Semantic spaces : measuring the distance between different subspaces

Semantic Space models, which provide a numerical representation of words’ meaning extracted from corpus of documents, have been formalized in terms of Hermitian operators over real valued Hilbert spaces by Bruza et al. [1]. The collapse of a word into a particular meaning has been investigated applying the notion of quantum collapse of superpositional states [2]. While the semantic association between words in a Semantic Space can be computed by means of the Minkowski distance [3] or the cosine of the angle between the vector representation of each pair of words, a new procedure is needed in order to establish relations between two or more Semantic Spaces. We address the question: how can the distance between different Semantic Spaces be computed? By representing each Semantic Space as a subspace of a more general Hilbert space, the relationship between Semantic Spaces can be computed by means of the subspace distance. Such distance needs to take into account the difference in the dimensions between subspaces. The availability of a distance for comparing different Semantic Subspaces would enable to achieve a deeper understanding about the geometry of Semantic Spaces which would possibly translate into better effectiveness in Information Retrieval tasks.

Toward an object-based semantic memory for long-term operation of mobile service robots

Throughout a lifetime of operation, a mobile service robot needs to acquire, store and update its knowledge of a working environment. This includes the ability to identify and track objects in different places, as well as using this information for interaction with humans. This paper introduces a long-term updating mechanism, inspired by the modal model of human memory, to enable a mobile robot to maintain its knowledge of a changing environment. The memory model is integrated with a hybrid map that represents the global topology and local geometry of the environment, as well as the respective 3D location of objects. We aim to enable the robot to use this knowledge to help humans by suggesting the most likely locations of specific objects in its map. An experiment using omni-directional vision demonstrates the ability to track the movements of several objects in a dynamic environment over an extended period of time.

Explanation of relevance judgement discrepancy with quantum interference

A key concept in many Information Retrieval (IR) tasks, e.g. document indexing, query language modelling, aspect and diversity retrieval, is the relevance measurement of topics, i.e. to what extent an information object (e.g. a document or a query) is about the topics. This paper investigates the interference of relevance measurement of a topic caused by another topic. For example, consider that two user groups are required to judge whether a topic q is relevant to a document d, and q is presented together with another topic (referred to as a companion topic). If different companion topics are used for different groups, interestingly different relevance probabilities of q given d can be reached. In this paper, we present empirical results showing that the relevance of a topic to a document is greatly affected by the companion topic’s relevance to the same document, and the extent of the impact differs with respect to different companion topics. We further analyse the phenomenon from classical and quantum-like interference perspectives, and connect the phenomenon to nonreality and contextuality in quantum mechanics. We demonstrate that quantum like model fits in the empirical data, could be potentially used for predicting the relevance when interference exists.

A case study of utilising concept knowledge in a topic specific document collection

The use of ‘topic’ concepts has shown improved search performance, given a query, by bringing together relevant documents which use different terms to describe a higher level concept. In this paper, we propose a method for discovering and utilizing concepts in indexing and search for a domain specific document collection being utilized in industry. This approach differs from others in that we only collect focused concepts to build the concept space and that instead of turning a user’s query into a concept based query, we experiment with different techniques of combining the original query with a concept query. We apply the proposed approach to a real-world document collection and the results show that in this scenario the use of concept knowledge at index and search can improve the relevancy of results.

Pattern-based topic modelling and its application for information filtering and information retrieval

This thesis targets on a challenging issue that is to enhance users' experience over massive and overloaded web information. The novel pattern-based topic model proposed in this thesis can generate high-quality multi-topic user interest models technically by incorporating statistical topic modelling and pattern mining. We have successfully applied the pattern-based topic model to both fields of information filtering and information retrieval. The success of the proposed model in finding the most relevant information to users mainly comes from its precisely semantic representations to represent documents and also accurate classification of the topics at both document level and collection level.

Towards robust automatic affective classification of images using facial expressions for practical applications

Affect is an important feature of multimedia content and conveys valuable information for multimedia indexing and retrieval. Most existing studies for affective content analysis are limited to low-level features or mid-level representations, and are generally criticized for their incapacity to address the gap between low-level features and high-level human affective perception. The facial expressions of subjects in images carry important semantic information that can substantially influence human affective perception, but have been seldom investigated for affective classification of facial images towards practical applications. This paper presents an automatic image emotion detector (IED) for affective classification of practical (or non-laboratory) data using facial expressions, where a lot of “real-world” challenges are present, including pose, illumination, and size variations etc. The proposed method is novel, with its framework designed specifically to overcome these challenges using multi-view versions of face and fiducial point detectors, and a combination of point-based texture and geometry. Performance comparisons of several key parameters of relevant algorithms are conducted to explore the optimum parameters for high accuracy and fast computation speed. A comprehensive set of experiments with existing and new datasets, shows that the method is effective despite pose variations, fast, and appropriate for large-scale data, and as accurate as the method with state-of-the-art performance on laboratory-based data. The proposed method was also applied to affective classification of images from the British Broadcast Corporation (BBC) in a task typical for a practical application providing some valuable insights.

