Dissertations / Theses on the topic 'Semantic classification of nouns'

PONTIFÍCIA UNIVERSIDADE CATÓLICA DO RIO DE JANEIRO
COORDENAÇÃO DE APERFEIÇOAMENTO DO PESSOAL DE ENSINO SUPERIOR
PROGRAMA DE SUPORTE À PÓS-GRADUAÇÃO DE INSTS. DE ENSINO
Este trabalho propõe uma classificação semântica dos compostos nominais técnicos em língua inglesa e a análise sintática e semântica das traduções para o português. Para tanto, faz-se um panorama da literatura sobre as relações semânticas dos compostos nominais em língua inglesa. A tipologia aqui proposta é, portanto, baseada em estudos clássicos sobre a semântica dos compostos nominais (Levi, 1978; Warren, 1978) e em pesquisas mais recentes — inseridas no escopo da Linguística Computacional e ou influenciadas pela Teoria do Léxico Gerativo, de Pustejovsky (1995) —, e adaptada para a natureza dos compostos nominais selecionados. A presente dissertação também analisa as traduções dos compostos nominais técnicos para o português, bem como a função das preposições nas estruturas com sintagmas preposicionados. O corpus foi retirado de um livro técnico da área de engenharia elétrica/eletrônica traduzido pela autora. Além da classificação semântica dos compostos nominais técnicos, propõe-se a criação de ontologias que contemplem os compostos com os mesmos núcleos ou modificadores, de modo a observar se núcleos ou modificadores iguais implicam a mesma categorização, e se é respeitada a relação de hiperonímia e hiponímia entre os compostos nominais inseridos na mesma ontologia.
The major purpose of this thesis is to suggest a semantic categorization of English technical noun compounds, as well as to analyze the semantics and syntax of the Portuguese renderings. First, the literature on semantic relations in English compound nouns is reviewed. The classification here suggested is therefore based on classic studies on the semantics of compound nouns (Levi, 1978; Warren, 1978) and on more recent research within the scope of Computational Linguistics, which are influenced by the Generative Lexicon Theory (Pustejovsky, 1995). The semantic categorization is also adapted to the data collected in this work. This thesis also analyzes the Portuguese translation of the English compound nouns, as well as the role of the prepositions in prepositional phrases. The data was taken from an electrical/electronics engineering book which was translated by the author. In addition to the semantic classification, the technical compound nouns are grouped together according to the head or modifiers of the structure, and assembled into ontologies. Compound nouns sharing a common head or modifier are grouped together, so as to investigate if there is a hypernym-hyponym relation among the compounds assembled in the same ontology.

The thesis constitutes a first description of the Joola language Kujireray. In addition to a grammatical sketch, it comprises an analysis of the noun classification system in Kujireray, including a detailed treatment of verbal nouns and their interaction with this system. The analysis takes place within a Cognitive Linguistics framework. The noun classification system is shown to be semantically motivated along such parameters as number and physical configuration. The semantic analysis is carried out at the level of the noun class paradigm, which approach is able to draw a more fine-grained picture of the structure/organization of the system. However, it is recognized that noun classification operates on three distinct but interdependent levels - the paradigm, the noun class prefix, and the agreement pattern - all of which contribute meaning. The analysis also encompasses a detailed treatment of verbal nouns, as they interact within the noun classification system. It is shown that the formation of verbal nouns in various noun class prefixes is semantically motivated just as in the nominal domain, and furthermore that analogies can be drawn between the semantic domains in the nominal domain and the verbal one. The analysis is situated within a Cognitive Linguistics framework, whereby notions of embodied experience, encyclopaedic knowledge and metaphorical thought are invoked to account for the semantic organization of noun classification system. It is shown that noun formation in Kujireray is constructional, with individual components possessing underspecified semantics which are elaborated in combination with each other. Furthermore, it is the property of underspecification which accounts for the parallels between the nominal and verbal domains.

Most of the work investigating the representation of concrete nouns in the brain has focused on the locations that code the information. We present a model to study the contributions of perceptual and semantic features to the neural code representing concepts over time and space. The model is evaluated using magnetoencephalography data from different paradigms and not only corroborates previous findings regarding a distributed code, but provides further details about how the encoding of different subcomponents varies in the space-time spectrum. The model also successfully generalizes to novel concepts that it has never seen during training, which argues for the combination of specific properties in forming the meaning of concrete nouns in the brain. The results across paradigms are in agreement when the main differences among the experiments (namely, the number of repetitions of the stimulus, the task the subjects performed, and the type of stimulus provided) were taken into consideration. More specifically, these results suggest that features specific to the physical properties of the stimuli, such as word length and right-diagonalness, are encoded in posterior regions of the brain in the first hundreds of milliseconds after stimulus onset. Then, properties inherent to the nouns, such as is it alive? and can you pick it up?, are represented in the signal starting at about 250 ms, focusing on more anterior parts of the cortex. The code for these different features was found to be distributed over time and space, and it was common for several regions to simultaneously code for a particular property. Moreover, most anterior regions were found to code for multiple features, and a complex temporal profile could be observed for the majority of properties. For example, some features inherent to the nouns were encoded earlier than others, and the extent of time in which these properties could be decoded varied greatly among them. These findings complement much of the work previously described in the literature, and offer new insights about the temporal aspects of the neural encoding of concrete nouns. This model provides a spatiotemporal signature of the representation of objects in the brain. Paired with data from carefully-designed paradigms, the model is an important tool with which to analyze the commonalities of the neural code across stimulus modalities and tasks performed by the subjects.

Not all combinations of nouns and adjectives are possible and some are clearly more fre- quent than other. With this in mind this study aims to construct semantic representations of the two types of parts-of-speech, based on how they occur with each other. By inves- tigating these ideas via automatic natural language processing paradigms the study aims to find evidence for a semantic mutuality between nouns and adjectives, this notion sug- gests that the semantics of a noun can be captured by its corresponding adjectives, and vice versa. Furthermore, a set of proposed categories of adjectives and nouns, based on the ideas of Gärdenfors (2014), is presented that hypothetically are to fall in line with the produced representations. Four evaluation methods were used to analyze the result rang- ing from subjective discussion of nearest neighbours in vector space to accuracy generated from manual annotation. The result provided some evidence for the hypothesis which suggests that further research is of value.

