Linear complexity context-free parsing pipelines via chart constraints

Abstract

In this paper, we extend methods from Roark and Hollingshead (2008) for reducing the worst-case complexity of a context-free parsing pipeline via hard constraints derived from finite-state tagging pre-processing. Methods from our previous paper achieved quadratic worst-case complexity. We prove here that alternate methods for choosing constraints can achieve either linear or O(N log 2 N) complexity. These worst-case bounds on processing are demonstrated to be achieved without reducing the parsing accuracy, in fact in some cases improving the accuracy. The new methods achieve observed performance comparable to the previously published quadratic complexity method. Finally, we demonstrate improved performance by combining complexity bounding methods with additional high precision constraints.

Improving Shift‐Reduce Phrase‐Structure Parsing with Constituent Boundary Information

“…Parser LR LP F1 SINGLE Charniak (2000) 79.6 82.1 80.8 Bikel (2004) 79.3 82.0 80.6 Zhu et al (2013) 81.9 84.3 83.2 Wang et al (2015) 83.2 Our Baseline 81.9 84.4 83.2 Petrov and Klein (2007) 81.9 84.8 83.3 Watanabe and Sumita (2015) 84.3 Roark and Hollingshead (2009) Zhu et al (2013) 84.4 86.8 85.6 Wang et al (2015) 86.6…”

“…To get semi‐supervised features, we first parse unlabeled data with our baseline parser, and then extract two lists from auto‐parsed data, based on which semi‐supervised features are designed. To the best of our knowledge, previous work on boundary prediction (Roark and Hollingshead ; Bodenstab et al ) does not consider semi‐supervised features.…”

“…Such boundary information was used to prune candidate parses in a chart‐based parser. In Roark and Hollingshead (), the idea was extended further to handle single‐word spans,and experimental results were reported on both English and Chinese. A similar idea was used in Zhang et al () to increase the efficiency of a CCG parser.…”

CSSG: Learning within NLP Pipelines for Scalable Data Mining and Information Extraction

“…In the first, recently graduated PhD student Kristy Hollingshead (now a post-doctoral fellow at the University of Maryland in College Park) collaborated with the PI and graduate student Nate Bodenstab on methods for using a finite-state tagger to achieve complexity improvements in context-free parsing pipelines. First, in 2009, Roark and Hollingshead [2] presented a method to guarantee linear complexity of context-free parsing pipelines given finite-state annotations, as well as related methods that provide large typical case speedups. Recent follow up work [3] showed that additional finite-state tagged constraints can yield further typical case speedups.…”

Cube pruning as heuristic search

“…Recent years have seen a flurry of innovative techniques designed to tackle this problem. These include cube pruning (Chiang, 2007), cube growing (Huang and Chiang, 2007), early pruning (Moore and Quirk, 2007), closing spans (Roark and Hollingshead, 2008;Roark and Hollingshead, 2009), coarse-to-fine methods (Petrov et al, 2008), pervasive laziness (Pust and Knight, 2009), and many more.…”