We discuss on how related stories can be compared by their characters. We investigate character graphs, or social networks, in order to measure evolution of character importance over time. To illustrate this, we chose the Siegfried-Sigurd myth that may come from a Merovingian king named Sigiberthus. The Nibelungenlied, the Völsunga saga and the History of the Franks are the three resources used.

For the study of certain linguistic phenomena and their development over time, large amounts of textual data must be enriched with relevant annotations. Since the manual creation of such annotations requires a lot of effort, automating the process with NLP methods would be convenient. But the required amounts of training data are usually not available for non-standard or historical language. The present study investigates whether models trained on modern newspaper text can be used to automatically identify topological fields, i.e. syntactic structures, in different modern and historical German texts. The evaluation shows that, in general, it is possible to transfer a parser model to other registers or time periods with overall F1-scores >92%. However, an error analysis makes clear that additional rules and domain-specific training data would be beneficial if sentence structures differ significantly from the training data, e.g. in the case of Early New High German.

Entity recognition provides semantic access to ancient materials in the Digital Humanities: it exposes people and places of interest in texts that cannot be read exhaustively, facilitates linking resources and can provide a window into text contents, even for texts with no translations. In this paper we present entity recognition for Coptic, the language of Hellenistic era Egypt. We evaluate NLP approaches to the task and lay out difficulties in applying them to a low-resource, morphologically complex language. We present solutions for named and non-named nested entity recognition and semi-automatic entity linking to Wikipedia, relying on robust dependency parsing, feature-based CRF models, and hand-crafted knowledge base resources, enabling high accuracy NER with orders of magnitude less data than those used for high resource languages. The results suggest avenues for research on other languages in similar settings.

We provide a comprehensive overview of existing systems for the computational generation of verbal humor in the form of jokes and short humorous texts. Considering linguistic humor theories, we analyze the systematic strengths and drawbacks of the different approaches. In addition, we show how the systems have been evaluated so far and propose two evaluation criteria: humorousness and complexity. From our analysis of the field, we conclude new directions for the advancement of computational humor generation.

We investigate the use of Iconclass in the context of neural machine translation for NL<->EN artwork titles. Iconclass is a widely used iconographic classification system used in the cultural heritage domain to describe and retrieve subjects represented in the visual arts. The resource contains keywords and definitions to encode the presence of objects, people, events and ideas depicted in artworks, such as paintings. We propose a simple concatenation approach that improves the quality of automatically generated title translations for artworks, by leveraging textual information extracted from Iconclass. Our results demonstrate that a neural machine translation system is able to exploit this metadata to boost the translation performance of artwork titles. This technology enables interesting applications of machine learning in resource-scarce domains in the cultural sector.

The quality of Optical Character Recognition (OCR) is a key factor in the digitisation of historical documents. OCR errors are a major obstacle for downstream tasks and have hindered advances in the usage of the digitised documents. In this paper we present a two-step approach to automatic OCR post-correction. The first component is responsible for detecting erroneous sequences in a set of OCRed texts, while the second is designed for correcting OCR errors in them. We show that applying the preceding detection model reduces both the character error rate (CER) compared to a simple one-step correction model and the amount of falsely changed correct characters.

In this paper we describe an approach for the computer-aided identification of Shakespearean intertextuality in a corpus of contemporary fiction. We present the Vectorian, which is a framework that implements different word embeddings and various NLP parameters. The Vectorian works like a search engine, i.e. a Shakespeare phrase can be entered as a query, the underlying collection of fiction books is then searched for the phrase and the passages that are likely to contain the phrase, either verbatim or as a paraphrase, are presented in a ranked results list. While the Vectorian can be used via a GUI, in which many different parameters can be set and combined manually, in this paper we present an ablation study that automatically evaluates different embedding and NLP parameter combinations against a ground truth. We investigate the behavior of different parameters during the evaluation and discuss how our results may be used for future studies on the detection of Shakespearean intertextuality.

We present Vital Records, a demonstrator based on deep-learning approaches to handwritten-text recognition, table processing and information extraction, which enables data from century-old documents to be parsed and analysed, making it possible to explore death records in space and time. This demonstrator provides a user interface for browsing and visualising data extracted from 80,000 handwritten pages of tabular data.

