We introduce (1) a novel neural network structure for bilingual modeling of sentence pairs that allows efficient capturing of bilingual relationship via biconstituent composition, (2) the concept of neural network biparsing, which applies to not only machine translation (MT) but also to a variety of other bilingual research areas, and (3) the concept of a biparsing-backpropagation training loop, which we hypothesize that can efficiently learn complex biparse tree patterns. Our work distinguishes from sequential attention-based models, which are more traditionally found in neural machine translation (NMT) in three aspects. First, our model enforces compositional constraints. Second, our model has a smaller search space in terms of discovering bilingual relationships from bilingual sentence pairs. Third, our model produces explicit biparse trees, which enable transparent error analysis during evaluation and external tree constraints during training.

Previous attempts at injecting semantic frame biases into SMT training for low resource languages failed because either (a) no semantic parser is available for the low resource input language; or (b) the output English language semantic parses excise relevant parts of the alignment space too aggressively. We present the first semantic SMT model to succeed in significantly improving translation quality across many low resource input languages for which no automatic SRL is available —consistently and across all common MT metrics. The results we report are the best by far to date for this type of approach; our analyses suggest that in general, easier approaches toward including semantics in training SMT models may be more feasible than generally assumed even for low resource languages where semantic parsers remain scarce. While recent proposals to use the crosslingual evaluation metric XMEANT during inversion transduction grammar (ITG) induction are inapplicable to low resource languages that lack semantic parsers, we break the bottleneck via a vastly improved method of biasing ITG induction toward learning more semantically correct alignments using the monolingual semantic evaluation metric MEANT. Unlike XMEANT, MEANT requires only a readily-available English (output language) semantic parser. The advances we report here exploit the novel realization that MEANT represents an excellent way to semantically bias expectationmaximization induction even for low resource languages. We test our systems on challenging languages including Amharic, Uyghur, Tigrinya and Oromo. Results show that our model influences the learning towards more semantically correct alignments, leading to better translation quality than both the standard ITG or GIZA++ based SMT training models on different datasets.

We introduce a new statistical machine translation approach specifically geared to learning translation from low resource languages, that exploits monolingual English semantic parsing to bias inversion transduction grammar (ITG) induction. We show that in contrast to conventional statistical machine translation (SMT) training methods, which rely heavily on phrase memorization, our approach focuses on learning bilingual correlations that help translating low resource languages, by using the output language semantic structure to further narrow down ITG constraints. This approach is motivated by previous research which has shown that injecting a semantic frame based objective function while training SMT models improves the translation quality. We show that including a monolingual semantic objective function during the learning of the translation model leads towards a semantically driven alignment which is more efficient than simply tuning loglinear mixture weights against a semantic frame based evaluation metric in the final stage of statistical machine translation training. We test our approach with three different language pairs and demonstrate that our model biases the learning towards more semantically correct alignments. Both GIZA++ and ITG based techniques fail to capture meaningful bilingual constituents, which is required when trying to learn translation models for low resource languages. In contrast, our proposed model not only improve translation by injecting a monolingual objective function to learn bilingual correlations during early training of the translation model, but also helps to learn more meaningful correlations with a relatively small data set, leading to a better alignment compared to either conventional ITG or traditional GIZA++ based approaches.

We discuss various improvements to our MEANT tuned system, previously presented at IWSLT 2013. In our 2014 system, we incorporate this year’s improved version of MEANT, improved Chinese word segmentation, Chinese named entity recognition and dedicated proper name translation, and number expression handling. This results in a significant performance jump compared to last year’s system. We also ran preliminary experiments on tuning to IMEANT, our new ITG based variant of MEANT. The performance of tuning to IMEANT is comparable to tuning on MEANT (differences are statistically insignificant). We are presently investigating if tuning on IMEANT can produce even better results, since IMEANT was actually shown to correlate with human adequacy judgment more closely than MEANT. Finally, we ran experiments applying our new architectural improvements to a contrastive system tuned to BLEU. We observed a slightly higher jump in comparison to last year, possibly due to mismatches of MEANT’s similarity models to our new entity handling.

