Developing natural language processing (NLP) systems that can work indistinctly with different input languages is a challenging task; yet, such a setup is useful for many real-world applications. One expensive solution is to annotate data for each input language and then to train a separate system for each one. Another option, which can be also costly, is to translate the input, e.g., using machine translation (MT), and then to work monolingually in the target languageHartrumpf et al. (2008); Lin and Kuo (2010); Ture and Boschee (2016). However, the machine-translated text can be of low quality, might lose some input signal, e.g., it can alter sentiment Mohammad et al. (2016), or may not be really needed Bouma et al. (2008); Pouran Ben Veyseh (2016). Using a unified cross-language representation of the input is a third, less costly option, which allows any combination of input languages during both training and testing.
In this paper, we take this last approach, i.e., combining languages during both training and testing, and we study the problem of question-question similarity reranking in community Question Answering (cQA), when the input question can be either in English or in Arabic, and the questions it is compared to are always in English. We start with a simple language-independent representation based on cross-language word embeddings, which we input into a feed-forward multilayer neural network to classify pairs of questions, (English, English) or (Arabic, English), regarding their similarity.
Furthermore, we explore the question of whether adversarial training can be used to improve the performance of the network when we have some unlabeled examples in the target language. In particular, we adapt the Domain Adversarial Neural Network model from Ganin et al. (2016), which was originally used for domain adaptation, to our cross-language setting. To the best of our knowledge, this is novel for cross-language question-question similarity reranking, as well as for natural language processing (NLP) in general; moreover, we are not aware of any previous work on cross-language question reranking for community Question Answering.
In our setup, the basic task-solving network is paired with another network that shares the internal representation of the input and tries to decide whether the input example comes from the source (English) or from the target (Arabic) language. The training of this language discriminator network is adversarial
with respect to the shared layers by using gradient reversal during backpropagation, which makes the training tomaximize the loss of the discriminator rather than to minimize it. The main idea is to learn a high-level abstract representation that is discriminative for the main classification task, but is invariant across the input languages.
We apply this method to an extension of the SemEval-2016 Task 3, subtask B benchmark dataset for question-question similarity reranking Nakov et al. (2016b). In particular, we hired professional translators to translate the original English questions to Arabic, and we further collected additional unlabeled questions in English, which we also got translated into Arabic. We show that using the unlabeled data for adversarial training allows us to improve the results by a sizable margin in both directions, i.e., when training on English and adapting the system with the Arabic unlabeled data, and vice versa. Moreover, the resulting performance is comparable to the best monolingual English systems at SemEval. We also compare our unsupervised model to a semi-supervised model, where we have some labeled data for the target language.
The remainder of this paper is organized as follows: Section 2 discusses some related work. Section 3 introduces our model for adversarial training for cross-language problems. Section 4 describes the experimental setup. Section 5 presents the evaluation results. Finally, Section 6 concludes and points to possible directions for future work.
2 Related Work
Below we discuss three relevant research lines: (a) adversarial training, (b) question-question similarity, and (c) cross-language learning.
of neural networks has shown a big impact recently, especially in areas such as computer vision, where generative unsupervised models have proved capable of synthesizing new imagesGoodfellow et al. (2014); Radford et al. (2016); Makhzani et al. (2016). One crucial challenge in adversarial training is to find the right balance between the two components: the generator and the adversarial discriminator. Thus, several methods have been proposed recently to stabilize training Metz et al. (2017); Arjovsky et al. (2017). Adversarial training has also been successful in training predictive models. More relevant to our work is the work of Ganin:2016:DTN:2946645.2946704, who proposed domain adversarial neural networks (DANN) to learn discriminative but at the same time domain-invariant representations, with domain adaptation as a target. Here, we use adversarial training to learn task-specific representations in a cross-language setting, which is novel for this task, to the best of our knowledge.
Question-question similarity was part of Task 3 on cQA at SemEval-2016/2017 Nakov et al. (2016b, 2017); there was also a similar subtask as part of SemEval-2016 Task 1 on Semantic Textual Similarity Agirre et al. (2016). Question-question similarity is an important problem with application to question recommendation, question duplicate detection, community question answering, and question answering in general. Typically, it has been addressed using a variety of textual similarity measures. Some work has paid attention to modeling the question topic, which can be done explicitly, e.g., using a graph of topic terms Cao et al. (2008), or implicitly, e.g., using LDA-based topic language model that matches the questions not only at the term level but also at the topic level Zhang et al. (2014). Another important aspect is syntactic structure, e.g., Wang et al. (2009) proposed a retrieval model for finding similar questions based on the similarity of syntactic trees, and Da San Martino et al. (2016) used syntactic kernels. Yet another emerging approach is to use neural networks, e.g., dos Santos et al. (2015)
used convolutional neural networks (CNNs),Romeo et al. (2016)
used long short-term memory (LSTMs) networks with neural attention to select the important part when comparing two questions, andLei et al. (2016) used a combined recurrent–convolutional model to map questions to continuous semantic representations. Finally, translation models have been popular for question-question similarity Jeon et al. (2005); Zhou et al. (2011). Unlike that work, here we are interested in cross-language adaptation for question-question similarity reranking. The problem was studied in Martino et al. (2017) using cross-language kernels and deep neural networks; however, they used no adversarial training.
