SimAlign: High Quality Word Alignments without Parallel Training Data using Static and Contextualized Embeddings

By Masoud Jalili Sabet, Philipp Dufter, Hinrich Schütze. 2020

280 citations recorded by the source corpus.

Word alignments are useful for tasks like statistical and neural machine translation (NMT) and cross-lingual annotation projection. Statistical word aligners perform well, as do methods that extract alignments jointly with translations in NMT. However, most approaches require parallel training data and quality decreases as less training data is available. We propose word alignment methods that require no parallel data. The key idea is to leverage multilingual word embeddings – both static and contextualized – for word alignment. Our multilingual embeddings are created from monolingual data only without relying on any parallel data or dictionaries. We find that alignments created from embeddings are superior for four and comparable for two language pairs compared to those produced by traditional statistical aligners – even with abundant parallel data; e.g., contextualized embeddings achieve a word alignment F1 for English-German that is 5 percentage points higher than eflomal, a high-quality statistical aligner, trained on 100k parallel sentences.

Subfields: Artificial Intelligence, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Natural Language Processing

Citations in the corpus

Survey of Hallucination in Natural Language Generation (2022, methodology): “[237] use an unsupervised model that extracts alignments from similarity matrices of word embeddings [162], and then predicts the target token as hallucinated if it is not aligned to the source.”
InfoXLM: An Information-Theoretic Framework for Cross-Lingual Language Model Pre-Training (2020, methodology): “Because previous analysis (Sabet et al., 2020; Dufter and Schutze, 2020) shows that the speciﬁc layers of MMLM learn more universal representations and work better on cross-lingual retrieval tasks than other layers.”
Detecting Hallucinated Content in Conditional Neural Sequence Generation (2020, methodology): “SimAlign is essentially used for crosslingual tasks, and we adapt it to summarization by using embeddings from the pretrained BERT-large (De-vlin et al., 2019).”
XLM-E: Cross-lingual Language Model Pre-training via ELECTRA (2021, neutral): “This trend is also found in other cross-lingual language models such as XLM-R and XLM-Align (Jalili Sabet et al., 2020; Chi et al., 2021c).”
Hallucinated but Factual! Inspecting the Factuality of Hallucinations in Abstractive Summarization (2021, methodology): “Baselines We compare our approach with four baselines: a teacher forcing trained summarizer (MLE), a RL-based summarizer (RL) (Pang and He, 2021) and a summarizer trained with the loss truncation technique from Kang and Hashimoto (2020). We also replace our factuality assessment model ENTFA with Fil”
Word Alignment by Fine-tuning Embeddings on Parallel Corpora (2021, methodology): “While language models can be used to produce reasonable word alignments even without any finetuning (Sabet et al., 2020), we propose objectives that further improve their alignment ability if we have access to parallel data.”