We investigate novel challenges involved in comparing model performance on the task of improvising responses to hip hop lyrics and discuss observations regarding inter-evaluator agreement on judging improvisation quality. We believe the analysis serves as a first step toward designing robust evaluation strategies for improvisation tasks, a relatively neglected area to date. Unlike most natural language processing tasks, improvisation tasks suffer from a high degree of subjectivity, making it difficult to design discriminative evaluation strategies to drive model development. We propose a simple strategy with fluency and rhyming as the criteria for evaluating the quality of generated responses, which we apply to both our inversion transduction grammar based FREESTYLE hip hop challenge-response improvisation system, as well as various contrastive systems. We report inter-evaluator agreement for both English and French hip hop lyrics, and analyze correlation with challenge length. We also compare the extent of agreement in evaluating fluency with that of rhyming, and quantify the difference in agreement with and without precise definitions of evaluation criteria.

We present analyses showing that HMEANT is a reliable, accurate and fine-grained semantic frame based human MT evaluation metric with high inter-annotator agreement (IAA) and correlation with human adequacy judgments, despite only requiring a minimal training of about 15 minutes for lay annotators. Previous work shows that the IAA on the semantic role labeling (SRL) subtask within HMEANT is over 70%. In this paper we focus on (1) the IAA on the semantic role alignment task and (2) the overall IAA of HMEANT. Our results show that the IAA on the alignment task of HMEANT is over 90% when humans align SRL output from the same SRL annotator, which shows that the instructions on the alignment task are sufficiently precise, although the overall IAA where humans align SRL output from different SRL annotators falls to only 61% due to the pipeline effect on the disagreement in the two annotation task. We show that instead of manually aligning the semantic roles using an automatic algorithm not only helps maintaining the overall IAA of HMEANT at 70%, but also provides a finer-grained assessment on the phrasal similarity of the semantic role fillers. This suggests that HMEANT equipped with automatic alignment is reliable and accurate for humans to evaluate MT adequacy while achieving higher correlation with human adequacy judgments than HTER.

We present the results of large-scale human semantic MT evaluation with HMEANT on the IWSLT 2013 German-English MT and SLT tracks and show that HMEANT evaluates the performance of the MT systems differently compared to BLEU and TER. Together with the references, all the translations are annotated by annotators who are native English speakers in both semantic role labeling stage and role filler alignment stage of HMEANT. We obtain high inter-annotator agreement and low annotation time costs which indicate that it is feasible to run a large-scale human semantic MT evaluation campaign using HMEANT. Our results also show that HMEANT is a robust and reliable semantic MT evaluation metric for running large-scale evaluation campaigns as it is inexpensive and simple while maintaining the semantic representational transparency to provide a perspective which is different from BLEU and TER in order to understand the performance of the state-of-the-art MT systems.

We present the first ever results showing that Chinese MT output is significantly improved by tuning a MT system against a semantic frame based objective function, MEANT, rather than an n-gram based objective function, BLEU, as measured across commonly used metrics and different test sets. Recent work showed that by preserving the meaning of the translations as captured by semantic frames in the training process, MT systems for translating into English on both formal and informal genres are constrained to produce more adequate translations by making more accurate choices on lexical output and reordering rules. In this paper we describe our experiments in IWSLT 2013 TED talk MT tasks on tuning MT systems against MEANT for translating into Chinese and English respectively. We show that the Chinese translation output benefits more from tuning a MT system against MEANT than the English translation output due to the ambiguous nature of word boundaries in Chinese. Our encouraging results show that using MEANT is a promising alternative to BLEU in both evaluating and tuning MT systems to drive the progress of MT research across different languages.

We present the first known experiments incorporating unsupervised bilingual nonterminal category learning within end-to-end fully unsupervised transduction grammar induction using matched training and testing models. Despite steady recent progress, such induction experiments until now have not allowed for learning differentiated nonterminal categories. We divide the learning into two stages: (1) a bootstrap stage that generates a large set of categorized short transduction rule hypotheses, and (2) a minimum conditional description length stage that simultaneously prunes away less useful short rule hypotheses, while also iteratively segmenting full sentence pairs into useful longer categorized transduction rules. We show that the second stage works better when the rule hypotheses have categories than when they do not, and that the proposed conditional description length approach combines the rules hypothesized by the two stages better than a mixture model does. We also show that the compact model learned during the second stage can be further improved by combining the result of different iterations in a mixture model. In total, we see a jump in BLEU score, from 17.53 for a standalone minimum description length baseline with no category learning, to 20.93 when incorporating category induction on a Chinese–English translation task.

