自然言語処理25_527

Size: px

Start display at page:

Download "自然言語処理25_527"

えみみうら
4 years ago
Views:

1 Between Reading Time and the Information Status of Noun Phrases Masayuki Asahara Japanese noun phrases are not marked by articles. The information status of Japanese noun phrases is not overt. Due to limited contextual information and world knowledge, it is difficult to estimate the information status, which is analyzed through the given/new status or indefinite/definite status. However, in Japanese language processing, the notion of the information status is yet to be understood. In this paper, we explain the information status of Japanese noun phrases. Then, we explore how the information status of Japanese noun phrases is estimated through the reading time. As a first step, we investigate the correlation between reading time and the information status of Japanese noun phrases. The statistical evaluation shows that readers information status affects reading time in Japanese. Key Words: Information Status, Reading Time, Annotation, Bayesian Linear Mixed Model 1, NINJAL, Japan

2 Vol. 25 No. 5 December 2018 (information status) (commonness) 2 discourse-old discourse-new hearer-old bridging hearer-new information status (definiteness) (specificity) (animacy) (sentience) (agentivity) ( 2016) 528

3 Götze (given)/ (accessible)/ (new) (aboutness/frame setting) (new-information focus)/ (contrastive focus) (Götze, Weskott, Endriss, Fiedler, Hinterwimmer, Petrova, Schwarz, Skopeteas, and Stoel 2007) Prasad Penn Discourse Treebank (Prasad, Dinesh, Lee, Miltsakaki, Robaldo, Joshi, and Webber 2008) PropBank (Palmer, Gildea, and Kingsbury 2005) (Prasad, Webber, Lee, Pradhan, and Joshi 2015) ( 2016) ( 2013) Dundee Eye Tracking Corpus (Kennedy and Pynte 2005) 10 Demberg Dundee Eye Tracking Corpus Gibson dependency locality theory (DLT)(Gibson 2008) Hale surprisal theory (Hale 2001) (Demberg and Keller 2008) Barret (Barrett, Bingel, Keller, and Søgaard 2016) Klerke (Klerke, Martínez Alonso, and Søgaard 2015) 529

4 Vol. 25 No. 5 December (Miyauchi, Asahara, Nakagawa, and Kato 2018; 2018) BCCWJ-Infostr 7 (1) a. (information status) (discourse-new) (discourse-old) b. (commonness) (hearer-new) (hearer-old) c. (definiteness) (definite) (indefinite) d. (specificity) (specific) (unspecific) e. (animacy) (animacy) (inanimate) f. (sentience) (sentient) (insentient) g. (agentivity) (agent) (patient/theme) (BCCWJ) (Maekawa, Yamazaki, Ogiso, Maruyama, Ogura, Kashino, Koiso, Yamaguchi, Tanaka, and Den 2014) 2 BCCWJ-DepParaPAS ( 2015; 2016) (referentiality) BCCWJ-DepParaPAS

5 3.1 (1a) BCCWJ (2) a. b. [BCCWJ:PN1c 00001] (2a) (2b) (2a) BCCWJ-DepParaPAS (1b) (hearer-old) (hearer-new) (bridging) (3) a b c [BCCWJ:PN4c 00001] 3 a. b. c. 3.2 (1c) (definite) 531

6 Vol. 25 No. 5 December 2018 (indefinite) 3 (4) a b [BCCWJ:PN4c 00001] (4) a. (4) 3 (4) b. (1d) (specific) (unspecific) 3 (5) a b [BCCWJ:PN4c 00001] 5 a. 30, b. 3.3 (1e) (animate) (inanimate) (1f) (sentient) (insentient) 532

7 - (6) a b [BCCWJ:PN4c 00001] (6) a. (6) b. 3.4 (1g) (agent) (patient/theme) (7) a. b. c. [BCCWJ:PN1d 00001] (7a) (7b) (7c)