La información de clase semántica de los nombres es fundamental para una amplia variedad de tareas del procesamiento del lenguaje natural (PLN), como la traducción automática, la discriminación de referentes en tareas como la detección y el seguimiento de eventos, la búsqueda de respuestas, el reconocimiento y la clasificación de nombres de entidades, la construcción y ampliación automática de ontologías, la inferencia textual, etc. Una aproximación para resolver la construcción y el mantenimiento de los léxicos de gran cobertura que alimentan los sistemas de PNL, una tarea muy costosa y lenta, es la adquisición automática de información léxica, que consiste en la inducción de una clase semántica relacionada con una palabra en concreto a partir de datos de su distribución obtenidos de un corpus. Precisamente, por esta razón, se espera que la investigación actual sobre los métodos para la producción automática de léxicos de alta calidad, con gran cantidad de información y con anotación de clase como el trabajo que aquí presentamos, tenga un gran impacto en el rendimiento de la mayoría de las aplicaciones de PNL. En esta tesis, tratamos la adquisición automática de información léxica como un problema de clasificación. Con este propósito, adoptamos métodos de aprendizaje automático para generar un modelo que represente los datos de distribución vectorial que, basados en ejemplos conocidos, permitan hacer predicciones de otras palabras desconocidas. Las principales preguntas de investigación que planteamos en esta tesis son: (i) si los datos de corpus proporcionan suficiente información para construir representaciones de palabras de forma eficiente y que resulten en decisiones de clasificación precisas y sólidas, y (ii) si la adquisición automática puede gestionar, también, los nombres polisémicos. Para hacer frente a estos problemas, realizamos una serie de validaciones empíricas sobre nombres en inglés. Nuestros resultados confirman que la información obtenida a partir de la distribución de los datos de corpus es suficiente para adquirir automáticamente clases semánticas, como lo demuestra un valor-F global promedio de 0,80 aproximadamente utilizando varios modelos de recuento de contextos y en datos de corpus de distintos tamaños. No obstante, tanto el estado de la cuestión como los experimentos que realizamos destacaron una serie de retos para este tipo de modelos, que son reducir la escasez de datos del vector y dar cuenta de la polisemia nominal en las representaciones distribucionales de las palabras. En este contexto, los modelos de word embedding (WE) mantienen la “semántica” subyacente en las ocurrencias de un nombre en los datos de corpus asignándole un vector. Con esta elección, hemos sido capaces de superar el problema de la escasez de datos, como lo demuestra un valor-F general promedio de 0,91 para las clases semánticas de nombres de sentido único, a través de una combinación de la reducción de la dimensionalidad y de números reales. Además, las representaciones de WE obtuvieron un rendimiento superior en la gestión de las ocurrencias asimétricas de cada sentido de los nombres de tipo complejo polisémicos regulares en datos de corpus. Como resultado, hemos podido clasificar directamente esos nombres en su propia clase semántica con un valor-F global promedio de 0,85. La principal aportación de esta tesis consiste en una validación empírica de diferentes representaciones de distribución utilizadas para la clasificación semántica de nombres junto con una posterior expansión del trabajo anterior, lo que se traduce en recursos léxicos y conjuntos de datos innovadores que están disponibles de forma gratuita para su descarga y uso.
La información de clase semántica de los nombres es fundamental para una amplia variedad de tareas del procesamiento del lenguaje natural (PLN), como la traducción automática, la discriminación de referentes en tareas como la detección y el seguimiento de eventos, la búsqueda de respuestas, el reconocimiento y la clasificación de nombres de entidades, la construcción y ampliación automática de ontologías, la inferencia textual, etc. Una aproximación para resolver la construcción y el mantenimiento de los léxicos de gran cobertura que alimentan los sistemas de PNL, una tarea muy costosa y lenta, es la adquisición automática de información léxica, que consiste en la inducción de una clase semántica relacionada con una palabra en concreto a partir de datos de su distribución obtenidos de un corpus. Precisamente, por esta razón, se espera que la investigación actual sobre los métodos para la producción automática de léxicos de alta calidad, con gran cantidad de información y con anotación de clase como el trabajo que aquí presentamos, tenga un gran impacto en el rendimiento de la mayoría de las aplicaciones de PNL. En esta tesis, tratamos la adquisición automática de información léxica como un problema de clasificación. Con este propósito, adoptamos métodos de aprendizaje automático para generar un modelo que represente los datos de distribución vectorial que, basados en ejemplos conocidos, permitan hacer predicciones de otras palabras desconocidas. Las principales preguntas de investigación que planteamos en esta tesis son: (i) si los datos de corpus proporcionan suficiente información para construir representaciones de palabras de forma eficiente y que resulten en decisiones de clasificación precisas y sólidas, y (ii) si la adquisición automática puede gestionar, también, los nombres polisémicos. Para hacer frente a estos problemas, realizamos una serie de validaciones empíricas sobre nombres en inglés. Nuestros resultados confirman que la información obtenida a partir de la distribución de los datos de corpus es suficiente para adquirir automáticamente clases semánticas, como lo demuestra un valor-F global promedio de 0,80 aproximadamente utilizando varios modelos de recuento de contextos y en datos de corpus de distintos tamaños. No obstante, tanto el estado de la cuestión como los experimentos que realizamos destacaron una serie de retos para este tipo de modelos, que son reducir la escasez de datos del vector y dar cuenta de la polisemia nominal en las representaciones distribucionales de las palabras. En este contexto, los modelos de word embedding (WE) mantienen la “semántica” subyacente en las ocurrencias de un nombre en los datos de corpus asignándole un vector. Con esta elección, hemos sido capaces de superar el problema de la escasez de datos, como lo demuestra un valor-F general promedio de 0,91 para las clases semánticas de nombres de sentido único, a través de una combinación de la reducción de la dimensionalidad y de números reales. Además, las representaciones de WE obtuvieron un rendimiento superior en la gestión de las ocurrencias asimétricas de cada sentido de los nombres de tipo complejo polisémicos regulares en datos de corpus. Como resultado, hemos podido clasificar directamente esos nombres en su propia clase semántica con un valor-F global promedio de 0,85. La principal aportación de esta tesis consiste en una validación empírica de diferentes representaciones de distribución utilizadas para la clasificación semántica de nombres junto con una posterior expansión del trabajo anterior, lo que se traduce en recursos léxicos y conjuntos de datos innovadores que están disponibles de forma gratuita para su descarga y uso.
Lexical semantic class information for nouns is critical for a broad variety of Natural Language Processing (NLP) tasks including, but not limited to, machine translation, discrimination of referents in tasks such as event detection and tracking, question answering, named entity recognition and classification, automatic construction and extension of ontologies, textual inference, etc. One approach to solve the costly and time-consuming manual construction and maintenance of large-coverage lexica to feed NLP systems is the Automatic Acquisition of Lexical Information, which involves the induction of a semantic class related to a particular word from distributional data gathered within a corpus. This is precisely why current research on methods for the automatic production of high- quality information-rich class-annotated lexica, such as the work presented here, is expected to have a high impact on the performance of most NLP applications. In this thesis, we address the automatic acquisition of lexical information as a classification problem. For this reason, we adopt machine learning methods to generate a model representing vectorial distributional data which, grounded on known examples, allows for the predictions of other unknown words. The main research questions we investigate in this thesis are: (i) whether corpus data provides sufficient distributional information to build efficient word representations that result in accurate and robust classification decisions and (ii) whether automatic acquisition can handle also polysemous nouns. To tackle these problems, we conducted a number of empirical validations on English nouns. Our results confirmed that the distributional information obtained from corpus data is indeed sufficient to automatically acquire lexical semantic classes, demonstrated by an average overall F1-Score of almost 0.80 using diverse count-context models and on different sized corpus data. Nonetheless, both the State of the Art and the experiments we conducted highlighted a number of challenges of this type of model such as reducing vector sparsity and accounting for nominal polysemy in distributional word representations. In this context, Word Embeddings (WE) models maintain the “semantics” underlying the occurrences of a noun in corpus data by mapping it to a feature vector. With this choice, we were able to overcome the sparse data problem, demonstrated by an average overall F1-Score of 0.91 for single-sense lexical semantic noun classes, through a combination of reduced dimensionality and “real” numbers. In addition, the WE representations obtained a higher performance in handling the asymmetrical occurrences of each sense of regular polysemous complex-type nouns in corpus data. As a result, we were able to directly classify such nouns into their own lexical-semantic class with an average overall F1-Score of 0.85. The main contribution of this dissertation consists of an empirical validation of different distributional representations used for nominal lexical semantic classification along with a subsequent expansion of previous work, which results in novel lexical resources and data sets that have been made freely available for download and use.

This degree project is mainly discussing about a web service classification approach based on suffix tree algorithm. Nowadays, Web Services are made up of WSDL web Service, RESTful web Service and many traditional component Services on Internet. The cost of manual classification cannot satisfy the increasing web services, so this paper proposes an approach to automatically classify web service because of this approach only relies on the textual description of service. Though semantic similarity calculation, we achieve web service classification automatically. Experimental evaluation results show that this approach has an acceptable and stable efficiency on precision and recall.

