Learning Methods and Algorithms for Semantic Text Classification across Multiple Domains

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.

Context-dependent interpretation of the conjunction 'und' in different age-groups

In contrast to formal semantics, the conjunction and is nonsymmetrical in pragmatics. The events in Marc went to bed and fell asleep seem to have occurred chronologically although no explicit time reference is given. As the temporal interpretation appears to be weaker in Mia ate chocolate and drank milk, it seems that the kind and nature of events presented in a context influences the interpretation of the conjunction. This work focuses on contextual influences on the interpretation of the German conjunction und (‘and’). A variety of theoretic approaches are concerned with whether and contributes to the establishment of discourse coherence via pragmatic processes or whether the conjunction has complex semantic meaning. These approaches are discussed with respect to how they explain the temporal and additive interpretation of the conjunction and the role of context in the interpre-tation process. It turned out that most theoretic approaches do not consider the importance of different kinds of context in the interpretation process.rnIn experimental pragmatics there are currently only very few studies that investigate the inter-pretation of the conjunction. As there are no studies that investigate contextual influences on the interpretation of und systematically or investigate preschoolers interpretation of the con-junction, research questions such as How do (preschool) children interpret ‘und’? and Does the kind of events conjoined influence children’s and adults’ interpretation? are yet to be answered. Therefore, this dissertation systematically investigates how different types of context influence children’s interpretation of und. Three auditory comprehension studies were conducted in German. Of special interest was whether and how the order of events introduced in a context contributes to the temporal read-ing of the conjunction und. Results indicate that the interpretation of und is – at least in Ger-man – context-dependent: The conjunction is interpreted temporally more often when events that typical occur in a certain order are connected (typical contexts) compared to events with-out typical event order (neutral contexts). This suggests that the type of events conjoined in-fluences the interpretation process. Moreover, older children and adults interpret the conjunc-tion temporally more often than the younger cohorts if the conjoined events typically occur in a certain order. In neutral contexts, additive interpretations increase with age. 5-year-olds reject reversed order statements more often in typical contexts compared to neutral contexts. However, they have more difficulties with reversed order statements in typical contexts where they perform at chance level. This suggests that not only the type of event but also other age-dependent factors such as knowledge about scripts influence children’s performance. The type of event conjoined influences children’s and adults’ interpretation of the conjunction. There-fore, the influence of different event types and script knowledge on the interpretation process does not only have to be considered in future experimental studies on language acquisition and pragmatics but also in experimental pragmatics in general. In linguistic theories, context has to be given a central role and a commonly agreed definition of context that considers the consequences arising from different event types has to be agreed upon.

Analysis of the relationships between Spartina maritima and nutrient enrichment in North Adriatic salt marshes

Salt marshes are coastal ecosystem in the upper intertidal zone between internal water and sea and are widely spread throughout Italy, from Friuli Venezia Giulia, in the North, to Sicily, in the South. These delicate environments are threatened by eutrophication, habitat conversion (for land reclaiming or agriculture) and climate change impacts such as sea level rise. The objectives of my thesis were to: 1) analyse the distribution and biomass of the perennial native cordgrass Spartina maritima (one of the most relevant foundation species in the low intertidal saltmarsh vegetation in the study region) at 7 sites along the Northern Adriatic coast and relate it to critical environmental parameters and 2) to carry out a nutrient manipulation experiment to detect nutrient enrichment effects on S. maritima biomass and vegetation characteristics. The survey showed significant differences among sites in biological response variables - i.e., live belowground, live aboveground biomass, above:belowground (R:S) biomass ratio, % cover, average height and stem density – which were mainly related to differences in nitrate, nitrite and phosphate contents in surface water. Preliminary results from the experiment (which is still ongoing) showed so far no significant effects of nutrient enrichment on live aboveground and belowground biomass, R:S ratio, leaf %Carbon, average height, stem density and random shoot height; however, a significantly higher (P=0.018) increase in leaf %Nitrogen content in treated plots indicated that nutrient uptake had occurred.