Cross-language Question Answering was the topic of several challenges, e.g., at CLEF 2008 Forner et al. (2008), at NTCIR-8 Mitamura et al. (2010), and at BOLT Soboroff et al. (2016). Cross-language QA methods typically use machine translation directly or adapt MT models to the QA setting Berger et al. (2000); Echihabi and Marcu (2003); Soricut and Brill (2006); Riezler et al. (2007); Hartrumpf et al. (2008); Lin and Kuo (2010); Surdeanu et al. (2011); Ture and Boschee (2016). They can also map terms across languages using Wikipedia links or BabelNet Bouma et al. (2008); Pouran Ben Veyseh (2016). However, adversarial training has not been tried in that setting.
3 Adversarial Training for Cross-Language Problems
We demonstrate our approach for cross-language representation learning with adversarial training on a cross-lingual extension of the question–question similarity reranking subtask of SemEval-2016 Task 3 on community Question Answering.
An example for the monolingual task is shown in Figure 1. We can see an original English input question and a list of several potentially similar questions from the Qatar Living111http://www.qatarliving.com/forum forum, retrieved by a search engine. The original question (also referred to as a new question) asks about how to tip in Qatar. Question is relevant with respect to it as it asks the same thing, and so is , which asks how much one should tip in a specific situation. However, and are irrelevant: the former asks about what to wear at business meetings, and the latter asks about how to tip a kind of person who does not normally receive tips.
In our case, the input question is in a different language (Arabic) than the language of the retrieved questions (English). The goal is to rerank a set of retrieved questions written in a source language (e.g., English) according to their similarity with respect to an input user question that comes in another (target) language, e.g., Arabic. For simplicity, henceforth we will use Arabic as target and English as source. However, in principle, our method generalizes to any source-target language pair.
3.1 Unsupervised Language Adaptation
We approach the problem as a classification task, where given a question pair , the goal is to decide whether the retrieved question is similar (i.e., relevant) to or not. Let
denote the class label: 1 for similar, and 0 for not similar. We use the posterior probabilityas a score for ranking all retrieved questions by similarity, where are the model parameters.
More formally, let denote the set of retrieved questions for a new question . Note that the questions in are always in English. We consider a training scenario where we have labeled examples for English , but we only have unlabeled examples for Arabic , with denoting the class label for the pair . We want to train a cross-language model that can classify any test example , where is in Arabic. This scenario is of practical importance, e.g., when an Arabic speaker wants to query the system in Arabic, and the database of related information is only in English. Here, we adapt the idea for adversarial training for domain adaptation as proposed by Ganin:2016:DTN:2946645.2946704.
Figure 2 shows the architecture of our cross-language adversarial neural network (CLANN) model. The input to the network is a pair
, which is first mapped to fixed-length vectors. To generate these word embeddings, one can use existing tools such as word2vec Mikolov et al. (2013) and monolingual data from the respective languages. Alternatively, one can use cross-language word embeddings, e.g., trained using the bivec model Luong et al. (2015). The latter can yield better initialization, which could be potentially crucial when the labeled data is too small to train the input representations with the end-to-end system.
The network then models the interactions between the input embeddings by passing them through two non-linear hidden layers, and . Additionally, the network considers pairwise features that go directly to the output layer, and also through the second hidden layer.
The following equations describe the transformations through the hidden layers:
where denotes concatenation of two column vectors, and are the weight matrices in the first and in the second hidden layer, and2010).
, we found it beneficial to use them directly to the output layer as well as through a hidden-layer transformation. The non-linear transformation allows us to learn high-level abstract features from the raw similarity measures, while the adversarial training, as we describe below, will make these abstract features language-invariant.
The output layer computes a sigmoid:
where are the output layer weights.
We train the network by minimizing the negative log-probability of the gold labels:
The network described so far learns the abstract features through multiple hidden layers that are discriminative for the classification task, i.e., similar vs. non-similar. However, our goal is also to make these features invariant across languages. To this end, we put a language discriminator, another neural network that takes the internal representation of the network (see Equation 2) as input, and tries to discriminate between English and Arabic inputs — in our case, whether the input comes from or from .
The language discriminator is again defined by a sigmoid function:
where denotes the language of (1 for English, and 0 for Arabic), are the final layer weights of the discriminator, and defines the hidden layer of the discriminator with being the layer weights and being the ReLU activations.
We use the negative log-probability as the discrimination loss:
The overall training objective of the composite model can be written as follows:
where , , and the hyper-parameter controls the relative strength of the two networks.
In training, we look for parameter values that satisfy a min-max optimization criterion as follows:
which involves a maximization (gradient ascent) with respect to and a minimization (gradient descent) with respect to . Note that maximizing with respect to is equivalent to minimizing the discriminator loss in Equation (6), which aims to improve the discrimination accuracy. In other words, when put together, the updates of the shared parameters for the two classifiers work adversarially with respect to each other.
In our gradient descent training, the above min-max optimization is performed by reversing the gradients of the language discrimination loss , when they are backpropagated to the shared layers. As shown in Figure 2, the gradient reversal is applied to layer and also to the layers that come before it.
