本論文では，計量的な手法を用いて川端康成の代筆問題と文体問題に取り組み，次に挙げたことを明らかにした。①小説『乙女の港』と『花日記』は川端康成と中里恒子の共同執筆である。②『コスモスの友』，『古都』，『眠れる美女』と『山の音』は代筆の可能性が低い。③泉鏡花，徳田秋聲と横光利一と比べ川端康成文体の存在が確認され，終戦の1945 年を境に川端康成の語彙の豊富さと，機能語の助詞，副詞，接続詞に変化が見られた。 / In this study, we revealed the following ghostwriting and writing style problem of Kawabata quantitatively. ①Otome no minato and Hana nikki were colloboratively written by Kawabata and Nakazato. ② Kosumosu no tomo, The Old Capital, House of the Sleeping Beauties, and The Sound of the Mountain were not written by the suspected ghostwriters. ③ Kawabata has his own writing style as compared to novels written by Izumi, Tokuda, and Yokomitsu. Changes were observed in vocabulary richness, postpositional particles, adverb, and conjunctions in Kawabata's novels after the second world war. / 博士(文化情報学) / Doctor of Culture and Information Science / 同志社大学 / Doshisha University

宇野浩二は精神病や挫折による執筆の空白期間を経るたびに異なる文体を編み出すと言われる作家である。本研究は、作品からデータを抽出し、計量的分析手法を用いて、宇野浩二の文体変化に関する諸問題を解明することを目的とする。文芸評論を踏まえ、病気後の文体変化、文体変化の時期、戦後の文体変化、同時代作家との比較の四つの問題を取り上げ、客観的なデータの分析結果に基づき、宇野浩二の文体特徴、文体変化の時期および戦後作品の独特性を明らかにした。 / Kouji Uno is a well-known Japanese littérateur, whose creative activity was interrupted twice because of a mental illness and frustrations. It was reported that his writing style had changed when he resumed writing. The study aims to investigated the differences of Kouji Uno's writing style in his different stages by using quantitative analysis. We demonstrated Uno's stylistic changes after the mental illness, the exact time of these changes, and the characteristics of works published after the War based on the literary criticism and statistical results. In addition, we clarified the uniqueness of his postwar works through a comparative analysis with contemporary writers. / 博士(文化情報学) / Doctor of Culture and Information Science / 同志社大学 / Doshisha University

Diese Arbeit untersucht vollständige Zeichenkettenfrequenzverteilungen natürlichsprachiger Texte auf ihren linguistischen und anwendungsbezogenen Gehalt. Im ersten Teil wird auf dieser Datengrundlage ein unüberwachtes Lernverfahren entwickelt, das Texte in Morpheme zerlegt. Die Zerlegung geht von der Satzebene aus und verwendet jegliche vorhandene Kontextinformation. Es ergibt sich ein sprachunabhängiger Algorithmus, der die gefundenen Morpheme teilweise zu Baumstrukturen zusammenordnet. Die Evaluation der Ergebnisse mit Hilfe statistischer Modelle ermöglicht die Identifizierung auch kleiner Performanzunterschiede. Diese sind einer linguistischen Interpretation zugänglich. Der zweite Teil der Arbeit besteht aus stilometrischen Untersuchungen anhand eines Textähnlichkeitsmaßes, das ebenfalls auf vollständigen Zeichenkettenfrequenzen beruht. Das Textähnlichkeitsmaß wird in verschiedenen Varianten definiert und anhand vielfältiger stilometrischer Fragestellungen und auf Grundlage unterschiedlicher Korpora ausgewertet. Dabei ist ein wiederholter Vergleich mit der Performanz bisheriger Forschungsansäzte möglich. Die Performanz moderner Maschinenlernverfahren kann mit dem hier vorgestellten konzeptuell einfacheren Verfahren reproduziert werden. Während die Segmentierung in Morpheme ein lokaler Vorgang ist, besteht Stilometrie im globalen Vergleich von Texten. Daher bietet die Untersuchung dieser zwei unverbunden scheinenden Fragestellungen sich gegenseitig ergänzende Perspektiven auf die untersuchten Häufigkeitsdaten. Darüber hinaus zeigt die Diskussion der rezipierten Literatur zu beiden Themen ihre Verbindungen durch verwandte Konzepte und Denkansätze auf. Aus der Gesamtheit der empirischen Untersuchungen zu beiden Fragestellungen kann abgeleitet werden, dass den längeren und damit selteneren Zeichenketten wesentlich mehr Informationsgehalt innewohnt, als in der bisherigen Forschung gemeinhin angenommen wird. / This thesis investigates the linguistic and application specific content of complete character substring frequency distributions of natural language texts. The first part develops on this basis an unsupervised learning algorithm for segmenting text into morphemes. The segmentation starts from the sentence level and uses all available context information. The result is a language independent algorithm which arranges the found morphemes partly into tree like structures. The evaluation of the output using advanced statistical modelling allows for identifying even very small performance differences. These are accessible to linguistic interpretation. The second part of the thesis consists of stylometric investigations by means of a text similarity measure also rooted in complete substring frequency statistics. The similarity measure is defined in different variants and evaluated for various stylometric tasks and on the basis of diverse corpora. In most of the case studies the presented method can be compared with publicly available performance figures of previous research. The high performance of modern machine learning methods is reproduced by the considerably simpler algorithm developed in this thesis. While the segmentation into morphemes is a local process, stylometry consists in the global comparison of texts. For this reason investigating of these two seemingly unconnected problems offers complementary perspectives on the explored frequency data. The discussion of the recieved litarature concerning both subjects additionally shows their connectedness by related concepts and approaches. It can be deduced from the totality of the empirical studies on text segmentation and stylometry conducted in this thesis that the long and rare character sequences contain considerably more information then assumed in previous research.

