Controlling Utterance Length in NMT-based Word Segmentation with Attention

One of the basic tasks of computational language documentation (CLD) is to identify\r\nword boundaries in an unsegmented phonemic stream. While several unsupervised\r\nmonolingual word segmentation algorithms exist in the literature,\r\nthey are challenged in real-world CLD settings by the small amount of available\r\ndata. A possible remedy is to take advantage of glosses or translation in a foreign,\r\nwell-resourced, language, which often exist for such data. In this paper, we explore and compare\r\nways to exploit neural machine translation models to perform unsupervised boundary detection with bilingual information, notably introducing a new loss function for jointly learning alignment and segmentation. We experiment with an actual under-resourced language, Mboshi, and show that these techniques can effectively control the output segmentation length.