More than Word Cooccurrence: Exploring Support and Opposition in International Climate Negotiations with Semantic Parsing

More than Word Cooccurrence: Exploring Support and Opposition in International Climate Negotiations with Semantic Parsing Pablo Ruiz Fabo, Clément Plancq, Thierry Poibeau LATTICE Lab (CNRS, Ecole normale supérieure, U Paris 3, PSL Research University, Université Sorbonne Paris Cité) 1, rue Maurice Arnoux, 92120 Montrouge, France E-mail: {pablo.ruiz.fabo,clement.plancq,thierry.poibeau}@ens.fr Abstract Text analysis methods widely used in digital humanities often involve word co-occurrence, e.g. concept co-occurrence networks. These methods provide a useful corpus overview, but cannot determine the predicates that relate co-occurring concepts. Our goal was identifying propositions expressing the points supported or opposed by participants in international climate negotiations. Word co-occurrence methods were not sufficient, and an analysis based on open relation extraction had limited coverage for nominal predicates. We present a pipeline which identifies the points that different actors support and oppose, via a domain model with support/opposition predicates, and analysis rules that exploit the output of semantic role labelling, syntactic dependencies and anaphora resolution. Entity linking and keyphrase extraction are also performed on the propositions related to each actor. A user interface allows examining the main concepts in points supported or opposed by each participant, which participants agree or disagree with each other, and about which issues. The system is an example of tools that digital humanities scholars are asking for, to render rich textual information (beyond word co-occurrence) more amenable to quantitative treatment. An evaluation of the tool was satisfactory. Keywords: climate diplomacy, semantic role labelling, digital humanities 1. Introduction Text-analysis methods widely used in social or political sciences often involve word co-occurrence. For instance, the concept co-occurrence networks surveyed in Venturini et al. (2012), or the clustering and topic modelling approaches surveyed in Grimmer and Stewart (2013). These methods are useful in order to arrive at an overview of the content of large corpora. However, these techniques do not identify which predicates relate co-occurring elements with each other. If an actor like France is mentioned in the same sentence as a concept, like stricter regulations, which is the verb mediating between both? Is France in favour of, or against stricter regulations? Several technologies can detect related elements in texts, and the predicate that indicates their relation. A recent approach is Open Relation Extraction (e.g. Mausam et al., 2012), where relations are identified without the need to previously specify a vocabulary of predicates or actors. The corpus we’re working on consists of summaries of international climate negotiations (ENB corpus) 1 . A single sentence in this corpus can contain several support and opposition predicates, which can be verbal or nominal (see Figure 1). For this corpus, the results of a workflow based on open relation extraction tools were uneven, particularly with nominal predicates. 2 To address these challenges, we developed an application with a domain model and analysis rules which operate on outputs for semantic role labelling, syntactic dependencies and anaphora resolution, provided by a natural language processing (NLP) pipeline. 1The Earth Negotiations Bulletin corpus (ENB), http://www.iisd.ca/vol12/ 2 Based on outputs from Open Information Extraction 4.0 (https://github.com/knowitall/openie) and OLLIE …

Content truncated to 500 words