Cross-lingual Alignment Methods for Multilingual BERT: A Comparative Study
Multilingual BERT (mBERT) has shown reasonable capability for zero-shot cross-lingual transfer when fine-tuned on downstream tasks. Since mBERT is not pre-trained with explicit cross-lingual supervision, transfer performance can further be improved by aligning mBERT with cross-lingual signal. Prior work proposes several approaches to align contextualised embeddings. In this paper we analyse how different forms of cross-lingual supervision and various alignment methods influence the transfer capability of mBERT in zero-shot setting. Specifically, we compare parallel corpora vs. dictionary-based supervision and rotational vs. fine-tuning based alignment methods. We evaluate the performance of different alignment methodologies across eight languages on two tasks: Name Entity Recognition and Semantic Slot Filling. In addition, we propose a novel normalisation method which consistently improves the performance of rotation-based alignment including a notable 3% F1 improvement for distant and typologically dissimilar languages. Importantly we identify the biases of the alignment methods to the type of task and proximity to the transfer language. We also find that supervision from parallel corpus is generally superior to dictionary alignments.
Multilingual BERT (mBERT) Devlin et al. (2019) is the BERT architecture trained on data from 104 languages where all languages are embedded in the same vector space. Due to the multilingual and contextual representation properties of mBERT, it has gained popularity in various multilingual and cross-lingual tasks Karthikeyan et al. (2020); Wu and Dredze (2019). In particular, it has demonstrated good zero-shot cross-lingual transfer performance on many downstream tasks, such as Document Classification, NLI, NER, POS tagging, and Dependency Parsing Wu and Dredze (2019), when the source and the target languages are similar.
Many experiments Ahmad et al. (2019) suggest that to achieve reasonable performance in the zero-shot setup, the source and the target languages need to share similar grammatical structure or lie in the same language family. In addition, since mBERT is not trained with explicit language signal, mBERT’s multilingual representations are less effective for languages with little lexical overlap Patra et al. (2019). One branch of work is therefore dedicated to improve the multilingual properties of mBERT by aligning the embeddings of different languages with cross-lingual supervision.
Broadly, two methods have been proposed in prior work to induce cross-lingual signals in contextual embeddings: 1) Rotation Alignment as described in Section 2 aims at learning a linear rotation transformation to project source language embeddings into their respective locations in the target language space Schuster et al. (2019b); Wang et al. (2019); Aldarmaki and Diab (2019); 2) Fine-tuning Alignment as explained in Section 3 internally aligns language sub-spaces in mBERT through tuning its weights such that distances between embeddings of word translations decrease while not losing the informativity of the embeddings Cao et al. (2020). Additionally, two sources of cross-lingual signal have been considered in literature to align languages: parallel corpora and bilingual dictionaries. While the choice of each alignment method and source of supervision have a variety of advantages and disadvantages, it is unclear how these affect the performance of the aligned spaces across languages and various tasks.
In this paper, we empirically investigate the effect of these cross-lingual alignment methodologies and applicable sources of cross-lingual supervision by evaluating their performance on zero-shot Named Entity Recognition (NER), a structured prediction task, and Semantic Slot-filling (SF), a semantic labelling task, across eight language pairs.
The motivation for choice of these tasks to evaluate are two-fold: 1. Prior work has already studied alignment methods on sentence level tasks. \newcitecao show the effectiveness of mBERT alignment methods on XNLI Conneau et al. (2018). 2. Word-level tasks do not benefit from more pre-training unlike other language tasks that improve by simply supplementing with more pre-training data. In experiments over the XTREME benchmark, \newcitextreme find that transfer performance improves across all tasks when multilingual language models are pre-trained with more data, with the sole exception of word-level tasks. They note that this indicates current deep pre-trained models do not fully exploit the pre-training data to transfer to word-level tasks. We believe that NER and Slot-filling tasks are strong candidate tasks to assess alignment methods due to limited cross-lingual transfer capacity of current models to these tasks.