Our optimization setup is related to the training method of Generative Adversarial Networks or GANs Goodfellow et al. (2014), where the goal is to build deep generative models that can generate realistic images. The discriminator in GANs tries to distinguish real images from model-generated images, and thus the training attempts to minimize the discrepancy between the two image distributions, i.e., empirical as in the training data vs. model-based as produced by the generator. When backpropagating to the generator network, they consider a slight variation of the reverse gradients with respect to the discriminator loss. In particular, if is the discriminator probability, instead of reversing the gradients of , they use the gradients of . Reversing the gradient is a different way to achieve the same goal.
shows pseudocode for the algorithm we use to train our model, which is based on stochastic gradient descent (SDG). We first initialize the model parameters by using samples from glorot-uniform distributionGlorot and Bengio (2010). We then form minibatches of size by randomly sampling labeled examples from and unlabeled examples from . For the labeled instances, both and losses are active, while only the loss is active for the unlabeled instances.
As mentioned above, the main challenge in adversarial training is to balance the two components of the network. If one component becomes smarter, its loss to the shared layer becomes useless, and the training fails to converge Arjovsky et al. (2017). Equivalently, if one component gets weaker, its loss overwhelms that of the other, causing training to fail. In our experiments, the language discriminator was weaker. This could be due to the use of cross-language word embeddings to generate input embedding representations for and . To balance the two components, we would want the error signals from the discriminator to be fairly weak initially, with full power unleashed only as the classification errors start to dominate. We follow the weighting schedule proposed by [p. 21]Ganin:2016:DTN:2946645.2946704, who initialize to , and then change it gradually to as training progresses. I.e., we start training the task classifier first, and we gradually add the discriminator’s loss.
3.2 Semi-supervised Extension
Above we considered an unsupervised adaptation scenario, where we did not have any labeled instance for the target language, i.e., when the new question is in Arabic. However, our method can be easily generalized to a semi-supervised setting, where we have access to some labeled instances in the target language, . In this case, each minibatch during training is formed by labeled instances from both and , and unlabeled instances from .
|System||Input||Discrim.||Target||Hyperparam. (b, d, , , )||MAP||MRR||AvgRec|
|FNN||en||–||ar||8, 0.2, 10, 100, 0.03||75.28||84.26||89.48|
|CLANN||en||en vs. ar’||ar||8, 0.2, 15, 100, 0.02||76.64||84.52||90.92|
|FNN||ar||–||en||8, 0.4, 20, 125, 0.03||75.32||84.17||89.26|
|CLANN||ar||ar vs. en’||en||8, 0.4, 15, 75, 0.02||76.70||84.52||90.61|
4 Experimental Setting
In this section, we describe the datasets we used, the generation of the input embeddings, the nature of the pairwise features, and the general training setup of our model.
SemEval-2016 Task 3 Nakov et al. (2016b), provides 267 input questions for training, 50 for development, and 70 for testing, and ten times as many potentially related questions retrieved by an IR engine for each input question: 2,670, 500, and 700, respectively. Based on this data, we simulated a cross-language setup for question-question similarity reranking. We first got the 387 original train+dev+test questions translated into Arabic by professional translators. Then, we used these Arabic questions as an input with the goal to rerank the ten related English questions. As an example, this is the Arabic translation of the original English question from Figure 1:
هل تعطون الاكراميات؟ ماذا تفعلون بهذا الشأن؛¿
اذا كانت الاجابة نعم ، ما هو قوة ما تتجنبونه؟¿
في بلادنا ، نترك من 15 إلى 20 بالمئة.¿
We further collected 221 additional original questions and 1,863 related questions as unlabeled data, and we got the 221 English questions translated to Arabic.222Our cross-language dataset and code are available at https://github.com/qcri/CLANN
4.2 Cross-language Embeddings
We used the TED Abdelali et al. (2014) and the OPUS parallel Arabic–English bi-texts Tiedemann (2012) to extract a bilingual dictionary, and to learn cross-language embeddings. We chose these bi-texts as they are conversational (TED talks and movie subtitles, respectively), and thus informal, which is close to the style of our community question answering forum.
We trained Arabic-English cross-language word embeddings from the concatenation of these bi-texts using bivec Luong et al. (2015), a bilingual extension of word2vec, which has achieved excellent results on semantic tasks close to ours Upadhyay et al. (2016). In particular, we trained 200-dimensional vectors using the parameters described in Upadhyay et al. (2016)
, with a context window of size 5 and iterating for 5 epochs. We then compute the representation for a question by averaging the embedding vectors of the words it contains. Using these cross-language embeddings allows us to compare directly representations of an Arabic or an English input questionto English potentially related questions .
4.3 Pairwise Features
In addition to the embeddings, we also used some pairwise features that model the similarity or some other relation between the input question and the potentially related questions.333This required translating the Arabic input question to English. For this, we used an in-house Arabic–English phrase-based statistical machine translation system, trained on the TED and on the OPUS bi-texts; for language modeling, it also used the English Gigaword corpus. These features were proposed in the previous literature for the question–question similarity problem, and they are necessary to obtain state-of-the-art results.