8 Vol. 25 No. 5 December BCCWJ(Maekawa et al. 2014) PN 1 (Asahara and Matsumoto 2016) (Matsumoto, Asahara, and Arita 2018) ( 2017) (Miyauchi et al. 2018; 2018) ( 2015) ( 2014) Dundee Eye Tracking Corpus Dundee Treebank (Kennedy, Hill, and Pynte 2003) BCCWJ-EyeTrack (Just, Carpenter, and Woolley 1982) Linger dr/linger/ 534

9 1 SR Research EyeLink ,000 Hz MS BCCWJ 2 20 A, B, C, D

10 Vol. 25 No. 5 December A B C D A B C D A B C D C D A B C D A B B A D C B A D C D C B A D C B A 5 (Maekawa et al. 2014) BCCWJ BCCWJ-EyeTrack (Asahara, Ono, and Miyamoto 2016) BCCWJ-Infostr (Miyauchi et al. 2018; 2018) 3 (Sorensen, Hohenstein, and Vasishth 2016) 5.1 BCCWJ-EyeTrack: Start Fixation Time End Fixation Time Fixation Time 5 (measures) (Van Gompel, Fischer, Murray, and Hill 2007) First Fixation Time (FFT) First-Pass Time (FPT) 536

11 Second-Pass Time (SPT) Regression Path Time (RPT) Total Time (TOTAL) First Fixation Time (FFT) FFT 5 3 (BCCWJ-EyeTrack) surface factor time int measure factor sample factor article factor metadata orig factor metadata factor space factor length int is first factor is last factor is second last factor 2 sessionn int articlen int screenn int linen int segmentn int subj factor dependent int (BCCWJ-Infostr) infostatus factor definiteness factor animacy factor commonness factor specificity factor sentience factor agentivity factor 2 537

12 Vol. 25 No. 5 December 2018 First-Pass Time (FPT) FPT 5, 6 Second-Pass Time (SPT) 2 SPT 9, 11 FPT + SPT Total Time Regression Path Time (RPT) RPT RPT 5, 6, 7, 8, 9 9 RPT 5, 6, 7, 8 Total Time (TOTAL) TOTAL 5, 6, 9, 11 CSV CSV 3 (surface: factor) (time: int) First Fixation Time (FFT), First-Pass Time (FPT), Second-Pass Time (SPT), Regression Path Time (RPT), Total Time (Total) 5 (measure: factor) { SelfPaced, EyeTrack:FFT, EyeTrack:FPT, EyeTrack:SPT, EyeTrack:RPT, EyeTrack:Total } 1 (logtime: num) time (length: int) surface (space: factor) (dependent: int) (Asahara and Matsumoto 2016) sample, article, metadata orig, metadata 4 (sample: factor) {A,B,C,D} 538

13 5 6 (article: factor) BCCWJ BCCWJ ID (metadata orig: factor) BCCWJ BCCWJ XML ancestor axis (metadata: factor) metadata orig {authorsdata caption listitem profile titleblock } BCCWJ (sessionn: int) 2 (articlen: int) (1 5) (screenn: int) (linen: int) (1 5) (bunsetsun: int) 0 2 (is first: bool) (is last: bool) 2 (is second last: bool) sample screen ID (subj: factor) 2 1 (rspan: num) (voc: num) 1,000 ( ) 5.2 BCCWJ-Infostr: BCCWJ BCCWJ-Infostr (Miyauchi et al. 2018; 2018) 3 7 (infostatus:is) (definiteness:def) (specificity:spec) 539

14 Vol. 25 No. 5 December 2018 (animacy:ani) (sentience:sent) (agentivity:ag) (commonnness:com) is:disc-new is:disc-old BCCWJ (BCCWJ-PAS) (Lyons 1999; Heim 2011) def:definite def:indefinite (von Heusinger 2011) spec:specific spec:unspecific 3 ani:animate ani:inanimate sent:sentient sent:insentient ag:agent ag:patient ag:neither 1 ag:both com:hearer-old com:hearer-new com:bridging 4 BCCWJ-EyeTrack *:either (def:either) 540