Over the past twenty years, we have attacked the historical methodological barriers between statistical machine translation and traditional models of syntax, semantics, and structure. In this tutorial, we will survey some of the central issues and techniques from each of these aspects, with an emphasis on `deeply theoretically integrated' models, rather than hybrid approaches such as superficial statistical aggregation or system combination of outputs produced by traditional symbolic components. On syntactic SMT, we will explore the trade-offs for SMT between learnability and representational expressiveness. After establishing a foundation in the theory and practice of stochastic transduction grammars, we will examine very recent new approaches to automatic unsupervised induction of various classes of transduction grammars. We will show why stochastic linear transduction grammars (LTGs and LITGs) and their preterminalized variants (PLITGs) are proving to be particularly intriguing models for the bootstrapping of inducing full-fledged stochastic inversion transduction grammars (ITGs). On semantic SMT, we will explore the trade-offs for SMT involved in applying various lexical semantics models. We will first examine word sense disambiguation, and discuss why traditional WSD models that are not deeply integrated within the SMT model tend, surprisingly, to fail. In contrast, we will show how a deeply embedded phrase sense disambiguation (PSD) approach succeeds where traditional WSD does not. We will then turn to semantic role labeling, and discuss the challenges of early approaches of applying SRL models to SMT. Finally, on semantic MT evaluation, we will explore some very new human and semi-automatic metrics based on semantic frame agreement. We show that by keeping the metrics deeply grounded within the theoretical framework of semantic frames, the new HMEANT and MEANT metrics can significantly outperform even the state-of-the-art expensive HTER and TER metrics, while at the same time maintaining the desirable characteristics of simplicity, inexpensiveness, and representational transparency.

We present the methodology that underlies mew metrics for semantic machine translation evaluation we are developing. Unlike widely-used lexical and n-gram based MT evaluation metrics, the aim of semantic MT evaluation is to measure the utility of translations. We discuss the design of empirical studies to evaluate the utility of machine translation output by assessing the accuracy for key semantic roles. These roles are from the English 5W templates (who, what, when, where, why) used in recent GALE distillation evaluations. Recent work by Wu and Fung (2009) introduced semantic role labeling into statistical machine translation to enhance the quality of MT output. However, this approach has so far only been evaluated using lexical and n-gram based SMT evaluation metrics like BLEU which are not aimed at evaluating the utility of MT output. Direct data analysis are still needed to understand how semantic models can be leveraged to evaluate the utility of MT output. In this paper, we discuss a new methodology for evaluating the utility of the machine translation output, by assessing the accuracy with which human readers are able to complete the English 5W templates.

We present new direct data analysis showing that dynamically-built context-dependent phrasal translation lexicons are more useful resources for phrase-based statistical machine translation (SMT) than conventional static phrasal translation lexicons, which ignore all contextual information. After several years of surprising negative results, recent work suggests that context-dependent phrasal translation lexicons are an appropriate framework to successfully incorporate Word Sense Disambiguation (WSD) modeling into SMT. However, this approach has so far only been evaluated using automatic translation quality metrics, which are important, but aggregate many different factors. A direct analysis is still needed to understand how context-dependent phrasal translation lexicons impact translation quality, and whether the additional complexity they introduce is really necessary. In this paper, we focus on the impact of context-dependent translation lexicons on lexical choice in phrase-based SMT and show that context-dependent lexicons are more useful to a phrase-based SMT system than a conventional lexicon. A typical phrase-based SMT system makes use of more and longer phrases with context modeling, including phrases that were not seen very frequently in training. Even when the segmentation is identical, the context-dependent lexicons yield translations that match references more often than conventional lexicons.

This paper describes the HKUST experiments in the IWSLT 2007 evaluation campaign on spoken language translation. Our primary objective was to compare the open-source phrase-based statistical machine translation toolkit Moses against Pharaoh. We focused on Chinese to English translation, but we also report results on the Arabic to English, Italian to English, and Japanese to English tasks.