In particular, we calculated the similarity between the two questions using machine translation evaluation metrics, as suggested inGuzmán et al. (2016). In particular, we used Bleu Papineni et al. (2002); NIST Doddington (2002); TER v0.7.25 Snover et al. (2006); Meteor v1.4 Lavie and Denkowski (2009) with paraphrases; Unigram Precision; Unigram Recall. We also used features that model various components of Bleu, as proposed in Guzmán et al. (2015): -gram precisions, -gram matches, total number of -grams (=1,2,3,4), hypothesis and reference length, length ratio, and brevity penalty.
We further used as features the cosine similarity between question embeddings. In particular, we used (i) 300-dimensional pre-trained Google News embeddings from Mikolov et al. (2013), (ii) 100-dimensional embeddings trained on the entire Qatar Living forum Mihaylov and Nakov (2016), and (iii) 25-dimensional Stanford neural parser embeddings Socher et al. (2013). The latter are produced by the parser internally, as a by-product.
Furthermore, we computed various task-specific features, most of them introduced in the 2015 edition of the SemEval task by nicosia-EtAl:2015:SemEval. This includes some question-level features: (1) number of URLs/images/emails/phone numbers; (2) number of tokens/sentences; (3) average number of tokens; (4) type/token ratio; (5) number of nouns/verbs/adjectives/adverbs/ pronouns; (6) number of positive/negative smileys; (7) number of single/double/ triple exclamation/interrogation symbols; (8) number of interrogative sentences (based on parsing); (9) number of words that are not in word2vec’s Google News vocabulary. Also, some question-question pair features: (10) count ratio in terms of sentences/tokens/nouns/verbs/ adjectives/adverbs/pronouns; (11) count ratio of words that are not in word2vec’s Google News vocabulary. Finally, we also have one meta feature: (12) reciprocal rank of the related question in the list of related questions.
4.4 Model settings
We trained our CLANN model by optimizing the objective in Equation (7) using ADAM Kingma and Ba (2015) with default parameters. For this, we used up to 200 epochs. In order to avoid overfitting, we used dropout Srivastava et al. (2014) of hidden units, regularization on weights, and early stopping by observing MAP on the development dataset —if MAP did not increase for consecutive epochs, we exited with the best model recorded so far. We optimized the values of the hyper-parameters using grid search: for minibatch (b) size in , for dropout (d) rate in , for layer size in , for layer size in , and for strength in . The fifth column in Table 1 shows the optimal hyper-parameter setting for the different models. Finally, we used the best model as found on the development dataset for the final evaluation on the test dataset.
5 Evaluation Results
Below we present the experimental results for the unsupervised and semi-supervised language adaptation settings. We compare our cross-language adversarial network (CLANN) to a feed forward neural network (FNN) that has no adversarial part.
5.1 Unsupervised Adaptation Experiments
Table 1 shows the main results for our cross-language adaptation experiments. Rows 1-2 present the results when the target language is Arabic and the system is trained with English input. Rows 3-4 show the reverse case, i.e., adaptation into English when training on Arabic. FNN stands for feed-forward neural network, and it is the upper layer in Figure 2, excluding the language discriminator. CLANN is the full cross-language adversarial neural network, training the discriminator with English inputs paired with random Arabic related questions from the unlabeled dataset. We show three ranking-oriented evaluation measures that are standard in the field of Information Retrieval: mean average precision (MAP), mean reciprocal rank (MRR), and average recall (AvgRec). We computed them using the official scorer from SemEval-2016 Task 3.444http://alt.qcri.org/semeval2016/task3/ Similarly to that task, we consider Mean Average Precision (MAP) as the main evaluation metric. The table also presents, for reproducibility, the values of the neural network hyper-parameters after tuning (in the fifth column).
|Monolingual (English) from SemEval-2016|
|Cross-language (Arabic into English)|
We can see that the MAP score for FNN with Arabic target is 75.28. When doing the adversarial adaptation with the unlabeled Arabic examples (CLANN), the MAP score is boosted to 76.64 (+1.36 points). Going in the reverse direction, with English as the target, yields very comparable results: MAP goes from 75.32 to 76.70 (+1.38).
To put these results into perspective, Table 2 shows the results for the top-2 best-performing systems from SemEval-2016 Task 3, which used a monolingual English setting. We can see that our FNN approach based on cross-language input embeddings is already not far from the best systems. Yet, when we consider the full adversarial network, in any of the two directions, we get performance that is on par with the best, in all metrics.
We conclude that the adversarial component in the network does the expected job, and improves the performance by focusing the language-independent features in the representation layer. The scatter plots in Figure 3 are computed by projecting the representation layer vectors of the first 500 test examples into two dimensions using t-SNE visualization van der Maaten and Hinton (2008). The first 250 are taken with Arabic input (blue), the second 250 are taken with English input (red). 0-1 are the class labels (similar vs. non-similar). The top plot corresponds to CLANN training with English and adapting with Arabic examples, while the second one covers the opposite direction. The plots look as expected. CLANN really mixes the blue and the red examples, as the adversarial part of the network pushes for learning shared abstract features that are language-insensitive. At the same time, the points form clusters with clear majorities of 0s or 1s, as the supervised part of the network learns how to classify them in these classes.
