Alignment-Augmented Consistent Translation for Multilingual Open Information Extraction

Abstract

Progress with supervised Open Information Extraction (OpenIE) has been primarily limited to English due to the scarcity of training data in other languages. In this paper, we explore techniques to automatically convert English text for training OpenIE systems in other languages. We introduce the Alignment-Augmented Consistent Translation (AACTRANS) model to translate English sentences and their corresponding extractions consistently with each other -with no changes to vocabulary or semantic meaning which may result from independent translations. Using the data generated with AACTRANS, we train a novel two-stage generative OpenIE model, which we call GEN2OIE, that outputs for each sentence: 1) relations in the first stage and 2) all extractions containing the relation in the second stage. GEN2OIE increases relation coverage using a training data transformation technique that is generalizable to multiple languages, in contrast to existing models that use an English-specific training loss. Evaluations on 5 languages -Spanish, Portuguese, Chinese, Hindi and Telugu -show that the GEN2OIE with AACTRANS data outperforms prior systems by a margin of 6-25% F1. 1 * denotes equal contribution 1 Code and models released at github.com:dair-iitd/moie

Exploiting Duality in Open Information Extraction with Predicate Prompt

“…IMoJIE [14] is also composed of a BERT encoder and an LSTM decoder, and it leverages an iterative generation mechanism to alleviate redundant extraction. Gen2OIE [15] is an OpenIE system based on two mT5 models. Given that ChatGPT is a powerful large language model (LLM) capable of producing contextually appropriate and logically connected responses, We designed the prompts of 0-shot, 3-shot and Chain-Of-Thought (COT) [24] to instruct ChatGPT to achieve extraction task.…”

“…It leverages an iterative generation mechanism to alleviate the redundancy in triplets generation. Gen2OIE [15] is composed of two mT5 [25] models, which reconstructs multiple inputs based on the first model, and then uses the second model to generate triplets.…”

“…The generative solutions model OpenIE as a Seq2Seq problem. For example, IMoJIE [14] adopts an iterative generation mechanism to alleviate redundancy in extractions.Gen2OIE [15] leverages two mT5 models [25] to extract triplets by reconstructing multiple inputs. However, the previous models' inadequate understanding of diverse relations between arguments and the sentence's intricate structure hinder them to extract the complicated triplets Table 1: A toy example of OpenIE where three triplets can be extracted from the input sentence.…”

IndIE: A Multilingual Open Information Extraction Tool For Indic Languages

“…Depending on the level of detail (granularity) in triples, two predicates that can be extracted among numerous possible predicates are "became" and "became the prime minister". Kolluru et al (2022) introduced a novel approach to multilingual Open Information Extraction that leverages Natural Language Generation (NLG) techniques and cross-lingual projections. Their method is capable of extracting overlapping relations (predicates) and triple arguments.…”

“…To compare the performance of our triple extractor (IndIE), we used the following five baselines: (i) M&K (Faruqui and Kumar, 2015), (ii) ArgOE (Gamallo and Garcia, 2015), (iii) PredPatt (White et al, 2016), (iv) Multi2OIE (Ro et al, 2020), and (v) Gen2OIE (Kolluru et al, 2022) , because of their multilingual nature. The source code for M&K is not publicly available, but the authors have released a dataset of sentences and the corresponding triples generated by their method 9 .…”