Downstream effects of biased training data have become a major concern of the NLP community. How this may impact the automated curation and annotation of cultural heritage material is currently not well known. In this work, we create an experimental framework to measure the effects of different types of stylistic and social bias within training data for the purposes of literary classification, as one important subclass of cultural material. Because historical collections are often sparsely annotated, much like our knowledge of history is incomplete, researchers often cannot know the underlying distributions of different document types and their various sub-classes. This means that bias is likely to be an intrinsic feature of training data when it comes to cultural heritage material. Our aim in this study is to investigate which classification methods may help mitigate the effects of different types of bias within curated samples of training data. We find that machine learning techniques such as BERT or SVM are robust against reproducing the different kinds of bias within our test data, except in the most extreme cases. We hope that this work will spur further research into the potential effects of bias within training data for other cultural heritage material beyond the study of literature.

Grammatical Error Correction (GEC) is the task of correcting different types of errors in written texts. To manage this task, large amounts of annotated data that contain erroneous sentences are required. This data, however, is usually annotated according to each annotator’s standards, making it difficult to manage multiple sets of data at the same time. The recently introduced Error Annotation Toolkit (ERRANT) tackled this problem by presenting a way to automatically annotate data that contain grammatical errors, while also providing a standardisation for annotation. ERRANT extracts the errors and classifies them into error types, in the form of an edit that can be used in the creation of GEC systems, as well as for grammatical error analysis. However, we observe that certain errors are falsely or ambiguously classified. This could obstruct any qualitative or quantitative grammatical error type analysis, as the results would be inaccurate. In this work, we use a sample of the FCE coprus (Yannakoudakis et al., 2011) for secondary error type annotation and we show that up to 39% of the annotations of the most frequent type should be re-classified. Our corrections will be publicly released, so that they can serve as the starting point of a broader, collaborative, ongoing correction process.

An increasing amount of historic data is now available in digital (text) formats. This gives quantitative researchers an opportunity to use distant reading techniques, as opposed to traditional close reading, in order to analyse larger quantities of historic data. Distant reading allows researchers to view overall patterns within the data and reduce researcher bias. One such data set that has recently been transcribed is a collection of over 500 Australian World War I (WW1) diaries held by the State Library of New South Wales. Here we apply distant reading techniques to this corpus to understand what soldiers wrote about and how they felt over the course of the war. Extracting dates accurately is important as it allows us to perform our analysis over time, however, it is very challenging due to the variety of date formats and abbreviations diarists use. But with that data, topic modelling and sentiment analysis can then be applied to show trends, for instance, that despite the horrors of war, Australians in WW1 primarily wrote about their everyday routines and experiences. Our results detail some of the challenges likely to be encountered by quantitative researchers intending to analyse historical texts, and provide some approaches to these issues.

Prose fiction typically consists of passages alternating between the narrator’s telling of the story and the characters’ direct speech in that story. Detecting direct speech is crucial for the downstream analysis of narrative structure, and may seem easy at first thanks to quotation marks. However, typographical conventions vary across languages, and as a result, almost all approaches to this problem have been monolingual. In contrast, the aim of this paper is to provide a multilingual method for identifying direct speech. To this end, we created a training corpus by using a set of heuristics to automatically find texts where quotation marks appear sufficiently consistently. We then removed the quotation marks and developed a sequence classifier based on multilingual-BERT which classifies each token as belonging to narration or speech. Crucially, by training the classifier with the quotation marks removed, it was forced to learn the linguistic characteristics of direct speech rather than the typography of quotation marks. The results in the zero-shot setting of the proposed model are comparable to the strong supervised baselines, indicating that this is a feasible approach.

This paper accompanies the corpus publication of EncycNet, a novel XML/TEI annotated corpus of 22 historical German encyclopedias from the early 18th to early 20th century. We describe the creation and annotation of the corpus, including the rationale for its development, suggested methodology for TEI annotation, possible use cases and future work. While many well-developed annotation standards for lexical resources exist, none can adequately model the encyclopedias at hand, and we therefore suggest how the TEI Lex-0 standard may be modified with additional guidelines for the annotation of historical encyclopedias. As the digitization and annotation of historical encyclopedias are settling on TEI as the de facto standard, our methodology may inform similar projects.

