Distributional semantics and machine learning for statistical machine translation

[EU]Lan honetan semantika distribuzionalaren eta ikasketa automatikoaren erabilera aztertzen dugu itzulpen automatiko estatistikoa hobetzeko. Bide horretan, erregresio logistikoan oinarritutako ikasketa automatikoko eredu bat proposatzen dugu hitz-segiden itzulpen- probabilitatea modu dinamikoan modelatzeko. Proposatutako eredua itzulpen automatiko estatistikoko ohiko itzulpen-probabilitateen orokortze bat dela frogatzen dugu, eta testuinguruko nahiz semantika distribuzionaleko informazioa barneratzeko baliatu ezaugarri lexiko, hitz-cluster eta hitzen errepresentazio bektorialen bidez. Horretaz gain, semantika distribuzionaleko ezagutza itzulpen automatiko estatistikoan txertatzeko beste hurbilpen bat lantzen dugu: hitzen errepresentazio bektorial elebidunak erabiltzea hitz-segiden itzulpenen antzekotasuna modelatzeko. Gure esperimentuek proposatutako ereduen baliagarritasuna erakusten dute, emaitza itxaropentsuak eskuratuz oinarrizko sistema sendo baten gainean. Era berean, gure lanak ekarpen garrantzitsuak egiten ditu errepresentazio bektorialen mapaketa elebidunei eta hitzen errepresentazio bektorialetan oinarritutako hitz-segiden antzekotasun neurriei dagokienean, itzulpen automatikoaz haratago balio propio bat dutenak semantika distribuzionalaren arloan.

Semantic-level Visual Content Descriptor combining visual and textual cues as collaterals

In this paper, we introduce a novel high-level visual content descriptor which is devised for performing semantic-based image classification and retrieval. The work can be treated as an attempt to bridge the so called “semantic gap”. The proposed image feature vector model is fundamentally underpinned by the image labelling framework, called Collaterally Confirmed Labelling (CCL), which incorporates the collateral knowledge extracted from the collateral texts of the images with the state-of-the-art low-level image processing and visual feature extraction techniques for automatically assigning linguistic keywords to image regions. Two different high-level image feature vector models are developed based on the CCL labelling of results for the purposes of image data clustering and retrieval respectively. A subset of the Corel image collection has been used for evaluating our proposed method. The experimental results to-date already indicates that our proposed semantic-based visual content descriptors outperform both traditional visual and textual image feature models.

Automatically constructing the semantic web

The storage and processing capacity realised by computing has lead to an explosion of data retention. We now reach the point of information overload and must begin to use computers to process more complex information. In particular, the proposition of the Semantic Web has given structure to this problem, but has yet realised practically. The largest of its problems is that of ontology construction; without a suitable automatic method most will have to be encoded by hand. In this paper we discus the current methods for semi and fully automatic construction and their current shortcomings. In particular we pay attention the application of ontologies to products and the particle application of the ontologies.

Networks of contextualised ontologies for the Semantic Web

Currently many ontologies are available for addressing different domains. However, it is not always possible to deploy such ontologies to support collaborative working, so that their full potential can be exploited to implement intelligent cooperative applications capable of reasoning over a network of context-specific ontologies. The main problem arises from the fact that presently ontologies are created in an isolated way to address specific needs. However we foresee the need for a network of ontologies which will support the next generation of intelligent applications/devices, and, the vision of Ambient Intelligence. The main objective of this paper is to motivate the design of a networked ontology (Meta) model which formalises ways of connecting available ontologies so that they are easy to search, to characterise and to maintain. The aim is to make explicit the virtual and implicit network of ontologies serving the Semantic Web.

What's in a name? Lexical knowledge drives infants' visual preferences in the absence of referential input

Seventeen-month-old infants were presented with pairs of images, in silence or with the non-directive auditory stimulus 'look!'. The images had been chosen so that one image depicted an item whose name was known to the infant, and the other image depicted an image whose name was not known to the infant. Infants looked longer at images for which they had names than at images for which they did not have names, despite the absence of any referential input. The experiment controlled for the familiarity of the objects depicted: in each trial, image pairs presented to infants had previously been judged by caregivers to be of roughly equal familiarity. From a theoretical perspective, the results indicate that objects with names are of intrinsic interest to the infant. The possible causal direction for this linkage is discussed and it is concluded that the results are consistent with Whorfian linguistic determinism, although other construals are possible. From a methodological perspective, the results have implications for the use of preferential looking as an index of early word comprehension.

xhRank: ranking entities on the Semantic Web

In general, ranking entities (resources) on the Semantic Web (SW) is subject to importance, relevance, and query length. Few existing SW search systems cover all of these aspects. Moreover, many existing efforts simply reuse the technologies from conventional Information Retrieval (IR), which are not designed for SW data. This paper proposes a ranking mechanism, which includes all three categories of rankings and are tailored to SW data.

