From "Strings" to "Things" for Personal Knowledge Graphs: Evaluating LLM Triple Extraction for Recommendation Systems

Abhirup Dasgupta; Fernando Spadea; Oshani Seneviratne

From "Strings" to "Things" for Personal Knowledge Graphs: Evaluating LLM Triple Extraction for Recommendation Systems

T0 review · 3 major / 0 minor · reviewed 2026-07-12 · grok-4.5

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Lightweight open-weight LLMs can turn recommendation chats into RDF preference triples that still drive useful recommendations.

desk verdict Solid, reproducible workshop paper: first systematic open-weight LLM extraction of RDF preference triples from ReDial with controlled FedTREK-LM transfer and a useful inverse-scaling observation; soft spots are missing baselines and single-downstream-system scope, not load-bearing errors. read the letter →

arxiv 2607.00003 v1 pith:7TVHSCNW submitted 2026-04-18 cs.IR cs.AI

Abhirup Dasgupta , Fernando Spadea , Oshani Seneviratne This is my paper

classification cs.IRcs.AI

keywords PersonalKnowledgeGraphConstructionLargeLanguageModelsTripleExtractionConversationalRecommendationSystemsPersonalizedAIDecentralizedPersonalizationRDFWikidatalinking

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Personal Knowledge Graphs give a structured, privacy-friendly way to store what a user likes, has seen, or has been offered, but building them from ordinary multi-turn dialogue has been hard. This paper shows a reproducible pipeline that feeds conversational movie-recommendation dialogues to small open-weight language models and asks them to emit RDF-compliant (user, relation, movie) triples linked to Wikidata identifiers. The authors measure both how faithfully those triples match human annotations and whether the resulting graphs still support a downstream recommendation model. Certain models, especially smaller instruction-tuned ones that extract a balanced mix of relations, produce graphs whose recommendation utility stays surprisingly high relative to their extraction F1. The practical claim is that decentralized, on-device PKG construction is already feasible with lightweight models rather than requiring massive centralized extractors.

What carries the argument

A prompt-based PKG construction pipeline that converts multi-turn dialogues into (User, relation, Item) triples under a fixed five-relation preference ontology, then links movie mentions to Wikidata IRIs before the graph is consumed by a federated recommendation model.

What would settle it

Re-run the same extraction and recommendation pipeline on a conversational dataset whose preference labels were collected independently of the dialogue text; if extraction F1 collapses or downstream recommendation utility no longer tracks extraction quality, the claim that dialogue alone yields faithful PKGs fails.

Watch

Extended reading notes

Core claim

Certain lightweight open-weight models extract usable RDF preference triples from conversational recommendation dialogues, and the Personal Knowledge Graphs they produce retain proportionally high downstream recommendation performance relative to their raw extraction F1; in particular, a small, balanced extractor can outperform a larger high-precision one when the recommender itself is small.

Load-bearing premise

The binary liked/seen/suggested labels already present in the dialogue dataset are treated as complete, recoverable ground-truth triples that an LLM can extract from the raw conversation text alone.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

This is a clean, usable empirical study that does what it claims. The new piece is the end-to-end pipeline: open-weight Qwen3 and Gemma-3 models extract RDF-compliant (User, relation, Item) triples grounded to Wikidata IRIs from ReDial dialogues, scored against the dataset’s own liked/seen/suggested annotations, then plugged into FedTREK-LM so you can see how extraction F1 actually moves recommendation F1. They ship the prompt templates, code, and full P/R/F1 tables by relation and shot count. That is real work, not vapor.

What they do well is the controlled transfer and the inverse-scaling observation. Gemma-3-12B wins raw extraction F1, yet Gemma-3-1B extractions give the small Qwen3-0.6B recommender the best downstream F1 (0.184). Precision-heavy sparse graphs hurt more than balanced denser ones when the recommender itself is small. That is a practical finding for on-device PKG construction and is supported by the numbers they report. The ontology is small and sensible; the Wikidata linking step is explicit; self-citation of FedTREK-LM is only as the evaluator, not as the source of labels.

Soft spots are real but proportional. No non-LLM extraction baseline, no error bars or significance tests on the F1 differences, and the “proportionally high utility” claim rests on a single recommender family (FedTREK-LM, Qwen3 only). If that system is unusually sensitive to density of “liked” vs “suggested” triples, the inverse-scaling result may not travel. The assumption that ReDial metadata are fully recoverable from dialogue text alone is also untested; if annotators used extra context, both extraction and transfer numbers are optimistic. None of these break the central feasibility claim.

