Support vector machines in spectral data classification

Työssä käydään läpi tukivektorikoneiden teoreettista pohjaa sekä tutkitaan eri parametrien vaikutusta spektridatan luokitteluun.

From the seeds of rediscovered texts come new fields of inquiry

Kirjallisuusarvostelu

Limits of Meritocracy. How Fields of Study and Gender Segregation Affect Social Stratification in Finland

This dissertation critically reviews the idea of meritocracy from both a theoretical and an empirical perspective. Based on a discussion of classical texts of social philosophy and sociology, it is argued that meritocracy as a concept for social stratification is best compatible with the sociological tradition of status attainment research: both frame social inequality in primarily individualistic terms, centring on the role of ascribed (e.g., gender, social background) and achieved (e.g., educational qualifications) characteristics for determining individuals’ socioeconomic rewards. This theoretical argument introduces the research problem at the core of this dissertation: to what extent can the individualistic conception of social stratification be maintained empirically? Fields of study and their interaction with educational attainment levels play a prominent role in the analysis of this question. Drawing on sociological versions of segmented labour market theory, it is assumed that fields of study may channel individuals into heterogeneous political-economic contexts on the labour market, which potentially modify the socioeconomic benefit individuals derive from their qualification levels. The focus on fields of study may also highlight economic differentials between men and women that derive from the persisting segregation of men’s and women’s occupational and educational specializations rather than direct gender discrimination on the labour market. The quantitative analyses in this dissertation consist of three research articles, which are based primarily on Finnish data, but occasionally extend the view to other European countries. The data sources include register-based macro- and microdata as well as survey data. Article I examines the extent and the patterns of gender segregation within the Finnish educational system between 1981 and 2005. The results show that differences between men’s and women’s field specializations have for the most part remained stable during this period, with particularly high levels of gender segregation observed at lower educational levels. The focus in Article II rests on the effects of gender-segregated fields of study on higher education graduates’ occupational status. It is shown that fields of study matter for accessing professional jobs and avoiding low-skilled positions in Finland: at the early career stage, particularly polytechnic graduates from female-dominated fields are less likely to work in professional positions. Finnish university graduates from male-dominated fields were more likely than their peers with different specializations to work as professionals, yet they also faced a greater risk of being sorted into lowskilled jobs if they failed to make use of this advantage. Article III proceeded to analyse the joint impact of educational qualification levels and fields of study on young adults’ median earnings in Finland between 1985 and 2005. The results show that qualification levels do not confer a consistent benefit in the process of earnings stratification. Advanced qualifications raise median earnings most clearly among individuals specializing in the same field of study. When comparing individuals with different field specializations, on the other hand, higher-level qualifications do not necessarily lead to higher median earnings. Overall, the findings of this dissertation reveal a heterogeneous effect of education for achieving social positions, which challenges individual-centred, meritocratic accounts of social stratification and underlines the problematic lack of structural and institutional dimensions in the dominant account of social status attainment.

Map of a part of eastern Asia to show the commercial position of the productive coal fields of Yesso, Japan

Kartta kuuluu A. E. Nordenskiöldin kokoelmaan

Sentence Compression with Supervised Machine Learning

The subject of the thesis is automatic sentence compression with machine learning, so that the compressed sentences remain both grammatical and retain their essential meaning. There are multiple possible uses for the compression of natural language sentences. In this thesis the focus is generation of television program subtitles, which often are compressed version of the original script of the program. The main part of the thesis consists of machine learning experiments for automatic sentence compression using different approaches to the problem. The machine learning methods used for this work are linear-chain conditional random fields and support vector machines. Also we take a look which automatic text analysis methods provide useful features for the task. The data used for machine learning is supplied by Lingsoft Inc. and consists of subtitles in both compressed an uncompressed form. The models are compared to a baseline system and comparisons are made both automatically and also using human evaluation, because of the potentially subjective nature of the output. The best result is achieved using a CRF - sequence classification using a rich feature set. All text analysis methods help classification and most useful method is morphological analysis. Tutkielman aihe on suomenkielisten lauseiden automaattinen tiivistäminen koneellisesti, niin että lyhennetyt lauseet säilyttävät olennaisen informaationsa ja pysyvät kieliopillisina. Luonnollisen kielen lauseiden tiivistämiselle on monta käyttötarkoitusta, mutta tässä tutkielmassa aihetta lähestytään television ohjelmien tekstittämisen kautta, johon käytännössä kuuluu alkuperäisen tekstin lyhentäminen televisioruudulle paremmin sopivaksi. Tutkielmassa kokeillaan erilaisia koneoppimismenetelmiä tekstin automaatiseen lyhentämiseen ja tarkastellaan miten hyvin erilaiset luonnollisen kielen analyysimenetelmät tuottavat informaatiota, joka auttaa näitä menetelmiä lyhentämään lauseita. Lisäksi tarkastellaan minkälainen lähestymistapa tuottaa parhaan lopputuloksen. Käytetyt koneoppimismenetelmät ovat tukivektorikone ja lineaarisen sekvenssin mallinen CRF. Koneoppimisen tukena käytetään tekstityksiä niiden eri käsittelyvaiheissa, jotka on saatu Lingsoft OY:ltä. Luotuja malleja vertaillaan Lopulta mallien lopputuloksia evaluoidaan automaattisesti ja koska teksti lopputuksena on jossain määrin subjektiivinen myös ihmisarviointiin perustuen. Vertailukohtana toimii kirjallisuudesta poimittu menetelmä. Tutkielman tuloksena paras lopputulos saadaan aikaan käyttäen CRF sekvenssi-luokittelijaa laajalla piirrejoukolla. Kaikki kokeillut teksin analyysimenetelmät auttavat luokittelussa, joista tärkeimmän panoksen antaa morfologinen analyysi.