Search

In this article, we provide an overview of machine learning as it is applied in computational literary studies, the field of computational analysis of literary texts and literature related phenomena. We survey a number of scientific publications for the machine learning methodology the scholars used and explain concepts of machine learning and natural language processing while discussing our findings. We establish that besides transformer-based language models, researchers still make frequent use of more traditional, feature-based machine learning approaches; possible reasons for this are to be found in the challenging application of modern methods to the literature domain and in the more transparent nature of traditional approaches. We shed light on how machine learning-based approaches are integrated into a research process, which often proceeds primarily from the non-quantitative, interpretative approaches of non-digital literary studies. Finally, we conclude that the application of large language models in the computational literary studies domain may simplify the application of machine learning methodology going forward, if adequate approaches for the analysis of literary texts are found. [ABSTRACT FROM AUTHOR]

This survey presents a comprehensive description of recent neural entity linking (EL) systems developed since 2015 as a result of the "deep learning revolution" in natural language processing. Its goal is to systemize design features of neural entity linking systems and compare their performance to the remarkable classic methods on common benchmarks. This work distills a generic architecture of a neural EL system and discusses its components, such as candidate generation, mention-context encoding, and entity ranking, summarizing prominent methods for each of them. The vast variety of modifications of this general architecture are grouped by several common themes: joint entity mention detection and disambiguation, models for global linking, domain-independent techniques including zero-shot and distant supervision methods, and cross-lingual approaches. Since many neural models take advantage of entity and mention/context embeddings to represent their meaning, this work also overviews prominent entity embedding techniques. Finally, the survey touches on applications of entity linking, focusing on the recently emerged use-case of enhancing deep pre-trained masked language models based on the Transformer architecture. [ABSTRACT FROM AUTHOR]

In this paper we present a pipeline for the detection of spelling variants, i.e., different spellings that represent the same word, in non-standard texts. For example, in Middle Low German texts in and ihn (among others) are potential spellings of a single word, the personal pronoun 'him'. Spelling variation is usually addressed by normalization, in which non-standard variants are mapped to a corresponding standard variant, e.g. the Modern German word ihn in the case of in. However, the approach to spelling variant detection presented here does not need such a reference to a standard variant and can therefore be applied to data for which a standard variant is missing. The pipeline we present first generates spelling variants for a given word using rewrite rules and surface similarity. Afterwards, the generated types are filtered. We present a new filter that works on the token level, i.e., taking the context of a word into account. Through this mechanism ambiguities on the type level can be resolved. For instance, the Middle Low German word in can not only be the personal pronoun 'him', but also the preposition 'in', and each of these has different variants. The detected spelling variants can be used in two settings for Digital Humanities research: On the one hand, they can be used to facilitate searching in non-standard texts. On the other hand, they can be used to improve the performance of natural language processing tools on the data by reducing the number of unknown words. To evaluate the utility of the pipeline in both applications, we present two evaluation settings and evaluate the pipeline on Middle Low German texts. We were able to improve the F1 score compared with previous work from 0.39 to 0.52 for the search setting and from 0.23 to 0.30 when detecting spelling variants of unknown words. [ABSTRACT FROM AUTHOR]

Depending on downstream applications, it is advisable to extend the notion of tokenization from low-level character-based token boundary detection to identification of meaningful and useful language units. This entails both identifying units composed of several single words that form a several single words that form a, as well as splitting single-word compounds into their meaningful parts. In this article, we introduce unsupervised and knowledge-free methods for these two tasks. The main novelty of our research is based on the fact that methods are primarily based on distributional similarity, of which we use two flavors: a sparse count-based and a dense neural-based distributional semantic model. First, we introduce DRUID, which is a method for detecting MWEs. The evaluation on MWE-annotated data sets in two languages and newly extracted evaluation data sets for 32 languages shows that DRUID compares favorably over previous methods not utilizing distributional information. Second, we present SECOS, an algorithm for decompounding close compounds. In an evaluation of four dedicated decompounding data sets across four languages and on data sets extracted from Wiktionary for 14 languages, we demonstrate the superiority of our approach over unsupervised baselines, sometimes even matching the performance of previous language-specific and supervised methods. In a final experiment, we show how both decompounding and MWE information can be used in information retrieval. Here, we obtain the best results when combining word information with MWEs and the compound parts in a bag-of-words retrieval set-up. Overall, our methodology paves the way to automatic detection of lexical units beyond standard tokenization techniques without language-specific preprocessing steps such as POS tagging. [ABSTRACT FROM AUTHOR]

This article presents a comparison of different Word Sense Induction (WSI) clustering algorithms on two novel pseudoword data sets of semantic-similarity and co-occurrence-based word graphs, with a special focus on the detection of homonymic polysemy. We follow the original definition of a pseudoword as the combination of two monosemous terms and their contexts to simulate a polysemous word. The evaluation is performed comparing the algorithm’s output on a pseudoword’s ego word graph (i.e., a graph that represents the pseudoword’s context in the corpus) with the known subdivision given by the components corresponding to the monosemous source words forming the pseudoword. The main contribution of this article is to present a self-sufficient pseudoword-based evaluation framework for WSI graph-based clustering algorithms, thereby defining a new evaluation measure (TOP2) and a secondary clustering process (hyperclustering). To our knowledge, we are the first to conduct and discuss a large-scale systematic pseudoword evaluation targeting the induction of coarse-grained homonymous word senses across a large number of graph clustering algorithms. [ABSTRACT FROM AUTHOR]