This is for people building privacy-preserving conversational recommenders or lightweight PKG pipelines who want concrete numbers and public artifacts rather than another architecture paper. It is workshop-scale, not a field-reorganizing result, but it is honest and reproducible. I would send it to peer review; a referee can demand the missing baseline and a second downstream system without killing the contribution. Worth reading if you care about on-device personalization graphs.

Referee Report

3 major / 0 minor

Summary. The paper proposes a reproducible pipeline that uses lightweight open-weight LLMs (Qwen3 and Gemma-3 families, multiple sizes, 0–10-shot prompting) to extract RDF-compliant (User, relation, Item) preference triples from multi-turn recommendation dialogues, grounds movie entities via Wikidata IRIs, and assembles them into Personal Knowledge Graphs under a small domain ontology (pkg:likes, pkg:dislikes, pkg:hasSeen, pkg:hasNotSeen, pkg:wasSuggested). Extraction fidelity is measured by precision/recall/F1 against ReDial metadata-derived ground-truth triples (Table 2, Figures 3–4). Downstream utility is assessed by training the FedTREK-LM recommender on the resulting augmented datasets and comparing recommendation F1 against models trained on the original ReDial annotations (Section 4.3, Section 5.1, Figure 5). The central empirical claim is that certain models (especially Gemma-3-12B for extraction F1 and Gemma-3-1B for balanced downstream utility) produce usable PKGs whose recommendation performance is proportionally high relative to extraction quality, supporting privacy-preserving, on-device PKG construction.

Significance. If the results hold, the work supplies a concrete, open, and reproducible bridge from conversational ‘strings’ to interoperable RDF ‘things’ for personal knowledge graphs, a missing piece for decentralized personalization. Strengths that should be credited include the public code and prompt templates, the explicit RDF/Wikidata grounding and ontology (Figure 2, Algorithm 1), the full relation-wise extraction tables, and the dual evaluation of extraction fidelity plus a real downstream recommender. The observation that a 1B extractor can yield higher downstream F1 for a small recommender than larger high-precision extractors is practically relevant for on-device settings. The contribution is therefore of clear interest to the Semantic Web / conversational recommendation community, provided the transfer claims are shown to be robust beyond the single evaluation framework used.

major comments (3)

Section 4.3 and Section 5.1 / Figure 5: The load-bearing claim that extracted PKGs yield ‘proportionally high’ downstream performance (and the inverse-scaling result that Gemma-3-1B extractions outperform larger extractors for a small recommender) rests exclusively on FedTREK-LM trained only with Qwen3-family recommenders. No other recommender architecture, no ablation of the recommender’s sensitivity to triple density/balance, and no non-LLM extraction baseline are reported. Without at least one additional downstream system or a controlled density ablation, it remains unclear whether the proportional-utility and inverse-scaling conclusions are properties of the extracted PKGs or artifacts of FedTREK-LM’s particular dependence on ‘liked’ vs. ‘suggested’ coverage.
Table 2 and Section 5: Extraction and downstream F1 differences are reported as point estimates with no error bars, confidence intervals, or statistical significance tests across dialogues or random seeds. Given that several key comparisons (e.g., 3-shot vs. 5-shot precision trade-off, Gemma-1B vs. Gemma-12B downstream F1) are numerically close and drive the narrative, the absence of uncertainty quantification weakens the strength of the relative-performance claims.
Section 1.2, Table 1 and Section 4.1: The evaluation treats ReDial metadata features (liked/seen/suggested) as complete, recoverable ground-truth triples that can be obtained solely from dialogue text after normalization and Wikidata linking. If the annotations encode annotator inference or context not present in the surface conversation, both the absolute extraction F1 numbers and the claimed transfer to recommendation utility are mis-calibrated. A short qualitative error analysis or an estimate of annotation–dialogue mismatch would make the ground-truth assumption falsifiable.

Circularity Check

1 steps flagged · score 2.0 of 10

Minor self-citation of authors' own FedTREK-LM as the sole downstream recommender; extraction F1 and labels remain independent of any fit or self-definition.

self citation load bearing [§4.3 Downstream Task Evaluation / §5.1]
"Specifically, we evaluate the effectiveness of the downstream task on the FedTREK-LM recommendation system [33, 34]. This recommendation system uses a local LLM in a federated system to generate recommendations using PKGs for context, so it fits nicely with our local, LLM-based, triple-extraction pipeline. ... we construct the FedTREK-LM dataset with our augmented dataset as the base, but then we still use the original ground-truth test dataset for evaluation."

