<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:dc="http://purl.org/dc/elements/1.1/" version="2.0">
  <channel>
    <title>OAR@UM Community:</title>
    <link>https://www.um.edu.mt/library/oar/handle/123456789/5429</link>
    <description />
    <pubDate>Tue, 10 Nov 2026 07:56:18 GMT</pubDate>
    <dc:date>2026-11-10T07:56:18Z</dc:date>
    <item>
      <title>Can language models learn analogical reasoning? Investigating training objectives and comparisons to human performance</title>
      <link>https://www.um.edu.mt/library/oar/handle/123456789/150139</link>
      <description>Title: Can language models learn analogical reasoning? Investigating training objectives and comparisons to human performance
Authors: Petersen, Molly R.; van der Plas, Lonneke
Abstract: While analogies are a common way to evaluate&#xD;
word embeddings in NLP, it is also of interest&#xD;
to investigate whether or not analogical reasoning&#xD;
is a task in itself that can be learned. In&#xD;
this paper, we test several ways to learn basic&#xD;
analogical reasoning, specifically focusing on&#xD;
analogies that are more typical of what is used&#xD;
to evaluate analogical reasoning in humans than&#xD;
those in commonly used NLP benchmarks. Our&#xD;
experiments find that models are able to learn&#xD;
analogical reasoning, even with a small amount&#xD;
of data. We additionally compare our models&#xD;
to a dataset with a human baseline, and find&#xD;
that after training, models approach human performance.</description>
      <pubDate>Fri, 01 Dec 2023 00:00:00 GMT</pubDate>
      <guid isPermaLink="false">https://www.um.edu.mt/library/oar/handle/123456789/150139</guid>
      <dc:date>2023-12-01T00:00:00Z</dc:date>
    </item>
    <item>
      <title>Cross-lingual transfer learning with Persian</title>
      <link>https://www.um.edu.mt/library/oar/handle/123456789/150062</link>
      <description>Title: Cross-lingual transfer learning with Persian
Authors: Mollanorozy, Sepideh; Tanti, Marc; Nissim, Malvina
Abstract: The success of cross-lingual transfer learning&#xD;
for POS tagging has been shown to be strongly&#xD;
dependent, among other factors, on the (typological&#xD;
and/or genetic) similarity of the lowresource&#xD;
language used for testing and the language(&#xD;
s) used in pre-training or to fine-tune&#xD;
the model. We further unpack this finding in&#xD;
two directions by zooming in on a single language,&#xD;
namely Persian. First, still focusing on&#xD;
POS tagging we run an in-depth analysis of&#xD;
the behaviour of Persian with respect to closely&#xD;
related languages and languages that appear&#xD;
to benefit from cross-lingual transfer with Persian.&#xD;
To do so, we also use the World Atlas of&#xD;
Language Structures to determine which properties&#xD;
are shared between Persian and other languages&#xD;
included in the experiments. Based on&#xD;
our results, Persian seems to be a reasonable&#xD;
potential language for Kurmanji and Tagalog&#xD;
low-resource languages for other tasks as well.&#xD;
Second, we test whether previous findings also&#xD;
hold on a task other than POS tagging to pull&#xD;
apart the benefit of language similarity and the&#xD;
specific task for which such benefit has been&#xD;
shown to hold. We gather sentiment analysis&#xD;
datasets for 31 target languages and through&#xD;
a series of cross-lingual experiments analyse&#xD;
which languages most benefit from Persian as&#xD;
the source. The set of languages that benefit&#xD;
from Persian had very little overlap across the&#xD;
two tasks, suggesting a strong task-dependent&#xD;
component in the usefulness of language similarity&#xD;
in cross-lingual transfer.</description>
      <pubDate>Mon, 01 May 2023 00:00:00 GMT</pubDate>
      <guid isPermaLink="false">https://www.um.edu.mt/library/oar/handle/123456789/150062</guid>
      <dc:date>2023-05-01T00:00:00Z</dc:date>
    </item>
    <item>
      <title>The phonetics and phonology of nasal vowels in Polish and transferability into English</title>
      <link>https://www.um.edu.mt/library/oar/handle/123456789/148522</link>
      <description>Title: The phonetics and phonology of nasal vowels in Polish and transferability into English
Abstract: There appears to be a lack of consensus regarding the status of nasal vowels /ɔ̃ / and /ɛ̃ / in Polish &#xD;
as phonemes functioning as separate from /ɔ/ and /ɛ/. Some of the existing research treats them &#xD;
as complex segments made up of an oral vowel and a nasal stop and/or limits their occurrence in &#xD;
their nasal form to fricative preceding positions. This study investigates the status of these &#xD;
vowels by examining their realisations across different phonological contexts. It seeks in &#xD;
particular to assess whether nasalisation of vowels is affected by orthography and whether it is &#xD;
transferred into the L2 English of L1 Polish speakers. An auditory-acoustic analysis was carried &#xD;
out using speaking and reading task data which relied on semi-controlled and controlled stimuli. &#xD;
A detailed analysis of variants was carried out, with more such variants identified than &#xD;
previously described in the literature, although the phonological contexts in which they occurred &#xD;
was found to align with previous research on this topic. Orthography was in fact found to affect &#xD;
the pronunciation of these nasal vowels in Polish. Evidence of phonological transfer of &#xD;
nasalisation into English was also found although this study brought important insights into the &#xD;
topic, more research is needed.
Description: B.A. (Hons)(Melit.)</description>
      <pubDate>Thu, 01 Jan 2026 00:00:00 GMT</pubDate>
      <guid isPermaLink="false">https://www.um.edu.mt/library/oar/handle/123456789/148522</guid>
      <dc:date>2026-01-01T00:00:00Z</dc:date>
    </item>
    <item>
      <title>Maltese first language acquisition of morphosyntax : a focus on nominal inflection and agreement</title>
      <link>https://www.um.edu.mt/library/oar/handle/123456789/148051</link>
      <description>Title: Maltese first language acquisition of morphosyntax : a focus on nominal inflection and agreement
Abstract: This study traces the development of morphosyntax in typically developing, Maltese-dominant bilingual children aged 18–60 months, examining how mean length of utterance in morphemes (MLUm), parts-of-speech (POS) category distribution, and nominal morphosyntax evolve and interact over time. To this end, a corpus comprising over 30 hours of Maltese child language data was collected. Drawing on longitudinal naturalistic data complemented by targeted elicitation tasks, and supported by statistical modelling, the study provides a detailed account of language acquisition in a bilingual environment. The results show that MLUm increases steadily with age at a rate comparable to previous studies. A higher proportion of nouns in early productions gradually gives way to increased use of verbs and function words, with mixed-effects modelling confirming a significant shift over time and in relation to MLUm. Analyses of nominal inflection reveal that concatenative plural forms are acquired earlier and with greater accuracy than templatic forms. In nominal agreement, accuracy increases with age, although earlier stabilisation of gender relative to number is not statistically confirmed. Integrative analyses indicate that the stabilisation of nominal morphosyntax coincides with the transition to multiword syntax and the attenuation of this early noun dominance. Collectively, these findings depict the development of nominal systems as gradual and interdependent, contributing new empirical evidence to cross-linguistic theories of morphosyntactic learnability and bilingual acquisition.
Description: Ph.D.(Melit.)</description>
      <pubDate>Wed, 01 Jan 2025 00:00:00 GMT</pubDate>
      <guid isPermaLink="false">https://www.um.edu.mt/library/oar/handle/123456789/148051</guid>
      <dc:date>2025-01-01T00:00:00Z</dc:date>
    </item>
  </channel>
</rss>