|(b, d, , , )|
|FNN||en||—||ar||8, 0.3, 10, 100, 0.03||74.69||83.79||88.16|
|CLANN-unsup||en||en vs. ar’||ar||12, 0.3, 15, 75, 0.02||75.93||84.15||89.63|
|CLANN-semisup||en+ar||en vs. ar||ar||8, 0.4, 15, 75, 0.02||76.65||84.52||90.84|
|en vs. ar’|
|FNN||ar||—||en||8, 0.2, 10, 75, 0.03||75.38||84.05||89.12|
|CLANN-unsup||ar||ar vs. en’||en||12, 0.2, 15, 75, 0.03||75.89||84.29||89.54|
|CLANN-semisup||ar+en||ar vs. en||en||8, 0.2, 10, 75, 0.03||76.63||84.52||90.82|
|ar vs. en’|
5.2 Semi-supervised Experiments
We now study the semi-supervised scenario when we also have some labeled data from the target language, i.e., where the original question is in the target language. This can be relevant in practical situations, as sometimes we might be able to annotate some data in the target language. It is also an exploration of training with data in multiple languages all together.
To simulate this scenario, we split the training set in two halves. We train with one half as the source language, and we use the other half with the target language as extra supervised data. At the same time, we also use the unlabeled examples as before. We introduced the semi-supervised model in subsection 3.2, which is a straightforward adaptation of the CLANN model.
Table 3 shows the main results of our cross-language semi-supervised experiments. The table is split into two blocks by source and target language (en-ar or ar-en). We also use the same notation as in Table 1. The suffixes -unsup and -semisup indicate whether CLANN is trained in unsupervised mode (same as in Table 1) or in semi-supervised mode. The language discriminator in this setting is trained to discriminate between labeled source and labeled target examples, and labeled source and unlabeled target examples. This is indicated in the Discrim. column using asterisk and prime symbols, respectively.
There are several interesting observations that we can make about Table 3. First, since here we are training with only 50% of the original training data, both FNN and CLANN-unsup yield lower results compared to before, i.e., compared to Table 1; this is to be expected. However, the unsupervised adaptation, i.e., using the CLANN-unsup model, still yields improvements over the FNN model by a sizable margin, according to all three evaluation measures. When we also train using the additional labeled examples in the target language, i.e., using the CLANN-semisup model, the results are boosted again to a final MAP score that is very similar to what we had obtained before with the full source-language training dataset. In the English into Arabic adaptation, the MAP score jumps from 74.69 to 76.65 (+1.96 points) when going from the FNN to the CLANN-semisup model, the MRR score goes from 83.79 to 84.52 (+0.73), and the AvgRec score is boosted from 88.16 to 90.84 (+2.68). The results in the opposite adaptation direction, i.e., from Arabic into English, follow a very similar pattern.
These results demonstrate the effectiveness and the flexibility of our general adversarial training framework within our CLANN architecture when applied to a cross-language setting for question-question similarity, taking advantage of the unlabeled examples in the target language (i.e., when using unsupervised adaptation) and also taking advantage of any labeled examples in the target language that we may have at our disposal (i.e., when using semi-supervised training with input examples in the two languages simultaneously).
We have studied the problem of cross-language adaptation for the task of question–question similarity reranking in community question answering, when the input question can be either in English or in Arabic with the objective to port a system trained on one input language to another input language given labeled data for the source language and only unlabeled data for the target language. We used a discriminative adversarial neural network, which we trained to learn task-specific representations directly. This is novel in a cross-language setting, and we have shown that it works quite well. The evaluation results have shown sizable improvements over a strong neural network model that uses simple projection with cross-language word embeddings.
In future work, we want to extend the present research in several directions. For example, we would like to start with monolingual word embeddings and to try to learn the shared cross-language representation directly as part of the end-to-end training of our neural network. We further plan to try LSTM and CNN for generating the initial representation of the input text (instead of simple averaging of word embeddings). We also want to experiment with more than two languages at a time. Another interesting research direction we want to explore is to try to adapt our general CLANN framework to other tasks, e.g., to answer ranking in community Question Answering Joty et al. (2016); Nakov et al. (2016a) in a cross-language setting, as well as to cross-language representation learning for words and sentences.
This research was performed by the Arabic Language Technologies group at Qatar Computing Research Institute, HBKU, within the Interactive sYstems for Answer Search project (Iyas).
- Abdelali et al. (2014) Ahmed Abdelali, Francisco Guzmán, Hassan Sajjad, and Stephan Vogel. 2014. The AMARA corpus: Building parallel language resources for the educational domain. In Proceedings of the 9th International Conference on Language Resources and Evaluation. Reykjavik, Iceland, LREC ’14, pages 1856–1862.
- Agirre et al. (2016) Eneko Agirre, Carmen Banea, Daniel Cer, Mona Diab, Aitor Gonzalez-Agirre, Rada Mihalcea, German Rigau, and Janyce Wiebe. 2016. SemEval-2016 task 1: Semantic textual similarity, monolingual and cross-lingual evaluation. In Proceedings of the 10th International Workshop on Semantic Evaluation. San Diego, CA, USA, SemEval ’16, pages 497–511.
- Arjovsky et al. (2017) Martín Arjovsky, Soumith Chintala, and Léon Bottou. 2017. Wasserstein GAN. CoRR abs/1701.07875.
- Berger et al. (2000) Adam Berger, Rich Caruana, David Cohn, Dayne Freitag, and Vibhu Mittal. 2000. Bridging the lexical chasm: Statistical approaches to answer-finding. In Proceedings of the 23rd Annual International ACM Conference on Research and Development in Information Retrieval. Athens, Greece, SIGIR ’00, pages 192–199.
