Neural metrics trained on human evaluations of MT tend to correlate well with human judgments, but their behavior is not fully understood. In this paper, we perform a controlled experiment and compare a baseline metric that has not been trained on human evaluations (Prism) to a trained version of the same metric (Prism+FT). Surprisingly, we find that Prism+FT becomes more robust to machine-translated references, which are a notorious problem in MT evaluation. This suggests that the effects of metric training go beyond the intended effect of improving overall correlation with human judgments.

Automatically highlighting words that cause semantic differences between two documents could be useful for a wide range of applications. We formulate recognizing semantic differences (RSD) as a token-level regression task and study three unsupervised approaches that rely on a masked language model. To assess the approaches, we begin with basic English sentences and gradually move to more complex, cross-lingual document pairs. Our results show that an approach based on word alignment and sentence-level contrastive learning has a robust correlation to gold labels. However, all unsupervised approaches still leave a large margin of improvement.

Being able to rank the similarity of short text segments is an interesting bonus feature of neural machine translation. Translation-based similarity measures include direct and pivot translation probability, as well as translation cross-likelihood, which has not been studied so far. We analyze these measures in the common framework of multilingual NMT, releasing the NMTScore library. Compared to baselines such as sentence embeddings, translation-based measures prove competitive in paraphrase identification and are more robust against adversarial or multilingual input, especially if proper normalization is applied. When used for reference-based evaluation of data-to-text generation in 2 tasks and 17 languages, translation-based measures show a relatively high correlation to human judgments.

Omission and addition of content is a typical issue in neural machine translation. We propose a method for detecting such phenomena with off-the-shelf translation models. Using contrastive conditioning, we compare the likelihood of a full sequence under a translation model to the likelihood of its parts, given the corresponding source or target sequence. This allows to pinpoint superfluous words in the translation and untranslated words in the source even in the absence of a reference translation. The accuracy of our method is comparable to a supervised method that requires a custom quality estimation model.

Simplified language news articles are being offered by specialized web portals in several countries. The thousands of articles that have been published over the years are a valuable resource for natural language processing, especially for efforts towards automatic text simplification. In this paper, we present SNIML, a large multilingual corpus of news in simplified language. The corpus contains 13k simplified news articles written in one of six languages: Finnish, French, Italian, Swedish, English, and German. All articles are shared under open licenses that permit academic use. The level of text simplification varies depending on the news portal. We believe that even though SNIML is not a parallel corpus, it can be useful as a complement to the more homogeneous but often smaller corpora of news in the simplified variety of one language that are currently in use.

Lexical disambiguation is a major challenge for machine translation systems, especially if some senses of a word are trained less often than others. Identifying patterns of overgeneralization requires evaluation methods that are both reliable and scalable. We propose contrastive conditioning as a reference-free black-box method for detecting disambiguation errors. Specifically, we score the quality of a translation by conditioning on variants of the source that provide contrastive disambiguation cues. After validating our method, we apply it in a case study to perform a targeted evaluation of sequence-level knowledge distillation. By probing word sense disambiguation and translation of gendered occupation names, we show that distillation-trained models tend to overgeneralize more than other models with a comparable BLEU score. Contrastive conditioning thus highlights a side effect of distillation that is not fully captured by standard evaluation metrics. Code and data to reproduce our findings are publicly available.

Minimal sentence pairs are frequently used to analyze the behavior of language models. It is often assumed that model behavior on contrastive pairs is predictive of model behavior at large. We argue that two conditions are necessary for this assumption to hold: First, a tested hypothesis should be well-motivated, since experiments show that contrastive evaluation can lead to false positives. Secondly, test data should be chosen such as to minimize distributional discrepancy between evaluation time and deployment time. For a good approximation of deployment-time decoding, we recommend that minimal pairs are created based on machine-generated text, as opposed to human-written references. We present a contrastive evaluation suite for English–German MT that implements this recommendation.