Genesis and expansion of metazoan transcription factor gene classes

We know little about the genomic events that led to the advent of a multicellular grade of organization in animals, one of the most dramatic transitions in evolution. Metazoan multicellularity is correlated with the evolution of embryogenesis, which presumably was underpinned by a gene regulatory network reliant on the differential activation of signaling pathways and transcription factors. Many transcription factor genes that play critical roles in bilaterian development largely appear to have evolved before the divergence of cnidarian and bilaterian lineages. In contrast, sponges seem to have a more limited suite of transcription factors, suggesting that the developmental regulatory gene repertoire changed markedly during early metazoan evolution. Using whole- genome information from the sponge Amphimedon queenslandica, a range of eumetazoans, and the choanoflagellate Monosiga brevicollis, we investigate the genesis and expansion of homeobox, Sox, T- box, and Fox transcription factor genes. Comparative analyses reveal that novel transcription factor domains ( such as Paired, POU, and T- box) arose very early in metazoan evolution, prior to the separation of extant metazoan phyla but after the divergence of choanoflagellate and metazoan lineages. Phylogenetic analyses indicate that transcription factor classes then gradually expanded at the base of Metazoa before the bilaterian radiation, with each class following a different evolutionary trajectory. Based on the limited number of transcription factors in the Amphimedon genome, we infer that the genome of the metazoan last common ancestor included fewer gene members in each class than are present in extant eumetazoans. Transcription factor orthologues present in sponge, cnidarian, and bilaterian genomes may represent part of the core metazoan regulatory network underlying the origin of animal development and multicellularity.

Frequency of word-use predicts rates of lexical evolution throughout Indo-European history

Greek speakers say "ovpa", Germans "schwanz'' and the French "queue'' to describe what English speakers call a 'tail', but all of these languages use a related form of 'two' to describe the number after one. Among more than 100 Indo-European languages and dialects, the words for some meanings (such as 'tail') evolve rapidly, being expressed across languages by dozens of unrelated words, while others evolve much more slowly-such as the number 'two', for which all Indo-European language speakers use the same related word-form(1). No general linguistic mechanism has been advanced to explain this striking variation in rates of lexical replacement among meanings. Here we use four large and divergent language corpora (English(2), Spanish(3), Russian(4) and Greek(5)) and a comparative database of 200 fundamental vocabulary meanings in 87 Indo-European languages(6) to show that the frequency with which these words are used in modern language predicts their rate of replacement over thousands of years of Indo-European language evolution. Across all 200 meanings, frequently used words evolve at slower rates and infrequently used words evolve more rapidly. This relationship holds separately and identically across parts of speech for each of the four language corpora, and accounts for approximately 50% of the variation in historical rates of lexical replacement. We propose that the frequency with which specific words are used in everyday language exerts a general and law-like influence on their rates of evolution. Our findings are consistent with social models of word change that emphasize the role of selection, and suggest that owing to the ways that humans use language, some words will evolve slowly and others rapidly across all languages.

Developmental trends in lexical diversity.

This article discusses issues in measuring lexical diversity, before outlining an approach based on mathematical modelling that produces a measure, D, designed to address these problems. The procedure for obtaining values for D directly from transcripts using software (vocd) is introduced, and then applied to thirty-two children from the Bristol Study of Language Development (Wells 1985) at ten different ages. A significant developmental trend is shown for D and an indication is given of the average scores and ranges to be expected between the ages of 18 and 42 months and at 5 years for these L1 English speakers. The meaning attributable to further ranges of values for D is illustrated by analysing the lexical diversity of academic writing, and its wider application is demonstrated with examples from specific language impairment, morphological development, and foreign/second language learning.