Semantic Query-by-example Speech Search Using Visual Grounding (2019-04-15T00:00:00.000000Z)

TL;DR

A segmental QbE approach where variable-duration speech segments (queries, search utterances) are mapped to fixed-dimensional embedding vectors and it is shown that aQbE system using an embedding function trained on visually grounded speech data outperforms a purely acoustic QBE system in terms of both exact and semantic retrieval performance.

Abstract

A number of recent studies have started to investigate how speech systems can be trained on untranscribed speech by leveraging accompanying images at training time. Examples of tasks include keyword prediction and within- and across-mode retrieval. Here we consider how such models can be used for query-by-example (QbE) search, the task of retrieving utterances relevant to a given spoken query. We are particularly interested in semantic QbE, where the task is not only to retrieve utterances containing exact instances of the query, but also utterances whose meaning is relevant to the query. We follow a segmental QbE approach where variable-duration speech segments (queries, search utterances) are mapped to fixed-dimensional embedding vectors. We show that a QbE system using an embedding function trained on visually grounded speech data outperforms a purely acoustic QbE system in terms of both exact and semantic retrieval performance.

Authors

Karen Livescu

10 papers

H. Kamper

4 papers

Aristotelis Anastassiou

1 papers

TL;DR

Abstract

Authors

References36 items

Learning Word Embeddings: Unsupervised Methods for Fixed-size Representations of Variable-length Speech Segments

Phonetic-and-Semantic Embedding of Spoken words with Applications in Spoken Content Retrieval

Visually grounded cross-lingual keyword spotting in speech

Unsupervised Cross-Modal Alignment of Speech and Text Embedding Spaces

Segmental Audio Word2Vec: Representing Utterances as Sequences of Vectors with Applications in Spoken Term Detection

Jointly Discovering Visual Objects and Spoken Words from Raw Sensory Input

Low-Resource Speech-to-Text Translation

Speech2Vec: A Sequence-to-Sequence Framework for Learning Word Embeddings from Speech

Linguistic Unit Discovery from Multi-Modal Inputs in Unwritten Languages: Summary of the “Speaking Rosetta” JSALT 2017 Workshop

Semantic Speech Retrieval With a Visually Grounded Model of Untranscribed Speech

Query-by-Example Search with Discriminative Neural Acoustic Word Embeddings

Visually Grounded Learning of Keyword Prediction from Untranscribed Speech

Representations of language in a model of visually grounded speech signal

Learning Word-Like Units from Joint Audio-Visual Analysis

End-to-end ASR-free keyword search from speech

Discriminative acoustic word embeddings: Tecurrent neural network-based approaches

Jointly Learning to Locate and Classify Words Using Convolutional Networks

Unsupervised Learning of Audio Segment Representations using Sequence-to-sequence Recurrent Neural Networks

Deep multimodal semantic embeddings for speech and images

Deep convolutional acoustic word embeddings using word-pair side information

Spoken Content Retrieval—Beyond Cascading Speech Recognition with Text Retrieval

Query-by-example keyword spotting using long short-term memory networks

Segmental acoustic indexing for zero resource keyword search

Learning Words from Images and Speech

Phonetics embedding learning with side information

SimLex-999: Evaluating Semantic Models With (Genuine) Similarity Estimation

Fixed-dimensional acoustic embeddings of variable-length segments in low-resource settings

Query-by-example spoken term detection using phonetic posteriorgram templates

Unsupervised spoken keyword spotting via segmental DTW on Gaussian posteriorgrams

A Study on Similarity and Relatedness Using Distributional and WordNet-based Approaches

Unsupervised Learning of Spoken Language with Visual Context

Breaking the Unwritten Language Barrier: The BULB Project

Word embeddings for speech recognition

Indexing Raw Acoustic Features for Scalable Zero Resource Search

Word-level Acoustic Modeling with Convolutional Vector Regression

Unsupervised Pattern Discovery in Speech

Field of Study

Journal Information

Name

Page

Venue Information

Name

Type

URL

Alternate Names