Data augmentation is a technique to generate new training data based on existing data. We evaluate the simple and cost-effective method of concatenating the original data examples to build new training instances. Continued training with such augmented data is able to improve off-the-shelf Transformer and Conformer models that were optimized on the original data only. We demonstrate considerable improvements on the LibriSpeech-960h test sets (WER 2.83 and 6.87 for test-clean and test-other), which carry over to models combined with shallow fusion (WER 2.55 and 6.27). Our method of continued training also leads to improvements of up to 0.9 WER on the ASR part of CoVoST-2 for four non-English languages, and we observe that the gains are highly dependent on the size of the original training data. We compare different concatenation strategies and found that our method does not need speaker information to achieve its improvements. Finally, we demonstrate on two datasets that our methods also works for speech translation tasks.

Direct speech translation describes a scenario where only speech inputs and corresponding translations are available. Such data are notoriously limited. We present a technique that allows cascades of automatic speech recognition (ASR) and machine translation (MT) to exploit in-domain direct speech translation data in addition to out-of-domain MT and ASR data. After pre-training MT and ASR, we use a feed-back cycle where the downstream performance of the MT system is used as a signal to improve the ASR system by self-training, and the MT component is fine-tuned on multiple ASR outputs, making it more tolerant towards spelling variations. A comparison to end-to-end speech translation using components of identical architecture and the same data shows gains of up to 3.8 BLEU points on LibriVoxDeEn and up to 5.1 BLEU points on CoVoST for German-to-English speech translation.

We propose an on-the-fly data augmentation method for automatic speech recognition (ASR) that uses alignment information to generate effective training samples. Our method, called Aligned Data Augmentation (ADA) for ASR, replaces transcribed tokens and the speech representations in an aligned manner to generate previously unseen training pairs. The speech representations are sampled from an audio dictionary that has been extracted from the training corpus and inject speaker variations into the training examples. The transcribed tokens are either predicted by a language model such that the augmented data pairs are semantically close to the original data, or randomly sampled. Both strategies result in training pairs that improve robustness in ASR training. Our experiments on a Seq-to-Seq architecture show that ADA can be applied on top of SpecAugment, and achieves about 9-23% and 4-15% relative improvements in WER over SpecAugment alone on LibriSpeech 100h and LibriSpeech 960h test datasets, respectively.

Neural approaches to learning term embeddings have led to improved computation of similarity and ranking in information retrieval (IR).So far neural representation learning has not been extended to meta-textual information that is readily available for many IR tasks, for example, patent classes in prior-art retrieval, topical information in Wikipedia articles, or product categories in e-commerce data.We present a framework that learns embeddings for meta-textual categories, and optimizes a pairwise ranking objective for improved matching based on combined embeddings of textual and meta-textual information.We show considerable gains in an experimental evaluation on cross-lingual retrieval in the Wikipedia domain for three language pairs, and in the Patent domain for one language pair.Our results emphasize that the mode of combining different types of information is crucial for model improvement.

We present an approach to improve statistical machine translation of image descriptions by multimodal pivots defined in visual space. The key idea is to perform image retrieval over a database of images that are captioned in the target language, and use the captions of the most similar images for crosslingual reranking of translation outputs. Our approach does not depend on the availability of large amounts of in-domain parallel data, but only relies on available large datasets of monolingually captioned images, and on state-of-the-art convolutional neural networks to compute image similarities. Our experimental evaluation shows improvements of 1 BLEU point over strong baselines.

System combination is an effective strategy to boost retrieval performance, especially in complex applications such as cross-language information retrieval (CLIR) where the aspects of translation and retrieval have to be optimized jointly. We focus on machine learning-based approaches to CLIR that need large sets of relevance-ranked data to train high-dimensional models. We compare these models under various measures of orthogonality, and present an experimental evaluation on two different domains (patents, Wikipedia) and two different language pairs (Japanese-English, German-English). We show that gains of over 10 points in MAP/NDCG can be achieved over the best single model by a linear combination of the models that contribute the most orthogonal information, rather than by combining the models with the best standalone retrieval performance.