A T5-based interpretable reading comprehension model with more accurate evidence training

Information Processing & Management, Год журнала: 2023, Номер 61(2), С. 103584 - 103584

Опубликована: Ноя. 22, 2023

Язык: Английский

---

A cross-guidance cross-lingual model on generated parallel corpus for classical Chinese machine reading comprehension

Information Processing & Management, Год журнала: 2023, Номер 61(2), С. 103607 - 103607

Опубликована: Дек. 16, 2023

Язык: Английский

---

A Token-based transition-aware joint framework for multi-span question answering

Information Processing & Management, Год журнала: 2024, Номер 61(3), С. 103678 - 103678

Опубликована: Фев. 14, 2024

Язык: Английский

---

ArQuAD: An Expert-Annotated Arabic Machine Reading Comprehension Dataset

Cognitive Computation, Год журнала: 2024, Номер 16(3), С. 984 - 1003

Опубликована: Март 11, 2024

Язык: Английский

---

MTMS: Multi-teacher Multi-stage Knowledge Distillation for Reasoning-Based Machine Reading Comprehension

Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval, Год журнала: 2024, Номер unknown, С. 1995 - 2005

Опубликована: Июль 10, 2024

Язык: Английский

---

Interoperable information modelling leveraging asset administration shell and large language model for quality control toward zero defect manufacturing

Journal of Manufacturing Systems, Год журнала: 2024, Номер 77, С. 678 - 696

Опубликована: Окт. 28, 2024

Язык: Английский

---

Ask and Ye shall be Answered: Bayesian tag-based collaborative recommendation of trustworthy experts over time in community question answering

Information Fusion, Год журнала: 2023, Номер 99, С. 101856 - 101856

Опубликована: Май 31, 2023

Several challenging issues have yet to be jointly addressed in the recommendation of experts for community question answering, including dynamicity, comprehensive profiling, incorporation auxiliary data, and manipulation heterogeneous information. We argue that a unified treatment these is beneficial improving effectiveness. In this paper, we introduce formalize new more thorough instance expert-recommendation task which conceived suitably account connections among targeted issues. Moreover, order carry out devised task, present an innovative Bayesian tag-based approach systematically handles all coherent seamlessly manner. At heart proposed unprecedented principled fusion various types The integrated information enables peculiar characterization members terms three inherent properties, i.e., their temporally-discounted interest, expertise, willingness respond. first property determined by looking into questions, while last two are from answers. Within generic answering community, properties its allow selectively routing any specifically set responders. These recommended as trustworthy repliers, who not only actually themes routed per tags, but also still interested such willing answer at time. An extensive empirical assessment involving real-world benchmark data domains reveals higher effectiveness presented compared state-of-the-art competitors.

Язык: Английский

---

Disentangled Retrieval and Reasoning for Implicit Question Answering

IEEE Transactions on Neural Networks and Learning Systems, Год журнала: 2022, Номер 35(6), С. 7804 - 7815

Опубликована: Ноя. 15, 2022

To date, most of the existing open-domain question answering (QA) methods focus on explicit questions where reasoning steps are mentioned xmlns:xlink="http://www.w3.org/1999/xlink">explicitly in question. In this article, we study xmlns:xlink="http://www.w3.org/1999/xlink">implicit QA not evident Implicit is challenging two aspects. First, evidence retrieval difficult since there little overlap between a and its required evidence. Second, answer inference strategy latent tackle implicit QA, propose systematic solution denoted as DisentangledQA, which disentangles topic, attribute, from to guide reasoning. Specifically, disentangle topic attribute information retrieval. For reasoning, disentangled model for prediction based retrieved well representation strategy. The framework empowers each module specific element question, thus, leads effective learning them. Experiments StrategyQA dataset demonstrate effectiveness our method questions, improving performance by 31.7% 4.5%, respectively, achieving best official leaderboard. addition, achieved EntityQuestions dataset, indicating general tasks.

Язык: Английский

---

Global information-aware argument mining based on a top-down multi-turn QA model

Information Processing & Management, Год журнала: 2023, Номер 60(5), С. 103445 - 103445

Опубликована: Июль 6, 2023

Argument mining (AM) aims to automatically generate a graph that represents the argument structure of document. Most previous AM models only pay attention single component (AC) classify type AC or pair ACs identify and argumentative relation (AR) between two ACs. These ignore impact global documents, which is important, especially in some highly structured genres such as scientific papers, where process argumentation relatively fixed. Inspired by this, we propose novel two-stage model leverages information support AM. The first stage uses multi-turn question-answering incrementally an initial identifies relations among At each turn, all related query are generated simultaneously, sibling answer considered. In addition, partially constructed used extension with additional After whole has been determined, second assigns semantic types both ARs them, leveraging from this information. We test proposed methods on datasets (one AbstRCT dataset including 659 abstracts about cancer research other SciARG consists 225 computer linguistic 285 biomedical abstracts) student essay PE 402 essays. Our experiments show our improves state-of-the-art performance for different subtasks, average improvements 1%, 2.41%, 1.1% ACC, ARI ARC task respectively dataset, 2.36%, 1.84%, 8.87% dataset. also achieves comparative results datasets: 87.7% F1 scores ACC task, 81.4% 78.8% task.

Язык: Английский

---

CBKI: A confidence-based knowledge integration framework for multi-choice machine reading comprehension

Knowledge-Based Systems, Год журнала: 2023, Номер 277, С. 110796 - 110796

Опубликована: Июль 19, 2023

Язык: Английский

---