UniKeyphrase: A Unified Extraction and Generation Framework for Keyphrase Prediction

UniKeyphrase: A Unified Extraction and Generation Framework for Keyphrase Prediction Huanqin Wu1∗, Wei Liu1∗, Lei Li2, Dan Nie1, Tao Chen1, Feng Zhang 1, Di Wang 1 1Tencent AI Platform Department, China 2Beijing University of Posts and Telecommunications {huanqinwu,thinkweeliu,kathynie,vitochen,jayzhang,diwang}@tencent.com leili@bupt.edu.cn Abstract Keyphrase Prediction (KP) task aims at pre- dicting several keyphrases that can summarize the main idea of the given document. Main- stream KP methods can be categorized into purely generative approaches and integrated models with extraction and generation. How- ever, these methods either ignore the diver- sity among keyphrases or only weakly cap- ture the relation across tasks implicitly. In this paper, we propose UniKeyphrase, a novel end-to-end learning framework that jointly learns to extract and generate keyphrases. In UniKeyphrase, stacked relation layer and bag- of-words constraint are proposed to fully ex- ploit the latent semantic relation between ex- traction and generation in the view of model structure and training process, respectively. Experiments on KP benchmarks demonstrate that our joint approach outperforms main- stream methods by a large margin.1 1 Introduction Keyphrases are several phrases that highlight core topics or information of a document. Given a doc- ument, the KP task focuses on automatically ob- taining a set of keyphrases. As a basic NLP task, keyphrase prediction is useful for numerous down- stream NLP tasks such as summarization (Wang and Cardie, 2013; Pasunuru and Bansal, 2018), doc- ument clustering (Hulth and Megyesi, 2006), infor- mation retrieval (Kim et al., 2013). Keyphrases of a document fall into two cate- gories: present keyphrase that appears continuously in the document, and absent keyphrase which does not exist in the document. Figure 1 shows an exam- ple of a document and its keyphrases. Traditional KP methods are mainly extractive, which have ∗ Equal contribution. 1Our code is available on https://github.com/thinkwee/UniKeyphrase Document: On selecting an optimal wavelet for detecting singularities in traffic and vehicular data. …… applications of wavelet transform s ( wts ) in traffic engineering have been introduced however , …… , second order difference , oblique cumulative curve , and short time fourier transform ) . it then mathematically describes wts ability to detect singularities in traffic data . …… , it is shown that selecting a suitable wavelet largely depends on the specific research topic , and that the mexican hat wavelet generally gives a satisfactory performance in detecting singularities in traffic and vehicular data . Present keyphrases: { wavelet transform, oblique cumulative curve, short time fourier, the mexican hat wavelet } Absent keyphrases: { singularity detection, traffic data analysis } Figure 1: An example of an input document and its expected keyphrases. Blue and red denote present and absent keyphrases, respectively. been extensively researched in past decades (Wit- ten et al., 2005; Nguyen and Kan, 2007; Medelyan et al., 2009; Lopez and Romary, 2010; Zhang et al., 2016; Alzaidy et al., 2019; Sun et al., 2020). These methods aim to select text spans or phrases directly in the document, which show promising results on present keyphrase prediction. However, extrac- tive …

Content truncated to 500 words