Model Transfer for Tagging Low-resource Languages using a Bilingual Dictionary
By Meng Fang, Trevor Cohn. 2017
74 citations recorded by the source corpus.
Cross-lingual model transfer is a compelling and popular method for predicting annotations in a low-resource language, whereby parallel corpora provide a bridge to a high-resource language, and its associated annotated corpora. However, parallel data is not readily available for many languages, limiting the applicability of these approaches. We address these drawbacks in our framework which takes advantage of cross-lingual word embeddings trained solely on a high coverage dictionary. We propose a novel neural network model for joint training from both sources of data based on cross-lingual word embeddings, and show substantial empirical improvements over baseline techniques. We also propose several active learning heuristics, which result in improvements over competitive benchmark methods.
Subfields: Artificial Intelligence
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
- Sebastian Ruder
- Ivan Vulic
- Anders Søgaard
- Jacob Eisenstein
- Afshin Rahimi
- Iryna Gurevych
- Trevor Cohn
- Noah A. Smith
Citations in the corpus
- A Survey of Cross-lingual Word Embedding Models (2017, methodology): “Fang and Cohn (2017) use data from the CoNLL-X datasets of European languages (Buchholz & Marsi, 2006), from CoNLL 200316 and from Das and Petrov (2011), the latter of which is also used by Gouws and Søgaard (2015).”
- Strong Baselines for Neural Semi-Supervised Learning under Domain Shift (2018, unknown): “, (Fang and Cohn, 2017).”
- Learning how to Active Learn: A Deep Reinforcement Learning Approach (2017, unknown): “, 2014), and structured prediction (Shen et al., 2004; Settles and Craven, 2008; Stratos and Collins, 2015; Fang and Cohn, 2017).”
- Massively Multilingual Transfer for NER (2019, methodology): “There are a few exceptions, which use transfer from several languages, ranging from multitask learning (Duong et al., 2015; Ammar et al., 2016; Fang and Cohn, 2017), and annotation projection from several languages (T¨ackstr¨om, 2012; Fang and Cohn, 2016; Plank and Agi´c, 2018).”
- Unsupervised Domain Adaptation of Contextualized Embeddings for Sequence Labeling (2019, background): “In addition to supervised and unsupervised domain adaptation, there are “distantly supervised” methods that construct noisy target domain instances, e.g., by using a bilingual dictionary (Fang and Cohn, 2017).”