Perfusion fMRI evidence for priming of shared feature-to-lexical connections during cumulative semantic interference in spoken word production

The speed at which target pictures are named increases monotonically as a function of prior retrieval of other exemplars of the same semantic category and is unaffected by the number of intervening items. This cumulative semantic interference effect is generally attributed to three mechanisms: shared feature activation, priming and lexical-level selection. However, at least two additional mechanisms have been proposed: (1) a 'booster' to amplify lexical-level activation and (2) retrieval-induced forgetting (RIF). In a perfusion functional Magnetic Resonance Imaging (fMRI) experiment, we tested hypotheses concerning the involvement of all five mechanisms. Our results demonstrate that the cumulative interference effect is associated with perfusion signal changes in the left perirhinal and middle temporal cortices that increase monotonically according to the ordinal position of exemplars being named. The left inferior frontal gyrus (LIFG) also showed significant perfusion signal changes across ordinal presentations; however, these responses did not conform to a monotonically increasing function. None of the cerebral regions linked with RIF in prior neuroimaging and modelling studies showed significant effects. This might be due to methodological differences between the RIF paradigm and continuous naming as the latter does not involve practicing particular information. We interpret the results as indicating priming of shared features and lexical-level selection mechanisms contribute to the cumulative interference effect, while adding noise to a booster mechanism could account for the pattern of responses observed in the LIFG.

Auditory context effects in picture naming investigated with event-related fMRI

Naming an object entails a number of processing stages, including retrieval of a target lexical concept and encoding of its phonological word form. We investigated these stages using the picture-word interference task in an fMRI experiment. Participants named target pictures in the presence of auditorily presented semantically related, phonologically related, or unrelated distractor words or in isolation. We observed BOLD signal changes in left-hemisphere regions associated with lexical-conceptual and phonological processing, including the midto-posterior lateral temporal cortex. However, these BOLD responses manifested as signal reductions for all distractor conditions relative to naming alone. Compared with unrelated words, phonologically related distractors showed further signal reductions, whereas only the pars orbitalis of the left inferior frontal cortex showed a selective reduction in response in the semantic condition. We interpret these findings as indicating that the word forms of lexical competitors are phonologically encoded and that competition during lexical selection is reduced by phonologically related distractors. Since the extended nature of auditory presentation requires a large portion of a word to be presented before its meaning is accessed, we attribute the BOLD signal reductions observed for semantically related and unrelated words to lateral inhibition mechanisms engaged after target name selection has occurred, as has been proposed in some production models.

Negative priming in naming of categorically related objects: An fMRI study

Ignoring an object slows subsequent naming responses to it, a phenomenon known as negative priming (NP). A central issue in NP research concerns the level of representation at which the effect occurs. As object naming is typically considered to involve access to abstract semantic representations, Tipper 1985 proposed that the NP effect occurred at this level of processing, and other researchers supported this proposal by demonstrating a similar result with categorically related objects (e.g., Allport et al., 1985; Murray, 1995), an effect referred to as semantic NP. However, objects within categories share more physical or structural features than objects from different categories. Consequently, the NP effect observed with categorically related objects might occur at a structural rather than semantic level of representation. We used event related fMRI interleaving overt object naming and image acquisition to demonstrate for the first time that the semantic NP effect activates the left posterior-mid fusiform and insular-opercular cortices. Moreover, both naming latencies and left posterior-mid fusiform cortex responses were influenced by the structural similarity of prime-probe object pairings in the categorically related condition, increasing with the number of shared features. None of the cerebral regions activated in a previous fMRI study of the identity NP effect (de Zubicaray et al., 2006) showed similar activation during semantic NP, including the left anterolateral temporal cortex, a region considered critical for semantic processing. The results suggest that the identity and semantic NP effects differ with respect to their neural mechanisms, and the label "semantic NP" might be a misnomer. We conclude that the effect is most likely the result of competition between structurally similar category exemplars that determines the efficiency of object name retrieval.

An fMRI investigation of semantic and phonological naming treatment in aphasia

Di�culty naming objects is one of the most common impairments in people with aphasia post-stroke, irrespective of aphasia classification (Goodglass & Wingfield, 1997). Thus, remediation of naming impairments is often a focus of treatment in the rehabilitation of language. Such treatments typically employ phonological or semantic approaches, or a combination of the two, in order to target the major cognitive components involved in word retrieval (Nickels,2002). Although individuals can show greater benefit from one approach over the other, the relationship between an individual’s locus of breakdown in word retrieval and response to a particular treatment approach remains unclear, and knowledge of the underlying neural mechanisms which may be responsible for successful treatment is scarce. The aim of this study was to examine brain activity associated with successful phonological and semantic based treatments for word retrieval using functional Magnetic Resonance Imaging fMRI).

Similarity-based birdcall retrieval from environmental audio

Automated digital recordings are useful for large-scale temporal and spatial environmental monitoring. An important research effort has been the automated classification of calling bird species. In this paper we examine a related task, retrieval of birdcalls from a database of audio recordings, similar to a user supplied query call. Such a retrieval task can sometimes be more useful than an automated classifier. We compare three approaches to similarity-based birdcall retrieval using spectral ridge features and two kinds of gradient features, structure tensor and the histogram of oriented gradients. The retrieval accuracy of our spectral ridge method is 94% compared to 82% for the structure tensor method and 90% for the histogram of gradients method. Additionally, this approach potentially offers a more compact representation and is more computationally efficient.