Cancer researchers and oncologists benefit greatly from text mining major knowledge sources in biomedicine such as PubMed. Fundamentally, text mining depends on accurate text classification. In conventional natural language processing (NLP), this requires experts to annotate scientific text, which is costly and time consuming, resulting in small labelled datasets. This leads to extensive feature engineering and handcrafting in order to fully utilise small labelled datasets, which is again time consuming, and not portable between tasks and domains. In this work, we explore emerging neural network methods to reduce the burden of feature engineering while outperforming the accuracy of conventional pipeline NLP techniques. We focus specifically on the cancer domain in terms of applications, where we introduce two NLP classification tasks and datasets: the first task is that of semantic text classification according to the Hallmarks of Cancer (HoC), which enables text mining of scientific literature assisted by a taxonomy that explains the processes by which cancer starts and spreads in the body. The second task is that of the exposure routes of chemicals into the body that may lead to exposure to carcinogens. We present several novel contributions. We introduce two new semantic classification tasks (the hallmarks, and exposure routes) at both sentence and document levels along with accompanying datasets, and implement and investigate a conventional pipeline NLP classification approach for both tasks, performing both intrinsic and extrinsic evaluation. We propose a new approach to classification using multilevel embeddings and apply this approach to several tasks; we subsequently apply deep learning methods to the task of hallmark classification and evaluate its outcome. Utilising our text classification methods, we develop and two novel text mining tools targeting real-world cancer researchers. The first tool is a cancer hallmark text mining tool that identifies association between a search query and cancer hallmarks; the second tool is a new literature-based discovery (LBD) system designed for the cancer domain. We evaluate both tools with end users (cancer researchers) and find they demonstrate good accuracy and promising potential for cancer research.

The  majority  of  computer  vision  algorithms  only  use  RGB  images  to  make  inferencesabout  the  state  of  the  world.  With  the  increasing  availability  of  RGB-D  cameras  it  is  im-portant  to  examine  ways  to  effectively  fuse  this  extra  modality  for  increased  effective-ness.  This  paper  examines  how  depth  can  be  fused  into  CNNs  to  increase  accuracy  in  thetasks  of  classification  and  semantic  segmentation,  as  well  as  examining  how  this  depthshould  best  be  effectively  encoded  prior  to  inclusion  in  the  network.  Concatenating  depthas  a  fourth  image  channel  and  modifying  the  dimension  of  the  initial  layer  of  a  pretrainedCNN  is  initially  examined.  Creating  a  separate  duplicate  network  to  train  depth  on,  andfusing  both  networks  in  later  stages  is  shown  to  be  an  effective  technique  for  both  tasks.The  results  show  that  depth  concatenation  is  an  ineffective  strategy  as  it  clamps  the  ac-curacy  to  the  lower  accuracy  of  the  two  modalities,  whilst  late  fusion  can  improve  thetask  accuracy  beyond  that  of  just  the  RGB  trained  network  for  both  tasks.  It  is  also  foundthat  methods  such  as  HHA  encoding  which  revolve  around  calculating  geometric  prop-erties  of  the  depth,  such  as  surface  normals,  are  a  superior  encoding  method  than  sim-pler  colour  space  transformations  such  as  HSV.  This  only  holds  true  when  these  depthimages  are  normalised  over  the  maximum  depth  of  the  dataset  as  opposed  to  the  maxi-mum  depth  of  each  individual  image,  thus  retaining  geometric  consistency  between  im-ages.  The  reverse  holds  true  for  simpler  colour  space  transformations.
Majoriteten av algoritmerna för datorseende använder bara färginformation för att dra sultsatser om hur världen ser ut. Med ökande tillgänglighet av RGB-D-kameror är det viktigt att undersöka sätt att effektivt kombinera färg- med djupinformation. I denna uppsats undersöks hur djup kan kombineras med färg i CNN:er för att öka presentandan i både klassificering och semantisk segmentering, så väl som att undersöka hur djupet kodas mest effektivt före dess inkludering i nätverket. Att lägga till djupet som en fjärde färgkanal och modifiera en förtränad CNN utreds inledningsvis. Sedan studeras att istället skapa en separat kopia av nätverket för att träna djup och sedan kombinera utdata från båda nätverken. Resultatet visar att det är ineffektivt att lägga till djup som en fjärde färgkanal då nätverket begränsas av den sämsta informationen från djup och färg. Fusion från två separata nätverk med färg och djup ökar prestanda bortom det som färg och djup erbjuder separat. Resultatet visar också att metoder så som HHA-kodning, är överlägsna jämfört med enklare transformationer så som HSV. Värt att notera är att detta endast gäller då djupbilderna är normaliserade över alla bilders maxdjup och inte i varje enskild bilds för sig. Motsatsen är sann för enklare transformationer.

The correction, authentication, validation and identification of the original texts in Hafez’s poetry among 16 or so old versions of his Divan has been a challenge for scholars. The semantic analysis of poetry with modern Digital Humanities techniques is also challenging. Analyzing latent semantics is more challenging in poetry than in prose for evident reasons, such as conciseness, imaginary and metaphorical constructions. Hafez’s poetry is, on the one hand, cryptic and complex because of his era’s restricting social properties and censorship impediments, and on the other hand, sophisticated because of his encapsulation of high-calibre world-views, mystical and philosophical attributes, artistically knitted within majestic decorations. Our research is strongly influenced by and is a continuation of, Mahmoud Houman’s instrumental and essential chronological classification of ghazals by Hafez. Houman’s chronological classification method (Houman, 1938), which we have adopted here, provides guidance to choose the correct version of Hafez’s poem among multiple manuscripts. Houman’s semantic analysis of Hafez’s poetry is unique in that the central concept of his classification is based on intelligent scrutiny of meanings, careful observation the evolutionary psychology of Hafez through his remarkable body of work. Houman’s analysis has provided the annotated data for the classification algorithms we will develop to classify the poems. We pursue to understand Hafez through the Houman’s perspective. In addition, we asked a contemporary expert to annotate Hafez ghazals (Raad, 2019). The rationale behind our research is also to satisfy the need for more efficient means of scholarly research, and to bring literature and computer science together as much as possible. Our research will support semantic analysis, and help with the design and development of tools for poetry research. We have developed a digital corpus of Hafez’s ghazals and applied proper word forms and punctuation. We digitized and extended chronological criteria to guide the correction and validation of Hafez’s poetry. To our knowledge, no automatic chronological classification has been conducted for Hafez poetry. Other than the meticulous preparation of our bilingual Hafez corpus for computational use, the innovative aspect of our classification research is two-fold. The first objective of our work is to develop semantic features to better train automatic classifiers for annotated poems and to apply the classifiers to unannotated poems, which is to classify the rest of the poems by applying machine learning (ML) methodology. The second task is to extract semantic information and properties to help design a visualization scheme to assist with providing a link between the prediction’s rationale and Houman’s perception of Hafez’s chronological properties of Hafez’s poetry. We identified and used effective Natural Language Processing (NLP) techniques such as classification, word-embedding features, and visualization to facilitate and automate semantic analysis of Hafez’s poetry. We defined and applied rigorous and repeatable procedures that can potentially be applied to other kinds of poetry. We showed that the chronological segments identified automatically were coherent. We presented and compared two independent chronological labellings of Hafez’s ghazals in digital form, pro- duced their ontologies and explained the inter-annotator-agreement and distributional semantic properties using relevant NLP techniques to help guide future corrections, authentication, and interpretation of Hafez’s poetry. Chronological labelling of the whole corpus not only helps better understand Hafez’s poetry, but it is a rigorous guide to better recognition of the correct versions of Hafez’s poems among multiple manuscripts. Such a small volume of complex poetic text required careful selection when choosing and developing appropriate ML techniques for the task. Through many classification and clustering experiments, we have achieved state-of-the-art prediction of chronological poems, trained and evaluated against our hand-made Hafez corpus. Our selected classification algorithm was a Support Vector Machine (SVM), trained with Latent Dirichlet Allocation (LDA)-based similarity features. We used clustering to produce an alternative perspective to classification. For our visualization methodology, we used the LDA features but also passed the results to a Principal Component Analysis (PCA) module to reduce the number of dimensions to two, thereby enabling graphical presentations. We believe that applying this method to poetry classifications, and showing the topic relations between poems in the same classes, will help us better understand the interrelated topics within the poems. Many of our methods can potentially be used in similar cases in which the intention is to semantically classify poetry.

