@inproceedings{feng-etal-2026-dual,
title = "Dual-Prompt-Driven Textual Knowledge Graph Generation With Heterogeneous Information Embedding for {CTR} Prediction",
author = "Feng, Xiling and
Liu, Han and
Pan, Yinghui and
Xu, Zhiwu",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.401/",
pages = "9213--9230",
ISBN = "979-8-89176-459-0",
abstract = "In click-through rate prediction tasks, learning high-quality feature embeddings from multi-field categorical data is an essential task. Early approaches typically need to transform categorical features into one-hot encoded vectors for producing embeddings. This transformation often results in loss of semantic information. In the era of large language models, it has become necessary to transform multi-field categorical data into text to achieve embedding of semantic information. Moreover, semantic information of each categorical feature can be useful to model the relationships among categories for enhancing the embedding quality. To address this point, in this paper, we propose an approach of dual-prompt-driven textual knowledge graph generation with heterogeneous information embedding for enhancing the representation of multi-field categorical data. Specifically, we design two prompt templates to transform multi-field categorical data into a textual knowledge graph for problem reformulation, and propose a heterogeneous information embedding approach that works by jointly representing the generated text and knowledge graph in the setting of cross-attention gated modality fusion with feature alignment. The experimental results indicate that our approach outperforms those baselines of click-through rate prediction. The results also indicate that each module of our approach is useful for improving the performance."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="feng-etal-2026-dual">
<titleInfo>
<title>Dual-Prompt-Driven Textual Knowledge Graph Generation With Heterogeneous Information Embedding for CTR Prediction</title>
</titleInfo>
<name type="personal">
<namePart type="given">Xiling</namePart>
<namePart type="family">Feng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Han</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yinghui</namePart>
<namePart type="family">Pan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhiwu</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>In click-through rate prediction tasks, learning high-quality feature embeddings from multi-field categorical data is an essential task. Early approaches typically need to transform categorical features into one-hot encoded vectors for producing embeddings. This transformation often results in loss of semantic information. In the era of large language models, it has become necessary to transform multi-field categorical data into text to achieve embedding of semantic information. Moreover, semantic information of each categorical feature can be useful to model the relationships among categories for enhancing the embedding quality. To address this point, in this paper, we propose an approach of dual-prompt-driven textual knowledge graph generation with heterogeneous information embedding for enhancing the representation of multi-field categorical data. Specifically, we design two prompt templates to transform multi-field categorical data into a textual knowledge graph for problem reformulation, and propose a heterogeneous information embedding approach that works by jointly representing the generated text and knowledge graph in the setting of cross-attention gated modality fusion with feature alignment. The experimental results indicate that our approach outperforms those baselines of click-through rate prediction. The results also indicate that each module of our approach is useful for improving the performance.</abstract>
<identifier type="citekey">feng-etal-2026-dual</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.401/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>9213</start>
<end>9230</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Dual-Prompt-Driven Textual Knowledge Graph Generation With Heterogeneous Information Embedding for CTR Prediction
%A Feng, Xiling
%A Liu, Han
%A Pan, Yinghui
%A Xu, Zhiwu
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F feng-etal-2026-dual
%X In click-through rate prediction tasks, learning high-quality feature embeddings from multi-field categorical data is an essential task. Early approaches typically need to transform categorical features into one-hot encoded vectors for producing embeddings. This transformation often results in loss of semantic information. In the era of large language models, it has become necessary to transform multi-field categorical data into text to achieve embedding of semantic information. Moreover, semantic information of each categorical feature can be useful to model the relationships among categories for enhancing the embedding quality. To address this point, in this paper, we propose an approach of dual-prompt-driven textual knowledge graph generation with heterogeneous information embedding for enhancing the representation of multi-field categorical data. Specifically, we design two prompt templates to transform multi-field categorical data into a textual knowledge graph for problem reformulation, and propose a heterogeneous information embedding approach that works by jointly representing the generated text and knowledge graph in the setting of cross-attention gated modality fusion with feature alignment. The experimental results indicate that our approach outperforms those baselines of click-through rate prediction. The results also indicate that each module of our approach is useful for improving the performance.
%U https://aclanthology.org/2026.emnlp-main.401/
%P 9213-9230
Markdown (Informal)
[Dual-Prompt-Driven Textual Knowledge Graph Generation With Heterogeneous Information Embedding for CTR Prediction](https://aclanthology.org/2026.emnlp-main.401/) (Feng et al., EMNLP 2026)
ACL