@inproceedings{jha-mihata-2021-geodesic,
title = "On Geodesic Distances and Contextual Embedding Compression for Text Classification",
author = "Jha, Rishi and
Mihata, Kai",
editor = "Panchenko, Alexander and
Malliaros, Fragkiskos D. and
Logacheva, Varvara and
Jana, Abhik and
Ustalov, Dmitry and
Jansen, Peter",
booktitle = "Proceedings of the Fifteenth Workshop on Graph-Based Methods for Natural Language Processing (TextGraphs-15)",
month = jun,
year = "2021",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.textgraphs-1.15",
doi = "10.18653/v1/2021.textgraphs-1.15",
pages = "144--149",
abstract = "In some memory-constrained settings like IoT devices and over-the-network data pipelines, it can be advantageous to have smaller contextual embeddings. We investigate the efficacy of projecting contextual embedding data (BERT) onto a manifold, and using nonlinear dimensionality reduction techniques to compress these embeddings. In particular, we propose a novel post-processing approach, applying a combination of Isomap and PCA. We find that the geodesic distance estimations, estimates of the shortest path on a Riemannian manifold, from Isomap{'}s k-Nearest Neighbors graph bolstered the performance of the compressed embeddings to be comparable to the original BERT embeddings. On one dataset, we find that despite a 12-fold dimensionality reduction, the compressed embeddings performed within 0.1{\%} of the original BERT embeddings on a downstream classification task. In addition, we find that this approach works particularly well on tasks reliant on syntactic data, when compared with linear dimensionality reduction. These results show promise for a novel geometric approach to achieve lower dimensional text embeddings from existing transformers and pave the way for data-specific and application-specific embedding compressions.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="jha-mihata-2021-geodesic">
<titleInfo>
<title>On Geodesic Distances and Contextual Embedding Compression for Text Classification</title>
</titleInfo>
<name type="personal">
<namePart type="given">Rishi</namePart>
<namePart type="family">Jha</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kai</namePart>
<namePart type="family">Mihata</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2021-06</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Fifteenth Workshop on Graph-Based Methods for Natural Language Processing (TextGraphs-15)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Alexander</namePart>
<namePart type="family">Panchenko</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Fragkiskos</namePart>
<namePart type="given">D</namePart>
<namePart type="family">Malliaros</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Varvara</namePart>
<namePart type="family">Logacheva</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Abhik</namePart>
<namePart type="family">Jana</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dmitry</namePart>
<namePart type="family">Ustalov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Peter</namePart>
<namePart type="family">Jansen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Mexico City, Mexico</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>In some memory-constrained settings like IoT devices and over-the-network data pipelines, it can be advantageous to have smaller contextual embeddings. We investigate the efficacy of projecting contextual embedding data (BERT) onto a manifold, and using nonlinear dimensionality reduction techniques to compress these embeddings. In particular, we propose a novel post-processing approach, applying a combination of Isomap and PCA. We find that the geodesic distance estimations, estimates of the shortest path on a Riemannian manifold, from Isomap’s k-Nearest Neighbors graph bolstered the performance of the compressed embeddings to be comparable to the original BERT embeddings. On one dataset, we find that despite a 12-fold dimensionality reduction, the compressed embeddings performed within 0.1% of the original BERT embeddings on a downstream classification task. In addition, we find that this approach works particularly well on tasks reliant on syntactic data, when compared with linear dimensionality reduction. These results show promise for a novel geometric approach to achieve lower dimensional text embeddings from existing transformers and pave the way for data-specific and application-specific embedding compressions.</abstract>
<identifier type="citekey">jha-mihata-2021-geodesic</identifier>
<identifier type="doi">10.18653/v1/2021.textgraphs-1.15</identifier>
<location>
<url>https://aclanthology.org/2021.textgraphs-1.15</url>
</location>
<part>
<date>2021-06</date>
<extent unit="page">
<start>144</start>
<end>149</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T On Geodesic Distances and Contextual Embedding Compression for Text Classification
%A Jha, Rishi
%A Mihata, Kai
%Y Panchenko, Alexander
%Y Malliaros, Fragkiskos D.
%Y Logacheva, Varvara
%Y Jana, Abhik
%Y Ustalov, Dmitry
%Y Jansen, Peter
%S Proceedings of the Fifteenth Workshop on Graph-Based Methods for Natural Language Processing (TextGraphs-15)
%D 2021
%8 June
%I Association for Computational Linguistics
%C Mexico City, Mexico
%F jha-mihata-2021-geodesic
%X In some memory-constrained settings like IoT devices and over-the-network data pipelines, it can be advantageous to have smaller contextual embeddings. We investigate the efficacy of projecting contextual embedding data (BERT) onto a manifold, and using nonlinear dimensionality reduction techniques to compress these embeddings. In particular, we propose a novel post-processing approach, applying a combination of Isomap and PCA. We find that the geodesic distance estimations, estimates of the shortest path on a Riemannian manifold, from Isomap’s k-Nearest Neighbors graph bolstered the performance of the compressed embeddings to be comparable to the original BERT embeddings. On one dataset, we find that despite a 12-fold dimensionality reduction, the compressed embeddings performed within 0.1% of the original BERT embeddings on a downstream classification task. In addition, we find that this approach works particularly well on tasks reliant on syntactic data, when compared with linear dimensionality reduction. These results show promise for a novel geometric approach to achieve lower dimensional text embeddings from existing transformers and pave the way for data-specific and application-specific embedding compressions.
%R 10.18653/v1/2021.textgraphs-1.15
%U https://aclanthology.org/2021.textgraphs-1.15
%U https://doi.org/10.18653/v1/2021.textgraphs-1.15
%P 144-149
Markdown (Informal)
[On Geodesic Distances and Contextual Embedding Compression for Text Classification](https://aclanthology.org/2021.textgraphs-1.15) (Jha & Mihata, TextGraphs 2021)
ACL