This paper describes the offline and simultaneous speech translation systems developed at AppTek for IWSLT 2021. Our offline ST submission includes the direct end-to-end system and the so-called posterior tight integrated model, which is akin to the cascade system but is trained in an end-to-end fashion, where all the cascaded modules are end-to-end models themselves. For simultaneous ST, we combine hybrid automatic speech recognition with a machine translation approach whose translation policy decisions are learned from statistical word alignments. Compared to last year, we improve general quality and provide a wider range of quality/latency trade-offs, both due to a data augmentation method making the MT model robust to varying chunk sizes. Finally, we present a method for ASR output segmentation into sentences that introduces a minimal additional delay.

AppTek and RWTH Aachen University team together to participate in the offline and simultaneous speech translation tracks of IWSLT 2020. For the offline task, we create both cascaded and end-to-end speech translation systems, paying attention to careful data selection and weighting. In the cascaded approach, we combine high-quality hybrid automatic speech recognition (ASR) with the Transformer-based neural machine translation (NMT). Our end-to-end direct speech translation systems benefit from pretraining of adapted encoder and decoder components, as well as synthetic data and fine-tuning and thus are able to compete with cascaded systems in terms of MT quality. For simultaneous translation, we utilize a novel architecture that makes dynamic decisions, learned from parallel data, to determine when to continue feeding on input or generate output words. Experiments with speech and text input show that even at low latency this architecture leads to superior translation results.

In simultaneous machine translation, the objective is to determine when to produce a partial translation given a continuous stream of source words, with a trade-off between latency and quality. We propose a neural machine translation (NMT) model that makes dynamic decisions when to continue feeding on input or generate output words. The model is composed of two main components: one to dynamically decide on ending a source chunk, and another that translates the consumed chunk. We train the components jointly and in a manner consistent with the inference conditions. To generate chunked training data, we propose a method that utilizes word alignment while also preserving enough context. We compare models with bidirectional and unidirectional encoders of different depths, both on real speech and text input. Our results on the IWSLT 2020 English-to-German task outperform a wait-k baseline by 2.6 to 3.7% BLEU absolute.

In this work, we customized a neural machine translation system for translation of subtitles in the domain of entertainment. The neural translation model was adapted to the subtitling content and style and extended by a simple, yet effective technique for utilizing inter-sentence context for short sentences such as dialog turns. The main contribution of the paper is a novel subtitle segmentation algorithm that predicts the end of a subtitle line given the previous word-level context using a recurrent neural network learned from human segmentation decisions. This model is combined with subtitle length and duration constraints established in the subtitling industry. We conducted a thorough human evaluation with two post-editors (English-to-Spanish translation of a documentary and a sitcom). It showed a notable productivity increase of up to 37% as compared to translating from scratch and significant reductions in human translation edit rate in comparison with the post-editing of the baseline non-adapted system without a learned segmentation model.

This work describes AppTek’s speech translation pipeline that includes strong state-of-the-art automatic speech recognition (ASR) and neural machine translation (NMT) components. We show how these components can be tightly coupled by encoding ASR confusion networks, as well as ASR-like noise adaptation, vocabulary normalization, and implicit punctuation prediction during translation. In another experimental setup, we propose a direct speech translation approach that can be scaled to translation tasks with large amounts of text-only parallel training data but a limited number of hours of recorded and human-translated speech.