- Bouma et al. (2008) Gosse Bouma, Jori Mur, and Gertjan van Noord. 2008. Question answering with Joost at CLEF 2008. In Proceedings of the 9th Workshop of the Cross-Language Evaluation Forum: Evaluating Systems for Multilingual and Multimodal Information Access. Aarhus, Denmark, CLEF ’08, pages 257–260.
- Cao et al. (2008) Yunbo Cao, Huizhong Duan, Chin-Yew Lin, Yong Yu, and Hsiao-Wuen Hon. 2008. Recommending questions using the MDL-based tree cut model. In Proceedings of the 17th International Conference on World Wide Web. Beijing, China, WWW ’08, pages 81–90.
- Da San Martino et al. (2016) Giovanni Da San Martino, Alberto Barrón-Cedeño, Salvatore Romeo, Antonio Uva, and Alessandro Moschitti. 2016. Learning to re-rank questions in community question answering using advanced features. In Proceedings of the 25th ACM International Conference on Information and Knowledge Management. Indianapolis, IN, USA, CIKM ’16, pages 1997–2000.
George Doddington. 2002.
Automatic evaluation of machine translation quality using n-gram co-occurrence statistics.In Proceedings of the 2nd International Conference on Human Language Technology Research. San Diego, CA, USA, HLT ’02, pages 138–145.
- dos Santos et al. (2015) Cicero dos Santos, Luciano Barbosa, Dasha Bogdanova, and Bianca Zadrozny. 2015. Learning hybrid representations to retrieve semantically equivalent questions. In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing. Beijing, China, ACL-IJCNLP ’15, pages 694–699.
- Echihabi and Marcu (2003) Abdessamad Echihabi and Daniel Marcu. 2003. A noisy-channel approach to question answering. In Proceedings of the 41st Annual Meeting of the Association for Computational Linguistics. Sapporo, Japan, ACL ’03, pages 16–23.
- Forner et al. (2008) Pamela Forner, Anselmo Peñas, Eneko Agirre, Iñaki Alegria, Corina Forascu, Nicolas Moreau, Petya Osenova, Prokopis Prokopidis, Paulo Rocha, Bogdan Sacaleanu, Richard F. E. Sutcliffe, and Erik F. Tjong Kim Sang. 2008. Overview of the CLEF 2008 Multilingual Question Answering Track. In Proceedings of the 9th Workshop of the Cross-Language Evaluation Forum: Evaluating Systems for Multilingual and Multimodal Information Access. Aarhus, Denmark, CLEF ’08, pages 262–295.
Ganin et al. (2016)
Yaroslav Ganin, Evgeniya Ustinova, Hana Ajakan, Pascal Germain, Hugo
Larochelle, François Laviolette, Mario Marchand, and Victor Lempitsky.
Domain-adversarial training of neural networks.
Journal of Machine Learning Research17(1):2096–2030.
Glorot and Bengio (2010)
Xavier Glorot and Yoshua Bengio. 2010.
Understanding the difficulty of training deep feedforward neural
JMLR W&CP: Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics. Sardinia, Italy, volume 9 of AISTATS ’10, pages 249–256.
- Goodfellow et al. (2014) Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. 2014. Generative adversarial nets. In Proceedings of Advances in Neural Information Processing Systems Conference 27. Montréal, Canada, NIPS ’14, pages 2672–2680.
Guzmán et al. (2015)
Francisco Guzmán, Shafiq Joty, Lluís Màrquez, and Preslav Nakov.
Pairwise neural machine translation evaluation.In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). Beijing, China, ACL-IJCNLP ’15, pages 805–814.
- Guzmán et al. (2016) Francisco Guzmán, Lluís Màrquez, and Preslav Nakov. 2016. Machine translation evaluation meets community question answering. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics. Berlin, Germany, ACL ’16, pages 460–466.
- Hartrumpf et al. (2008) Sven Hartrumpf, Ingo Glöckner, and Johannes Leveling. 2008. Efficient question answering with question decomposition and multiple answer streams. In Proceedings of the 9th Workshop of the Cross-Language Evaluation Forum: Evaluating Systems for Multilingual and Multimodal Information Access. Aarhus,Denmark, CLEF ’08, pages 421–428.
- Jeon et al. (2005) Jiwoon Jeon, W. Bruce Croft, and Joon Ho Lee. 2005. Finding similar questions in large question and answer archives. In Proceedings of the 14th ACM International Conference on Information and Knowledge Management. Bremen, Germany, CIKM ’05, pages 84–90.
- Joty et al. (2016) Shafiq Joty, Lluís Màrquez, and Preslav Nakov. 2016. Joint learning with global inference for comment classification in community question answering. In Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics. San Diego, CA, USA, NAACL-HLT ’16.
- Kingma and Ba (2015) Diederik P. Kingma and Jimmy Ba. 2015. Adam: A method for stochastic optimization. In Proceedings of the 3rd International Conference for Learning Representations. San Diego, CA, USA, ICLR ’15.
- Lavie and Denkowski (2009) Alon Lavie and Michael Denkowski. 2009. The METEOR metric for automatic evaluation of machine translation. Machine Translation 23(2–3):105–115.
