Maximal Clique Based Non-Autoregressive Open Information Extraction

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 9696–9706 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 9696 Maximal Clique Based Non-Autoregressive Open Information Extraction Bowen Yu1,2∗, Yucheng Wang1,2∗, Tingwen Liu1,2†, Hongsong Zhu1,2, Limin Sun1,2, Bin Wang3 1Institute of Information Engineering, Chinese Academy of Sciences 2School of Cyber Security, University of Chinese Academy of Sciences 3Xiaomi AI Lab, Xiaomi Inc., Beijing, China {yubowen,wangyucheng,liutingwen}@iie.ac.cn {zhuhongsong,sunlimin}@iie.ac.cn wangbin11@xiaomi.com Abstract Open Information Extraction (OpenIE) aims to discover textual facts from a given sentence. In essence, the facts contained in plain text are unordered. However, the popular Ope- nIE systems usually output facts sequentially in the way of predicting the next fact condi- tioned on the previous decoded ones, which enforce an unnecessary order on the facts and involve the error accumulation between autore- gressive steps. To break this bottleneck, we propose MacroIE, a novel non-autoregressive framework for OpenIE. MacroIE firstly con- structs a fact graph based on the table filling scheme, in which each node denotes a fact ele- ment, and an edge links two nodes that belong to the same fact. Then OpenIE can be refor- mulated as a non-parametric process of finding maximal cliques from the graph. It directly outputs the final set of facts in one go, thus getting rid of the burden of predicting fact or- der, as well as the error propagation between facts. Experiments conducted on two bench- mark datasets show that our proposed model significantly outperforms current state-of-the- art methods, beats the previous systems by as much as 5.7 absolute gain in F1 score. 1 Introduction Open Information Extraction (OpenIE) aims to con- vert natural text to semi-structured knowledge, by mining facts in the form of n-ary tuples r = (sub- ject, predicate, object1, · · · , objectm), composed of a single subject and predicate as well as m ob- jects. While traditional IE systems require peo- ple to pre-specify the set of interested relations and provide per-relation training data, OpenIE is built on the principles of ontology-free (Tang et al., 2020), making it possible to adapt to various do- mains and applications, such as extending knowl- edge bases (Dong et al., 2014), and facilitating question answering systems (Fader et al., 2014). ∗ The two authors contribute equally. † Corresponding author. John is the premier and first minister of British Columbia Subject Predicate Object John premier of British Columbia John first minister of British Columbia Figure 1: An example of OpenIE in a sentence. Starting from rule-based systems to neural net- works, OpenIE has attracted increasing attention in recent years but remains challenging (Niklaus et al., 2018), due to the intrinsic difficulty in identifying complicated facts, including: (1) Overlapping, one fact element (either subject, predicate, or object) may belong to multiple facts in a sentence. For ex- ample, in Figure 1, the entity pairs of two facts are identical but the predicates are different; (2) Dis- continuous, one fact element can consist of spans that are separated by intervals, as the …

Content truncated to 500 words