TY - JOUR AU - Montalvão, Jugurta PY - 2022/10/18 Y2 - 2024/03/29 TI - On the Information Content of Predictions in Word Analogy Tests JF - Journal of Communication and Information Systems JA - Journal of Communication and Information Systems VL - 37 IS - 1 SE - Regular Papers DO - 10.14209/jcis.2022.18 UR - https://jcis.sbrt.org.br/jcis/article/view/839 SP - 175-181 AB - <pre>An approach is proposed to quantify, in bits of information, the actual relevance of analogies in analogy tests. The main component of this approach is a <em>soft</em> accuracy estimator that also yields entropy estimates with compensated biases. Experimental results obtained with pre-trained GloVe 300-D vectors and two public analogy test sets show that proximity hints are much more relevant than analogies in analogy tests, from an information content perspective. Accordingly, a simple word embedding model is used to predict that analogies carry about two bits of information, which is experimentally corroborated.</pre> ER -