To the authors’ knowledge, this is the first paper exploring the comparison of alignment methods for contextual embedding spaces: rotation vs. fine-tuning alignment and two sources of cross-lingual supervision: dictionary vs. parallel corpus supervision on a set of tasks of structural and semantic nature over a wide range of languages. From the results, we find that parallel corpora are better suited for aligning contextual embeddings. In addition, we find that rotation alignment is more robust for primarily structural NER downstream tasks while the fine-tuning alignment is found to improve performance across semantic SF tasks. In addition, we propose a novel normalisation procedure which consistently improves rotation alignment, motivated by the structure of mBERT space and how languages are distributed across it. We also find the effect of language proximity on transfer improvement for these alignment methods.
2 Rotation-based Alignment\newcite
mikolov proposed to learn a linear transformation which would project an embedding in the source language to its translation in the target language space , by minimising the distances between the projected source embeddings and their corresponding target embeddings:
and are matrices of size where is the dimensionality of embeddings and is the number of parallel words from word-aligned corpora, or word pairs from a bilingual dictionary between the source and target languages. Further work \newcitexing demonstrated that restricting to a purely rotational transform improves cross-lingual transfer across similar languages. The orthogonality assumption reduces Eq.(1) into the so-called Procrustes problem with the closed form solution:
and the SVD operator stands for Singular Value Decomposition.
2.1 Language Centering Normalization
A purely rotational transformation can align two embedding spaces only if the two spaces are roughly isometric and are distributed about the same mean. In case the two embedding distributions are not centered around the same mean, meaning the two spaces have little overlap and are shifted by a translation offset in the space, they cannot be aligned solely through rotation.
Since the linear transformation derived from solving the Procrustus problem only rotates the vector space, it assumes the embeddings of two languages are zero-centered. However \newcitelibovick2019languageneutral observe that languages distributions in mBERT have distinct and separable centroids and different language families have well separated sub-spaces in the mBERT embedding vector space. To address this discrepancy, we propose a new normalisation mechanism which entails:
Step 1. Normalising the embeddings of both languages so that they have zero mean:
where and are centroids of source and target embeddings and ; and and are mean-centered source and target language embeddings their rows correspond to word translations. Next, and are used to compute the transformation matrix by solving Eq.(2) and Eq.(3).
Step 2. During training a downstream task, embedding of a source language word needs to be re-centered, rotated and finally translated to the target language subspace to derive the projection :
This helps the task specific model, particularly in zero-shot setting, by projecting the source language task data to the same locality as the target language.
|(a) Alignment with dictionary.||(b) Alignment with parallel sentences.|
2.2 Supervision Signals for Rotation Alignment
In this section we describe how existing work utilises two different cross-lingual signals, bilingual dictionaries and parallel corpora, to supervise rotation alignment. Additionally, we analyse the advantages and disadvantages of the two choices.
Bilingual Dictionary Supervision
In order to utilise a bilingual dictionary to supervise the embedding alignment, each word in the dictionary needs to have a single representation. However the same word can have many representations in the contextualised language model vector space depending on the context it occurs in. \newciteschuster2019 observes that the contextual embeddings of the same word form a tight cluster - word cloud, the centroid of this word cloud is distinct and separable for individual words. They further propose that centroid of a word cloud can be considered as the context-independent representation of a word, called average word anchor. These word anchors are computed by averaging embeddings over all occurrences of a word in a monolingual corpora, where words occur in a variety of contexts. Formally the mBERT embedding of a source language word in context is denoted as . If this word occurs a total of times in the monolingual corpus, that is in contexts , the anchor word embedding for word across all the contexts is the average:
Average word anchor pair , where is the mBERT layer, for all word pairs from the dictionary form the rows of matrices and respectively, which are then used to solve Eq.(2) and Eq.(3), resulting in an alignment transformation matrix .
However, there are limitations to this approach. \newcitemultisense found that the word cloud of multi-sense words, such as the word “bank”, which can mean either the financial institution or the edge of a river depending on the context, are further composed of clearly separable clusters, for every word sense. Averaging over multiple contextual embeddings infers losing certain degree of contextual information at both the source and target language words. Figure 1a visualises word anchor calculation and also highlights this limitation. On the other hand, one of the advantages of this method is that bilingual dictionaries are available for even very low resource languages.
