This paper presents a supervised method for a novel task, namely, detecting elements of narration in passages of dialogue in prose fiction. The method achieves an F1-score of 80.8%, exceeding the best baseline by almost 33 percentage points. The purpose of the method is to enable a more fine-grained analysis of fictional dialogue than has previously been possible, and to provide a component for the further analysis of narrative structure in general.

This study aimed to investigate concerns of validity and reliability in subjective ratings of age-of-acquisition (AoA), through exploring characteristics of the individual rater. An additional aim was to validate the obtained AoA ratings against two corpora – one of child speech and one of adult speech – specifically exploring whether words over-represented in the child-speech corpus are rated with lower AoA than words characteristic of the adult-speech corpus. The results show that less than one-third of participating informants’ ratings are valid and reliable. However, individuals with high familiarity with preschool-aged children provide more valid and reliable ratings, compared to individuals who do not work with or have children of their own. The results further show a significant, age-adjacent difference in rated AoA for words from the two different corpora, thus strengthening their validity. The study provides AoA data, of high specificity, for 100 child-specific and 100 adult-specific Swedish words.

Annotation of second-language learner text is a cumbersome manual task which in turn requires interpretation to postulate the intended meaning of the learner’s language. This paper describes SVALA, a tool which separates the logical steps in this process while providing rich visual support for each of them. The first step is to pseudonymize the learner text to fulfil the legal and ethical requirements for a distributable learner corpus. The second step is to correct the text, which is carried out in the simplest possible way by text editing. During the editing, SVALA automatically maintains a parallel corpus with alignments between words in the learner source text and corrected text, while the annotator may repair inconsistent word alignments. Finally, the actual labelling of the corrections (the postulated errors) is performed. We describe the objectives, design and workflow of SVALA, and our plans for further development.

A neural language model trained on a text corpus can be used to induce distributed representations of words, such that similar words end up with similar representations. If the corpus is multilingual, the same model can be used to learn distributed representations of languages, such that similar languages end up with similar representations. We show that this holds even when the multilingual corpus has been translated into English, by picking up the faint signal left by the source languages. However, just as it is a thorny problem to separate semantic from syntactic similarity in word representations, it is not obvious what type of similarity is captured by language representations. We investigate correlations and causal relationships between language representations learned from translations on one hand, and genetic, geographical, and several levels of structural similarity between languages on the other. Of these, structural similarity is found to correlate most strongly with language representation similarity, whereas genetic relationships—a convenient benchmark used for evaluation in previous work—appears to be a confounding factor. Apart from implications about translation effects, we see this more generally as a case where NLP and linguistic typology can interact and benefit one another.

Error coding of second-language learner text, that is, detecting, correcting and annotating errors, is a cumbersome task which in turn requires interpretation of the text to decide what the errors are. This paper describes a system with which the annotator corrects the learner text by editing it prior to the actual error annotation. During the editing, the system automatically generates a parallel corpus of the learner and corrected texts. Based on this, the work of the annotator consists of three independent tasks that are otherwise often conflated in error coding: correcting the learner text, repairing inconsistent alignments, and performing the actual error annotation.

We present an open-access analytic tool, which allows researchers to simultaneously control for and combine language data from the child, the caregiver, multiple languages, and across multiple time points to make inferences about the social and cognitive factors driving the shape of language development. We demonstrate how the tool works in three domains of language learning and across six languages. The results demonstrate the usefulness of this approach as well as providing deeper insight into three areas of language production and acquisition: egocentric language use, the learnability of nouns versus verbs, and imageability. We have made the Frequency Filter tool freely available as an R-package for other researchers to use at https://github.com/rosemm/FrequencyFilter.

This paper describes an approach to identifying speakers and addressees in dialogues extracted from literary fiction, along with a dataset annotated for speaker and addressee. The overall purpose of this is to provide annotation of dialogue interaction between characters in literary corpora in order to allow for enriched search facilities and construction of social networks from the corpora. To predict speakers and addressees in a dialogue, we use a sequence labeling approach applied to a given set of characters. We use features relating to the current dialogue, the preceding narrative, and the complete preceding context. The results indicate that even with a small amount of training data, it is possible to build a fairly accurate classifier for speaker and addressee identification across different authors, though the identification of addressees is the more difficult task.

This paper reports on the status of learner corpus anonymization for the ongoing research infrastructure project SweLL. The main project aim is to deliver and make available for research a well-annotated corpus of essays written by second language (L2) learners of Swedish. As the practice shows, annotation of learner texts is a sensitive process demanding a lot of compromises between ethical and legal demands on the one hand, and research and technical demands, on the other. Below, is a concise description of the current status of pseudonymization of language learner data to ensure anonymity of the learners, with numerous examples of the above-mentioned compromises.

