Integrated semantic processing of complex pictures and spoken sentences - Evidence from event-related potentials

It has been suggested that semantic information processing is modularized according to the input form (e.g., visual, verbal, non-verbal sound). A great deal of research has concentrated on detecting a separate verbal module. Also, it has traditionally been assumed in linguistics that the meaning of a single clause is computed before integration to a wider context. Recent research has called these views into question. The present study explored whether it is reasonable to assume separate verbal and nonverbal semantic systems in the light of the evidence from event-related potentials (ERPs). The study also provided information on whether the context influences processing of a single clause before the local meaning is computed. The focus was on an ERP called N400. Its amplitude is assumed to reflect the effort required to integrate an item to the preceding context. For instance, if a word is anomalous in its context, it will elicit a larger N400. N400 has been observed in experiments using both verbal and nonverbal stimuli. Contents of a single sentence were not hypothesized to influence the N400 amplitude. Only the combined contents of the sentence and the picture were hypothesized to influence the N400. The subjects (n = 17) viewed pictures on a computer screen while hearing sentences through headphones. Their task was to judge the congruency of the picture and the sentence. There were four conditions: 1) the picture and the sentence were congruent and sensible, 2) the sentence and the picture were congruent, but the sentence ended anomalously, 3) the picture and the sentence were incongruent but sensible, 4) the picture and the sentence were incongruent and anomalous. Stimuli from the four conditions were presented in a semi-randomized sequence. Their electroencephalography was simultaneously recorded. ERPs were computed for the four conditions. The amplitude of the N400 effect was largest in the incongruent sentence-picture -pairs. The anomalously ending sentences did not elicit a larger N400 than the sensible sentences. The results suggest that there is no separate verbal semantic system, and that the meaning of a single clause is not processed independent of the context.

Data mining for telecommunications network log analysis

Telecommunications network management is based on huge amounts of data that are continuously collected from elements and devices from all around the network. The data is monitored and analysed to provide information for decision making in all operation functions. Knowledge discovery and data mining methods can support fast-pace decision making in network operations. In this thesis, I analyse decision making on different levels of network operations. I identify the requirements decision-making sets for knowledge discovery and data mining tools and methods, and I study resources that are available to them. I then propose two methods for augmenting and applying frequent sets to support everyday decision making. The proposed methods are Comprehensive Log Compression for log data summarisation and Queryable Log Compression for semantic compression of log data. Finally I suggest a model for a continuous knowledge discovery process and outline how it can be implemented and integrated to the existing network operations infrastructure.

Aerosol Particle Formation: Meteorological and Synoptic Processes behind the Event

Aerosol particles in the atmosphere are known to significantly influence ecosystems, to change air quality and to exert negative health effects. Atmospheric aerosols influence climate through cooling of the atmosphere and the underlying surface by scattering of sunlight, through warming of the atmosphere by absorbing sun light and thermal radiation emitted by the Earth surface and through their acting as cloud condensation nuclei. Aerosols are emitted from both natural and anthropogenic sources. Depending on their size, they can be transported over significant distances, while undergoing considerable changes in their composition and physical properties. Their lifetime in the atmosphere varies from a few hours to a week. New particle formation is a result of gas-to-particle conversion. Once formed, atmospheric aerosol particles may grow due to condensation or coagulation, or be removed by deposition processes. In this thesis we describe analyses of air masses, meteorological parameters and synoptic situations to reveal conditions favourable for new particle formation in the atmosphere. We studied the concentration of ultrafine particles in different types of air masses, and the role of atmospheric fronts and cloudiness in the formation of atmospheric aerosol particles. The dominant role of Arctic and Polar air masses causing new particle formation was clearly observed at Hyytiälä, Southern Finland, during all seasons, as well as at other measurement stations in Scandinavia. In all seasons and on multi-year average, Arctic and North Atlantic areas were the sources of nucleation mode particles. In contrast, concentrations of accumulation mode particles and condensation sink values in Hyytiälä were highest in continental air masses, arriving at Hyytiälä from Eastern Europe and Central Russia. The most favourable situation for new particle formation during all seasons was cold air advection after cold-front passages. Such a period could last a few days until the next front reached Hyytiälä. The frequency of aerosol particle formation relates to the frequency of low-cloud-amount days in Hyytiälä. Cloudiness of less than 5 octas is one of the factors favouring new particle formation. Cloudiness above 4 octas appears to be an important factor that prevents particle growth, due to the decrease of solar radiation, which is one of the important meteorological parameters in atmospheric particle formation and growth. Keywords: Atmospheric aerosols, particle formation, air mass, atmospheric front, cloudiness