15 4 is def spec anim sent ag com is:disc-new is:disc-old 1, def:definite def:indefinite def:either 1, spec:specific spec:unspecific spec:either spec:neither 1, anim:animate anim:inanimate anim:either 342 1,680 1 sent:sentient sent:insentient sent:either 337 1,678 8 ag:agent ag:patient ag:both ag:neither ,491 com:hearer-new com:hearer-old com:bridging com:neither 494 1, (spec:either) (ani:either) (sent:either) (ag:either) neither (spec:neither) (ag:neither) (com:neither) 3 (Asahara 2017) (is:*) (def:*) (ani:*) (com:*) 4 ag:neither *:NIL *:neither *:neither 5.3 Bayesian Linear Mixed Model: BCCWJ-EyeTrack BCCWJ-Infostr (Sorensen et al. 2016) 541

16 Vol. 25 No. 5 December 2018 metadata 3 {authorsdata, caption, listitem, profile, titleblock} time lognormal time lognormal(µ, σ) µ = α + β length length(x) + β space χ space (x) + β dependent dependent(x) + β sessionn sessionn(x) + β articlen articlen(x) + β screenn screenn(x) + β linen linen(x) + β segmentn segmentn(x) + β is first χ is first (x) + β is last χ is last (x) + β is second last χ is second last (x) + β is: χ is: (x) + β def: χ def: (x) + β ani: χ ani: (x) is: def: ani: + β com: χ com: (x) + γ article=a(x) + γ subj=s(x). com: a(x) A s(x) S time lognormal rstan σ lognormal µ lognormal α β length length(x) β space χ space (x) 4 β sessionn, β articlen, β screenn, β linen, β segmentn sessionn(x), articlen(x), screenn(x), linen(x), segmentn(x) β is first, β is last, β is second last χ is first (x), χ is last (x), χ is second last (x) is: βis: χ is: (x) def: βdef: χ def: (x) ani: βani: χ ani: (x) com: βcom: χ com: (x) a(x) A γarticle=a(x) x a(x) s(x) S γsubj=s(x) x s(x) 100 5,000 4 chains (Asahara 2017) 3 BCCWJ { (metadata orig) 4 1 if x A, χ A χ A (x) = 0 if x A. 542

17 AIC Total Time 50% β length β space β dependent β sessionn, β screenn, β linen, β segmentn β articlen γ article=a(x) β is first, β is last, β is second last 2 (Asahara et al. 2016) 4 7 Total Time β :NIL (Asahara and Kato 2017) β info:disc-new, β com:hearer-new β info:disc-old, β com:hearer-old β com:bridging β def:definite β def:indefinite 543

18 Vol. 25 No. 5 December 2018 β ani:animate β ani:inanimate Total Time First Fixation Time, First Path Time, Second Pass Time, Regression Path Time First Fixation Time 5 8 First Pass Time 6 9 Total Time Second Pass Time Total Time: 4 Total Time: 5 First Fixation Time: 6 First Pass Time: 544

19 Regresion Path Time Second Pass Time: 8 Regression Path Time: 5 Total First Fixation First Pass Second Pass Regression Path info:disc-new SD SD SD SD SD info:disc-old SD SD SD SD SD def:definite SD SD SD SD SD def:indefinite SD SD SD SD SD ani:animate SD SD SD SD SD ani:inanimate SD SD SD SD SD com:hearer-new SD SD SD SD SD com:hearer-old SD SD SD SD SD com:hearer-new SD SD SD SD SD com:bridging SD SD SD SD SD

20 Vol. 25 No. 5 December First Pass Time Regression Path Time 1 (com:hearer-new, com:hearer-old) First Fixation Total Time First Pass Time Second Pass Time 1 (com:hearer-new, com:bridging) Total Time First Pass Time

21 JSPS , 17H00917, 18H05521 The 31st Pacific Asia Conference on Language, Information and Computation PACLIC 31(2017) Between Reading Time and Information Structure Asahara, M. (2017). Between Reading Time and Information Structure. In Proceedings of the 31st Pacific Asia Conference on Language, Information and Computation, pp The National University (Phillippines). 547

