tensorkelechi/arxiv-cstext · Datasets at Hugging Face

text stringlengths 0 89.3k
Amazon
qzhuamznamazoncomXiang Song
Amazon
xiangsxamazoncomSoji Adeshina
Amazon
adesojiaamazoncom
Christos Faloutsos
Carnegie Mellon University
Amazon
christoscscmueduGeorge Karypis
Amazon
gkarypisamazoncomYizhou Sun
University of California Los Angeles
Amazon
yzsuncsuclaedu
ABSTRACT
Textrich graphs prevalent in data mining contexts like ecommerce
and academic graphs consist of nodes with textual features linked
by various relations Traditional graph machine learning models
such as Graph Neural Networks GNNs excel in encoding the
graph structural information but have limited capability in han
dling rich text on graph nodes Large Language Models LLMs
noted for their superior text understanding abilities offer a solution
for processing the text in graphs but face integration challenges
due to their limitation for encoding graph structures and their
computational complexities when dealing with extensive text in
large neighborhoods of interconnected nodes This paper intro
duces Hierarchical Compression HiCom a novel method to
align the capabilities of LLMs with the structure of textrich graphs
HiCom processes text in a nodes neighborhood in a structured
manner by organizing the extensive textual information into a
more manageable hierarchy and compressing node text step by
step Therefore HiCom not only preserves the contextual richness
of the text but also addresses the computational challenges of LLMs
which presents an advancement in integrating the text processing
power of LLMs with the structural complexities of textrich graphs
Empirical results show that HiCom can outperform both GNNs and
LLM backbones for node classification on ecommerce and citation
graphs HiCom is especially effective for nodes from a dense re
gion in a graph where it achieves a 348 average performance
improvement on five datasets while being more efficient than LLM
backbones
Work done while being an intern at Amazon Web Services
Permission to make digital or hard copies of all or part of this work for personal or
classroom use is granted without fee provided that copies are not made or distributed
for profit or commercial advantage and that copies bear this notice and the full citation
on the first page Copyrights for components of this work owned by others than the
authors must be honored Abstracting with credit is permitted To copy otherwise or
republish to post on servers or to redistribute to lists requires prior specific permission
andor a fee Request permissions from permissionsacmorg
Conference acronym XX June 0305 2018 Woodstock NY
2018 Copyright held by the ownerauthors Publication rights licensed to ACM
ACM ISBN 97814503XXXXX1806
httpsdoiorgXXXXXXXXXXXXXXCCS CONCEPTS
Computing methodologies Neural networks Natural
language processing Mathematics of computing Graph
algorithms
KEYWORDS
Machine Learning Graphs Compression Language Models
ACM Reference Format
Shichang Zhang Da Zheng Jiani Zhang Qi Zhu Xiang Song Soji Adeshina
Christos Faloutsos George Karypis and Yizhou Sun 2018 Hierarchical
Compression of TextRich Graphs via Large Language Models In Proceedings
of Make sure to enter the correct conference title from your rights confirmation
emai Conference acronym XX ACM New York NY USA 12 pages https
doiorgXXXXXXXXXXXXXX
1 INTRODUCTION
Textrich graphs have become prevalent and increasingly important
in data mining These graphs combine textual features with graph
structures offering a unique representation that captures complex
entity interactions For instance academic graphs represent publi
cations as nodes linked by citations with paper titles and abstracts
providing rich textual node features 51 Similarly ecommerce
graphs represent products as nodes connected by customer viewing
and purchasing histories enriched with detailed textual product de
scriptions 26 These graphs are characterized by both substantial
textual information and complex relations
Studies have pointed out that both textual features and graph
structures are crucial for data mining tasks associated with these
graphs eg node classification 4552 Figure 1 presents two
productcoviewing graph snippets demonstrating that product
category classifications such as Kitchen Dining orSports
Outdoors become clearer by considering the surrounding context
of neighboring nodes rather than relying solely on product descrip
tions For instance the category of a water bottle leans towards
Kitchen Dining when linked with items such as Glass Tum
blers and Strawberry Popping Boba suggesting a usage scenario
aligned with home and kitchen settings Conversely a water bot
tle with a similar description is more likely to be classified underarXiv240611884v1 csSI 13 Jun 2024Conference acronym XX June 0305 2018 Woodstock NY Zhang et al
Sports Outdoors if it is often coviewed with other water bot
tles for travel outdoors and gym as well as Health and Fitness
Trackers Thus integrating textual descriptions with the relational
information from a graph structure captures the comprehensive
neighborhood context and enhances predictive accuracy
To leverage both textual features and graph structures for solving
problems on textrich graphs Graph Neural Networks GNNs have
recently gained popularity as they can encode both information via
message passing 142138 While GNNs are good at capturing
structural information and have achieved great results on many
benchmarks 162532 they are not inherently equipped to handle