Semantic memory involvement in the default mode network: a functional neuroimaging study using independent component analysis

The Default Mode Network (DMN) is a higher order functional neural network that displays activation during passive rest and deactivation during many types of cognitive tasks. Accordingly, the DMN is viewed to represent the neural correlate of internally-generated self-referential cognition. This hypothesis implies that the DMN requires the involvement of cognitive processes, like declarative memory. The present study thus examines the spatial and functional convergence of the DMN and the semantic memory system. Using an active block-design functional Magnetic Resonance Imaging (fMRI) paradigm and Independent Component Analysis (ICA), we trace the DMN and fMRI signal changes evoked by semantic, phonological and perceptual decision tasks upon visually-presented words. Our findings show less deactivation during semantic compared to the two non-semantic tasks for the entire DMN unit and within left-hemispheric DMN regions, i.e., the dorsal medial prefrontal cortex, the anterior cingulate cortex, the retrosplenial cortex, the angular gyrus, the middle temporal gyrus and the anterior temporal region, as well as the right cerebellum. These results demonstrate that well-known semantic regions are spatially and functionally involved in the DMN. The present study further supports the hypothesis of the DMN as an internal mentation system that involves declarative memory functions.

Eliminating exposure to aqueous solvents is necessary for the early detection and ultrastructural elemental analysis of sites of calcium and phosphorus enrichment in mineralizing UMR106-01 osteoblastic cultures

The mechanism underlying the mineralization of bone is well studied and yet it remains controversial. Inherent difficulties of imaging mineralized tissues and the aqueous solubility of calcium and phosphate, the 2 ions which combine to form bone mineral crystals, limit current analyses of labile diffusible, amorphous, and crystalline intermediates by electron microscopy. To improve the retention of calcium and phosphorus, we developed a pseudo nonaqueous processing approach and used it to characterize biomineralization foci, extracellular sites of hydroxyapatite deposition in osteoblastic cell cultures. Since mineralization of UMR106-01 osteoblasts is temporally synchronized and begins 78 h after plating, we used these cultures to evaluate the effectiveness of our method when applied to cells just prior to the formation of the first mineral crystals. Our approach combines for the first time 3 well-established methods with a fourth one, i.e. dry ultrathin sectioning. Dry ultrathin sectioning with an oscillating diamond knife was used to produce electron spectroscopic images of mineralized biomineralization foci which were high-pressure frozen and freeze substituted. For comparison, cultures were also treated with conventional processing and wet sectioning. The results show that only the use of pseudo nonaqueous processing was able to detect extracellular sites of early calcium and phosphorus enrichment at 76 h, several hours prior to detection of mineral crystals within biomineralization foci.

Semantic priming of familiar songs

We explored the functional organization of semantic memory for music by comparing priming across familiar songs both within modalities (Experiment 1, tune to tune; Experiment 3, category label to lyrics) and across modalities (Experiment 2, category label to tune; Experiment 4, tune to lyrics). Participants judged whether or not the target tune or lyrics were real (akin to lexical decision tasks). We found significant priming, analogous to linguistic associative-priming effects, in reaction times for related primes as compared to unrelated primes, but primarily for within-modality comparisons. Reaction times to tunes (e.g., "Silent Night") were faster following related tunes ("Deck the Hall") than following unrelated tunes ("God Bless America"). However, a category label (e.g., Christmas) did not prime tunes from within that category. Lyrics were primed by a related category label, but not by a related tune. These results support the conceptual organization of music in semantic memory, but with potentially weaker associations across modalities.

Semantic Categorization of Unfamiliar Tunes in Older and Younger Adults

The process of learning the categories of new tunes in older and younger adults was examined for this study. Tunes were presented either one or three times along with a category name to see if multiple repetitions aid in category memory. Additionally, toexamine if an association may help some listeners, especially older ones, to better remember category information, some tunes were presented with a short associative fact; this fact was either neutral or emotional. Participants were tested on song recognition,fact recognition, and category memory. For all tasks, there was a benefit of three presentations. There were no age differences in fact recognition. For both song recognition and categorization, the memory burden of a neutral association was lessened when the association was emotional.

