Search

Neural models for text generation are often designed in an end-to-end fashion, typically with zero control over intermediate computations, limiting their practical usability in downstream applications. In this work, we incorporate explicit means into neural models to ensure topical continuity, informativeness and content diversity of generated radiology reports. For the purpose we propose a method to compute image representations specific to each sentential context and eliminate redundant content by exploiting diverse sentence states. We conduct experiments to generate radiology reports from medical images of chest x-rays using MIMIC-CXR. Our model outperforms baselines by up to 18% and 29% respective in the evaluation for informativeness and content ordering respectively, relative on objective metrics and 16% on human evaluation.

Customized translation need pay spe-cial attention to the target domain ter-minology especially the named-entities for the domain. Adding linguistic features to neural machine translation (NMT) has been shown to benefit translation in many studies. In this paper, we further demonstrate that adding named-entity (NE) feature with named-entity recognition (NER) into the source language produces better translation with NMT. Our experiments show that by just including the different NE classes and boundary tags, we can increase the BLEU score by around 1 to 2 points using the standard test sets from WMT2017. We also show that adding NE tags using NER and applying in-domain adaptation can be combined to further improve customized machine translation.

Syntactic analysis influences the way in which the source sentence is translated. Previous efforts add syntactic constraints to phrase-based translation by directly rewarding/punishing a hypothesis whenever it matches/violates source-side constituents. We present a new model that automatically learns syntactic constraints, including but not limited to constituent matching/violation, from training corpus. The model brackets a source phrase as to whether it satisfies the learnt syntactic constraints. The bracketed phrases are then translated as a whole unit by the decoder. Experimental results and analysis show that the new model outperforms other previous methods and achieves a substantial improvement over the baseline which is not syntactically informed.

Recently confusion network decoding shows the best performance in combining outputs from multiple machine translation (MT) systems. However, overcoming different word orders presented in multiple MT systems during hypothesis alignment still remains the biggest challenge to confusion network-based MT system combination. In this paper, we compare four commonly used word alignment methods, namely GIZA++, TER, CLA and IHMM, for hypothesis alignment. Then we propose a method to build the confusion network from intersection word alignment, which utilizes both direct and inverse word alignment between the backbone and hypothesis to improve the reliability of hypothesis alignment. Experimental results demonstrate that the intersection word alignment yields consistent performance improvement for all four word alignment methods on both Chinese-to-English spoken and written language tasks.

In this paper, we present a feature-based method to align documents with similar content across two sets of bilingual comparable corpora from daily news texts. We evaluate the contribution of each individual feature and investigate the incorporation of these diverse statistical and heuristic features for the task of bilingual document alignment. Experimental results on the English-Chinese and English-Malay comparable news corpora show that our proposed Discrete Fourier Transform-based term frequency distribution feature is very effective. It contributes 4.1% and 8% to performance improvement over Pearson's correlation method on the two comparable corpora. In addition, when more heuristic and statistical features as well as a bilingual dictionary are utilized, our method shows an absolute performance improvement of 23.2% and 15.3% on the two sets of bilingual corpora when comparing with a prior information retrieval-based method.

This paper studies three techniques that improve the quality of N-best hypotheses through additional regeneration process. Unlike the multi-system consensus approach where multiple translation systems are used, our improvement is achieved through the expansion of the N-best hypotheses from a single system. We explore three different methods to implement the regeneration process: redecoding, n-gram expansion, and confusion network-based regeneration. Experiments on Chinese-to-English NIST and IWSLT tasks show that all three methods obtain consistent improvements. Moreover, the combination of the three strategies achieves further improvements and outperforms the baseline by 0.81 BLEU-score on IWSLT'06, 0.57 on NIST'03, 0.61 on NIST'05 test set respectively.

Short Messaging Service (SMS) texts behave quite differently from normal written texts and have some very special phenomena. To translate SMS texts, traditional approaches model such irregularities directly in Machine Translation (MT). However, such approaches suffer from customization problem as tremendous effort is required to adapt the language model of the existing translation system to handle SMS text style. We offer an alternative approach to resolve such irregularities by normalizing SMS texts before MT. In this paper, we view the task of SMS normalization as a translation problem from the SMS language to the English language and we propose to adapt a phrase-based statistical MT model for the task. Evaluation by 5-fold cross validation on a parallel SMS normalized corpus of 5000 sentences shows that our method can achieve 0.80702 in BLEU score against the baseline BLEU score 0.6958. Another experiment of translating SMS texts from English to Chinese on a separate SMS text corpus shows that, using SMS normalization as MT preprocessing can largely boost SMS translation performance from 0.1926 to 0.3770 in BLEU score.