2。2 Previous Research

2。2。1 Research Targets using UN Speeches

Researchers have rarely targeted linguistic usages of speeches delivered in UN meetings before (Wang, 2012)。 The recently-rose Model United Nations activity accelerated the study, pioneering a few researches on summarizing linguistic features unique in these speeches。 Sun & Chen (2003) concluded speeches in such UN meetings as a special language variant that features increased range of sentence length, enriched sentence types, complexed sentence structure, more adaptive word choice as well as other lexical and semantic characteristics。 Sun & Chen observed a minor change in

3 Interactive aspect is not discussed in this paper as speeches in UN meetings are lengthy and cannot be interrupted by other members, thus have very little interactive meaning。

the usage of verbs that less phrasal but more formal verbs are applied。 Other research (Wang & Zhang, 2011) also pointed out a potential increase in positional subjectivity in verbal usage, however, no statistical proof was given。 The latter two will be the focus of this paper。文献综述

2。2。2 Research in Generic Political Linguistics

Researchers in the generic stream have conducted their studies with multiple approaches on the usage of interpersonal metafunction in both political linguistics and other related fields。 Small corpus- based researches with manual discourse analysis are common (He, 2009; Xiao, 2012; Yan & Zhao, 2009)。 These researches usually could conclude the language features of the observed genre or context but fail short in making their findings computable。

Statisticians have conducted probability-theory-based studies in various manipulation, from earlier identifying the author of the Federalist Papers (Mosteller & Wallace, 1963) to recent modality calculating in US presidents’ inaugural addresses (Sun & Zhao, 2014)。 The theoretical progress in combining Functional Grammar and Machine Learning, especially the recent Neural network-based Deep Learning technology, significantly improved theories and practices in linguistic computing。 For example, Tearle et al。 (2008) were able to reproduce manual-processed Mosteller & Wallace’s experiment result entirely with machine learning computers。

3 The Speech Corpus

Few research-ready linguistic corpora on United Nations documents, mostly parallel corpora on resolutions passed by General Assembly or Security Council for translation research (Rafalovitch & Dale, 2009), exists。 As this research focus on addresses given by members of the Security Council, a linguistic corpus featuring speech data is needed to set up using UN meetings’ provisional records。

3。1 The Nature of Provisional Records

The United Nations provides provisional records for Security Council’s meetings since 1994。 These documents contain minutes of speeches given by speakers in English and delegation- interpreted4   and approved verbatim translations for speech in other languages。来*自~优|尔^论:文+网www.youerw.com +QQ752018766*

A provisional record of a session of such meeting consists of the following parts:

1。 Information regarding the session and the provisional record itself, including a symbol to identify the record, e。g。 S/PV。7354 indicating the record of the 7354th meeting of the Security Council; the topic and agenda; list of member states and speakers attended the meeting。

2。 Speeches given by countries and speakers, as well as the president5, in the following format chronologically:

Speaker’s name (Country) (spoke in Language): Speech content