Search

Tässä artikkelissa haastattelimme neljää kieltenopettajaa selvittääksemme heidän ensivaikutelmiaan puheen automaattiseen arviointiin kehitetystä työkalusta. Esittelimme opettajille suomen ja ruotsin oppijoille suunnattua pilottiversiota DigiTala-tutkimushankkeen kehittämästä Moodle-työkalusta, joka ensin nauhoittaa oppijan puheen, muuttaa sen tekstiksi, arvioi sen automaattisesti ja lopulta näyttää tulokset oppijalle. Pilottiversiossa oppijalle näytettävä automaattinen palaute on muotoiltu ihmisarvioijien käyttämien analyyttisten ja holististen arviointikriteerien pohjalta. Tutkimuksemme aineisto koostuu neljästä Zoomissa toteutetusta puolistrukturoidusta yksilöhaastattelusta, joiden aikana kielten opettajille esiteltiin työkalua. Haastattelut litteroitiin ja niille tehtiin sisällönanalyysi. Havaitsimme, että automaattista palautetta voidaan hyödyntää merkittävästi kielenoppimisessa ja -opettamisessa. Opettajat kokivat, että työkalun antama arviointipalaute oli ymmärrettävää ja käyttökelpoista, vaikka palaute ei pystykään kattamaan kaikkia puhetuotoksen ulottuvuuksia. Tutkimuksemme tuloksista on hyötyä erityisesti automaattisen arvioinnin sekä suullisen kielitaidon tutkimuksen ja sovellusten saralla.

Eduskunnan täysistunnoissa on vuosina 1907–2022 pidetty lähes miljoona puheenvuoroa, ja yhteensä puhujia on ollut noin 2 800. Uusi Parlamenttisampo-palvelu tarjoaa kaiken tämän aineiston yhtenäisenä datana, mikä mahdollistaa kansanedustajien sekä poliittisen kielen, kulttuurin ja verkostojen tutkimisen entistä helpommalla tavalla.

Non

The paper discusses a language operation that we call context restriction. This operation is closely associated with context restriction rules (Kosken- niemi, 1983; Kiraz, 2000), right-arrow rules or implication rules (Kosken- niemi et al., 1992; Voutilainen, 1997) and the restriction operator (Beesley and Karttunen, 2003). The operation has been used in finite-state phonology and morphology in certain limited ways. A more general setting involves re- stricting overlapping occurrences of a center language under context condi- tions. Recently, star-free regular languages (and all regular languages) have been shown to be closed under context restrictions with such “overlapping centers” (Yli-Jyra ̈, 2003), but the construction involved is overly complex and becomes impractical when the number of operands grows. In this paper, we improve this recent result by presenting a more practical construction. This construction is not only simpler but it also leads to a generalization where contexts and centers may appear as conditions at both sides of the implication arrow (⇒): licensing conditions on the right-hand side specify the restriction and triggering conditions on the left-hand side regulate activation of the restriction. One application of the generalization is to facilitate splitting certain context restriction rules in grammars into a conjunction of separate rules.

This paper presents an application of Choice Network Analysis (CNA; proposed by Stuart Campbell in 2000) to analyse Machine Translation (MT) post-editing (PE) data. The data were collected from 33 translation students in a PE experiment where the students post-edited a text machine translated from English to Finnish. The text was a combination of neural (NMT), rule-based (RBMT) and statistical machine translation (SMT) output. CNA has been presented as a method for collecting data on the mental process of translating (or post-editing), to be used either instead of or in addition to experimental process studies. The method consists of collecting translations of the same text from different translators and building a network of the different options they have used to translate a particular item. The basic assumption is that when an item is translated in the same or a similar way, it requires less cognitive effort than an item that produced several different translations.In this paper, we use CNA to analyse a subset of the PE data, items that are repeated in the post-edited text, to compare differences both between editors and between MT engines. We also discuss two different approaches to presenting the choice networks.

Intralingual subtitles or subtitles for deaf and hard of hearing people are targeted at heterogenous audiences with differing possibilities of reception. The article presents results of two reception studies in which viewers’ opinions of intralingual Finnish subtitles were examined. In the first study, two groups, one consisting of deaf persons and the other of Finnish language learners, watched a subtitled movie and discussed the subtitles. The results were compared with the findings of a study carried out in 2014 in which the participants were late deafened and hard of hearing persons. The second study was a user experiment and tested subtitles produced with four different techniques: 1) speech recognition; 2) respeaking; 3) manual transcription of the speech; 4) traditional (the final subtitles of the programme). The data were tested and presented in a seminar for deafened and hard of hearing people, and the participants’ opinions were gathered with a questionnaire. Additionally, a group of Finnish learners watched the same data and discussed it. All the groups preferred verbatim subtitles, although for different reasons.