Parallel Corpus Supervision
Word-aligned parallel sentences can be utilised as a source of cross-lingual signal to align contextual embeddings Aldarmaki and Diab (2019); Wang et al. (2019). Given a parallel corpora, and are aligned source and the target language words appearing in context and , respectively. The parallel word embedding matrices and for mBERT layer- are composed from the contextual embeddings and respectively, and are used to solve Eq.(2) and Eq.(3) to derive an alignment transformation matrix .
Figure 1a and 1b illustrate how parallel supervision is more suited to align contextual embeddings compared to dictionary supervision where multiple senses of a word are compressed into a single word anchor. However, parallel corpora rarely come with word-alignment annotations that are often automatically generated by off-the-shelf tools such as fast_align Dyer et al. (2013), which can be noisy. It is worth noting that word alignment error rate of an off-the-shelf tool drops when number of parallel sentences increases, therefore parallel corpus supervision is favourable for languages where more parallel data is available.
3 Fine-tuning Alignment with Parallel Corpora
Rotation alignment has a strong assumption that the two language spaces (or sub-spaces in case of mBERT) are approximately isometric Søgaard et al. (2018). \newcitepatra reported that the geometry of language embeddings becomes dissimilar for distant languages, and the isometry assumption degrades the alignment performance in such cases. In addition, as explained in Section 2.1 rotation alignment alone cannot achieve effective mapping when two languages spaces have separate centroids. Therefore, next we consider existing work to non-linearly align two language spaces.
cao proposed to directly align languages within mBERT model through fine-tuning. The objective of the fine-tuning is to minimise the distance between the two contextual representations of an aligned word pair in parallel corpora:
However, fine-tuning with only the above objective would led to lose the semantic information in mBERT learnt during pre-training, since a trivial solution to the Eq.(7) can be simply to make all the embeddings equal. To deal with this, \newcitecao also proposed a regularisation loss that does not allow the embedding of a source language word to stray too far away from its original location in the pre-trained mBERT model, namely:
Note that is generated from a copy of the original pre-trained mBERT model where the parameters are kept frozen. Both of the alignment and the regularization losses are combined and jointly optimised in order to align the two language subspaces while maintaining informativity of embeddings:
Here to is the range of mBERT layers aligned. We experimented with two variants of the fine-tuning approach: 1) moving target language towards source language while keeping the source embeddings approximately fixed through the regularization term in Eq.(8); 2) moving the source language embeddings towards the target space while keeping the target language space relatively fixed, then the regularisation loss changes to:
4 Experimental Setup
In this section, we firstly describe the resources and implementation details of the alignment methods followed by the zero-shot NER and SF tasks used to evaluate the alignments. In addition, we briefly explain the datasets used in the experiments.
4.1 Learning Alignments
Our baseline model is a pre-trained mBERT
We use the bilingual dictionaries provided with the MUSE framework Lample et al. (2018) as the source for dictionary supervision. As for the parallel corpus supervision, since none of the collected parallel sentences contains word-level alignment information, we utilise fast_align Dyer et al. (2013) to automatically derive word alignment signals.
For the rotation alignment, we compute four independent transformation matrices for each of the last four transformer layers similar to \newcitezwang. We use RotateAlign and NormRotateAlign to refer the rotation alignment learnt without and with the proposed language centering normalisation, respectively. To be consistent, for the fine-tuning alignment we align the word representations in the last four transformer layers of the mBERT model, denoted as FineTuneAlign.
|Datasets||Task||Translated||# Language & Train/Dev/Test Size||# Slot Types||Domains|
|11||Weather, Alarm, Reminder|
|CoNLL-NER||[U.N.] official [Ekeus] heads for [Baghdad].|
|ATIS-SF||show the [latest] flight from [denver] to [boston]|
|FB-SF||do you have [wednesday’s] [weather forecast] for [half moon bay]|
|SNIPS-SF||add this [track] to [my] [global funk]|
4.2 Evaluation of the Alignments
We evaluate the learnt alignments using two downstream tasks: Named Entity Recognition (NER) and Semantic Slot Filling (SF), both of which aim to predict a label for each token in a sentence. NER is a more structural task with fewer entity types and involves less semantic understanding of the context compared to SF. Examples of the tasks can be found in Table 2.
