Summary_Скоринкин_14.12.2018 (1137486), страница 3
Текст из файла (страница 3)
Our markup allows easy extraction of networks using bothapproaches, so we were able to compare them and demonstrate the difference.At the first stage we built two networks — a co-occurrence one and a conversational one — forthe entire novel. Both were too big to analyze visually, so we extracted the core of each networkusing network centrality measures.Fig.
6. The top 10 characters by eigenvector centrality in conversational network (left) and cooccurrence network (right)10It turned out that regardless of centrality measure conversational network seems to downgrademilitary and political commanders, which constitute the ‘historical’ dimension of War and Peace,in terms of centralities. Napoleon, Kutuzov, Alexander the I tend to be much more central in theco-occurrence network. This applies to all three centralities that we used.The same effect was also observed when we analyzed the networks dynamically. In book 10 ofWar and Peace, where the Borodino battle is being prepared and then takes place, the distributionof centralities is very explicitly different — see figures 7 and 8.Fig 7.
Co-occurrence network for book 10 of War and Peace. Nodes proportional to eigenvectorcentrality.11Fig 8. Conversational network for book 10 of War and Peace. Nodes proportional to eigenvectorcentrality.Obviously, the co-occurrence network is better at depicting the central collision of this part of thenovel between the Russian army headed by Kutuzov and the French invaders led by Napoleon.Russian emperor Alexander I is also among the central characters. On the other hand, theconversational network places the highest centrality on Pierre — probably the most importantspectator of Borodino battle in Tolstoy’s novel. From this we may speculate on how networkcentralities in different networks may reflect different functions of characters in a complex plotlike the one of War and Peace.We have also performed additional statistical comparative testing of the two networks usingnetwork density measure (the ratio of the number of edges in the network to the maximum possiblenumber of edges, i.e.
the measure of how densely the network is interconnected). It turns out thatdensities in the co-occurrence network tend to reflect the change of settings between war-relatedand peace-related ones. The conversational network is less sensitive to these changes, but itsdensity peaks in the epilogue, reflecting the wrap-up of the events of the novel as two families areshown in (almost) happy co-existence.We conclude the thesis by highlighting the major findings and results:121. We have analyzed the existing body of research on computational and pre-computationalmodeling of characters and character spaces.
The analysis of contemporary digital studiesrevealed a problem of data extraction from the text and a related problem of researchreproducibility. This problem can be tackled with the creation of semantic markup.2. We have created and published semantic markup for Leo Tolstoy’s War and Peace, that isdesigned for computational modeling and analysis of characters in the novel.
The markupcontains 25,6 thousand identified character mentions and 6,3 thousand instances of directspeech with speaker and addressee identified.3. The markup has been tested as the basis for character space modeling using the analysis ofdirect speech and network analysis.4. The results of the analysis comply with certain findings of literary scholars who wereresearching War and Peace without the help of computer. In our view, this gives certaincredibility to the applied method.5. For each approach we have compared different methods and showed how they differ fromeach other while reflecting the same character system. These differences have not beenreported previously by other researchers.The main findings of the research were discussed in the following papers:● Skorinkin D.
Extracting Character Networks to Explore Literary Plot Dynamics //Komp'juternaja lingvistika i intellektual'nye tehnologii: po materialam ezhegodnojmezhdunarodnoj konferencii «Dialog». Vyp. 16 (23): V 2 t. 2017. P. 257-270.Bonch-Osmolovskaya А., Skorinkin D. Text mining War and Peace: Automatic extractionof character traits from literary pieces// Digital Scholarship in the Humanities, 2017Volume 32, Issue supplement 1, p. i17–i24● Skorinkin D. A., Bonch-Osmolovskaja A. A. «Osobye primety» v rechi hudozhestvennyhpersonazhej: kolichestvennyj analiz dialogov v «Vojne i mire» L.
N. Tolstogo //Jelektronnyj nauchno-obrazovatel'nyj zhurnal «Istorija» 2017. T. 7. # 7 (51)● Skorinkin D. A. Jelektronnoe predstavlenie teksta s pomoshh'ju standarta razmetki TEI //Vestnik Moskovskogo universiteta. Serija 9: Filologija. 2016. # 5. S. 90-108● Skorinkin D., Mozhaev E. TEI markup for the 90-volume edition of Leo Tolstoy’scomplete works, in: TEI Conference and Members’ Meeting 2016 Book of Abstracts.Wien: Austrian Centre for Digital Humanities, 2016. P.
107-109● Skorinkin D. Digital Edition of the Complete Works of Leo Tolstoy // 6th AIUCDConference Book of Abstracts. Rome, 2017. P. 264-26713● Bonch-Osmolovskaya А., Skorinkin D., Sidorova E. Verbal Identity of a FictionalCharacter: a Quantitative Study with a Machine Learning Experiment // Digital Humanities2016.
Conference Abstracts, 11–16 July 2016. Kraków: Jagiellonian University, 2016. P.747-749.14.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