- Lei et al. (2016) Tao Lei, Hrishikesh Joshi, Regina Barzilay, Tommi S. Jaakkola, Kateryna Tymoshenko, Alessandro Moschitti, and Lluís Màrquez. 2016. Semi-supervised question retrieval with gated convolutions. In Proceedings of the 15th Annual Conference of the North American Chapter of the Association for Computational Linguistics. San Diego, CA, USA, NAACL-HLT ’16, pages 1279–1289.
- Lin and Kuo (2010) Chuan-Jie Lin and Yu-Min Kuo. 2010. Description of the NTOU complex QA system. In Proceedings of the 8th NTCIR Workshop Meeting on Evaluation of Information Access Technologies: Information Retrieval, Question Answering, and Cross-Lingual Information Access. Tokyo, Japan, NTCIR ’10, pages 47–54.
- Luong et al. (2015) Thang Luong, Hieu Pham, and Christopher D. Manning. 2015. Bilingual word representations with monolingual quality in mind. In Proceedings of the 1st Workshop on Vector Space Modeling for Natural Language Processing. Denver, CO, USA, pages 151–159.
Makhzani et al. (2016)
Alireza Makhzani, Jonathon Shlens, Navdeep Jaitly, and Ian J. Goodfellow. 2016.
Adversarial autoencoders.In Proceedings of the International Conference on Learning Representations 2016. San Juan, Puerto Rico, ICLR ’16.
- Martino et al. (2017) Giovanni Da San Martino, Salvatore Romeo, Alberto Barrón Cedeño, Shafiq Joty, Lluís Màrquez, Alessandro Moschitti, and Preslav Nakov. 2017. Cross-language question re-ranking. In Proceedings of the 40th ACM SIGIR Conference on Research and Development in Information Retrieval. Tokyo, Japan, SIGIR ’17.
- Metz et al. (2017) Luke Metz, Ben Poole, David Pfau, and Jascha Sohl-Dickstein. 2017. Unrolled generative adversarial networks. In 5th International Conference on Learning Representations. Toulon, France, ICLR ’17.
- Mihaylov and Nakov (2016) Todor Mihaylov and Preslav Nakov. 2016. SemanticZ at SemEval-2016 Task 3: Ranking relevant answers in community question answering using semantic similarity based on fine-tuned word embeddings. In Proceedings of the 10th International Workshop on Semantic Evaluation. San Diego, CA, USA, SemEval ’16, pages 879–886.
- Mikolov et al. (2013) Tomas Mikolov, Wen-tau Yih, and Geoffrey Zweig. 2013. Linguistic regularities in continuous space word representations. In Proceedings of the 2013 Conference of the North American Chapter of the Association for Computational Linguistics. Atlanta, GA, USA, NAACL-HLT ’13, pages 746–751.
- Mitamura et al. (2010) Teruko Mitamura, Hideki Shima, Tetsuya Sakai, Noriko Kando, Tatsunori Mori, Koichi Takeda, Chin-Yew Song Ruihua Lin, Chuan-Jie Lin, and Cheng-Wei Lee. 2010. Overview of the NTCIR-8 ACLIA tasks: Advanced cross-lingual information access. In Proceedings of the 8th NTCIR Workshop Meeting on Evaluation of Information Access Technologies: Information Retrieval, Question Answering, and Cross-Lingual Information Access. Tokyo, Japan, NTCIR ’10, pages 15–24.
- Mohammad et al. (2016) Saif M. Mohammad, Mohammad Salameh, and Svetlana Kiritchenko. 2016. How translation alters sentiment. Journal of Artificial Intelligence Research 55(1):95–130.
Nair and Hinton (2010)
Vinod Nair and Geoffrey E. Hinton. 2010.
Rectified linear units improve restricted Boltzmann machines.In Proceedings of the 27th International Conference on Machine Learning. Haifa, Israel, ICML ’10, pages 807–814.
- Nakov et al. (2017) Preslav Nakov, Doris Hoogeveen, Lluís Màrquez, Alessandro Moschitti, Hamdy Mubarak, Timothy Baldwin, and Karin Verspoor. 2017. SemEval-2017 task 3: Community question answering. In Proceedings of the 11th International Workshop on Semantic Evaluation. Vancouver, Canada, SemEval ’17.
- Nakov et al. (2016a) Preslav Nakov, Lluís Màrquez, and Francisco Guzmán. 2016a. It takes three to tango: Triangulation approach to answer ranking in community question answering. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing. Austin, TX, USA, EMNLP ’16, pages 1586–1597.
- Nakov et al. (2016b) Preslav Nakov, Lluís Màrquez, Alessandro Moschitti, Walid Magdy, Hamdy Mubarak, Abed Alhakim Freihat, Jim Glass, and Bilal Randeree. 2016b. SemEval-2016 task 3: Community question answering. In Proceedings of the 10th International Workshop on Semantic Evaluation. San Diego, CA, USA, SemEval ’16, pages 525–545.