The Vector Space Model (VSM) of text representation suffers a number of limitations for text classification. Firstly, the VSM is based on the Bag-Of-Words (BOW) assumption where terms from the indexing vocabulary are treated independently of one another. However, the expressiveness of natural language means that lexically different terms often have related or even identical meanings. Thus, failure to take into account the semantic relatedness between terms means that document similarity is not properly captured in the VSM. To address this problem, semantic indexing approaches have been proposed for modelling the semantic relatedness between terms in document representations. Accordingly, in this thesis, we empirically review the impact of semantic indexing on text classification. This empirical review allows us to answer one important question: how beneficial is semantic indexing to text classification performance. We also carry out a detailed analysis of the semantic indexing process which allows us to identify reasons why semantic indexing may lead to poor text classification performance. Based on our findings, we propose a semantic indexing framework called Relevance Weighted Semantic Indexing (RWSI) that addresses the limitations identified in our analysis. RWSI uses relevance weights of terms to improve the semantic indexing of documents. A second problem with the VSM is the lack of supervision in the process of creating document representations. This arises from the fact that the VSM was originally designed for unsupervised document retrieval. An important feature of effective document representations is the ability to discriminate between relevant and non-relevant documents. For text classification, relevance information is explicitly available in the form of document class labels. Thus, more effective document vectors can be derived in a supervised manner by taking advantage of available class knowledge. Accordingly, we investigate approaches for utilising class knowledge for supervised indexing of documents. Firstly, we demonstrate how the RWSI framework can be utilised for assigning supervised weights to terms for supervised document indexing. Secondly, we present an approach called Supervised Sub-Spacing (S3) for supervised semantic indexing of documents. A further limitation of the standard VSM is that an indexing vocabulary that consists only of terms from the document collection is used for document representation. This is based on the assumption that terms alone are sufficient to model the meaning of text documents. However for certain classification tasks, terms are insufficient to adequately model the semantics needed for accurate document classification. A solution is to index documents using semantically rich concepts. Accordingly, we present an event extraction framework called Rule-Based Event Extractor (RUBEE) for identifying and utilising event information for concept-based indexing of incident reports. We also demonstrate how certain attributes of these events e.g. negation, can be taken into consideration to distinguish between documents that describe the occurrence of an event, and those that mention the non-occurrence of that event.

Semantic approaches, including semantic feature analysis (SFA), are commonly used to treat individuals with anomia (word-finding difficulties) due to nondegenerative chronic aphasia. Research has traditionally targeted nouns, with relatively few published studies targeting verbs in isolation or in comparison to nouns. Yet, verbs are essential for higher-level communications, and some evidence suggests that treating higher-level word types may have crossover benefits. Generalization to untrained words and discourse have also varied across studies.

Thus, the aim of this study was to determine if a modified SFA treatment could be effective for both nouns and verbs, to assess generalization, and to investigate potential crossover benefits. Results revealed that the treatment did improve spontaneous production of trained nouns and verbs as well as semantic retrieval of untrained words, with an unexpected result of untrained verbs achieving a higher level of spontaneous production than untrained nouns. Implications and avenues for future studies are also discussed.

The growth of multimedia content in recent years motivated the research on audio classification and content retrieval area. In this thesis, a general environmental audio classification and retrieval approach is proposed in which higher level semantic classes (outdoor, nature, meeting and violence) are obtained from lower level acoustic classes (emergency alarm, car horn, gun-shot, explosion, automobile, motorcycle, helicopter, wind, water, rain, applause, crowd and laughter). In order to classify an audio sample into acoustic classes, MPEG-7 audio features, Mel Frequency Cepstral Coefficients (MFCC) feature and Zero Crossing Rate (ZCR) feature are used with Hidden Markov Model (HMM) and Support Vector Machine (SVM) classifiers. Additionally, a new classification method is proposed using Genetic Algorithm (GA) for classification of semantic classes. Query by Example (QBE) and keyword-based query capabilities are implemented for content retrieval.

This study aims to investigate if there is a specific grammar for proper names, in particular in the field of morphology, or more precisely, in nominal derivation. We will concentrate on a class of derived nouns in a language with an open corpus, i.e. Italian (from now on It.), and in a language with a closed corpus, i.e. Ancient Greek (from now on AG). At stake here are the nouns formed from proper names with the suffixes -ismo/‑ista in Italian and respectively with ‑ismós/‑istḗs in AG. In both languages, these suffixes are highly productive (see below Section 3). Furthermore, they combine not only with proper names but also with common nouns, adjectives and other lexical categories.

The correct semantic interpretation of mathematical formulae in electronic mathematical documents is an important prerequisite for advanced tasks such as search, accessibility or computational processing. Especially in advanced maths, the meaning of characters and symbols is highly domain dependent, and only limited information can be gained from considering individual formulae and their structures. Although many approaches have been proposed for semantic interpretation of mathematical formulae, most of them rely on the limited semantics from maths representation languages whereas very few use maths context as a source of information. This thesis presents a novel approach for principal extraction of semantic information of mathematical formulae from their context in documents. We utilised different supervised machine learning (SML) techniques (i.e. Linear-Chain Conditional Random Fields (CRF), Maximum Entropy (MaxEnt) and Maximum Entropy Markov Models (MEMM) combined with Rprop- and Rprop+ optimisation algorithms) to detect definitions of simple and compound mathematical expressions, thereby deriving their meaning. The learning algorithms demand annotated corpus which its development considered as one of this thesis contributions. The corpus has been developed utilising a novel approach to extract desired maths expressions and sub-formulae and manually annotated by two independent annotators employing a standard measure for inter-annotation agreement. The thesis further developed a new approach to feature representation depending on the definitions' templates that extracted from maths documents to defeat the restraint of conventional window-based features. All contributions were evaluated by various techniques including employing the common metrics recall, precision, and harmonic F-measure.

This thesis presents an approach to automatically infer facts about cooking recipes.The analysis focuses on two aspects: recognizing words with special semantic meaningin a recipe such as ingredients and tools; and detecting temporal dependencies betweensteps in the cooking procedure. This is done by use of machine learning classicationand several dierent algorithms are benchmarked against each other.

The Semantic Web enables people and computers to interact and exchange information. Based on Semantic Web technologies, different machine learning applications have been designed. Particularly to emphasize is the possibility to create complex metadata descriptions for any problem domain, based on pre-defined ontologies. In this paper we evaluate the use of a semantic similarity measure based on pre-defined ontologies as an input for a classification analysis. A link prediction between actors of a social network is performed, which could serve as a recommendation system. We measure the prediction performance based on an ontology-based metadata modeling as well as a feature vector modeling. The findings demonstrate that the prediction accuracy based on ontology-based metadata is comparable to traditional approaches and shows that data mining using ontology-based metadata can be considered as a very promising approach.

The Semantic Web enables people and computers to interact and exchange information. Based on Semantic Web technologies, different machine learning applications have been designed. Particularly to emphasize is the possibility to create complex metadata descriptions for any problem domain, based on pre-defined ontologies. In this paper we evaluate the use of a semantic similarity measure based on pre-defined ontologies as an input for a classification analysis. A link prediction between actors of a social network is performed, which could serve as a recommendation system. We measure the prediction performance based on an ontology-based metadata modeling as well as a feature vector modeling. The findings demonstrate that the prediction accuracy based on ontology-based metadata is comparable to traditional approaches and shows that data mining using ontology-based metadata can be considered as a very promising approach.

