Part-of-speech (PoS) tagging of non-standard language with models developed for standard language is known to suffer from a significant decrease in accuracy. Two methods are typically used to improve it: word normalisation, which decreases the out-of-vocabulary rate of the PoS tagger, and domain adaptation where the tagger is made aware of the non-standard language variation, either through supervision via non-standard data being added to the tagger’s training set, or via distributional information calculated from raw texts. This paper investigates the two approaches, normalisation and domain adaptation, on carefully constructed data sets encompassing historical and user-generated Slovene texts, in particular focusing on the amount of labour necessary to produce the manually annotated data sets for each approach and comparing the resulting PoS accuracy. We give quantitative as well as qualitative analyses of the tagger performance in various settings, showing that on our data set closed and open class words exhibit significantly different behaviours, and that even small inconsistencies in the PoS tags in the data have an impact on the accuracy. We also show that to improve tagging accuracy, it is best to concentrate on obtaining manually annotated normalisation training data for short annotation campaigns, while manually producing in-domain training sets for PoS tagging is better when a more substantial annotation campaign can be undertaken. Finally, unsupervised adaptation via Brown clustering is similarly useful regardless of the size of the training data available, but improvements tend to be bigger when adaptation is performed via in-domain tagging data.

Some languages have very few NLP resources, while many of them are closely related to better-resourced languages. This paper explores how the similarity between the languages can be utilised by porting resources from better- to lesser-resourced languages. The paper introduces a way of building a representation shared across related languages by combining cross-lingual embedding methods with a lexical similarity measure which is based on the weighted Levenshtein distance. One of the outcomes of the experiments is a Panslavonic embedding space for nine Balto-Slavonic languages. The paper demonstrates that the resulting embedding space helps in such applications as morphological prediction, named-entity recognition and genre classification.

Text normalization is the task of mapping noncanonical language, typical of speech transcription and computer-mediated communication, to a standardized writing. This task is especially important for languages such as Swiss German, with strong regional variation and no written standard. In this paper, we propose a novel solution for normalizing Swiss German WhatsApp messages using the encoder–decoder neural machine translation (NMT) framework. We enhance the performance of a plain character-level NMT model with the integration of a word-level language model and linguistic features in the form of part-of-speech (POS) tags. The two components are intended to improve the performance by addressing two specific issues: the former is intended to improve the fluency of the predicted sequences, whereas the latter aims at resolving cases of word-level ambiguity. Our systematic comparison shows that our proposed solution results in an improvement over a plain NMT system and also over a comparable character-level statistical machine translation system, considered the state of the art in this task till recently. We perform a thorough analysis of the compared systems’ output, showing that our two components produce indeed the intended, complementary improvements.

This paper describes experiments in which I tried to distinguish between Flemish and Netherlandic Dutch subtitles, as originally proposed in the VarDial 2018 Dutch–Flemish Subtitle task. However, rather than using all data as a monolithic block, I divided them into two non-overlapping domains and then investigated how the relation between training and test domains influences the recognition quality. I show that the best estimate of the level of recognizability of the language varieties is derived when training on one domain and testing on another. Apart from the quantitative results, I also present a qualitative analysis, by investigating in detail the most distinguishing features in the various scenarios. Here too, it is with the out-of-domain recognition that some genuine differences between Flemish and Netherlandic Dutch can be found.

Twitter and other social media platforms are often used for sharing interest in products. The identification of purchase decision stages, such as in the AIDA model (Awareness, Interest, Desire, and Action), can enable more personalized e-commerce services and a finer-grained targeting of advertisements than predicting purchase intent only. In this paper, we propose and analyze neural models for identifying the purchase stage of single tweets in a user’s tweet sequence. In particular, we identify three challenges of purchase stage identification: imbalanced label distribution with a high number of non-purchase-stage instances, limited amount of training data, and domain adaptation with no or only little target domain data. Our experiments reveal that the imbalanced label distribution is the main challenge for our models. We address it with ranking loss and perform detailed investigations of the performance of our models on the different output classes. In order to improve the generalization of the models and augment the limited amount of training data, we examine the use of sentiment analysis as a complementary, secondary task in a multitask framework. For applying our models to tweets from another product domain, we consider two scenarios: for the first scenario without any labeled data in the target product domain, we show that learning domain-invariant representations with adversarial training is most promising, while for the second scenario with a small number of labeled target examples, fine-tuning the source model weights performs best. Finally, we conduct several analyses, including extracting attention weights and representative phrases for the different purchase stages. The results suggest that the model is learning features indicative of purchase stages and that the confusion errors are sensible.

Current automatic deception detection approaches tend to rely on cues that are based either on specific lexical items or on linguistically abstract features that are not necessarily motivated by the psychology of deception. Notably, while approaches relying on such features can do well when the content domain is similar for training and testing, they suffer when content changes occur. We investigate new linguistically defined features that aim to capture specific details, a psychologically motivated aspect of truthful versus deceptive language that may be diagnostic across content domains. To ascertain the potential utility of these features, we evaluate them on data sets representing a broad sample of deceptive language, including hotel reviews, opinions about emotionally charged topics, and answers to job interview questions. We additionally evaluate these features as part of a deception detection classifier. We find that these linguistically defined specific detail features are most useful for cross-domain deception detection when the training data differ significantly in content from the test data, and particularly benefit classification accuracy on deceptive documents. We discuss implications of our results for general-purpose approaches to deception detection.