Representation learning in discourse parsing: A survey
- PDF / 378,769 Bytes
- 26 Pages / 612 x 792 pts (letter) Page_size
- 63 Downloads / 217 Views
Print-CrossMark
https://doi.org/10.1007/s11431-020-1685-2
Special Topic: Natural Language Processing Technology
. Review .
Representation learning in discourse parsing: A survey SONG Wei* & LIU LiZhen* College of Information Engineering and Academy for Multidisciplinary Studies, Capital Normal University, Beijing 100056, China Received March 9, 2020; accepted July 8, 2020; published online September 16, 2020
Neural network based deep learning methods aim to learn representations of data and have produced state-of-the-art results in many natural language processing (NLP) tasks. Discourse parsing is an important research topic in discourse analysis, aiming to infer the discourse structure and model the coherence of a given text. This survey covers text-level discourse parsing, shallow discourse parsing and coherence assessment. We first introduce the basic concepts and traditional approaches, and then focus on recent advances in discourse structure oriented representation learning. We also introduce a trend of discourse structure aware representation learning that is to exploit discourse structures or discourse objectives for learning representations of sentences and documents for specific applications or for general purpose. Finally, we present a brief summary of the progress and discuss several future directions. discourse analysis, discourse parsing, discourse relation, coherence assessment, representation learning Citation:
Song W, Liu L Z. Representation learning in discourse parsing: A survey. Sci China Tech Sci, 2020, 63, https://doi.org/10.1007/s11431-020-1685-2
1 Introduction Discourse commonly comprises multiple clauses or sentences. Within a discourse, its parts relate to one another as a coherent whole to clearly express its meaning. Discourse analysis is an important research area in NLP, studies linguistic phenomena in discourse and aims to model discourse coherence and discourse structures, which are essential in many downstream NLP applications such as text summarization, machine translation, question answering, sentiment analysis, and language assessment. Early work has developed various theories to describe discourse structures. The most influential ones include lexical cohesion [1], intentional structure theory [2], coherence relation theory [3], rhetorical structure theory (RST) [4], centering theory [5], discourse representation theory (DRT) [6], and lexicalized tree adjoining grammar for discourse (DLTAG) theory [7]. These theories explain the structure of a *Corresponding authors (email: [email protected]; liz [email protected])
text through discourse relations or entity relations [8–10]. Inspired by the theories, computational discourse analysis aims to automatically recover discourse structures from raw or parsed text, which can be represented as a tree, a graph, topical or functional segments. The typical tasks include discourse relation recognition, coherence assessment, topic segmentation, coreference resolution, argumentative zoning, etc. For decades, statistical machine learning
Data Loading...
