The encoder-decoder framework for neural machine translation (NMT) has been shown effective in large data scenarios, but is much less effective for low-resource languages. We present a transfer learning method that significantly improves BLEU scores across a range of low-resource languages. Our key idea is to first train a high-resource language pair (the parent model), then transfer some of the learned parameters to the low-resource pair (the child model) to initialize and constrain training. Using our transfer learning method we improve baseline NMT models by an average of 5.6 BLEU on four low-resource language pairs. Ensembling and unknown word replacement add another 2 BLEU which brings the NMT performance on low-resource machine translation close to a strong syntax based machine translation (SBMT) system, exceeding its performance on one language pair. Additionally, using the transfer learning model for re-scoring, we can improve the SBMT system by an average of 1.3 BLEU, improving the state-of-the-art on low-resource machine translation.

Automatic storytelling is challenging since it requires generating long, coherent natural language to describes a sensible sequence of events. Despite considerable efforts on automatic story generation in the past, prior work either is restricted in plot planning, or can only generate stories in a narrow domain. In this paper, we explore open-domain story generation that writes stories given a title (topic) as input. We propose a plan-and-write hierarchical generation framework that first plans a storyline, and then generates a story based on the storyline. We compare two planning strategies. The dynamic schema interweaves story planning and its surface realization in text, while the static schema plans out the entire storyline before generating stories. Experiments show that with explicit storyline planning, the generated stories are more diverse, coherent, and on topic than those generated without creating a full plan, according to both automatic and human evaluations.

We present a syntax-based statistical translation model. Our model transforms a source-language parse tree into a target-language string by applying stochastic operations at each node. These operations capture linguistic differences such as word order and case marking. Model parameters are estimated in polynomial time using an EM algorithm. The model produces word alignments that are better than those produced by IBM Model 5.

When humans produce summaries of documents, they do not simply extract sentences and concatenate them. Rather, they create new sentences that are grammatical, that cohere with one another, and that capture the most salient pieces of information in the original document. Given that large collections of text/abstract pairs are available online, it is now possible to envision algorithms that are trained to mimic this process. In this paper, we focus on sentence compression, a simpler version of this larger challenge. We aim to achieve two goals simultaneously: our compressions should be grammatical, and they should retain the most important pieces of information. These two goals can conflict. We devise both a noisy-channel and a decision-tree approach to the problem, and we evaluate results against manual compressions and a simple baseline.

We propose a theory that gives formal semantics to word-level alignments defined over parallel corpora. We use our theory to introduce a linear algorithm that can be used to derive from word-aligned, parallel corpora the minimal set of syntactically motivated transformation rules that explain human translation data.

Statistical MT has made great progress in the last few years, but current translation models are weak on reordering and target language fluency. Syntactic approaches seek to remedy these problems. In this paper, we take the framework for acquiring multi-level syntactic translation rules of (Galley et al., 2004) from aligned tree-string pairs, and present two main extensions of their approach: first, instead of merely computing a single derivation that minimally explains a sentence pair, we construct a large number of derivations that include contextually richer rules, and account for multiple interpretations of unaligned words. Second, we propose probability estimates and a training procedure for weighting these rules. We contrast different approaches on real examples, show that our estimates based on multiple derivations favor phrasal re-orderings that are linguistically better motivated, and establish that our larger rules provide a 3.63 BLEU point increase over minimal rules.

We investigate whether a neural, encoderdecoder translation system learns syntactic information on the source side as a by-product of training. We propose two methods to detect whether the encoder has learned local and global source syntax. A fine-grained analysis of the syntactic structure learned by the encoder reveals which kinds of syntax are learned and which are missing.

The ambitious goal of this work is to develop a cross-lingual name tagging and linking framework for 282 languages that exist in Wikipedia. Given a document in any of these languages, our framework is able to identify name mentions, assign a coarse-grained or fine-grained type to each mention, and link it to an English Knowledge Base (KB) if it is linkable. We achieve this goal by performing a series of new KB mining methods: generating "silver-standard" annotations by transferring annotations from English to other languages through crosslingual links and KB properties, refining annotations through self-training and topic selection, deriving language-specific morphology features from anchor links, and mining word translation pairs from crosslingual links. Both name tagging and linking results for 282 languages are promising on Wikipedia data and on-Wikipedia data. All the data sets, resources and systems for 282 languages are made publicly available as a new benchmark 1 .