The present paper focuses on the segmentation of two-word phrases containing two closely competing lexical hypotheses. It is hypothesized that the bottom-up information, which also includes a mechanism called the Possible-Word Constraint, is explored first in segmenting these phrases. Non-sensory sentential information influences this process at a later stage and only shows an effect if the bot...

In this paper, we suggest a list of high-level features and study their applicability in detection of cyberpedophiles. We used a corpus of chats downloaded from www.perverted-justice.com and two negative datasets of different nature: cybersex logs available online and the NPS chat corpus. The SVM classification results show that the NPS data and the pedophiles’ conversations can be accurately d...

Knowledge-based numeric open caption recognition is proposed that can recognize numeric captions generated by character generator (CG) and automatically superimpose a modified caption using the recognized text only when a valid numeric caption appears in the aimed specific region of a live sportscast scene produced by other broadcasting stations. In the proposed method, mesh features are extrac...

Many connectionist language processing models have now reached a level of detail at which more realistic representations of semantics are required. In this paper we discuss the extraction of semantic representations from the word co-occurrence statistics of large text corpora and present a preliminary investigation into the validation and optimisation of such representations. We find that there...

Video content can be automatically analysed and indexed using trained classifiers which map low-level features to semantic concepts. Such classifiers need training data consisting of sets of images which contain such concepts and recently it has been discovered that such training data can be located using text-based search to image databases on the internet. Formulating the text queries which l...

This session on corpora and evaluation was composed of two distinct parts. Before the break, four papers dealing with a range of important aspects of evaluation of written language systems and spoken language systems were presented. A printed version of each of these papers is included in the conference proceedings. After the break, a series of informal reports (not included as proceedings pape...

This paper deals with the training phase of a Markov-type linguistic model that is based on transition probabilities between pvirs and triplets of syntactic categories. To determine the o?timal level of detail for a set of syntactic classes we developed a systetn that uses a set-theoretical formalism to defiue such sets mid has some measm~s to comp~uce and c,ptimize them fildividually. In secti...

This paper presents a method for an AAC system to predict a whole response given features of the previous utterance from the interlocutor. It uses a large corpus of scripted dialogs, computes a variety of lexical, syntactic and whole phrase features for the previous utterance, and predicts features that the response should have, using an entropy-based measure. We evaluate the system on a held-o...