We use the same model architecture and hyper-parameters as \newcitezwang, two BiLSTM layers followed by a CRF layer, where learning rate is set to for European languages and for the other languages determined by the validation set. In order to measure the effectiveness of a learnt alignment, all the experiments are conducted with zero-shot settings similar to \newcitezwang, where the source language data is first transformed to the target language space and then used to train a BiLSTM-CRF model. The target language validation set is used for hyper-parameter tuning and reporting the evaluation results. For each experiment we report F1 scores averaged over 5 runs.
4.3 NER and SF Datasets
CoNLL-NER: This includes CoNLL 2002, 2003 NER benchmark task Tjong Kim Sang (2002); Tjong Kim Sang and De Meulder (2003) containing entity annotations for news articles in English, German, Spanish and Dutch. We also include in this family PioNER
ATIS-SF: ATIS Corpus Price (1990) is an English dataset containing conversational queries about flight booking. \newciteatis-google manually translated a subset of the data into two languages, Turkish and Hindi, along with crowd-sourced phrase-level annotations.
FB-SF: \newcitefb introduced Multilingual Task-Oriented Dialog Corpus in English, Spanish and Thai across three domains: weather, alarm and reminders, where Spanish and Thai data were manually translated and annotated from a subset of the English data.
SNIPS-SF: A multi-domain slot-filling dataset in English released by \newcitesnips. \newcitesnips-it automatically translated this dataset into Italian, and then manually labelled the translation where entities were substituted by Italian entities collected from the Web.
|Transfer Pair||en to de||en to nl||en to es||en to hy||en to hi||en to tk||en to es||en to th||en to it|
|Baselines from Literature|
|mBERT Wu and Dredze (2019)||69.56||77.75||74.96||-||-||-||-||-||-|
|mBERT Rotation Alignment: Parallel Wang et al. (2019)||70.54||79.03||75.77||-||-||-||-||-||-|
|BERT, 1400 Target Language Train Bellomaria et al. (2019)||-||-||-||-||-||-||-||-||83.04|
|Non-contextual Zero-shot Baseline Upadhyay et al. (2018)||-||-||-||-||40||40||-||-||-|
|Translate train Schuster et al. (2019a)||-||-||-||-||-||-||72.87||55.43||-|
5 Results and Analysis
The evaluation results of each alignment method on the downstream NER and SF tasks are reported in Table 3 and Figure 2. In addition to the mBERT Baseline and for comparison purposes, we also list relevant results found in literature Wu and Dredze (2019); Wang et al. (2019); Upadhyay et al. (2018); Schuster et al. (2019a); Bellomaria et al. (2019) that have been evaluated on the same datasets.
5.1 mBERT Baseline and Language Proximity
mBERT Baseline numbers can be indicative of how well languages are already aligned in the mBERT space. High zero-shot scores for German, Dutch, Spanish and Italian indicate that European languages are extremely well aligned to English in mBERT. However, distant languages such as Thai and Turkish, which belong to different language families (KraâDai and Turkic) than English, have poor alignment with low F1 scores of 9.58 and 21.15, respectively. Finally, moderately distant languages such as Armenian and Hindi, which fall within the larger Indo-European language family, have moderate alignment with English with scores of 62.38 and 50.84, respectively.
5.2 mBERT Baseline vs./ Rotation Alignment
RotateAlign improves performance by 19% absolute for ATIS-Turkish, going from baseline of 21.15 to 38.18 in F1 score. For ATIS-Hindi the performance improves from 50.84 to 57.86 F1 (7 points), and 4% absolute for the PioNER-Armenian from 62.38 to 66.56. These numbers show how RotateAlign can improve performance over mBERT Baseline for moderately-close languages such as Hindi, Turkish and Armenian, while there is only around 1 point improvement for European languages. This implies that Hindi, Turkish and Armenian subspaces are geometrically similar to English, however they are misaligned in terms of rotation in mBERT Baseline.
