… and retrieve audio segments or textual descriptions that … audio metadata, and retrieve the
corresponding audio [1, 2]. An alternative approach is content-based retrieval, where an audio …

… embeddings. Specifically, given a collection of N audio samples with corresponding textual
… }, we aim to learn embedding functions, ψa and ψt, that project each audio sample ai and text …

… In this work we propose a method for allowing the textual generalization of cross-modal …
wider range of applications, such as cross-modal retrieval or zero-shot learning. Future work will …

… audio retrieval is introduced into as Subtask 6B, which aims to inspire further research into
audio retrieval with unconstrained textual … The final audio embedding is calculated by averag…

… Our main innovation is two-fold: i) we use logits as basic audio embeddings [3], … audio
recordings. We propose to combine the basic audio logit embeddings with the textual embeddings …

… [7] proposed a tag-based audio retrieval system using traditional machine learning … align
audio and textual features to a joint embedding space. Although these tag-based audio retrieval …

… valid texts that match a specific audio clip. Therefore, we would expect the same audio to
be retrieved for any of these queries and a retrieval … the value-projected frame embedding: …

… in the joint space by embedding visual and textual features into a … audio in learning the
embedding improves the result slightly. However, as the retrieval performance of individual audio …

… audio features by a fusion strategy for e cient retrieval. We also present a modi ed pairwise
loss to better learn the joint embedding… joint embedding between visual input and textual input…

… embeddings of keywords for which no audio samples are available but only their textual …
To obtain phone embeddings from audio we have implemented an audio encoder similarly …