Denna rapport analyserar och testar olika metoder som syftar till att särskiljamänskligt genererade lösningar på uppgifter och texter från de som genereras avartificiell intelligens. På senare tid har användningen av artificiell intelligens setten betydande ökning, särskilt bland studenter. Syftet med denna studie är attavgöra om det för närvarande är möjligt att upptäcka fusk från högskolestudenterinom elektroteknik som använder sig av AI. I rapporten testas lösningar påuppgifter och texter genererade av programmet ChatGPT med hjälp av en generellmetod och externa AI-verktyg. Undersökningen omfattar områdena matematik,programmering och skriven text. Resultatet av undersökningen tyder på att detinte är möjligt att upptäcka fusk med hjälp av AI i ämnena matematik ochprogrammering. Dock när det gäller text kan i viss utsträckning fusk vidanvändning av en AI upptäckas. / This report analyzes and tests various methods aimed at distinguishinghuman-generated solutions to tasks and texts from those generated by artificialintelligence. Recently the use of artificial intelligence has seen a significantincrease, especially among students. The purpose of this study is to determinewhether it is currently possible to detect if a college student in electricalengineering is using AI to cheat. In this report, solutions to tasks and textsgenerated by the program ChatGPT are tested using a general methodology andexternal AI-based tools. The research covers the areas of mathematics,programming and written text. The results of the investigation suggest that it is notpossible to detect cheating with the help of an AI in the subjects of mathematicsand programming. In the case of text, cheating by using an AI can be detected tosome extent.

Classic literature is rich, be it linguistically, historically, or culturally, making it valuable for future studies. Consequently, this project chose a set of 48 classic books to conduct a stylometric analysis on the defined set of books, adopting an approach used by a related work to divide the books into text segments, quantify the resulting text segments, and analyze the books using the quantified values to understand the linguistic attributes of the books. Apart from the latter, this project conducted different classification tasks for other objectives. In one respect, the study used the quantified values of the text segments of the books for classification tasks using advanced models like LightGBM and TabNet to assess the application of this approach in authorship attribution. From another perspective, the study utilized a State-Of-The-Art model, namely, RoBERTa for classification tasks using the segmented texts of the books instead to evaluate the performance of the model in authorship attribution. The results uncovered the characteristics of the books to a reasonable degree. Regarding the authorship attribution tasks, the results suggest that segmenting and quantifying text using stylometric analysis and supervised machine learning algorithms is practical in such tasks. This approach, while showing promise, may still require further improvements to achieve optimal performance. Lastly, RoBERTa demonstrated high performance in authorship attribution tasks.

The doctoral thesis compares the style of the “Romance of Horn”, “King Horn”, and “Horn Childe and Maiden Rimnild”. Each of them is a different version of the same story; the first is an Insular French romance, the second and third are Middle English romances. The stylistic analysis is both quantitative and qualitative, and it is developed employing the lemmatized edition of each witness of each version. The work begins with an introduction to the concept of style. The first chapter focuses on a review of the literature regarding each version, considering specifically the witnesses and textual criticism, metre and genre, language, date, and style. For the Insular French version, a specific review of the literature about the author is offered. The second chapter introduces stylometry and stylistic analysis with a brief literature review. Then, the protocols for the edition and lemmatization of each witness are described. Finally, the methods adopted for the stylistic analysis are explained. The third chapter develops the study of descriptions, anaphors, and formulas in each version. The collected data and their interpretations are considered altogether through the lens of a group of theoretical concepts: connotation, attribution, horizon of expectations and discursive tradition. Via these concepts, a trajectory of stylistic mutations is traced, and a link between style and socio-cultural context is displayed. The conclusions sum up all the information and inferences, suggesting further possibilities for new research. / La tesi sviluppa lo studio comparativo dello stile come osservato nel “Roman de Horn”, in “King Horn” e in “Horn Childe and Maiden Rimnild”, tre versioni della medesima vicenda, la prima in anglonormanno, le altre due in inglese medio. Nella tesi l’indagine stilistica è svolta in maniera qualitativa e quantitativa, lavorando sulle edizioni lemmatizzate secondo codifica TEI dei singoli testimoni di ciascuna versione. Dopo un’introduzione sul concetto di stile, il primo capitolo è dedicato a un’introduzione ai testi e ai testimoni delle tre versioni. Dopo un cappello introduttivo sui rapporti genetici tra le versioni, di ciascuna di esse si opera un'escussione della bibliografia esistente intorno a testimoni ed eventuale stemmatica, metro e genere, lingua, datazione, stile dell’opera; per la versione anglonormanna si aggiunge uno specifico approfondimento sull’autore. Nel secondo capitolo, si inquadrano stilistica e stilometria nel panorama accademico attuale, quindi si espongono i protocolli di edizione e lemmatizzazione adottata, infine i metodi di analisi adottati e le ragioni per cui circoscrivere l’indagine ai fenomeni di descrizione, anafora e formula. Nel terzo capitolo si procede ad analizzare in ogni versione le descrizioni e poi anafore e formule. Per le descrizioni, dopo aver offerto un quadro specifico per versione, si opera una sintesi sulla base dei concetti di connotazione e attribuzione, chiarendo una possibile traiettoria del mutamento stilistico. Parallelamente, dopo un’analisi di anafore e formule in ciascuna versione si offre una visione d’insieme alla luce dei concetti di orizzonte d’attesa e tradizione discorsiva, ponendo così in relazione stile e contesto socioculturale dell’Inghilterra medievale. Nelle conclusioni, si riassumono gli approdi dell’analisi, valutando metodi e risultati e proponendo possibili aperture a lavori futuri.