However, in the case of Thai, which is a distant language from English, RotateAlign does not improve performance over the mBERT Baseline. This suggests that Thai and English’s embedding spaces are structurally dissimilar.
5.3 Rotation Alignment with vs./ without Language Centering Normalisation
Applying the proposed language centering normalisation in Section 2.1 before performing the rotation alignment, namely NormRotateAlign in Table 3, is found to further improve downstream performance across all tasks and languages. The improvement over RotateAlign is up to 3% absolute F1 for Thai, around 1% absolute for moderately closer languages like Hindi, Turkish and Armenian, and around 0.5% absolute F1 for closer target languages such as German. Note that Thai, which does not benefit from rotation alignment alone, improves by an average of 2.3 points after applying the normalisation. These results corroborate that language families that are further away from each other have more separable sub-spaces in the mBERT Baseline, and bringing the language distributions closer helps the downstream task’s performance.
5.4 Parallel Corpus vs./ Dictionary Supervision
Amongst the cases where RotateAlign improves performance over the mBERT Baseline, parallel-corpus supervised RotateAlign is superior to dictionary supervision, with the exception of Hindi. This could be explained by the fact that word anchors are independent of multiple word senses, thereby the cross-lingual signal is poorer compared to parallel word alignments. This is in line with observations from \newcitemultisense.
5.5 Rotation vs./ Fine-tuning Alignment
From Table 3 and Figure 2 we can see that FineTuneAlign explained in Section 3 improves performance over RotateAlign for semantic tasks (SF), with the only exception of ATIS-Hindi. On the other hand, FineTuneAlign underperforms RotateAlign for structural tasks (NER), and in some cases even fall behind mBERT Baseline. Note that we notice no clear trend between FineTuneAlign and FineTuneAlign.
FineTuneAlign improves over the best rotation alignment NormRotateAlign by 7.8% absolute for the ATIS-Turkish task from 38.18 to 45.98. It significantly outperforms mBERT Baseline by 24 points. For FB-Thai FineTuneAlign surpasses NormRotateAlign by 8.39% absolute F1 from 12.38 to 20.77, 11 points higher than mBERT Baseline. For FB-Spanish we observe an improvement from 74.73 to 80.90 (6% absolute) compared to RotateAlign and similarly +6 points compared to mBERT Baseline. For SNIPS-Italian, FineTuneAlign improves performance over NormRotateAlign from 77.87 to 80.21 (2.5 points) and is 3.5 points better than mBERT Baseline.
All SF tasks considered are generated by translation from the source language data. This may indicate that the fine-tuning approach performs better than rotation-based methods for translated datasets, where there is high correlation between utterance structure of training data in source language and evaluation data in target language. On the other hand, rotation-based alignments generalise better when the downstream target sentence distribution is dissimilar from the source sentence distribution, as is the case for non-translated NER tasks.
5.6 Aligned Source Language vs./ Target Language Training
FineTuneAlign achieves top F1 score of 80.21 on SNIPS-Italian dataset which is not far from the score of 83 from a BERT-based model trained on 1400 manually-annotated Italian utterances Bellomaria et al. (2019). Also, our best alignment score of 80.90 for FB-Spanish (FineTuneAlign) surpasses translate-train baseline Schuster et al. (2019a) where the annotations are automatically inferred from a NMT model. This suggests that for closer target languages, fine-tuning based alignment are not far behind from unaligned models trained on additional target language labelled examples.
Performance improvement from fine-tuning alignment for translated datasets should not be attributed to superficial transfer of entity information from source language. An evidence to support this claim is the strong performance on the SNIPS Italian-SF dataset, which has been translated from SNIPS dataset Bellomaria et al. (2019), where English entities have been replaced with Italian entities collected from the Web during dataset preparation. Therefore, during validation, the model came across utterances with similar structure but different entities, which shows that improvement from fine-tuning alignment is largely independent of language specific entity memorisation.