Electrophysiological and behavioral correlates of stable automatic semantic retrieval in aging

Previous studies have shown both declining and stable semantic-memory abilities during healthy aging. There is consistent evidence that semantic processes involving controlled mechanisms weaken with age. In contrast, results of aging studies on automatic semantic retrieval are often inconsistent, probably due to methodological limitations and differences. The present study therefore examines age-related alterations in automatic semantic retrieval and memory structure with a novel combination of critical methodological factors, i.e., the selection of subjects, a well-designed paradigm, and electrophysiological methods that result in unambiguous signal markers. Healthy young and elderly participants performed lexical decisions on visually presented word/non-word pairs with a stimulus onset asynchrony (SOA) of 150 ms. Behavioral and electrophysiological data were measured, and the N400-LPC complex, an event-related potential component sensitive to lexical-semantic retrieval, was analyzed by power and topographic distribution of electrical brain activity. Both age groups exhibited semantic priming (SP) and concreteness effects in behavioral reaction time and the electrophysiological N400-LPC complex. Importantly, elderly subjects did not differ significantly from the young in their lexical decision and SP performances as well as in the N400-LPC SP effect. The only difference was an age-related delay measured in the N400-LPC microstate. This could be attributed to existing age effects in controlled functions, as further supported by the replicated age difference in word fluency. The present results add new behavioral and neurophysiological evidence to earlier findings, by showing that automatic semantic retrieval remains stable in global signal strength and topographic distribution during healthy aging.

Semantic network disconnection in formal thought disorder

Structural and functional findings in schizophrenic patients with formal thought disorder (FTD) show abnormalities within left-side semantic areas. The present study investigate the network function of the involved brain regions as a function of FTD severity.

Coping personality type and environmental enrichment affect aggression at weaning in pigs

This study investigated the effects of different environmental treatments and personality types on aggression at mixing of newly weaned domestic piglets. From birth to weaning, 16 litters were housed with their dams in either barren (B) or larger, substrate-enriched (E) environments. At 15 days old, piglets were classified as 'high' (HR) or low resistant' (LR) in a manual restraint test (backtest), which is thought to identify proactive (HR) and reactive (LR) stress coping strategies that may reflect different personality types. At 30 days old, 128 piglets were weaned, relocated and mixed into 32 pens comprising two HR and two LR unfamiliar pigs, balanced for sex and weaning weight. Eight B and eight E groups changed environmental condition whereas the others remained in the same type of environment. Number and duration of fights. fight outcomes and unilateral fighting were scored for 5 h post-mixing and skin lesions were counted before and 5 h, 1 day and 2 days after mixing. On the day following weaning, fighting and also exploratory and oral manipulative behaviours were measured for 6 h. Generalized Linear Mixed Model analyses suggested interactions between pre-weaning environment, post-weaning environment and personality type. Overall, pre-weaning E pigs had longer fights at weaning and mixing (P=0.01) and fought for longer on the next day (P=0.02) than pre-weaning B pigs, and inflicted more skin lesions (P=0.02). Post-weaning enrichment did not affect fighting at mixing but reduced the time spent fighting the next day (P=0.03). Personality had subtle and environment-dependent effects on fighting, and influenced the "structure" rather than the amount of aggressive behaviour. HR pigs, for instance, bullied (i.e. chased surrendering pigs) more often (P=0.009) and their fighting behaviour was less affected by their relative body weight than that of LR pigs. Post-weaning E pigs showed relatively higher levels of exploratory behaviour (P=0.02) and less oral manipulative behaviour (P=0.04) than post-weaning B pigs. In particular, switching from a good quality environment (E) to a worse quality one (B) at weaning decreased exploratory behaviour on the next day, especially for LR pigs, who also tended to fight with and orally manipulate their pen mates more in that condition, and seemed to be more affected by a deterioration of the environment. Overall, pre-weaning enrichment increased aggression after weaning whereas post-weaning enrichment reduced it, and personality type related to some aspects of fighting behaviour. (C) 2011 Elsevier B.V. All rights reserved.