Intelligence tools have been developed and applied widely in many different areas in engineering, business and management. Many commercialized tools for business intelligence are available in the market. However, no practically useful tools for technology intelligence are available at this time, and very little academic research in technology intelligence methods has been conducted to date. Patent databases are the most important data source for technology intelligence tools, but patents inherently contain unstructured data. Consequently, extracting text data from patent databases, converting that data to meaningful information and generating useful knowledge from this information become complex tasks. These tasks are currently being performed very ineffectively, inefficiently and unreliably by human experts. This deficiency is particularly vexing in product planning, where awareness of market needs and technological capabilities is critical for identifying opportunities for new products and services. Total nescience of the text of patents, as well as inadequate, unreliable and untimely knowledge derived from these patents, may consequently result in missed opportunities that could lead to severe competitive disadvantage and potentially catastrophic loss of revenue. The research performed in this dissertation tries to correct the abovementioned deficiency with an approach called patent mining. The research is conducted at Finex, an iron casting company that produces traditional kitchen skillets. To 'mine' pertinent patents, experts in new product development at Finex modeled one ontology for the required product features and another for the attributes of requisite metallurgical enabling technologies from which new product opportunities for skillets are identified by applying natural language processing, information retrieval, and machine learning (classification) to the text of patents in the USPTO database. Three main scenarios are examined in my research. Regular classification (RC) relies on keywords that are extracted directly from a group of USPTO patents. Ontological classification (OC) relies on keywords that result from an ontology developed by Finex experts, which is evaluated and improved by a panel of external experts. Ontological semantic classification (OSC) uses these ontological keywords and their synonyms, which are extracted from the WordNet database. For each scenario, I evaluate the performance of three classifiers: k-Nearest Neighbor (k-NN), random forest, and Support Vector Machine (SVM). My research shows that OSC is the best scenario and SVM is the best classifier for identifying product planning opportunities, because this combination yields the highest score in metrics that are generally used to measure classification performance in machine learning (e.g., ROC-AUC and F-score). My method also significantly outperforms current practice, because I demonstrate in an experiment that neither the experts at Finex nor the panel of external experts are able to search for and judge relevant patents with any degree of effectiveness, efficiency or reliability. This dissertation provides the rudiments of a theoretical foundation for patent mining, which has yielded a machine learning method that is deployed successfully in a new product planning setting (Finex). Further development of this method could make a significant contribution to management practice by identifying opportunities for new product development that have been missed by the approaches that have been deployed to date.

Le problème de représentation d’image est au cœur du domaine de vision. Le choix de représentation d’une image change en fonction de la tâche que nous voulons étudier. Un problème de recherche d’image dans des grandes bases de données exige une représentation globale compressée, alors qu’un problème de segmentation sémantique nécessite une carte de partitionnement de ses pixels. Les techniques d’apprentissage statistique sont l’outil principal pour la construction de ces représentations. Dans ce manuscrit, nous abordons l’apprentissage des représentations visuels dans trois problèmes différents : la recherche d’image, la correspondance sémantique et classification d’image. Premièrement, nous étudions la représentation vectorielle de Fisher et sa dépendance sur le modèle de mélange Gaussien employé. Nous introduisons l’utilisation de plusieurs modèles de mélange Gaussien pour différents types d’arrière-plans, e.g., différentes catégories de scènes, et analyser la performance de ces représentations pour objet classification et l’impact de la catégorie de scène en tant que variable latente. Notre seconde approche propose une extension de la représentation l’exemple SVM pipeline. Nous montrons d’abord que, en remplaçant la fonction de perte de la SVM par la perte carrée, on obtient des résultats similaires à une fraction de le coût de calcul. Nous appelons ce modèle la « square-loss exemplar machine », ou SLEM en anglais. Nous introduisons une variante de SLEM à noyaux qui bénéficie des même avantages computationnelles mais affiche des performances améliorées. Nous présentons des expériences qui établissent la performance et l’efficacité de nos méthodes en utilisant une grande variété de représentations de base et de jeux de données de recherche d’images. Enfin, nous proposons un réseau neuronal profond pour le problème de l’établissement sémantique correspondance. Nous utilisons des boîtes d’objets en tant qu’éléments de correspondance pour construire une architecture qui apprend simultanément l’apparence et la cohérence géométrique. Nous proposons de nouveaux scores géométriques de cohérence adaptés à l’architecture du réseau de neurones. Notre modèle est entrainé sur des paires d’images obtenues à partir des points-clés d’un jeu de données de référence et évaluées sur plusieurs ensembles de données, surpassant les architectures d’apprentissage en profondeur récentes et méthodes antérieures basées sur des caractéristiques artisanales. Nous terminons la thèse en soulignant nos contributions et en suggérant d’éventuelles directions de recherche futures
The problem of image representation is at the heart of computer vision. The choice of feature extracted of an image changes according to the task we want to study. Large image retrieval databases demand a compressed global vector representing each image, whereas a semantic segmentation problem requires a clustering map of its pixels. The techniques of machine learning are the main tool used for the construction of these representations. In this manuscript, we address the learning of visual features for three distinct problems: Image retrieval, semantic correspondence and image classification. First, we study the dependency of a Fisher vector representation on the Gaussian mixture model used as its codewords. We introduce the use of multiple Gaussian mixture models for different backgrounds, e.g. different scene categories, and analyze the performance of these representations for object classification and the impact of scene category as a latent variable. Our second approach proposes an extension to the exemplar SVM feature encoding pipeline. We first show that, by replacing the hinge loss by the square loss in the ESVM cost function, similar results in image retrieval can be obtained at a fraction of the computational cost. We call this model square-loss exemplar machine, or SLEM. Secondly, we introduce a kernelized SLEM variant which benefits from the same computational advantages but displays improved performance. We present experiments that establish the performance and efficiency of our methods using a large array of base feature representations and standard image retrieval datasets. Finally, we propose a deep neural network for the problem of establishing semantic correspondence. We employ object proposal boxes as elements for matching and construct an architecture that simultaneously learns the appearance representation and geometric consistency. We propose new geometrical consistency scores tailored to the neural network’s architecture. Our model is trained on image pairs obtained from keypoints of a benchmark dataset and evaluated on several standard datasets, outperforming both recent deep learning architectures and previous methods based on hand-crafted features. We conclude the thesis by highlighting our contributions and suggesting possible future research directions

Els algoritmes de clustering desenvolupats fins al moment s’han centrat en el processat de dades numèriques i categòriques, no considerant dades textuals. Per manegar adequadament aquestes dades, es necessari interpretar el seu significat a nivell semàntic. En aquest treball es presenta un nou mètode de clustering que es capaç d’interpretar, de forma integrada, dades numèriques, categòriques i textuals. Aquest últims es processaran mitjançant mesures de similitud semàntica basades en 1) la utilització del coneixement taxonòmic contingut en una o diferents ontologies i 2) l’estimació de la distribució de la informació dels termes a la Web. Els resultats mostren que una interpretació precisa de la informació textual a nivell semàntic millora els resultats del clustering i facilita la interpretació de les classificacions.
Clustering algorithms have focused on the management of numerical and categorical data. However, in the last years, textual information has grown in importance. Proper processing of this kind of information within data mining methods requires an interpretation of their meaning at a semantic level. In this work, a clustering method aimed to interpret, in an integrated manner, numerical, categorical and textual data is presented. Textual data will be interpreted by means of semantic similarity measures. These measures calculate the alikeness between words by exploiting one or several knowledge sources. In this work we also propose two new ways of compute semantic similarity based on 1) the exploitation of the taxonomical knowledge available on one or several ontologies and 2) the estimation of the information distribution of terms in the Web. Results show that a proper interpretation of textual data at a semantic level improves clustering results and eases the interpretability of the classifications

This dissertation proposes a new methodology for building and using structured knowledge models for automatic image annotation. Specifically, our first proposals deal with the automatic building of explicit and structured knowledge models, such as semantic hierarchies and multimedia ontologies, dedicated to image annotation. Thereby, we propose a new approach for building semantic hierarchies faithful to image semantics. Our approach is based on a new image-semantic similarity measure between concepts and on a set of rules that allow connecting the concepts with higher relatedness till the building of the final hierarchy. Afterwards, we propose to go further in the modeling of image semantics through the building of explicit knowledge models that incorporate richer semantic relationships between image concepts. Therefore, we propose a new approach for automatically building multimedia ontologies consisting of subsumption relationships between concepts, and also other semantic relationships such as contextual and spatial relations. Fuzzy description logics are used as a formalism to represent our ontology and to deal with the uncertainty and the imprecision of concept relationships. In order to assess the effectiveness of the built structured knowledge models, we propose subsequently to use them in a framework for image annotation. We propose therefore an approach, based on the structure of semantic hierarchies, to effectively perform hierarchical image classification. Furthermore, we propose a generic approach for image annotation combining machine learning techniques, such as hierarchical image classification, and fuzzy ontological-reasoning in order to achieve a semantically relevant image annotation. Empirical evaluations of our approaches have shown significant improvement in the image annotation accuracy.

