Graph- and surface-level sentence chunking

Abstract

The computing cost of many NLP tasks increases faster than linearly with the length of the representation of a sentence. For parsing the representation is tokens, while for operations on syntax and semantics it will be more complex. In this paper we propose a new task of sentence chunking: splitting sentence representations into coherent substructures. Its aim is to make further processing of long sentences more tractable. We investigate this idea experimentally using the Dependency Minimal Recursion Semantics (DMRS) representation.

Parsing transcripts of speech

“…We might also introduce a heuristic to deal with long speechunits, which are particularly troublesome for existing parsers. One technique we can adopt is that of 'clause splitting', or 'chunking', which subdivides long strings for the purpose of higher quality analysis over small units (Tjong et al, 2001;Muszyńska, 2016). We hypothesise that such a step would play to the strength of existing parsers, namely their robustness over short inputs.…”

Ensemble Model for Chunking

“…Named Entity Recognition was initially performed using extensive knowledge base systems, its orthographic features, ontological and lexicon rules [4][2][14] [15]. However, the new trend has shifted towards neural network-based structures to define entity relations [5] Chunking has been done using machine learning-based models such as HMM(Hidden Markov Model) [7] [17] and Maximum Entropy model and has gradually seen a shift to Statistical models such as Support Vector Machines and Boosting [8], [3], [7]. In more recent times, Neural Models have been on a rise as a tool for chunking.…”

“…Chunking is the process of splitting the words of a sentence into tokens and then grouping the tokens in a meaningful way. These chunks are our point of interest which are used to solve our relevant NLP tasks [3]. It labels every word of the sentence suitably and thus lays out a basic framework for bigger tasks such as question answering, information extraction, topic modeling, etc [16].…”

Realization of long sentences using chunking

“…This is because of the presence of grammatical structures not covered by the chunking algorithm, which lead to incorrect subgraphs. Limitations of the chunking algorithm are discussed in detail elsewhere (Muszyńska, 2016).…”

“…In this paper we propose chunking (Muszyńska, 2016) as a way to reduce memory and time cost of realization. The general idea of chunking is that strings and semantic representations can be divided into smaller parts which can be processed independently and then recombined without a loss of information.…”