Accenture at CheckThat! 2021: Interesting claim identification and ranking with contextually sensitive lexical training data augmentation

This paper discusses the approach used by the Accenture Team for CLEF2021\nCheckThat! Lab, Task 1, to identify whether a claim made in social media would\nbe interesting to a wide audience and should be fact-checked. Twitter training\nand test data were provided in English, Arabic, Spanish, Turkish, and\nBulgarian. Claims were to be classified (check-worthy/not check-worthy) and\nranked in priority order for the fact-checker. Our method used deep neural\nnetwork transformer models with contextually sensitive lexical augmentation\napplied on the supplied training datasets to create additional training\nsamples. This augmentation approach improved the performance for all languages.\nOverall, our architecture and data augmentation pipeline produced the best\nsubmitted system for Arabic, and performance scales according to the quantity\nof provided training data for English, Spanish, Turkish, and Bulgarian. This\npaper investigates the deep neural network architectures for each language as\nwell as the provided data to examine why the approach worked so effectively for\nArabic, and discusses additional data augmentation measures that should could\nbe useful to this problem.\n