Unsupervised named entity transliteration using temporal and phonetic correlation

Abstract

In this paper we investigate unsupervised name transliteration using comparable corpora, corpora where texts in the two languages deal in some of the same topics -and therefore share references to named entities -but are not translations of each other. We present two distinct methods for transliteration, one approach using an unsupervised phonetic transliteration method, and the other using the temporal distribution of candidate pairs. Each of these approaches works quite well, but by combining the approaches one can achieve even better results. We believe that the novelty of our approach lies in the phonetic-based scoring method, which is based on a combination of carefully crafted phonetic features, and empirical results from the pronunciation errors of second-language learners of English. Unlike previous approaches to transliteration, this method can in principle work with any pair of languages in the absence of a training dictionary, provided one has an estimate of the pronunciation of words in text.

Proceedings of The 20th SIGNLL Conference on Computational Natural Language Learning

“…Tao et al (2006) show improvement in transliteration mining performance using phonetic feature vectors resembling the ones we have used. Jagarlamudi and Daumé III (2012) use phonemic representa- Figure 1: Overview of Proposed Approach tion based interlingual projection for multilingual transliteration mining.…”

Entity Translation Mining from Comparable Corpora: Combining Graph Mapping with Corpus Latent Features

“…Tao [15] combines E and temporal feature T. To compute E, Tao constructs a cost matrix using string-alignment and alignment-scoring techniques based on phonological features and phonetic similarity. T builds on observation that entities that refer the same entity often have correlated frequency patterns.…”

“…Corpus latent features. We extract three new latent features from the corpus coreference, temporal distribution, and spatial distribution, as similarly explored in [15], [16]. First, we extract different entity names that are commonly referred to as, or "coreferenced," as the identical entity in documents (e.g., "Beckham" and "David Beckham").…”

Named entity transliteration with comparable corpora

“…The work we report is ongoing. We are investigating transliterations among several language pairs, and are extending these methods to Korean, Arabic, Russian and Hindi -see (Tao et al, 2006).…”