Reference Resolution for Pedestrian Wayfinding Systems

Götze, Jana; Boye, Johan

doi:10.1007/978-3-319-56759-4_4

Cited by 1 publication

(6 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Baseline WAC SpaceRefNet Prec. p Table 2: Performance of different methods for solving spatial reference resolution (sRR) task on the test set ("(p)" stands for positives, "(n)" stands for negatives), "WAC" stands for words-as-classifiers method (results reported in (Götze and Boye, 2017)).…”

Section: Metricmentioning

confidence: 99%

“…try to identify objects in the images taken from different locations around the world. (Götze and Boye, 2017) deal with reference resolution in a complex city environment. (Chen et al, 2019) present a TOUCH-DOWN dataset, where the agents navigate in a real-life visual urban environment trying to find a hidden object based on a number of cues formulated in a natural language.…”

Section: Introductionmentioning

confidence: 99%

“…The candidate set for every referring expression was set to contain all object segmentations of the given image and every candidate is encoded using a deep convolutional neural network augmented with a number of extra features. Similarly, (Götze and Boye, 2017) have dealt with a constantly changing candidate set of diverse geographical objects in a pedestrian's vicinity. Each geographical object was then represented by a pedestrian's position and a number of properties inferred from Open-StreetMap (OSM) (Haklay and Weber, 2008).…”

Section: Introductionmentioning

confidence: 99%

“…As mentioned above, most of the studies on eRR have been conducted in an unrealistically small toy domain, hence REs can be identi-fied manually, as in (Engonopoulos et al, 2013), (Funakoshi et al, 2012) or (Kennington and Schlangen, 2015). and (Götze and Boye, 2017) addressed RR in a realistic domain, but all REs were manually annotated as well. (Schutte et al, 2010) worked on resolving REs in simple manipulation instructions, e.g.…”

Section: Introductionmentioning

confidence: 99%

“…where U is a representation of an utterance constituting RE, S is the search space of possible referents, O is an object in the search space, O * is the predicted referent. Such a stochastic approach is pursued, for instance, in (Kennington and Schlangen, 2015), (Engonopoulos et al, 2013), (Matuszek et al, 2014), (Funakoshi et al, 2012), , (Götze and Boye, 2017).…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

SpaceRefNet: a neural approach to spatial reference resolution in a real city environment

Kalpakchi¹,

Boye²

2019

Proceedings of the 20th Annual SIGdial Meeting on Discourse and Dialogue

Self Cite

View full text Add to dashboard Cite

Adding interactive capabilities to pedestrian wayfinding systems in the form of spoken dialogue will make them more natural to humans. Such an interactive wayfinding system needs to continuously understand and interpret pedestrian's utterances referring to the spatial context. Achieving this requires the system to identify exophoric referring expressions in the utterances, and link these expressions to the geographic entities in the vicinity. This exophoric spatial reference resolution problem is difficult, as there are often several dozens of candidate referents. We present a neural network-based approach for identifying pedestrian's references (using a network called RefNet) and resolving them to appropriate geographic objects (using a network called Space-RefNet). Both methods show promising results beating the respective baselines and earlier reported results in the literature.

show abstract

Section: Metricmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%