This dissertation investigates and introduces novel algorithms, theories, and supporting frameworks to significantly improve the growing problem of Internet security. A distributed firewall and active response architecture is introduced that enables any device within a cyber environment to participate in the active discovery and response of cyber attacks. A theory of semantic association systems is developed for the general problem of knowledge discovery in data. The theory of semantic association systems forms the basis of a novel semantic path merger packet classification algorithm. The theoretical aspects of the semantic path merger packet classification algorithm are investigated, and the algorithm's hardware-based implementation is evaluated along with comparative analysis versus content addressable memory. Experimental results show that the hardware implementation of the semantic path merger algorithm significantly outperforms content addressable memory in terms of energy consumption and operational timing.

Information is nowadays a key resource: machine learning and data mining techniques have been developed to extract high-level information from great amounts of data. As most data comes in form of unstructured text in natural languages, research on text mining is currently very active and dealing with practical problems. Among these, text categorization deals with the automatic organization of large quantities of documents in priorly defined taxonomies of topic categories, possibly arranged in large hierarchies. In commonly proposed machine learning approaches, classifiers are automatically trained from pre-labeled documents: they can perform very accurate classification, but often require a consistent training set and notable computational effort. Methods for cross-domain text categorization have been proposed, allowing to leverage a set of labeled documents of one domain to classify those of another one. Most methods use advanced statistical techniques, usually involving tuning of parameters. A first contribution presented here is a method based on nearest centroid classification, where profiles of categories are generated from the known domain and then iteratively adapted to the unknown one. Despite being conceptually simple and having easily tuned parameters, this method achieves state-of-the-art accuracy in most benchmark datasets with fast running times. A second, deeper contribution involves the design of a domain-independent model to distinguish the degree and type of relatedness between arbitrary documents and topics, inferred from the different types of semantic relationships between respective representative words, identified by specific search algorithms. The application of this model is tested on both flat and hierarchical text categorization, where it potentially allows the efficient addition of new categories during classification. Results show that classification accuracy still requires improvements, but models generated from one domain are shown to be effectively able to be reused in a different one.

Le point de départ de cette thèse est la classe « P » (« verbes psychologiques ») du dictionnaire électronique Les Verbes français de J. Dubois et F. Dubois-Charlier, et porte plus précisément sur les verbes de sentiment. Les critères de cette classification n’étant pas toujours explicites, l’objectif est, dans un premier temps, d’en comprendre les paramètres syntactico sémantiques puis de l'affiner afin d'obtenir des classes plus petites, mais homogènes du point de vue du sentiment exprimé, et d’y intégrer les noms et / ou adjectifs morphologiquement apparentés aux verbes et partageant les mêmes propriétés syntactico-sémantique. Nos analyses, appuyées sur un corpus d’énoncés attestés, aboutissent à la caractérisation des prédicats de « nostalgie », de « regret »,de « dés)agrément » et d’« étonnement »
The starting point of this work is “P” class (“verbes psychologiques”) of Les Verbes françaisof J. Dubois and F. Dubois-Charlier's electronic dictionary. It deals, more precisely, with feeling verbs.Because of the classification's criterions which are not that explicit, our main target is to try tounderstand their syntactic and semantic parameters, then to refine it in order to make small classes,but sharing in common the feeling expressed. Nouns and / or adjectives, morphologically connectedto verbs, and sharing the same syntactic and semantic properties, will be integrated to them. Ouranalysis, based on a large attested Corpus, lead to predicates characterization of “nostalgia”, “regret”,“annoyance” and “astonishment”

Semantic role labeling is an important stage in systems for Natural Language Understanding. The basic problem is one of identifying who did what to whom for each predicate in a sentence. Thus labeling is a two-step process: identify constituent phrases that are arguments to a predicate, then label those arguments with appropriate thematic roles. Existing systems for semantic role labeling use machine learning methods to assign roles one-at-a-time to candidate arguments. There are several drawbacks to this general approach. First, more than one candidate can be assigned the same role, which is undesirable. Second, the search for each candidate argument is exponential with respect to the number of words in the sentence. Third, single-role assignment cannot take advantage of dependencies known to exist between semantic roles of predicate arguments, such as their relative juxtaposition. And fourth, execution times for existing algorithm are excessive, making them unsuitable for real-time use. This thesis seeks to obviate these problems by approaching semantic role labeling as a multi-argument classification process. It observes that the only valid arguments to a predicate are unembedded constituent phrases that do not overlap that predicate. Given that semantic role labeling occurs after parsing, this thesis proposes an algorithm that systematically traverses the parse tree when looking for arguments, thereby eliminating the vast majority of impossible candidates. Moreover, instead of assigning semantic roles one at a time, an algorithm is proposed to assign all labels simultaneously; leveraging dependencies between roles and eliminating the problem of duplicate assignment. Experimental results are provided as evidence to show that a combination of the proposed argument identification and multi-argument classification algorithms outperforms all existing systems that use the same syntactic information.

Thesis (M.S.) -- Central Connecticut State University, 2008.
Thesis advisor: Roger Bilisoly. "... in partial fulfillment of the requirements for the degree of Master of Science in Data Mining." Includes bibliographical references (leaves 122-127). Also available via the World Wide Web.

As plataformas de Mídias Sociais se tornaram um meio essencial para a disponibilização de informações. Dentre elas, o Twitter tem se destacado, devido ao grande volume de mensagens que são compartilhadas todos os dias, principalmente mencionando eventos ao redor do mundo. Tais mensagens são uma importante fonte de informação e podem ser utilizadas em diversas aplicações. Contudo, a classificação de texto em tweets é uma tarefa não trivial. Além disso, não há um consenso quanto à quais tarefas devem ser executadas para Identificação e Classificação de Eventos em tweets, uma vez que as abordagens existentes trabalham com tipos específicos de eventos e determinadas suposições, que dificultam a reprodução e a comparação dessas abordagens em eventos de natureza distinta. Neste trabalho, nós elaboramos um framework para a classificação de eventos de natureza distinta. O framework possui os seguintes elementos chave: a) enriquecimento externo a partir da exploração de páginas web relacionadas, como uma forma de complementar a extração de features conceituais do conteúdo dos tweets; b) enriquecimento semântico utilizando recursos da Linked Open Data cloud para acrescentar features semânticas relacionadas; e c) técnica de poda para selecionar as features semânticas mais discriminativas Nós avaliamos o framework proposto através de um vasto conjunto de experimentos, que incluem: a) sete eventos alvos de natureza distinta; b) diferentes combinações das features conceituais propostas (i.e. entidades, vocabulário, e a combinação de ambos); c) estratégias distintas para a extração de features (i.e. a partir do conteúdo dos tweets e das páginas web); d) diferentes métodos para a seleção das features semânticas mais relevantes de acordo com o domínio (i.e. poda, seleção de features, e a combinação de ambos); e) dois algoritmos de classificação. Nós também comparamos o desempenho do framework em relação a outro método utilização para o enriquecimento contextual, o qual tem como base word embeddings. Os resultados mostraram as vantagens da utilização do framework proposto e que a nossa solução é factível e generalizável, dando suporte a classificação de diferentes tipos de eventos.
Social Media platforms have become key as a means of spreading information, opinions or awareness about real-world events. Twitter stands out due to the huge volume of messages about all sorts of topics posted every day. Such messages are an important source of useful information about events, presenting many useful applications (e.g. the detection of breaking news, real-time awareness, updates about events). However, text classification on Twitter is by no means a trivial task that can be handled by conventional Natural Language Processing techniques. In addition, there is no consensus about the definition of which kind of tasks are executed in the Event Identification and Classification in tweets, since existing approaches often focus on specific types of events, based on specific assumptions, which makes it difficult to reproduce and compare these approaches in events of distinct natures. In this work, we aim at building a unifying framework that is suitable for the classification of events of distinct natures. The framework has as key elements: a) external enrichment using related web pages for extending the conceptual features contained within the tweets; b) semantic enrichment using the Linked Open Data cloud to add related semantic features; and c) a pruning technique that selects the semantic features with discriminative potential We evaluated our proposed framework using a broad experimental setting, that includes: a) seven target events of different natures; b) different combinations of the conceptual features proposed (i.e. entities, vocabulary and their combination); c) distinct feature extraction strategies (i.e. from tweet text and web related documents); d) different methods for selecting the discriminative semantic features (i.e. pruning, feature selection, and their combination); and e) two classification algorithms. We also compared the proposed framework against another kind of contextual enrichment based on word embeddings. The results showed the advantages of using the proposed framework, and that our solution is a feasible and generalizable method to support the classification of distinct event types.