Deep neural networks have advanced the state of the art in numerous fields, but they generally suffer from low computational efficiency and the level of improvement compared to more efficient machine learning models is not always significant. We perform a thorough PoS tagging evaluation on the Universal Dependencies treebanks, pitting a state-of-the-art neural network approach against UDPipe and our sparse structured perceptron-based tagger, efselab. In terms of computational efficiency, efselab is three orders of magnitude faster than the neural network model, while being more accurate than either of the other systems on 47 of 65 treebanks.

We use automatic processing of 120,000 sign videos in 31 different sign languages to show a cross-linguistic pattern for two types of iconic form–meaning relationships in the visual modality. First, we demonstrate that the degree of inherent plurality of concepts, based on individual ratings by non-signers, strongly correlates with the number of hands used in the sign forms encoding the same concepts across sign languages. Second, we show that certain concepts are iconically articulated around specific parts of the body, as predicted by the associational intuitions by non-signers. The implications of our results are both theoretical and methodological. With regard to theoretical implications, we corroborate previous research by demonstrating and quantifying, using a much larger material than previously available, the iconic nature of languages in the visual modality. As for the methodological implications, we show how automatic methods are, in fact, useful for performing large-scale analysis of sign language data, to a high level of accuracy, as indicated by our manual error analysis.

Child-directed spoken data is the ideal source of support for claims about children’s linguistic environments. However, phonological transcriptions of child-directed speech are scarce,compared to sources like adult-directed speech or text data. Acquiring reliable descriptions of children’s phonological environments from more readily accessible sources would mean considerable savings of time and money. The first step towards this goal is to quantify the reliability of descriptions derived from such secondary sources. We investigate how phonological distributions vary across different modalities (spoken vs. written), and across the age of the intended audience (children vs. adults). Using a previously unseen collection of Swedish adult- and child-directed spoken and written data, we combine lexicon look-up and grapheme-to-phonemeconversion to approximate phonological characteristics. The analysis shows distributional differences across datasets both for single phonemes and for longer phoneme sequences. Some of these are predictably attributed to lexical and contextual characteristics of text vs. speech.The generated phonological transcriptions are remarkably reliable. The differences in phonological distributions between child-directed speech and secondary sources highlight a need for compensatory measures when relying on written data or onadult-directed spoken data, and/or for continued collection ofactual child-directed speech in research on children’s language environments.

In earlier work, we have shown that articulation rate in Swedish child-directed speech (CDS) increases as a function of the age of the child, even when utterance length and differences in articulation rate between subjects are controlled for. In this paper we show on utterance level in spontaneous Swedish speech that i) for the youngest children, articulation rate in CDS is lower than in adult-directed speech (ADS), ii) there is a significant negative correlation between articulation rate and surprisal (the negative log probability) in ADS, and iii) the increase in articulation rate in Swedish CDS as a function of the age of the child holds, even when surprisal along with utterance length and differences in articulation rate between speakers are controlled for. These results indicate that adults adjust their articulation rate to make it fit the linguistic capacity of the child.

Most existing models for multilingual natural language processing (NLP) treat language as a discrete category, and make predictions for either one language or the other. In contrast, we propose using continuous vector representations of language. We show that these can be learned efficiently with a character-based neural language model, and used to improve inference about language varieties not seen during training. In experiments with 1303 Bible translations into 990 different languages, we empirically explore the capacity of multilingual language models, and also show that the language vectors capture genetic relationships between languages.

Assessing the semantic similarity between sentences in different languages is challenging. We approach this problem by leveraging multilingual distributional word representations, where similar words in different languages are close to each other. The availability of parallel data allows us to train such representations on a large amount of languages. This allows us to leverage semantic similarity data for languages for which no such data exists. We train and evaluate on five language pairs, including English, Spanish, and Arabic. We are able to train wellperforming systems for several language pairs, without any labelled data for that language pair.

In this paper, we describe a method for mapping the phonological feature location of Swedish Sign Language (SSL) signs to the meanings in the Swedish semantic dictionary SALDO. By doing so, we observe clear differences in the distribution of meanings associated with different locations on the body. The prominence of certain locations for specific meanings clearly point to iconic mappings between form and meaning in the lexicon of SSL, which pinpoints modalityspecific properties of the visual modality.

We use n-gram language models to investigate how far language approximates an optimal code for human communication in terms of Information Theory [1], and what differences there are between Learner proficiency levels. Although the language of lower level learners is simpler, it is less optimal in terms of information theory, and as a consequence more difficult to process.