6 Related Work\newcite
diab propose to align ELMo embeddings Peters et al. (2018) with word-level and sentence-level alignments. They compare the aligned ELMo with static character-level embeddings with similar alignments.
cao originally proposed fine-tuning alignment of mBERT language sub-spaces. They claim these methods are strictly stronger to rotation alignments methods based solely on zero-shot experimentation on XNLI task Conneau et al. (2018), a semantic sentence-level classification task generated through translation from source language. On the contrary, we observe that fine-tuning does not improve performance across all tasks, particularly structural tasks, where utterance structure changes and there is higher incidence of domain shift. This raises the question whether translated datasets are biased to fine-tuning alignment, and whether such datasets are a good evaluation test-bed for general cross-lingual transfer.
zwang applies rotational alignment to mBERT and reports results on CoNLL NER tasks, however the main focus of their work is on the overlap of static bilingual embeddings. They do not extend similar analysis on contextualised embeddings. In our work, drawing from the observations made by \newcitelibovick2019languageneutral on the distribution of languages in mBERT space, we propose a normalization mechanism to increase the overlap of two languages distributions prior to computing rotational alignment.
schuster2019 originally proposed dictionary supervision to align ELMo with rotational transform. They claim supervision from dictionary is superior to using parallel word aligned corpora, however they do not substantiate these through comparative experiments. We observe that parallel corpus supervision is stronger than dictionary supervision possibly because of considering contextual alignment.
In this paper, we investigate cross-lingual alignment methods for multilingual BERT. We empirically evaluate their effect on zero-shot transfer for downstream tasks of two types: structural NER and semantic Slot-filling, across a set of diverse languages. Specifically, we compare rotation alignment and fine-tuning cross-lingual alignment. We compare the effect of dictionary and parallel corpora supervision across all tasks. We also propose a novel normalisation technique that improves state-of-the-art performance on zero-shot NER and Semantic Slot-filling downstream tasks, motivated by how languages are distributed across the mBERT space. Our experimental settings cover four datasets families (one for NER and three for SF) across eight language pairs.
Key findings of this paper are as follows: (1) rotation-based alignments show large performance improvements (up to +19% absolute for Turkish ATIS-SF) on moderately close languages, only a small improvement for very close target languages and no improvement for very distant languages; (2) we propose a novel normalisation which centers language distributions prior to learning rotation maps and is consistently shown to improve rotation alignment across all tasks particularly for Thai, by up to 3% absolute; (3) rotational alignments are more robust and generalise well for structural tasks such as NER which may have higher utterance variability and domain shift; (4) supervision from parallel corpus generally leads to better alignment than dictionary-based, since it offers the possibility of generating contextualised alignments; (5) fine-tuning alignment improves performance for semantic tasks such as slot-filling where the source language data has minimal shift in utterance structure or domain from target language data and particularly improves performance for extremely distant languages (up to +8.39% absolute higher for Thai FB-SF) compared to rotation alignment; (6) for close languages and tasks with similar utterance structure, zero-shot fine-tuning alignment is competitive versus unaligned models trained on additional annotated data in target language.
This work aims to pave the way for optimising language transfer capability in contextual multilingual models. In the future, we would like to further investigate patterns in the embedding space and apply alignment methods into specific regions of the multilingual hyperspace to obtain more tailor-suited alignments between language pairs. We would also like to evaluate zero-shot capabilities of alignments when applied to other language tasks.
- Available for download at: https://github.com/google-research/bert/blob/master/multilingual.md
- PioNER data only has PER, LOC and ORG labels and does not contain MISC.
- The AMARA Corpus: Building Parallel Language Resources for the Educational Domain. In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC’14), Reykjavik, Iceland, pp. 1856–1862. External Links: Cited by: §4.1.