Le travail de cette thèse se situe dans le domaine de la sémantique nominale et porte sur un échantillon de noms dénotant les humains (NH) pendant les différentes phases de vie : bébé, enfant, adolescent, adulte, vieillard (N-[âge]). Elle poursuit un triple objectif. D'abord, il s'agit d'offrir une description linguistique fine d'un ensemble de N-[âge] dont le fonctionnement est resté méconnu jusqu'à présent. En démontrant leur caractère hybride - les N-[âge] dénotent à la fois des propriétés référentiellement essentielles et intrinsèquement transitoires - nous mettons en évidence des paramètres inédits pour la description des NH en général. Enfin, notre thèse explore la possibilité d'un élargissement notionnel du domaine aspectuel vers la sémantique nominale en interrogeant la notion de phase et en démontrant que l'ensemble des N-[âge] bénéficie d'une structure phasale.

Computer and Information Science
Ph.D.
Data is a critical component in a supervised machine learning system. Many successful applications of learning systems on various tasks are based on a large amount of labeled data. For example, deep convolutional neural networks have surpassed human performance on ImageNet classification, which consists of millions of labeled images. However, one challenge in conventional supervised learning systems is their generalization ability. Once a model is trained on a specific dataset, it can only perform the task on those \emph{seen} classes and cannot be used for novel \emph{unseen} classes. In order to make the model work on new classes, one has to collect and label new data and then re-train the model. However, collecting data and labeling them is labor-intensive and costly, in some cases, it is even impossible. Also, there is an enormous amount of different tasks in the real world. It is not applicable to create a dataset for each of them. These problems raise the need for Transfer Learning, which is aimed at using data from the \emph{source} domain to improve the performance of a model on the \emph{target} domain, and these two domains have different data or different tasks. One specific case of transfer learning is Zero-Shot Learning. It deals with the situation where \emph{source} domain and \emph{target} domain have the same data distribution but do not have the same set of classes. For example, a model is given animal images of `cat' and `dog' for training and will be tested on classifying 'tiger' and 'wolf' images, which it has never seen. Different from conventional supervised learning, Zero-Shot Learning does not require training data in the \emph{target} domain to perform classification. This property gives ZSL the potential to be broadly applied in various applications where a system is expected to tackle unexpected situations. In this dissertation, we develop algorithms that can help a model effectively transfer visual and semantic knowledge learned from \emph{source} task to \emph{target} task. More specifically, first we develop a model that learns a uniform visual representation of semantic attributes, which help alleviate the domain shift problem in Zero-Shot Learning. Second, we develop an ensemble network architecture with a progressive training scheme, which transfers \emph{source} domain knowledge to the \emph{target} domain in an end-to-end manner. Lastly, we move a step beyond ZSL and explore Label-less Classification, which transfers knowledge from pre-trained object detectors into scene classification tasks. Our label-less classification takes advantage of word embeddings trained from unorganized online text, thus eliminating the need for expert-defined semantic attributes for each class. Through comprehensive experiments, we show that the proposed methods can effectively transfer visual and semantic knowledge between tasks, and achieve state-of-the-art performances on standard datasets.
Temple University--Theses

North Ambrym, an Oceanic language spoken in Vanuatu, exhibits the two common Oceanic possessive construction types: direct and indirect. This thesis focuses on the indirect construction which occurs when the possessed noun refers to a semantically alienable item. In North Ambrym the indirect possessive construction is marked by one of a set of possessive classifiers. The theory within Oceanic linguistics is that the possessive classifiers do not classify a property of the possessed noun but the relation between possessor and possessed (Lichtenberk 1983b). Thus, it is the intentional use of the possessed by the possessor that is encoded by the possessive classifier, such that an ‘edible’ classifier will be used if the possessor intends to eat the possessed or the ‘drinkable’ classifier will be used if the possessed is intended to be drunk. This thesis challenges this theory and instead proposes that the classifiers act like possessed classifiers in North Ambrym and characterise a functional property of the possessed noun. Several experiments were conducted that induced different contextual uses of possessions, however this did not result in classifier change, which would be expected in the relational classifier theory. Each classifier has a large amount of seemingly semantically disparate members and they do not all share the semantic features of the central members, thus an analysis using the classical theory of classification is untenable. Instead the classifier categories are best analysed using prototype theory as certain semantic groups of possessions are considered to be more central members. This hypothesis is supported by further experimentation into classification which helps define the centrality of classifier category members. Finally an analysis using cognitive linguistic theory proposes that non-central members are linked to central members via semantic chains using notions of metaphor and metonymy. All languge data from this project has been deposited at the Endangered Language Archive (ELAR) at SOAS,University of London.

Understanding road surface conditions is an important component in active vehicle safety. Estimations can be achieved through image classification using increasingly popular convolutional neural networks (CNNs). In this paper, we explore the effects of multi-task learning by creating CNNs capable of simultaneously performing the two tasks road surface condition classification (RSCC) and road scene semantic segmentation (RSSS). A multi-task network, containing a shared feature extractor (VGG16, ResNet-18, ResNet-101) and two taskspecific network branches, is built and trained using the Road-Conditions and Cityscapes datasets. We reveal that utilizing task-dependent homoscedastic uncertainty in the learning process improvesmulti-task model performance on both tasks. When performing task adaptation, using a small set of additional data labeled with semantic information, we gain considerable RSCC improvements on complex models. Furthermore, we demonstrate increased model generalizability in multi-task models, with up to 12% higher F1-score compared to single-task models.

Gujjolaay Eegimaa (G.E.), an Atlantic language of the Niger-Congo phylum spoken in the Basse-Casamance area in Senegal, exhibits a system of nominal classification known as a "gender/ noun class system". In this type of nominal classification system which is prevalent in Niger-Congo languages, there is controversy as to whether the obligatory classification of all nouns into a finite number of classes has semantic motivations. In addition to the disputed issue of the semantic basis of the nominal classification, the formal criteria for assigning nouns into classes are also disputed in Joola languages and in G.E. In this PhD thesis, I propose an investigation of the formal and semantic properties of the nominal classification system of Gujjolaay Eegimaa (G.E). Based on cross-linguistic and language-specific research, I propose formal criteria whose application led to the discovery of fifteen noun classes in G.E. Here, I argue that the G.E. noun class system has semantic motivations. I show that some nouns in this language may be classified or categorized on the basis of shared properties as stipulated in the classical theory of categorization. However, most of the classification of the G.E. nouns is based on prototypicality and extension of such prototypes by family resemblance, chaining process, metaphor and metonymy, as argued in the prototype theory from cognitive semantics. The parameters of categorization that fruitfully account for the semantic basis of the G.E. nominal classification system are both universal and cultural-specific. Primary data constitutes the material used in this research and include lexical (including loanwords), textual as well as experimental data using picture stimuli. The collected data comprise different types of communicative events recorded in audio and video formats and also in written format through participant observation.

