Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
文献紹介 ( 02/18/2016 )
Search
yoshii25
February 18, 2016
0
190
文献紹介 ( 02/18/2016 )
yoshii25
February 18, 2016
Tweet
Share
More Decks by yoshii25
See All by yoshii25
文献紹介 (05/13/2016)
yoshii25
0
120
文献紹介 ( 2015年7月8日(水) )
yoshii25
0
180
文献紹介 ( 2015年6月10日(水) )
yoshii25
0
190
文献紹介 ( 2015年5月12日(火) )
yoshii25
0
190
文献紹介 ( 2015年4月1日(水) )
yoshii25
0
220
B3 勉強会 第12回目 ( 2015年3月26日(木) )
yoshii25
0
180
B3 勉強会 第10回目 ( 2015年3月13日(金) )
yoshii25
0
180
B3 勉強会 第8回目 ( 2015年2月20日(金) )
yoshii25
0
210
B3 勉強会 第6回目 ( 2015年2月10日(火) )
yoshii25
0
220
Featured
See All Featured
Crafting Experiences
bethany
1
48
We Analyzed 250 Million AI Search Results: Here's What I Found
joshbly
1
700
Data-driven link building: lessons from a $708K investment (BrightonSEO talk)
szymonslowik
1
910
Docker and Python
trallard
47
3.7k
Self-Hosted WebAssembly Runtime for Runtime-Neutral Checkpoint/Restore in Edge–Cloud Continuum
chikuwait
0
320
svc-hook: hooking system calls on ARM64 by binary rewriting
retrage
1
99
Designing for humans not robots
tammielis
254
26k
Marketing to machines
jonoalderson
1
4.6k
Sam Torres - BigQuery for SEOs
techseoconnect
PRO
0
180
Joys of Absence: A Defence of Solitary Play
codingconduct
1
290
Let's Do A Bunch of Simple Stuff to Make Websites Faster
chriscoyier
508
140k
Being A Developer After 40
akosma
91
590k
Transcript
ழມ ܚथ Ԭٕज़Պֶେֶ ࣗવݴޠॲཧݚڀࣨ จݙհ(2016/02/18)
հจݙ Arti Ramesh, Shachi H. Kumar, James Foulds, and Lise
Getoor. 2015. Weakly Supervised Models of Aspect-Sentiment for Online Course Discussion Forums. In Proceed- ings of the Annual Meeting of the Association for Computational Linguistics (ACL). 2
֓ཁ • هࣄͷτϐοΫऀͷײΛಛఆ͢Δ͜ͱʹ ΑΓɺγεςϜͷͷରॲੜెͷؔ৺ͷ ্ʹཱͭ. • طଘͷϞσϧʹϧʔϧΛՃ͑Δ͜ͱʹΑΓɺ͞ ΒͳΔهࣄͷྨ͕Մೳʹͳͬͨ. 3
MOOCs • Massive Open Online Courses. • Πϯλʔωοτ্ͰडߨͰ͖ΔڭҭγεςϜ. • ੈքதͰར༻͞Ε͓ͯΓɺඇৗʹଟ͘ͷίʔε͕
ଘࡏ͢Δ. • ࢀՃऀಉ࢜ͰͷٞΛߦ͏͜ͱ͕ग़དྷΔ. 4
MOOC aspects 5 Table2 coarse and fine aspects
Seeded LDA Model • τϐοΫɾײਪఆͷͨΊʹ༻͍ΒΕΔϞσϧ. • LDA ͷҰछͰ͋ΓɺγʔυͱݺΕΔਪఆͷࢦ ඪͱͳΔ୯ޠΛख͕͔ΓʹਪఆΛߦ͏. •
ࠓճਓखͰͷใ༩Λߦ͍ɺγʔυͷσʔλ ηοτΛ४උͨ͠ɻ 6
σʔληοτ • ΫϥυιʔγϯάΛ༻͍ͯɺ୯ޠʹใΛ༩ ͢Δ. • ຊ࣮ݧͰ Crowdflower Λ༻. • ֤هࣄʹগͳ͘ͱ
3 ਓ͕࡞ۀʹͨΔ. • ৴པΛܭࢉ͠ɺ৴པੑ͕͍ใ࠾༻͠ͳ ͍. 7
γʔυ 8 Table3 Seed words for coarse aspects Table4 Seed
words for fine aspects
γʔυ Table5 Seed words for sentiment Table6 Seed words for
sentiment specific to online courses 9
࣮ݧ • NLTK toolkit ʹΑΓɺهࣄͷτʔΫϯԽͱες ϜΛߦ͏. • The topic model
Dirichlet hyper-parameter. • α = 0.01, β = 0.01. • Table 3, 4, 5 ͷγʔυͰɺτϐοΫͱಉ͡ ͷγʔυτϐοΫΛ༻. • Table 6, 3 ͷγʔυͰɺ10 τϐοΫΛ༻. 10
݁Ռ 11
݁Ռ 12
ग़ྗ 13