The aim of the present study is to further the understanding of the relationship between perceptual categorization and exposure to different frequency distributions of sounds. Previous studies have shown that speech sound discrimination proficiency is in- fluenced by exposure to different distributions of speech sound continua varying along one or several acoustic dimensions, both in adults and in infants. In the current study, adults were presented with either a bimodal or a unimodal frequency distri- bution of spectrally rotated sounds along a continuum (a vowel continuum before rotation). Categorization of the sounds, quantified as amplitude of the event-related potential (ERP) component mismatch negativity (MMN) in response to two of the sounds, was measured before and after exposure. It was expected that the bimodal group would have a larger MMN amplitude after exposure whereas the unimodal group would have a smaller MMN amplitude after exposure. Contrary to expectations, the MMN amplitude was smaller overall after exposure, and no difference was found between groups. This suggests that either the previously reported sensitivity to frequency distributions of speech sounds is not present for non-speech sounds, or the MMN amplitude is not a sensitive enough measure of categorization to detect an influence from passive exposure, or both.

Non-verbal cues from speakers, such as eye gaze and hand positions, play an important role in word learning. This is consistent with the notion that for meaning to be reconstructed, acoustic patterns need to be linked to time-synchronous patterns from at least one other modality. In previous studies of a multimodally annotated corpus of parent–child interaction, we have shown that parents interacting with infants at the early word-learning stage (7–9 months) display a large amount of time-synchronous patterns, but that this behaviour tails off with increasing age of the children. Furthermore, we have attempted to quantify the informativeness of the different nonverbal cues, that is, to what extent they actually help to discriminate between different possible referents, and how critical the timing of the cues is. The purpose of this paper is to generalise our earlier model by quantifying informativeness resulting from non-verbal cues occurring both before and after their associated verbal references.

We present the RUG-SU team's submission at the Native Language Identification Shared Task 2017. We combine several approaches into an ensemble, based on spelling error features, a simple neural network using word representations, a deep residual network using word and character features, and a system based on a recurrent neural network. Our best system is an ensemble of neural networks, reaching an F1 score of 0.8323. Although our system is not the highest ranking one, we do outperform the baseline by far.

For the second year in a row we brought two related themes of NLP for Computer-Assisted Language Learning and NLP for Language Acquisition together. The goal of organizing joint workshops is to provide a meeting place for researchers working on language learning issues including both empirical and experimental studies and NLP-based applications. The resulting volume covers a variety of topics from the two fields and - hopefully - showcases the challenges and achievements in the field.

The seven papers in this volume cover native language identification in learner writings, using syntactic complexity development in language learner language to identify reading comprehension texts of appropriate level, exploring the potential of parallel corpora to predict mother-language specific problem areas for learners of another language, tools for learning languages - both well-resourced ones such as English as well as endangered or under-resourced ones such as Yakut and Võro, as well as exploring the potential of automatically identifying and correcting word-level errors in Swedish learner writing.

This paper describes the Stockholm University/University of Groningen (SU-RUG) system for the SIGMORPHON 2017 shared task on morphological inflection. Our system is based on an attentional sequence-to-sequence neural network model using Long Short-Term Memory (LSTM) cells, with joint training of morphological inflection and the inverse transformation, i.e. lemmatization and morphological analysis. Our system outperforms the baseline with a large margin, and our submission ranks as the 4th best team for the track we participate in (task 1, high resource).

The CLIN27 shared task evaluates the effect of translating historical text to modern text with the goal of improving the quality of the output of contemporary natural language processing tools applied to the text. We focus on improving part-of-speech tagging analysis of seventeenth-century Dutch. Eight teams took part in the shared task. The best results were obtained by teams employing character-based machine translation. The best system obtained an error reduction of 51% in comparison with the baseline of tagging unmodified text. This is close to the error reduction obtained by human translation (57%).

We introduce the Helsinki Neural Machine Translation system (HNMT) and how it is applied in the news translation task at WMT 2017, where it ranked first in both the human and automatic evaluations for English–Finnish. We discuss the successof English–Finnish translations and the overall advantage of NMT over a strong SMT baseline. We also discuss our sub-missions for English–Latvian, English–Chinese and Chinese–English.

We present a very simple model for text quality assessment based on a deep convolutional neural network, where the only supervision required is one corpus of user-generated text of varying quality, and one contrasting text corpus of consistently high quality. Our model is able to provide local quality assessments in different parts of a text, which allows visual feedback about where potentially problematic parts of the text are located, as well as a way to evaluate which textual features are captured by our model. We evaluate our method on two corpora: a large corpus of manually graded student essays and a longitudinal corpus of language learner written production, and find that the text quality metric learned by our model is a fairly strong predictor of both essay grade and learner proficiency level.