Multilingual Sentiment Analysis Using Latent Semantic Indexing and Machine Learning faces smoke angry his five anger kings news laughter months crown scare man sting angel fallen fun paradise Brett Bader, Digital Globe, [email protected]Philip Kegelmeyer, Sandia National Laboratories, [email protected]Peter Chew, Galisteo Consulting Group Inc, [email protected]Sandia is a multiprogram laboratory operated by Sandia Corporation, a Lockheed Martin Company, for the United States Department of Energys National Nuclear Security Administration under contract DE-AC04-94AL85000. SENTIRE, December 11, 2011
26
Embed
Multilingual Sentiment Analysis Using Latent Semantic Indexing and ...
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Multilingual Sentiment Analysis
Using Latent Semantic Indexing and Machine Learningfaces!
Psalm 126:2–4Then was our mouth filled withlaughter, and our tongue withsinging: then said they amongthe heathen, The LORD hathdone great things for them. TheLORD hath done great things forus; whereof we are glad. Turnagain our captivity, O LORD, asthe streams in the south.
Revelation 9:18–19By these three was the third partof men killed, by the fire, and bythe smoke, and by the brimstone,which issued out of their mouths.For their power is in their mouth,and in their tails: for their tailswere like unto serpents, and hadheads, and with them they dohurt.
Salmos 126:2–4Entonces nuestra boca sehenchira de risa, Y nuestralengua de alabanza; Entoncesdiran entre las gentes: Grandescosas ha hecho Jehova con estos.Grandes cosas ha hecho Jehovacon nosotros; Estaremos alegres.Haz volver nuestra cautividad ohJehova, Como los arroyos en elaustro.
Apocalipsis 9:18–19De estas tres plagas fue muerta latercera parte de los hombres: delfuego, y del humo, y del azufre,que salan de la boca de ellos.Porque su poder esta en su bocay en sus colas: porque sus colaseran semejantes serpientes, ytenıan cabezas, y con ellas danan.
Revelation 9:1–12 (a demonic plague of locusts) scored positive.
1 The fifth angel sounded his trumpet , and I saw a star that had fallen from the sky to the earth .
The star was given the key to the shaft of the Abyss. 2 When he opened the Abyss, smoke rose from it
like the smoke from a gigantic furnace. The sun and sky were darkened by the smoke from the Abyss. 3
And out of the smoke locusts came down upon the earth and were given power like that of scorpions of
the earth . 4 They were told not to harm the grass of the earth or any plant or tree , but only those
people who did not have the seal of God on their foreheads. 5 They were not given power to kill them,
but only to torture them for five months. And the agony they suffered was like that of the sting of a
scorpion when it strikes a man. 6 During those days men will seek death , but will not find it; they will long
to die, but death will elude them.
7 The locusts looked like horses prepared for battle. On their heads they wore something like crowns of
gold , and their faces resembled human faces. 8 Their hair was like women’s hair, and their teeth were like
lions’ teeth. 9 They had breastplates like breastplates of iron , and the sound of their wings was like thethundering of many horses and chariots rushing into battle. 10 They had tails and stings like scorpions, and
in their tails they had power to torment people for five months. 11 They had as king over them the
angel of the Abyss, whose name in Hebrew is Abaddon, and in Greek, Apollyon.
12 The first woe is past; two other woes are yet to come. . . .
• We have a demonstrated a supervised machine learning approach todetermine sentiment in multilingual documents.
– Does not require translation
– Uses a sentiment lexicon only for bootstrapping sentiment labels
– Uses LSA to project documents into a language-independent space.
– Uses machine learning on these features to build a predictive model
Extensions:
– Could easily be used with other topic models, such as LDA or NMF.
– Could be applied to other emotional dimensions or meta-properties,such as “framing language”; prior similar application has been seenin characterizing ideology[8] in multilingual text.
References[1] S. Deerwester, “Improving Information Retrieval with Latent Semantic Indexing,” in Proceedings of the 51st
ASIS Annual Meeting (ASIS ’88), C. L. Borgman and E. Y. H. Pai, Eds., vol. 25. Atlanta, Georgia:American Society for Information Science, Oct. 1988.
[2] P. A. Chew, B. W. Bader, S. Helmreich, A. Abdelali, and S. J. Verzi, “An information-theoretic,vector-space model approach to cross-language information retrieval,” Journal of Natural Language
Engineering, 2010.[3] B. Bader and P. Chew, Text Mining: Applications and Theory. Wiley, 2010, ch. Algebraic Techniques for
Multilingual Document Clustering.[4] M. M. Bradley and P. J. Lang, “Affective norms for English words (ANEW): Instruction manual and
affective ratings,” Technical Report C-1, The Center for Research in Psychophysiology University, 1999.[5] K. Denecke, “Using SentiWordNet for multilingual sentiment analysis,” in ICDE Workshops. IEEE
Computer Society, 2008, pp. 507–512.[6] R. Mihalcea, C. Banea, and J. Wiebe, “Learning multilingual subjective language via cross-lingual
projections,” in Proceedings of the 45th Annual Meeting of the Association of Computational Linguistics, 2007, pp.976–983.
[7] A.-L. Ginsca, E. Boros, A. Iftene, D. Trandabat, M. Toader, M. Corici, C.-A. Perez, and D. Cristea,“Sentimatrix — multilingual sentiment analysis service,” in Proceedings of the 2nd Workshop on Computational
Approaches to Subjectivity and Sentiment Analysis (WASSA 2.011). Portland, Oregon: Association forComputational Linguistics, June 2011, pp. 189–195.
[8] P. Chew, P. Kegelmeyer, B. Bader, and A. Abdelali, “The knowledge of good and evil: Multilingualideology classification with PARAFAC2 and machine learning,” Language Forum, vol. 34, no. 1, pp. 37–52,2008.
[9] B. Bader and P. Chew, “Enhancing multilingual latent semantic analysis with term alignmentinformation,” in COLING 2008, 2008.
[10] P. Chew and A. Abdelali, “Benefits of the massively parallel Rosetta Stone: Cross-language informationretrieval with over 30 languages,” in Proceedings of the Association for Computational Linguistics, 2007, pp.872–879.