22 Vol. 25 No. 5 December 2018 Asahara, M. and Kato, S. (2017). Between Reading Time and Syntactic/Semantic Categories. In Proceedings of the The 8th International Joint Conference on Natural Language Processing, pp Asahara, M. and Matsumoto, Y. (2016). BCCWJ-DepPara: A Syntactic Annotation Treebank on the Balanced Corpus of Contemporary Written Japanese. In Proceedings of the 12th Workshop on Asian Langauge Resources (ALR12), pp (2016). BCCWJ-DepParaPAS:. 22, pp Asahara, M., Ono, H., and Miyamoto, E. T. (2016). Reading-Time Annotations for Balanced Corpus of Contemporary Written Japanese. In Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pp Barrett, M., Bingel, J., Keller, F., and Søgaard, A. (2016). Weakly Supervised Part-of-speech Tagging Using Eye-tracking Data. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pp Demberg, V. and Keller, F. (2008). Data from Eye-tracking Corpora as Evidence for Theories of Syntactic Processing Complexity. Cognition, 109 (2), pp Gibson, E. (2008). Linguistic Complexity: Locality of Syntactic Dependencies. Cognition, 68, pp Götze, M., Weskott, T., Endriss, C., Fiedler, I., Hinterwimmer, S., Petrova, S., Schwarz, A., Skopeteas, S., and Stoel, R. (2007). Information Structure. In Dipper, S., Götze, M., and Skopeteas, S. (Eds.), Information Structure in Cross-linguistic Corpora: Annotation Guidelines for Phonology, Morphology, Syntax, Semantics and Information Structure, Vol. 7, pp Universitätsverlag Potsdam. Hale, J. (2001). A Probabilistic Earley Parser as a Psycholinguistic Model. In Proceedings of the 2nd Conference of the North American Chapter of the Association for Computational Linguistics, Vol. 2, pp Heim, I. (2011). Definiteness and Indefiniteness. In von Heusinger, K., Maienborn, C., and Portner, P. (Eds.), Semantics: An International Handbook of Natural Language Meaning, Vol. 2, pp Mouton de Gruyter. Just, M. A., Carpenter, P. A., and Woolley, J. D. (1982). Paradigms and Processes in Reading Comprehension. Journal of Experimental Psychology: General, 3, pp (2017).. 23, pp

23 Kennedy, A., Hill, R., and Pynte, J. (2003). The Dundee Corpus. In Proceedings of the 12th European Conference on Eye Movement. Kennedy, A. and Pynte, J. (2005). Parafoveal-on-foveal Effects in Normal Reading. Vision Research, 45, pp Klerke, S., Martínez Alonso, H., and Søgaard, A. (2015). Looking Hard: Eye Tracking for Detecting Grammaticality of Automatically Compressed Sentences. In Proceedings of the 20th Nordic Conference of Computational Linguistics (NODALIDA 2015), pp Lyons, C. (1999). Definiteness. Cambridge University Press, Cambridge. Maekawa, K., Yamazaki, M., Ogiso, T., Maruyama, T., Ogura, H., Kashino, W., Koiso, H., Yamaguchi, M., Tanaka, M., and Den, Y. (2014). Balanced Corpus of Contemporary Written Japanese. Language Resources and Evaluation, 48, pp Matsumoto, S., Asahara, M., and Arita, S. (2018). Japanese Clause Classification Annotation on the Balanced Corpus of Contemporary Written Japanese. In Proceedings of Workshop on Asian Language Resources 13 (ALR13), pp (2014).., 21 (2), pp (2018).., 16. Miyauchi, T., Asahara, M., Nakagawa, N., and Kato, S. (2018). Annotation of Information Structure on The Balanced Corpus of Contemporary Written Japanese. In Computational Linguistics: 15th International Conference of the Pacific Association for Computational Linguistics, PACLING 2017, Yangon, Myanmar, August 16 18, 2017, Revised Selected Papers, pp (2016).. 22, pp Palmer, M., Gildea, D., and Kingsbury, P. (2005). The Proposition Bank: An Annotated Corpus of Semantic Roles. Conputational Linguistics, 31 (1), pp Prasad, R., Dinesh, N., Lee, A., Miltsakaki, E., Robaldo, L., Joshi, A., and Webber, B. (2008). The Penn Discourse TreeBank 2.0. In Proceedings of the 6th International Conference on Language Resources and Evaluation (LREC 08), pp Prasad, R., Webber, B., Lee, A., Pradhan, S., and Joshi, A. (2015). Bridging Sentential and Discourse-level Semantics through Clausal Adjuncts. In Proceedings of the 1st Workshop on Linking Computational Models of Lexical, Sentential and Discourse-level Semantics, pp Sorensen, T., Hohenstein, S., and Vasishth, S. (2016). Bayesian Linear Mixed Models using Stan: 549