Towards Optimal Testing of Auditory Memory : Methodological development of recording of the mismatch negativity (MMN) of the auditory event-related potential (ERP)

The overlapping sound pressure waves that enter our brain via the ears and auditory nerves must be organized into a coherent percept. Modelling the regularities of the auditory environment and detecting unexpected changes in these regularities, even in the absence of attention, is a necessary prerequisite for orientating towards significant information as well as speech perception and communication, for instance. The processing of auditory information, in particular the detection of changes in the regularities of the auditory input, gives rise to neural activity in the brain that is seen as a mismatch negativity (MMN) response of the event-related potential (ERP) recorded by electroencephalography (EEG). --- As the recording of MMN requires neither a subject s behavioural response nor attention towards the sounds, it can be done even with subjects with problems in communicating or difficulties in performing a discrimination task, for example, from aphasic and comatose patients, newborns, and even fetuses. Thus with MMN one can follow the evolution of central auditory processing from the very early, often critical stages of development, and also in subjects who cannot be examined with the more traditional behavioural measures of auditory discrimination. Indeed, recent studies show that central auditory processing, as indicated by MMN, is affected in different clinical populations, such as schizophrenics, as well as during normal aging and abnormal childhood development. Moreover, the processing of auditory information can be selectively impaired for certain auditory attributes (e.g., sound duration, frequency) and can also depend on the context of the sound changes (e.g., speech or non-speech). Although its advantages over behavioral measures are undeniable, a major obstacle to the larger-scale routine use of the MMN method, especially in clinical settings, is the relatively long duration of its measurement. Typically, approximately 15 minutes of recording time is needed for measuring the MMN for a single auditory attribute. Recording a complete central auditory processing profile consisting of several auditory attributes would thus require from one hour to several hours. In this research, I have contributed to the development of new fast multi-attribute MMN recording paradigms in which several types and magnitudes of sound changes are presented in both speech and non-speech contexts in order to obtain a comprehensive profile of auditory sensory memory and discrimination accuracy in a short measurement time (altogether approximately 15 min for 5 auditory attributes). The speed of the paradigms makes them highly attractive for clinical research, their reliability brings fidelity to longitudinal studies, and the language context is especially suitable for studies on language impairments such as dyslexia and aphasia. In addition I have presented an even more ecological paradigm, and more importantly, an interesting result in view of the theory of MMN where the MMN responses are recorded entirely without a repetitive standard tone. All in all, these paradigms contribute to the development of the theory of auditory perception, and increase the feasibility of MMN recordings in both basic and clinical research. Moreover, they have already proven useful in studying for instance dyslexia, Asperger syndrome and schizophrenia.

Discriminative learning of Bayesian networks via factorized conditional log-likelihood

We propose an efficient and parameter-free scoring criterion, the factorized conditional log-likelihood (ˆfCLL), for learning Bayesian network classifiers. The proposed score is an approximation of the conditional log-likelihood criterion. The approximation is devised in order to guarantee decomposability over the network structure, as well as efficient estimation of the optimal parameters, achieving the same time and space complexity as the traditional log-likelihood scoring criterion. The resulting criterion has an information-theoretic interpretation based on interaction information, which exhibits its discriminative nature. To evaluate the performance of the proposed criterion, we present an empirical comparison with state-of-the-art classifiers. Results on a large suite of benchmark data sets from the UCI repository show that ˆfCLL-trained classifiers achieve at least as good accuracy as the best compared classifiers, using significantly less computational resources.