Multi-label toxicity detection is highly prominent, with many research groups, companies, and individuals engaging with it through shared tasks and dedicated venues. This paper describes a cross-lingual approach to annotating multi-label text classification on a newly developed Dutch language dataset, using a model trained on English data. We present an ensemble model of one Transformer model and an LSTM using Multilingual embeddings. The combination of multilingual embeddings and the Transformer model improves performance in a cross-lingual setting.
We report on an inter-annotator agreement experiment involving instances of text reuse focusing on the well-known case of biblical intertextuality in medieval literature. We target the application use case of literary scholars whose aim is to document instances of biblical references in the ‘apparatus fontium’ of a prospective digital edition. We develop a Bayesian implementation of Cohen’s kappa for multiple annotators that allows us to assess the influence of various contextual effects on the inter-annotator agreement, producing both more robust estimates of the agreement indices as well as insights into the annotation process that leads to the estimated indices. As a result, we are able to produce a novel and nuanced estimation of inter-annotator agreement in the context of intertextuality, exploring the challenges that arise from manually annotating a dataset of biblical references in the writings of Bernard of Clairvaux. Among others, our method was able to unveil the fact that the obtained agreement depends heavily on the biblical source book of the proposed reference, as well as the underlying algorithm used to retrieve the candidate match.
This paper describes the system developed by the Antwerp Centre for Digital humanities and literary Criticism [UAntwerp] for toxic span detection. We used a stacked generalisation ensemble of five component models, with two distinct interpretations of the task. Two models attempted to predict binary word toxicity based on ngram sequences, whilst 3 categorical span based models were trained to predict toxic token labels based on complete sequence tokens. The five models’ predictions were ensembled within an LSTM model. As well as describing the system, we perform error analysis to explore model performance in relation to textual features. The system described in this paper scored 0.6755 and ranked 26th.
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.
Fictional prose can be broadly divided into narrative and discursive forms with direct speech being central to any discourse representation (alongside indirect reported speech and free indirect discourse). This distinction is crucial in digital literary studies and enables interesting forms of narratological or stylistic analysis. The difficulty of automatically detecting direct speech, however, is currently under-estimated. Rule-based systems that work reasonably well for modern languages struggle with (the lack of) typographical conventions in 19th-century literature. While machine learning approaches to sequence modeling can be applied to solve the task, they typically face a severed skewness in the availability of training material, especially for lesser resourced languages. In this paper, we report the result of a multilingual approach to direct speech detection in a diverse corpus of 19th-century fiction in 9 European languages. The proposed method finetunes a transformer architecture with multilingual sentence embedder on a minimal amount of annotated training in each language, and improves performance across languages with ambiguous direct speech marking, in comparison to a carefully constructed regular expression baseline.
Lemmatization of standard languages is concerned with (i) abstracting over morphological differences and (ii) resolving token-lemma ambiguities of inflected words in order to map them to a dictionary headword. In the present paper we aim to improve lemmatization performance on a set of non-standard historical languages in which the difficulty is increased by an additional aspect (iii): spelling variation due to lacking orthographic standards. We approach lemmatization as a string-transduction task with an Encoder-Decoder architecture which we enrich with sentence information using a hierarchical sentence encoder. We show significant improvements over the state-of-the-art by fine-tuning the sentence encodings to jointly optimize a bidirectional language model loss. Crucially, our architecture does not require POS or morphological annotations, which are not always available for historical corpora. Additionally, we also test the proposed model on a set of typologically diverse standard languages showing results on par or better than a model without fine-tuned sentence representations and previous state-of-the-art systems. Finally, to encourage future work on processing of non-standard varieties, we release the dataset of non-standard languages underlying the present study, which is based on openly accessible sources.
The detection of allusive text reuse is particularly challenging due to the sparse evidence on which allusive references rely — commonly based on none or very few shared words. Arguably, lexical semantics can be resorted to since uncovering semantic relations between words has the potential to increase the support underlying the allusion and alleviate the lexical sparsity. A further obstacle is the lack of evaluation benchmark corpora, largely due to the highly interpretative character of the annotation process. In the present paper, we aim to elucidate the feasibility of automated allusion detection. We approach the matter from an Information Retrieval perspective in which referencing texts act as queries and referenced texts as relevant documents to be retrieved, and estimate the difficulty of benchmark corpus compilation by a novel inter-annotator agreement study on query segmentation. Furthermore, we investigate to what extent the integration of lexical semantic information derived from distributional models and ontologies can aid retrieving cases of allusive reuse. The results show that (i) despite low agreement scores, using manual queries considerably improves retrieval performance with respect to a windowing approach, and that (ii) retrieval performance can be moderately boosted with distributional semantics.
This paper addresses Hip-Hop lyric generation with conditional Neural Language Models. We develop a simple yet effective mechanism to extract and apply conditional templates from text snippets, and show—on the basis of a large-scale crowd-sourced manual evaluation—that these templates significantly improve the quality and realism of the generated snippets. Importantly, the proposed approach enables end-to-end training, targeting formal properties of text such as rhythm and rhyme, which are central characteristics of rap texts. Additionally, we explore how generating text at different scales (e.g. character-level or word-level) affects the quality of the output. We find that a hybrid form—a hierarchical model that aims to integrate Language Modeling at both word and character-level scales—yields significant improvements in text quality, yet surprisingly, cannot exploit conditional templates to their fullest extent. Our findings highlight that text generation models based on Recurrent Neural Networks (RNN) are sensitive to the modeling scale and call for further research on the observed differences in effectiveness of the conditioning mechanism at different scales.
Recent applications of neural language models have led to an increased interest in the automatic generation of natural language. However impressive, the evaluation of neurally generated text has so far remained rather informal and anecdotal. Here, we present an attempt at the systematic assessment of one aspect of the quality of neurally generated text. We focus on a specific aspect of neural language generation: its ability to reproduce authorial writing styles. Using established models for authorship attribution, we empirically assess the stylistic qualities of neurally generated text. In comparison to conventional language models, neural models generate fuzzier text, that is relatively harder to attribute correctly. Nevertheless, our results also suggest that neurally generated text offers more valuable perspectives for the augmentation of training data.
Although in recent years numerous forms of Internet communication ― such as e-mail, blogs, chat rooms and social network environments ― have emerged, balanced corpora of Internet speech with trustworthy meta-information (e.g. age and gender) or linguistic annotations are still limited. In this paper we present a large corpus of Flemish Dutch chat posts that were collected from the Belgian online social network Netlog. For all of these posts we also acquired the users' profile information, making this corpus a unique resource for computational and sociolinguistic research. However, for analyzing such a corpus on a large scale, NLP tools are required for e.g. automatic POS tagging or lemmatization. Because many NLP tools fail to correctly analyze the surface forms of chat language usage, we propose to normalize this anomalous' input into a format suitable for existing NLP solutions for standard Dutch. Additionally, we have annotated a substantial part of the corpus (i.e. the Chatty subset) to provide a gold standard for the evaluation of future approaches to automatic (Flemish) chat language normalization.