“The Ridge Went North”: Did the Observer Go as Well? Corpus-driven Investigation of Fictive Motion

Abstract

Fictive motion ("The ridge went north") can refer to both dynamic (observer is moving) and static (observer is visually scanning) scenes. Using a corpus of alpine narratives, we extract fictive motion constructions and compare those representing static and dynamic scenes. According to our findings, some of the verbs appear exclusively in static (or dynamic) scenes, while others can be found in both and thus require broader context for the correct annotation. The results can be seen as a step towards the automatic identification of the role of geographic objects in text.

Fictive motion extraction and classification

“…In this article, we set out to examine the way these differences are reflected in the way spatial information is treated and represented linguistically. Pragmatically, our goal included the development of a set of rules for the automated extraction and classification of FM into these three types of spatial descriptions using the corpus of annotated FM provided by Egorova et al (2016). Our contributionthe set of concepts and their linguistic operationalization, as well as resulting rules, is thus valuable for several lines of work, including the development of spatial annotation schemes (Pustejovsky 2017), automated reconstruction of spatial information from text (Vasardani et al 2013, Moncla et al 2016, as well as, more generally, research on spatial description and thinking strategies (Landau and Jackendoff 1993, Denis et al 1999, Zlatev 2007.…”

“…Untravellability can be also constructed through the semantics of other linguistic featuresverbs, adjectives, adverbs with strong connotation, especially in those vista descriptions that convey the sense of place (Egorova et al 2016). In examples (13a-b), the strong connotation of 'precipitously' and 'plunged' excludes the possibility of FM to represent motion of the observer:…”

“…Transducer 4 deals with polysemy in order to avoid false positives as described in Section 4.1. Transducer 5 annotates motion verbs based on the list used by Egorova et al (2016). Transducer 6 implements heuristics to exclude false positives related to complex noun phrases where the spatial entity is not the subject, as described in Section 4.1.…”

Fictive motion in the context of mountaineering

“…In our preliminary study (Egorova, Boo & Purves, 2016), we reported the first results of the extraction and classification of FM in a corpus consisting of 1484 texts (6356,455 words) from the digitized Alpine Journal between 1968 and 2008 (Bubenhofer et al, 2015). In this paper, we use these data to explore more closely the types of scenes described with FM and spatial properties that are highlighted as relevant in the description of mountaineering experience.…”

“…In Section 2, we describe in detail the process of extraction and classification of FM that is outlined in Egorova et al (2016), including interrater reliability measures. In Section 3-5, we explore the three research questions separately; each section has its own description of methods, results, and discussion.…”