Amazon

qzhuamznamazoncomXiang Song

Amazon

xiangsxamazoncomSoji Adeshina

Amazon

adesojiaamazoncom

Christos Faloutsos

Carnegie Mellon University

Amazon

christoscscmueduGeorge Karypis

Amazon

gkarypisamazoncomYizhou Sun

University of California Los Angeles

Amazon

yzsuncsuclaedu

ABSTRACT

Textrich graphs prevalent in data mining contexts like ecommerce

and academic graphs consist of nodes with textual features linked

by various relations Traditional graph machine learning models

such as Graph Neural Networks GNNs excel in encoding the

graph structural information but have limited capability in han

dling rich text on graph nodes Large Language Models LLMs

noted for their superior text understanding abilities offer a solution

for processing the text in graphs but face integration challenges

due to their limitation for encoding graph structures and their

computational complexities when dealing with extensive text in

large neighborhoods of interconnected nodes This paper intro

duces Hierarchical Compression HiCom a novel method to

align the capabilities of LLMs with the structure of textrich graphs

HiCom processes text in a nodes neighborhood in a structured

manner by organizing the extensive textual information into a

more manageable hierarchy and compressing node text step by

step Therefore HiCom not only preserves the contextual richness

of the text but also addresses the computational challenges of LLMs

which presents an advancement in integrating the text processing

power of LLMs with the structural complexities of textrich graphs

Empirical results show that HiCom can outperform both GNNs and

LLM backbones for node classification on ecommerce and citation

graphs HiCom is especially effective for nodes from a dense re

gion in a graph where it achieves a 348 average performance

improvement on five datasets while being more efficient than LLM

backbones

Work done while being an intern at Amazon Web Services

Permission to make digital or hard copies of all or part of this work for personal or

classroom use is granted without fee provided that copies are not made or distributed

for profit or commercial advantage and that copies bear this notice and the full citation

on the first page Copyrights for components of this work owned by others than the

authors must be honored Abstracting with credit is permitted To copy otherwise or

republish to post on servers or to redistribute to lists requires prior specific permission

andor a fee Request permissions from permissionsacmorg

Conference acronym XX June 0305 2018 Woodstock NY

2018 Copyright held by the ownerauthors Publication rights licensed to ACM

ACM ISBN 97814503XXXXX1806

httpsdoiorgXXXXXXXXXXXXXXCCS CONCEPTS

Computing methodologies Neural networks Natural

language processing Mathematics of computing Graph

algorithms

KEYWORDS

Machine Learning Graphs Compression Language Models

ACM Reference Format

Shichang Zhang Da Zheng Jiani Zhang Qi Zhu Xiang Song Soji Adeshina

Christos Faloutsos George Karypis and Yizhou Sun 2018 Hierarchical

Compression of TextRich Graphs via Large Language Models In Proceedings

of Make sure to enter the correct conference title from your rights confirmation

emai Conference acronym XX ACM New York NY USA 12 pages https

doiorgXXXXXXXXXXXXXX

1 INTRODUCTION

Textrich graphs have become prevalent and increasingly important

in data mining These graphs combine textual features with graph

structures offering a unique representation that captures complex

entity interactions For instance academic graphs represent publi

cations as nodes linked by citations with paper titles and abstracts

providing rich textual node features 51 Similarly ecommerce

graphs represent products as nodes connected by customer viewing

and purchasing histories enriched with detailed textual product de

scriptions 26 These graphs are characterized by both substantial

textual information and complex relations

Studies have pointed out that both textual features and graph

structures are crucial for data mining tasks associated with these

graphs eg node classification 4552 Figure 1 presents two

productcoviewing graph snippets demonstrating that product

category classifications such as Kitchen Dining orSports

Outdoors become clearer by considering the surrounding context

of neighboring nodes rather than relying solely on product descrip

tions For instance the category of a water bottle leans towards

Kitchen Dining when linked with items such as Glass Tum

blers and Strawberry Popping Boba suggesting a usage scenario

aligned with home and kitchen settings Conversely a water bot

tle with a similar description is more likely to be classified underarXiv240611884v1 csSI 13 Jun 2024Conference acronym XX June 0305 2018 Woodstock NY Zhang et al

Sports Outdoors if it is often coviewed with other water bot

tles for travel outdoors and gym as well as Health and Fitness

Trackers Thus integrating textual descriptions with the relational

information from a graph structure captures the comprehensive

neighborhood context and enhances predictive accuracy

To leverage both textual features and graph structures for solving

problems on textrich graphs Graph Neural Networks GNNs have

recently gained popularity as they can encode both information via

message passing 142138 While GNNs are good at capturing

structural information and have achieved great results on many

benchmarks 162532 they are not inherently equipped to handle