- On Difficulties of Cross-Lingual Transfer with Order Differences: A Case Study on Dependency Parsing. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), Minneapolis, Minnesota, pp. 2440–2452. External Links: Cited by: §1.
- Context-Aware Cross-Lingual Mapping. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), Minneapolis, Minnesota, pp. 3906–3911. External Links: Cited by: §1, §2.2.2.
- Almawave-SLU: A new dataset for SLU in Italian. External Links: Cited by: Table 1, Table 3, §5.6, §5.6, §5.
- Multilingual alignment of contextual word representations. arXiv preprint arXiv:2002.03518. Cited by: §1.
- XNLI: Evaluating Cross-lingual Sentence Representations. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, Cited by: §1, §6.
- Snips voice platform: an embedded spoken language understanding system for private-by-design voice interfaces. arXiv preprint arXiv:1805.10190. Cited by: Table 1.
- BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), Minneapolis, Minnesota, pp. 4171–4186. External Links: Cited by: §1.
- A Simple, Fast, and Effective Reparameterization of IBM Model 2. In Proceedings of the 2013 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Atlanta, Georgia, pp. 644–648. External Links: Cited by: §2.2.2, §4.1.
- pioNER: Datasets and Baselines for Armenian Named Entity Recognition. 2018 Ivannikov Ispras Open Conference (ISPRAS), pp. 56–61. Cited by: §4.3, Table 1.
- Cross-Lingual Ability of Multilingual BERT: An Empirical Study. ArXiv abs/1912.07840. Cited by: §1.
- Europarl: A Parallel Corpus for Statistical Machine Translation. In Conference Proceedings: the tenth Machine Translation Summit, Phuket, Thailand, pp. 79–86. External Links: Cited by: §4.1.
- Word translation without parallel data. Cited by: §4.1.
- OpenSubtitles2015: Extracting Large Parallel Corpora from Movie and TV Subtitles. In International Conference on Language Resources and Evaluation, Cited by: §4.1.
- Bilingual Lexicon Induction with Semi-supervision in Non-Isometric Embedding Spaces. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, Florence, Italy, pp. 184–193. External Links: Cited by: §1.
- Deep Contextualized Word Representations. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers), New Orleans, Louisiana, pp. 2227–2237. External Links: Cited by: §6.
- Evaluation of Spoken Language Systems: the ATIS Domain. In Speech and Natural Language: Proceedings of a Workshop Held at Hidden Valley, Pennsylvania, June 24-27,1990, External Links: Cited by: §4.3, Table 1.
- Cross-lingual Transfer Learning for Multilingual Task Oriented Dialog. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), Minneapolis, Minnesota, pp. 3795–3805. External Links: Cited by: Table 1, Table 3, §5.6, §5.
- Cross-Lingual Alignment of Contextual Word Embeddings, with Applications to Zero-shot Dependency Parsing. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), Minneapolis, Minnesota, pp. 1599–1613. External Links: Cited by: §1.
- On the Limitations of Unsupervised Bilingual Dictionary Induction. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp. 778–788. Cited by: §3.
- Introduction to the CoNLL-2003 Shared Task: Language-Independent Named Entity Recognition. In Proceedings of the Seventh Conference on Natural Language Learning at HLT-NAACL 2003, pp. 142–147. External Links: Cited by: §4.3, Table 1.
- Introduction to the CoNLL-2002 Shared Task: Language-Independent Named Entity Recognition. In COLING-02: The 6th Conference on Natural Language Learning 2002 (CoNLL-2002), External Links: Cited by: §4.3, Table 1.
- (Almost) Zero-Shot Cross-Lingual Spoken Language Understanding. In 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Vol. , pp. 6034–6038. Cited by: Table 1, Table 3, §5.
- Cross-lingual Alignment vs Joint Training: A Comparative Study and A Simple Unified Framework. External Links: Cited by: §1, §2.2.2, Table 3, §5.
- Beto, Bentz, Becas: The Surprising Cross-Lingual Effectiveness of BERT. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), Hong Kong, China, pp. 833–844. External Links: Cited by: §1, Table 3, §5.