Automatic identification of discourse markers in dialogues: An in-depth study of like and well

The lexical items like and well can serve as discourse markers (DMs), but can also play numerous other roles, such as verb or adverb. Identifying the occurrences that function as DMs is an important step for language understanding by computers. In this study, automatic classifiers using lexical, prosodic/positional and sociolinguistic features are trained over transcribed dialogues, manually annotated with DM information. The resulting classifiers improve state-of-the-art performance of DM identification, at about 90% recall and 79% precision for like (84.5% accuracy, κ = 0.69), and 99% recall and 98% precision for well (97.5% accuracy, κ = 0.88). Automatic feature analysis shows that lexical collocations are the most reliable indicators, followed by prosodic/positional features, while sociolinguistic features are marginally useful for the identification of DM like and not useful for well. The differentiated processing of each type of DM improves classification accuracy, suggesting that these types should be treated individually.

Towards Automatic Identification of Discourse Markers in Dialogs: the Case of 'Like'

This article discusses the detection of discourse markers (DM) in dialog transcriptions, by human annotators and by automated means. After a theoretical discussion of the definition of DMs and their relevance to natural language processing, we focus on the role of like as a DM. Results from experiments with human annotators show that detection of DMs is a difficult but reliable task, which requires prosodic information from soundtracks. Then, several types of features are defined for automatic disambiguation of like: collocations, part-of-speech tags and duration-based features. Decision-tree learning shows that for like, nearly 70% precision can be reached, with near 100% recall, mainly using collocation filters. Similar results hold for well, with about 91% precision at 100% recall.

Automatic extraction of femur contours from calibrated fluoroscopic images

Automatic identification and extraction of bone contours from X-ray images is an essential first step task for further medical image analysis. In this paper we propose a 3D statistical model based framework for the proximal femur contour extraction from calibrated X-ray images. The automatic initialization is solved by an estimation of Bayesian network algorithm to fit a multiple component geometrical model to the X-ray data. The contour extraction is accomplished by a non-rigid 2D/3D registration between a 3D statistical model and the X-ray images, in which bone contours are extracted by a graphical model based Bayesian inference. Preliminary experiments on clinical data sets verified its validity

Extended CT scale overcomes restoration caused streak artifacts for dental identification in CT--3D color encoded automatic discrimination of dental restorations

OBJECTIVE: Besides DNA, dental radiographs play a major role in the identification of victims in mass casualties or in corpses with major postmortem alterations. Computed tomography (CT) is increasingly applied in forensic investigations and is used to scan the dentition of deceased persons within minutes. We investigated different restoration materials concerning their radiopacity in CT for dental identification purposes. METHODS: Extracted teeth with different filling materials (composite, amalgam, ceramic, temporary fillings) were CT scanned. Radiopacities of the filling materials were analyzed in extended CT scale images. RESULTS: Radiopacity values ranged from 6000-8500HU (temporary fillings), 4500-17000HU (composite fillings) and >30710HU (Amalgam and Gold). The values were used to define presets for a 3D colored volume rendering software. CONCLUSIONS: The effects of filling material caused streak artifacts could be distinctively reduced for the assessment of the dental status and a postprocessing algorithm was introduced that allows for 3D color encoded visualization and discrimination of different dental restorations based on postmortem CT data.

The examination and identification of bite marks in foods using 3D scanning and 3D comparison methods

Bite mark analysis offers the opportunity to identify the biter based on the individual characteristics of the dentitions. Normally, the main focus is on analysing bite mark injuries on human bodies, but also, bite marks in food may play an important role in the forensic investigation of a crime. This study presents a comparison of simulated bite marks in different kinds of food with the dentitions of the presumed biter. Bite marks were produced by six adults in slices of buttered bread, apples, different kinds of Swiss chocolate and Swiss cheese. The time-lapse influence of the bite mark in food, under room temperature conditions, was also examined. For the documentation of the bite marks and the dentitions of the biters, 3D optical surface scanning technology was used. The comparison was performed using two different software packages: the ATOS modelling and analysing software and the 3D studio max animation software. The ATOS software enables an automatic computation of the deviation between the two meshes. In the present study, the bite marks and the dentitions were compared, as well as the meshes of each bite mark which were recorded in the different stages of time lapse. In the 3D studio max software, the act of biting was animated to compare the dentitions with the bite mark. The examined food recorded the individual characteristics of the dentitions very well. In all cases, the biter could be identified, and the dentitions of the other presumed biters could be excluded. The influence of the time lapse on the food depends on the kind of food and is shown on the diagrams. However, the identification of the biter could still be performed after a period of time, based on the recorded individual characteristics of the dentitions.

Time course based artifact identification for independent components of resting-state FMRI

In functional magnetic resonance imaging (fMRI) coherent oscillations of the blood oxygen level-dependent (BOLD) signal can be detected. These arise when brain regions respond to external stimuli or are activated by tasks. The same networks have been characterized during wakeful rest when functional connectivity of the human brain is organized in generic resting-state networks (RSN). Alterations of RSN emerge as neurobiological markers of pathological conditions such as altered mental state. In single-subject fMRI data the coherent components can be identified by blind source separation of the pre-processed BOLD data using spatial independent component analysis (ICA) and related approaches. The resulting maps may represent physiological RSNs or may be due to various artifacts. In this methodological study, we propose a conceptually simple and fully automatic time course based filtering procedure to detect obvious artifacts in the ICA output for resting-state fMRI. The filter is trained on six and tested on 29 healthy subjects, yielding mean filter accuracy, sensitivity and specificity of 0.80, 0.82, and 0.75 in out-of-sample tests. To estimate the impact of clearly artifactual single-subject components on group resting-state studies we analyze unfiltered and filtered output with a second level ICA procedure. Although the automated filter does not reach performance values of visual analysis by human raters, we propose that resting-state compatible analysis of ICA time courses could be very useful to complement the existing map or task/event oriented artifact classification algorithms.

Use of Extended Characteristics of Locomotion and Feeding Behavior for Automated Identification of Lame Dairy Cows.

This study was carried out to detect differences in locomotion and feeding behavior in lame (group L; n = 41; gait score ≥ 2.5) and non-lame (group C; n = 12; gait score ≤ 2) multiparous Holstein cows in a cross-sectional study design. A model for automatic lameness detection was created, using data from accelerometers attached to the hind limbs and noseband sensors attached to the head. Each cow's gait was videotaped and scored on a 5-point scale before and after a period of 3 consecutive days of behavioral data recording. The mean value of 3 independent experienced observers was taken as a definite gait score and considered to be the gold standard. For statistical analysis, data from the noseband sensor and one of two accelerometers per cow (randomly selected) of 2 out of 3 randomly selected days was used. For comparison between group L and group C, the T-test, the Aspin-Welch Test and the Wilcoxon Test were used. The sensitivity and specificity for lameness detection was determined with logistic regression and ROC-analysis. Group L compared to group C had significantly lower eating and ruminating time, fewer eating chews, ruminating chews and ruminating boluses, longer lying time and lying bout duration, lower standing time, fewer standing and walking bouts, fewer, slower and shorter strides and a lower walking speed. The model considering the number of standing bouts and walking speed was the best predictor of cows being lame with a sensitivity of 90.2% and specificity of 91.7%. Sensitivity and specificity of the lameness detection model were considered to be very high, even without the use of halter data. It was concluded that under the conditions of the study farm, accelerometer data were suitable for accurately distinguishing between lame and non-lame dairy cows, even in cases of slight lameness with a gait score of 2.5.