- Nicosia et al. (2015) Massimo Nicosia, Simone Filice, Alberto Barrón-Cedeño, Iman Saleh, Hamdy Mubarak, Wei Gao, Preslav Nakov, Giovanni Da San Martino, Alessandro Moschitti, Kareem Darwish, Lluís Màrquez, Shafiq Joty, and Walid Magdy. 2015. QCRI: Answer selection for community question answering - experiments for Arabic and English. In Proceedings of the 9th International Workshop on Semantic Evaluation. Denver, CO,USA, SemEval ’15, pages 203–209.
- Papineni et al. (2002) Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. BLEU: a method for automatic evaluation of machine translation. In Proceedings of 40th Annual Meting of the Association for Computational Linguistics. Philadelphia, PA, USA, ACL ’02, pages 311–318.
- Pouran Ben Veyseh (2016) Amir Pouran Ben Veyseh. 2016. Cross-lingual question answering using common semantic space. In Proceedings of the 2016 Workshop on Graph-based Methods for Natural Language Processing. San Diego, CA, USA, TextGraphs ’16, pages 15–19.
- Radford et al. (2016) Alec Radford, Luke Metz, and Soumith Chintala. 2016. Unsupervised representation learning with deep convolutional generative adversarial networks. In Proceedings of the 4th International Conference on Learning Representations. San Juan, Puerto Rico, ICLR ’16.
- Riezler et al. (2007) Stefan Riezler, Alexander Vasserman, Ioannis Tsochantaridis, Vibhu Mittal, and Yi Liu. 2007. Statistical machine translation for query expansion in answer retrieval. In Proceedings of the 45th Annual Meeting of the Association of Computational Linguistics. Prague, Czech Republic, ACL ’07, pages 464–471.
- Romeo et al. (2016) Salvatore Romeo, Giovanni Da San Martino, Alberto Barrón-Cedeño, Alessandro Moschitti, Yonatan Belinkov, Wei-Ning Hsu, Yu Zhang, Mitra Mohtarami, and James Glass. 2016. Neural attention for learning to rank questions in community question answering. In Proceedings of the 26th International Conference on Computational Linguistics. Osaka, Japan, COLING ’16, pages 1734–1745.
- Snover et al. (2006) Matthew Snover, Bonnie Dorr, Richard Schwartz, Linnea Micciulla, and John Makhoul. 2006. A study of translation edit rate with targeted human annotation. In Proceedings of the 7th Conference of the Association for Machine Translation in the Americas. Cambridge, MA, USA, AMTA ’06, pages 223–231.
- Soboroff et al. (2016) Ian Soboroff, Kira Griffitt, and Stephanie Strassel. 2016. The BOLT IR test collections of multilingual passage retrieval from discussion forums. In Proceedings of the 39th International Conference on Research and Development in Information Retrieval. Pisa, Italy, SIGIR ’16, pages 713–716.
- Socher et al. (2013) Richard Socher, John Bauer, Christopher D. Manning, and Ng Andrew Y. 2013. Parsing with compositional vector grammars. In Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics. Sofia, Bulgaria, ACL ’13, pages 455–465.
- Soricut and Brill (2006) Radu Soricut and Eric Brill. 2006. Automatic question answering using the web: Beyond the factoid. Information Retrieval 9(2):191–206.
- Srivastava et al. (2014) Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. 2014. Dropout: A simple way to prevent neural networks from overfitting. Journal of Machine Learning Research 15(1):1929–1958.
- Surdeanu et al. (2011) Mihai Surdeanu, Massimiliano Ciaramita, and Hugo Zaragoza. 2011. Learning to rank answers to non-factoid questions from web collections. Computational Linguistics 37(2):351–383.
- Tiedemann (2012) Jörg Tiedemann. 2012. Parallel data, tools and interfaces in OPUS. In Proceedings of the 8th International Conference on Language Resources and Evaluation. Istanbul, Turkey, LREC ’12, pages 2214–2218.
- Ture and Boschee (2016) Ferhan Ture and Elizabeth Boschee. 2016. Learning to translate for multilingual question answering. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing. Austin, TX, USA, EMNLP ’16, pages 573–584.
- Upadhyay et al. (2016) Shyam Upadhyay, Manaal Faruqui, Chris Dyer, and Dan Roth. 2016. Cross-lingual models of word embeddings: An empirical comparison. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics. Berlin, Germany, ACL ’16, pages 1661–1670.
van der Maaten and Hinton (2008)
Laurens van der Maaten and Geoffrey Hinton. 2008.
Visualizing high-dimensional data using t-SNE.Journal of Machine Learning Research 9:2579 –2605.
- Wang et al. (2009) Kai Wang, Zhaoyan Ming, and Tat-Seng Chua. 2009. A syntactic tree matching approach to finding similar questions in community-based QA services. In Proceedings of the 32nd international ACM SIGIR conference on Research and development in information retrieval. Boston, MA, USA, SIGIR ’09, pages 187–194.
- Zhang et al. (2014) Kai Zhang, Wei Wu, Haocheng Wu, Zhoujun Li, and Ming Zhou. 2014. Question retrieval with high quality answers in community question answering. In Proceedings of the 23rd ACM International Conference on Information and Knowledge Management. Shanghai, China, CIKM ’14, pages 371–380.
- Zhou et al. (2011) Guangyou Zhou, Li Cai, Jun Zhao, and Kang Liu. 2011. Phrase-based translation model for question retrieval in community question answer archives. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics. Portland, OR, USA, ACL ’11, pages 653–662.