma GLSA (Generalized Latent Sematic Analysis) dan perbandingannya dengan algortima LSA = Performaci analysis of GLSA (Generalized Latent Semantic Analysis) based automated essay garding system compared to LSA based automated essay garnding system |
| publisher |
Fakultas Teknik Universitas Indonesia |
| publishDate |
2012 |
| topic |
Programming languages (Electronic computers) --Semantics |
| url |
http://lib.ui.ac.id/file?file=digital/20307789-S42481-Analisa untuk.pdf |
| contents |
[<b>ASBTRAK</b><br>
Representasi dokumen sebagai vektor GLSA pada beberapa percobaan seperti uji
sinonim, klasifikasi dokumen, dan clustering terbukti mampu menghasilkan
tingkat akurasi yang lebih baik daripada sistem sejenis yang berbasis algoritma
LSA akan tetapi GLSA belum pernah diujikan pada sistem penilai esay otomatis.
Percobaan ini meneliti pengaruh implementasi GLSA pada sistem penilai esay
otomatis dan perbandingan unjuk kerjanya dengan sistem penilai esay otomatis
berbasis LSA. Unjuk kerja sistem penilai esai otomatis berbasis GLSA lebih
unggul daripada sistem berbasis LSA. Dari 60 kali pengujian, GLSA
menghasilkan nilai yang lebih akurat pada 47 kali pengujian atau 78,3% total
pengujian sedangkan LSA hanya unggul pada 9 kali pengujian atau 15% total
pengujian dan sisanya 4 kali pengujian atau 6,7% total pengujian menghasilkan
nilai dengan tingkat akurasi yang sama. Nilai Pearson Product Moment
Correlation pada percobaan menggunakan sistem LSA 0.57775-0.85868
sedangkan pada GLSA sebesar 0.73335-0.76971. Hal ini mengindikasikan bahwa
sistem berbasis LSA dan GLSA yang diujikan layak pakai karena memiliki
performa yang sama baiknya dengan performa yang dilakukan oleh manusia.
Ditinjau dari waktu proses yang dibutuhkan, LSA unggul pada soal 1 dan 2
dengan rataan 0,07466 detik dan 0,2935 detik sedangkan pada GLSA rataan
waktu proses soal 1 dan 2 sebesar 1,32329 detik dan 17,3641 detik. Waktu proses
yang dibutuhkan sistem penilai esai otomatis berbasis GLSA lebih lama
dibandingkan dengan LSA. Akan tetapi karena GLSA menunjukkan kinerja yang
amat baik, amat dipercaya bahwa manfaatnya lebih besar daripada biaya
komputasi.
<hr>
<b>ABSTRACT</b><br>
Document representation as GLSA vectors were shown to improve performance
on different tasks such as synonymy test, document classification, and clustering
compared to LSA based systems, however GLSA performance has never been
tested on automated essay grading system. This experiment examines the effect of
GLSA implementation on automated essay grading system and evaluates its
performance compared to LSA based system. GLSA performance was shown to
outperform LSA based automated essay grading system. From 60 samples, GLSA
outperform LSA 47 times (78,3%), LSA outperform GLSA 9 times (15%), and 4
times (6,7%) resulted the same score accuracy. Pearson Product Moment
Correlation Value resulted from the experiment using LSA based system is
0.57775-0.85868 and 0.73335-0.76971 for GLSA based system. This result
incidates LSA and GLSA based system used on this experiment are ready to be
used as human rater replacement because both of the system deliver similar
performance with human rater. Processing time of LSA based system is faster
with average processing time consecutively 0,07466 second and 0,2935 second
compared to GLSA consecutively 1,32329 second and 17,3641 second. GLSA
requires more processing time than LSA based system because GLSA based
system has more calculation steps than LSA. However GLSA showed better
performance, therefore it?s believed that its benefits outweigh the computational
cost., Document representation as GLSA vectors were shown to improve performance
on different tasks such as synonymy test, document classification, and clustering
compared to LSA based systems, however GLSA performance has never been
tested on automated essay grading system. This experiment examines the effect of
GLSA implementation on automated essay grading system and evaluates its
performance compared to LSA based system. GLSA performance was shown to
outperform LSA based automated essay grading system. From 60 samples, GLSA
outperform LSA 47 times (78,3%), LSA outperform GLSA 9 times (15%), and 4
times (6,7%) resulted the same score accuracy. Pearson Product Moment
Correlation Value resulted from the experiment using LSA based system is
0.57775-0.85868 and 0.73335-0.76971 for GLSA based system. This result
incidates LSA and GLSA based system used on this experiment are ready to be
used as human rater replacement because both of the system deliver similar
performance with human rater. Processing time of LSA based system is faster
with average processing time consecutively 0,07466 second and 0,2935 second
compared to GLSA consecutively 1,32329 second and 17,3641 second. GLSA
requires more processing time than LSA based system because GLSA based
system has more calculation steps than LSA. However GLSA showed better
performance, therefore it?s believed that its benefits outweigh the computational
cost.] |
| id |
IOS18064.20307789 |
| institution |
Universitas Indonesia |
| institution_id |
51 |
| institution_type |
library:university library |
| library |
Perpustakaan Universitas Indonesia |
| library_id |
492 |
| collection |
Repository Skripsi (open) Universitas Indonesia |
| repository_id |
18064 |
| city |
KOTA DEPOK |
| province |
JAWA BARAT |
| repoId |
IOS18064 |
| first_indexed |
2022-12-13T08:49:18Z |
| last_indexed |
2022-12-13T08:49:18Z |
| recordtype |
dc |
| merged_child_boolean |
1 |
| _version_ |
1752181564481994753 |
| score |
17.13294 |