Die Hochschulbibliothek Zwickau nahm die Aktualität des Semantic Web zum Anlass, die Hauptakteure des Diskurses in Bibliotheken zu einem Vortrag einzuladen. Die Schwerpunkte des Vortrages lagen nicht nur in der Einführung des Semantic Web und der Open Linked Data, sondern dass Bibliotheken in Datenstrukturierung und Datenspeicherung wichtige Akteure sind oder werden können.

This thesis presents a robust face recognition method and a combination of methods for gender identification and age group classification for semantic annotation of videos. Local binary pattern histogram which has 256 bins and pixel intensity differences are used as extracted facial features for gender classification. DCT Mod2 features and edge detection results around facial landmarks are used as extracted facial features for age group classification. In gender classification module, a Random Trees classifier is trained with LBP features and an adaboost classifier is trained with pixel intensity differences. DCT Mod2 features are used for training of a Random Trees classifier and LBP features around facial landmark points are used for training another Random Trees classifier in age group classification module. DCT Mod2 features of the detected faces morped by two dimensional face morphing method based on Active Appearance Model and Barycentric Coordinates are used as the inputs of the nearest neighbor classifier with weights obtained from the trained Random Forest classifier in face identification module. Different feature extraction methods are tried and compared and the best achievements in the face recognition module to be used in the method chosen. We compared our classification results with some successful earlier works results in our experiments performed with same datasets and got satisfactory results.

This dissertation addresses the difficulties of semantic segmentation when dealing with an extensive collection of images and 3D point clouds. Due to the ubiquity of digital cameras that help capture the world around us, as well as the advanced scanning techniques that are able to record 3D replicas of real cities, the sheer amount of visual data available presents many opportunities for both academic research and industrial applications. But the mere quantity of data also poses a tremendous challenge. In particular, the problem of distilling useful information from such a large repository of visual data has attracted ongoing interests in the fields of computer vision and data mining. Structural Semantics are fundamental to understanding both natural and man-made objects. Buildings, for example, are like languages in that they are made up of repeated structures or patterns that can be captured in images. In order to find these recurring patterns in images, I present an unsupervised frequent visual pattern mining approach that goes beyond co-location to identify spatially coherent visual patterns, regardless of their shape, size, locations and orientation. First, my approach categorizes visual items from scale-invariant image primitives with similar appearance using a suite of polynomial-time algorithms that have been designed to identify consistent structural associations among visual items, representing frequent visual patterns. After detecting repetitive image patterns, I use unsupervised and automatic segmentation of the identified patterns to generate more semantically meaningful representations. The underlying assumption is that pixels capturing the same portion of image patterns are visually consistent, while pixels that come from different backdrops are usually inconsistent. I further extend this approach to perform automatic segmentation of foreground objects from an Internet photo collection of landmark locations. New scanning technologies have successfully advanced the digital acquisition of large-scale urban landscapes. In addressing semantic segmentation and reconstruction of this data using LiDAR point clouds and geo-registered images of large-scale residential areas, I develop a complete system that simultaneously uses classification and segmentation methods to first identify different object categories and then apply category-specific reconstruction techniques to create visually pleasing and complete scene models.

The central part of this thesis (chapter 3) consists of a classification of 9071 lexical items comprising the semantic field Good and Evil. This classified semantic field, with minor alterations, will form part of the Historical Thesaurus of English currently being compiled in the English Language Department of Glasgow University. Some significant features of the Good and Evil classification system, devised and explained in this thesis, have also been adopted by the Historical Thesaurus. Chapter 1 places the thesis in a wider academic context. It explains briefly the Historical Thesaurus project, and describes how the classification of Good and Evil contributes to this. It also relates the thesis to linguistics, semantics, and especially to semantic theory, lexicography, and semantic classification. Chapter 2 defines the semantic field Good and Evil and discusses how the field was assembled. It provides details of those areas which were either rejected or extended in order to form the semantic field. It then describes in some detail the classification system devised for Good and Evil. The structure of the classification is explained, the use of the parts of speech as a valuable classificatory device is justified, and the contribution of other classificatory work is acknowledged. The chapter also discusses some particular problems and features of the Old English corpus. It ends with lists of stylistic and other conventions. Chapter 3 contains the Good and Evil classification, and chapter 4 consists of detailed notes on the classification. These notes discuss points relating to dating. Old English material, classificatory devices, closely connected categories, and some problems of dictionary definitions, among other things. Chapter 5 conducts a number of studies based on historical and etymological information drawn from the classification. The relative numbers of accessions and losses in different centuries in the categories are presented and discussed. The range of sources of origin of a limited number of categories arc detailed. The patterns of change, and the extent and rate of influence of different languages in different centuries, are then commented on and compared. Chapter 6 selects one area of vocabulary from Good and Evil - animal names used as names for people - and subjects this area to a detailed examination. The variety of animal names, and the range of people to whom they arc applied, arc discussed, and various statistics and comparisons arc drawn up. Also considered is the time gap between the first literal use of an animal name and the first figurative or metaphorical application of the same term to a person. In the process some interesting and, on occasion, unproven points about animal metaphor are brought to light. The thesis ends with three appendices. The first contains extra Good and Evil material not in the main classification, the second details 19th century obsolescences, and the third gives a numerical distribution of items in each category by part of speech

Semantic Web technologies such as RDF, OWL, and SPARQL can be successfully used to bridge complementary musicological information. In this paper, we describe, compare, and evaluate the datasets and workflows used to create two such aggregator projects: In Collaboration with In Concert, and JazzCats, both of which bring together a cluster of smaller projects containing concert and performance metadata.
Semantische Web-Technologien wie RDF, OWL und SPARQL ermöglichen die Verknüpfung von komplementären musikwissenschaftlichen Daten. In diesem Artikel beschreiben, vergleichen und bewerten wir die Datensätze und Workflows, die zur Erstellung zweier solcher Aggregationsprojekte verwendet wurden: In Collaboration with In Concert und JazzCats, die jeweils Sammlungen kleinerer Projekte mit Konzert- und Performance-Metadaten zusammenführen.

This thesis proposes a fine-grained semantic classification of what is traditionally named agent nouns in French; it is thus a study of semantic derivation. Our classification was elaborated starting from 1573 agent nouns extracted from the dictionary Le Nouveau Petit Robert. The definitions of several agent nouns in Le Nouveau Petit Robert represent paraphrastic reformulations that are sufficiently close, which allows us to distribute these nouns into semantically more specific subclasses. We have identified 22 subclasses of agent nouns; these subclasses, along with the corresponding lexical units, were described by means of the formalism of lexical functions proposed by Meaning Text theory. We also elaborated definition templates for lexical units of each subclass as well as generalized government patterns (? subcategorization frames) for them. The interest of our work lies in the fact that the suggested classification allows for a more uniform and a more coherent global description of agent nouns.
Le présent travail porte sur une classification sémantique détaillée de ce qu’on appelle traditionnellement les noms d’agent [= S1] en français. Notre classification a été élaborée à partir de 1573 noms d’agent extraits du Nouveau Petit Robert. Les définitions de plusieurs noms d’agent dans le Nouveau Petit Robert représentent des reformulations paraphrastiques suffisamment proches, ce qui permet de repartir ces noms en sous-ensembles sémantiquement plus spécifiques. Nous avons identifié 22 sous-classes de noms d’agent ; ces classes et les lexies correspondantes ont été décrites au moyen du formalisme des fonctions lexicales proposé par la théorie linguistique Sens Texte. Nous avons également élaboré des définitions lexicographiques généralisées pour les lexies de chaque sous-classe, ainsi que leurs schémas de régime (? cadres de sous-catégorisations) généralisés. L’intérêt de notre travail réside dans le fait que la classification proposée mène vers une description globale plus uniforme et plus cohérente des noms d’agent.