Automatic identification of widely-known facts not only plays a crucial role in automatic curation of databases of common-sense knowledge, but also benefits many applications of natural language processing such as searching, question answering, and common-sense reasoning. For better performance of such automatic identification, it is important to understand the linguistic characteristics of the writings about widely-known facts, if there are such characteristics. In order to investigate these linguistic characteristics, we juxtapose the sentences that report an event and the database entries that contain information about that event. We designed a task of identifying the number of databases in which the event of interest appears. From the experiment, we found that the number of databases to which an event is registered is strongly correlated with the writing style in which the event is described in the literature. This means that the style of writing reveals multi-level understanding of the individual authors on the status of knowledge prominence. To the best our knowledge, our work is the first that suggests an extrinsic evidence supporting the existence of such multi-level cognition. (KAIST)

According to Biber (1993) and Biber and Conrad (2009), multiple dimensional approach on several bi-polar features can classify genres founds in texts. Problem is that the process between feature extraction and statistical model building is exhaustively applicable when the machine learning system is dependent on frequency based language models. Following the machine learning approach of Kanaris and Stmatatos (2007), nearly 8,000 frequency-based models are used to induce genre distinction. If their whole process is not successful, they have to exhaustively redo the classification process. Upon such problem, we used a word embedding language model in order to deal feature extraction and model building at the same time. Among several word embedding models, our research is based on Doc2Vec model which captures paragraph feature of texts. We used the genre distinctions from the Project Gutenberg, dividing text databases into three parts in revealing distributional characteristics of linguistic features. Our method for detecting text genre is convenient in process as well as accurate in capturing feature distribution of text genre.

Since Ross (1969) and Rosen (1976), there has been a consensus that Swiping (= P-stranded Sprouting Sluicing) is a subtype of Sluicing. A question is then why wh-the hells are allowed in Swiping, but are not in both Merger and Pied-piped Sluicing, as first documented by Merchant (2002). Regarding this question, Sprouse (2006) argues that the difference between Sluicing and Swiping with the hell is solely due to phonological reasons. However, we argue that the difference between them results from phonological (as a proclitic) and semantic (as a polarity item) deficiencies of the hell.

In this paper, I attempt to find the answers to the questions of what a speech act comes from and how a MP is interpreted. To do this, I discuss whether Mood Phrases (= MPs) can be conjoined or disjoined, and why they can or cannot, in matrix and embedded clauses. If a MP is used as a matrix one, it becomes a speech act. If some phenomena are observed only in matrix MPs, they are phenomena of speech acts. If some phenomena are observed in embedded MPs, they are phenomena of MPs. Based on these criteria, I claim that a speech act comes from the utterance of a MP as a separate sentence, not from the MP itself. This is also supported by the fact that there is no conjunction connective that combines two MPs. Based on these observations, I also claim that kuliko 'and' and hokun 'or' are coordinators of speech acts. Speech acts can be conjoined relatively freely, but they can be disjoined only if a preceding speech act can be re-interpreted as contributing to a larger speech act. And there is only one coordinator -kena 'or' that disjoins embedded MPs and there is no conjunction connective of embedded MPs. I explain all the observations based on the claim that the use of a MP presupposes an event or state that involves the meaning of the MP. It is an utterance event when a matrix MP is used and it is an event or state introduced in the context when an embedded MP is used. This allows us to explain that MPs can be disjoined only if the disjunction connective takes wide scope over a predicate that introduces an event or state that involves the MPs, and leads to the meaning of conjunction. Hence no conjunction connective of MPs is necessary. (Hongik University)