Multi-passage extraction-based machine reading comprehension based on verification sorting. (March 2023)

• Record Type: Journal Article
• Authors: Dong, Runlu; Wang, Xirong; Dong, Lihong; Zhang, Zexuan
• Is Part Of: Computers & electrical engineering. Volume 106(2023)
• Publication Date: 2023-03
• Subjects: Application scenarios -- Multi-passage -- Machine reading comprehension -- Passage retrieval -- Verification sorting; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2023.108576
• Languages: English
• ISSNs: 0045-7906

Abstract

Highlights: This paper proposes a sequencing-based multi-passage reading comprehension model. Compared with the traditional machine reading model, the innovation of this paper is to combine the retriever with the machine reading model, avoiding the retriever and reader being connected in series in the pipeline model. This paper constructs a unified loss function, which includes the losses of paragraph retrieval, answer extraction and answer verification and ranking, and constitutes an end-to-end multi paragraph retrieval reading comprehension model. The proposed multi-channel reading comprehension model based on ranking is applied to complex scenarios. The experiment shows that compared with the traditional single channel machine reading comprehension, the multi-channel machine reading comprehension is applicable to more complex application scenarios, and the accuracy is improved by 9%. Abstract: For traditional single-passage machine reading comprehension, the text data of a single passage does not well reflect the complexity of practical application scenarios. Many researchers have shifted their research goals to study multi-passage machine reading comprehension. To solve the problem of multi-passage machine reading comprehension, this paper proposes a unified multi-module end-to-end reading comprehension model for passage retrieval, answer extraction, and multi-answer verification ranking using a pre-trained model. In this paper, the passage retrieval module selects the passage fragments with the highest probability of survival, and the answer extraction component extracts the possible candidate answers in each passage. The multi-answer verification ranking component uses an attention mechanism to fuse multiple candidate answer feature representations, obtains the score of each candidate answer and selects the candidate answer with the highest score as the final answer. Finally, through experimental validation, the proposed model achieves scores of 49.59 and 46.28 on the evaluation metrics BLUE-4 and ROUGH-L on the DuReader dataset, and scores of 44.78 and 46.45 on the metrics BLUE-1 and ROUGH-L on the MS-MARCO dataset, respectively. Graphical abstract: Image, graphical abstract