This resource paper introduces a dataset for multi-scale rating inference of film review scores based upon review summaries. The dataset and task are unique in pairing a text regression problem with ratings given on multiple scales, e.g. the A-F letter scale and the 4-point star scale. It retains entity identifiers such as film and reviewer names. The paper describes the construction of the dataset before exploring potential baseline architectures for the task, and evaluating their performance. Baselines based on classifier-per-scale, affine-per-scale, and ordinal regression models are presented and evaluated with the BERT-base backbone. Additional experiments are used to ground a discussion of the different architectures’ merits and drawbacks with regards to explainability and model interpretation.
Previous work concerning measurement of second language learners has tended to focus on the knowledge of small numbers of words, often geared towards measuring vocabulary size. This paper presents a “tall” dataset containing information about a few learners’ knowledge of many words, suitable for evaluating Vocabulary Inventory Prediction (VIP) techniques, including those based on Computerised Adaptive Testing (CAT). In comparison to previous comparable datasets, the learners are from varied backgrounds, so as to reduce the risk of overfitting when used for machine learning based VIP. The dataset contains both a self-rating test and a translation test, used to derive a measure of reliability for learner responses. The dataset creation process is documented, and the relationship between variables concerning the participants, such as their completion time, their language ability level, and the triangulated reliability of their self-assessment responses, are analysed. The word list is constructed by taking into account the extensive derivation morphology of Finnish, and infrequent words are included in order to account for explanatory variables beyond word frequency.
We present a COVID-19 news dashboard which visualizes sentiment in pandemic news coverage in different languages across Europe. The dashboard shows analyses for positive/neutral/negative sentiment and moral sentiment for news articles across countries and languages. First we extract news articles from news-crawl. Then we use a pre-trained multilingual BERT model for sentiment analysis of news article headlines and a dictionary and word vectors -based method for moral sentiment analysis of news articles. The resulting dashboard gives a unified overview of news events on COVID-19 news overall sentiment, and the region and language of publication from the period starting from the beginning of January 2020 to the end of January 2021.
The aim of vocabulary inventory prediction is to predict a learner’s whole vocabulary based on a limited sample of query words. This paper approaches the problem starting from the 2-parameter Item Response Theory (IRT) model, giving each word in the vocabulary a difficulty and discrimination parameter. The discrimination parameter is evaluated on the sub-problem of question item selection, familiar from the fields of Computerised Adaptive Testing (CAT) and active learning. Next, the effect of the discrimination parameter on prediction performance is examined, both in a binary classification setting, and in an information retrieval setting. Performance is compared with baselines based on word frequency. A number of different generalisation scenarios are examined, including generalising word difficulty and discrimination using word embeddings with a predictor network and testing on out-of-dataset data.
This paper describes the automatic construction of FinnMWE: a lexicon of Finnish Multi-Word Expressions (MWEs). In focus here are syntactic frames: verbal constructions with arguments in a particular morphological form. The verbal frames are automatically extracted from FinnWordNet and English Wiktionary. The resulting lexicon interoperates with dependency tree searching software so that instances can be quickly found within dependency treebanks. The extraction and enrichment process is explained in detail. The resulting resource is evaluated in terms of its coverage of different types of MWEs. It is also compared with and evaluated against Finnish PropBank.