24 Vol. 25 No. 5 December 2018 A Tutorial for Psychologists, Linguists, and Cognitive Scientists. Quantitative Methods for Psychology, 12, pp (2013).. 19, pp (2015).. 8, pp Van Gompel, R. P., Fischer, M. H., Murray, W. S., and Hill, R. L. (2007). Eye-movement Research: An Overview of Current and Past Developments. Eye Movements: A Window on Mind and Brain, pp von Heusinger, K. (2011). Specificity. In von Heusinger, K., Maienborn, C., and Portner, P. (Eds.), Semantics: An International Handbook of Natural Language Meaning, Vol. 2, pp Mouton de Gruyter. A 6, 7, 8, 9, 10, 11 Parameter: Rhat: 1.1 n eff: mean: sd: se mean: 2.5%: 2.5% 50%: 97.5%: 97.5% 550

25 6 Total Time Parameter Rhat n eff mean sd se mean 2.5% 50% 97.5% alpha , beta length , beta space , beta dependent , beta sessionn , beta articlen , beta screenn , beta linen , beta segmentn , beta is first , beta is last , beta is second last , sigma , sigma article , sigma subj , sigma info log-posterior , Total Time Parameter Rhat n eff mean sd se mean 2.5% 50% 97.5% beta info:nil , beta info:disc-new , beta info:disc-old , beta def:nil , beta def:definite , beta def:indefinite , beta ani:nil , beta ani:animate , beta ani:inanimate , beta com:nil , beta com:bridging , beta com:hearer-new , beta com:hearer-old , beta com:neither ,

26 Vol. 25 No. 5 December First Fixation Time Parameter Rhat n eff mean sd se mean 2.5% 50% 97.5% beta info:nil beta info:disc-new , beta info:disc-old , beta def:nil beta def:definite beta def:indefinite beta ani:nil , beta ani:animate beta ani:inanimate beta com:nil beta com:bridging , beta com:hearer-new , beta com:hearer-old , beta com:neither , First Pass Time Parameter Rhat n eff mean sd se mean 2.5% 50% 97.5% beta info:nil , beta info:disc-new , beta info:disc-old , beta def:nil , beta def:definite , beta def:indefinite , beta ani:nil , beta ani:animate , beta ani:inanimate , beta com:nil , beta com:bridging , beta com:hearer-new , beta com:hearer-old , beta com:neither ,

27 10 Second Pass Time Parameter Rhat n eff mean sd se mean 2.5% 50% 97.5% beta info:nil , beta info:disc-new , beta info:disc-old , beta def:nil , beta def:definite , beta def:indefinite , beta ani:nil , beta ani:animate , beta ani:inanimate , beta com:nil , beta com:bridging , beta com:hearer-new , beta com:hearer-old , beta com:neither , Regression Path Time Parameter Rhat n eff mean sd se mean 2.5% 50% 97.5% beta info:nil , beta info:disc-new , beta info:disc-old , beta def:nil , beta def:animate , beta def:inanimate , beta ani:nil , beta ani:animate , beta ani:inanimate , beta com:nil , beta com:bridging , beta com:hearer-new , beta com:hearer-old , beta com:neither ,

28 Vol. 25 No. 5 December

Microsoft PowerPoint LRW.pptx

Microsoft PowerPoint LRW.pptx Reading Time Balanced Corpus Corpus Annotation Data 文節文画面 A 470 66 19 B 455 67 21 C 355 44 16 D 363 41 15 Group 視線走査法自己ペース読文法 1 A 境界なし B 境界あり C 境界なし D 境界あり 2 A 境界あり B 境界なし C 境界あり D 境界なし 3 C 境界なし