It is an open question to what extent perceptions of literary quality are derived from text-intrinsic versus social factors. While supervised models can predict literary quality ratings from textual factors quite successfully, as shown in the Riddle of Literary Quality project (Koolen et al., 2020), this does not prove that social factors are not important, nor can we assume that readers make judgments on literary quality in the same way and based on the same information as machine learning models. We report the results of a pilot study to gauge the effect of textual features on literary ratings of Dutch-language novels by participants in a controlled experiment with 48 participants. In an exploratory analysis, we compare the ratings to those from the large reader survey of the Riddle in which social factors were not excluded, and to machine learning predictions of those literary ratings. We find moderate to strong correlations of questionnaire ratings with the survey ratings, but the predictions are closer to the survey ratings. Code and data: https://github.com/andreasvc/litquest

The paper investigates the impact of using geometric deep learning models on the performance of a character name linking system. The neural models that contain graph convolutional layers are confronted with the models that include conventional fully connected layers. The evaluation is performed with respect to the perfect name boundaries obtained from the test set and in a more demanding end-to-end setting where the character name linking system is preceded by a named entity recognizer.

In this paper, we describe OuPoCo, a system producing new sonnets by recombining verses from existing sonnets, following an idea that Queneau described in his book “Cent Mille Milliards de poèmes, Gallimard”, 1961. We propose to demonstrate different outputs of our implementation (a Web site, a Twitter bot and a specifically developed device, called ‘La Boîte à poésie’) based on a corpus of 19th century French poetry. Our goal is to make people interested in poetry again, by giving access to automatically produced sonnets through original and entertaining channels and devices.

Recent advancements in NLP and machine learning have created unique challenges and opportunities for digital humanities research. In particular, there are ample opportunities for NLP and machine learning researchers to analyze data from literary texts and to broaden our understanding of human sentiment in classical Greek tragedy. In this paper, we will explore the challenges and benefits from the human and machine collaboration for sentiment analysis in Greek tragedy and address some open questions related to the collaborative annotation for the sentiments in literary texts. We focus primarily on (i) an analysis of the challenges in sentiment analysis tasks for humans and machines, and (ii) whether consistent annotation results are generated from the multiple human annotators and multiple machine annotators. For human annotators, we have used a survey-based approach with about 60 college students. We have selected three popular sentiment analysis tools for machine annotators, including VADER, CoreNLP’s sentiment annotator, and TextBlob. We have conducted a qualitative and quantitative evaluation and confirmed our observations on sentiments in Greek tragedy.

Environmental factors determine the smells we perceive, but societal factors factors shape the importance, sentiment and biases we give to them. Descriptions of smells in text, or as we call them ‘smell experiences’, offer a window into these factors, but they must first be identified. To the best of our knowledge, no tool exists to extract references to smell experiences from text. In this paper, we present two variations on a semi-supervised approach to identify smell experiences in English literature. The combined set of patterns from both implementations offer significantly better performance than a keyword-based baseline.

Creating a story is difficult. Professional writers often experience a writer’s block. Thus, providing automatic support to writers is crucial but also challenging. Recently, in the field of generating and understanding stories, story completion (SC) has been proposed as a method for generating missing parts of an incomplete story. Despite this method’s usefulness in providing creative support, its applicability is currently limited because it requires the user to have prior knowledge of the missing part of a story. Writers do not always know which part of their writing is flawed. To overcome this problem, we propose a novel approach called “missing position prediction (MPP).” Given an incomplete story, we aim to predict the position of the missing part. We also propose a novel method for MPP and SC. We first conduct an experiment focusing on MPP, and our analysis shows that highly accurate predictions can be obtained when the missing part of a story is the beginning or the end. This suggests that if a story has a specific beginning or end, they play significant roles. We conduct an experiment on SC using MPP, and our proposed method demonstrates promising results.

TL-Explorer is a digital humanities tool for mapping and analyzing translated literature, encompassing the World Map and the Translation Dashboard. The World Map displays collected literature of different languages, locations, and cultures and establishes the foundation for further analysis. It comprises three global maps for spatial and temporal interpretation. A further investigation into an individual point on the map leads to the Translation Dashboard. Each point represents one edition or translation. Collected translations are processed in order to build multilingual parallel corpora for a large number of under-resourced languages as well as to highlight the transnational circulation of knowledge. Our first rendition of TL-Explorer was conducted on the well-traveled American novel, Adventures of Huckleberry Finn, by Mark Twain. The maps currently chronicle nearly 400 translations of this novel. And the dashboard supports over 30 collected translations. However, the TL-Explore is easily extended to other works of literature and is not limited to type of texts, such as academic manuscripts or constitutional documents to name a few.