The claim that extracted PKGs yield 'proportionally high' recommendation utility (and the inverse-scaling observation that Gemma-1B extractions beat larger models) is demonstrated exclusively inside the authors' own prior FedTREK-LM system. While extraction labels and the test set remain external, the transfer-utility conclusion has no independent recommender baseline and therefore rests on a self-cited evaluator.

full rationale

The paper is an empirical evaluation of LLM triple extraction against the external ReDial human annotations (Table 1, §4.1). Precision/recall/F1 are computed by exact match of (subject, predicate, object) after Wikidata linking; nothing is fitted and then re-predicted. The downstream utility experiments replace only the training triples while always evaluating on the original ground-truth test set (§4.3, §5.1). The sole self-citation is the use of the authors' prior FedTREK-LM framework as the recommendation engine that consumes the extracted PKGs. That citation supplies an evaluation tool rather than a uniqueness theorem, ansatz, or definitional identity that forces the reported numbers. No equation, prompt, or metric reduces by construction to its own inputs. The inverse-scaling and 'proportionally high' claims are therefore framework-specific observations, not circular derivations. Score 2 reflects the single non-load-bearing self-citation; the central extraction results stand independently.

Assumptions & free parameters 3 free parameters · 3 assumptions · 1 invented entities

Empirical IR paper; free parameters are ordinary experimental knobs (shot count, match threshold, model selection). Axioms are standard domain assumptions about conversational annotations and RDF interoperability. No exotic physical or mathematical entities are postulated; the preference ontology is a lightweight schema definition, not a new ontological primitive claimed to exist independently.

free parameters (3)

few-shot count (0/1/3/5/10) = 5-shot selected as primary for downstream
Hand-chosen discrete values; performance peaks then collapses, so the reported optimum is sensitive to this discrete grid.
Wikidata fuzzy-match confidence threshold
Algorithm 1 gates IRI replacement on an unspecified numeric threshold; entity-linking errors directly affect triple match counts.
model-size sampling
Closest available sizes across families chosen by hand; absolute parameter counts are not matched, affecting scale comparisons.

assumptions (3)

domain assumption ReDial metadata features (liked/seen/suggested) map one-to-one onto the five pkg: relations and constitute complete ground truth recoverable from dialogue text alone
Table 1 and §4.1; all precision/recall numbers rest on this mapping.
domain assumption RDF triples grounded in Wikidata IRIs are a suitable intermediate representation for both evaluation and downstream recommendation
§3.1 ontology design; assumed without ablation against non-RDF alternatives.
domain assumption FedTREK-LM F1 on the original test set is a valid utility measure of extracted PKG quality
§4.3; the transfer claim depends on this evaluator being unbiased.

invented entities (1)

User Movie Preference Ontology (pkg:likes, pkg:dislikes, pkg:hasSeen, pkg:hasNotSeen, pkg:wasSuggested)
purpose: Provides a fixed, type-safe relation vocabulary for serializing extracted triples into RDF
Defined in Figure 2 and §3.1 solely for this pipeline; no independent existence claimed outside the paper’s schema.

how reviews work

0 comments

Cite this review

Pith. "Pith review of From "Strings" to "Things" for Personal Knowledge Graphs: Evaluating LLM Triple Extraction for Recommendation Systems." pith.science (2026). https://pith.science/paper/7TVHSCNW

@misc{pith2026260700003,
  author       = {Pith},
  title        = {Pith review of: From "Strings" to "Things" for Personal Knowledge Graphs: Evaluating LLM Triple Extraction for Recommendation Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7TVHSCNW}},
  note         = {Machine review of arXiv:2607.00003}
}

read the original abstract

Personal Knowledge Graphs (PKGs) offer a privacy-preserving framework for modeling user preferences, yet constructing them from unstructured, decentralized conversational data remains a challenge. This paper bridges the gap between conversational "strings" and semantic "things" by presenting a reproducible pipeline for extracting structured user-preference triples using lightweight Large Language Models. We evaluate Qwen- and Gemma-based models on their ability to extract RDF-compliant triples linked to Wikidata identifiers from conversational data for PKG construction. Our evaluation assesses both the semantic extraction fidelity and the utility of the resulting graphs in a downstream recommendation task. We found that certain models performed well and had proportionally high downstream performance relative to their triple extraction performance.

Figures

Figures reproduced from arXiv: 2607.00003 by the authors.

**Figure 1.** Triple Extraction Pipeline: A natural language conversation between a user and a recommender agent is processed by an LLM to extract structured preference triples (e.g., (User, likes, Up), (User, unseen, Wall-E)). The model is guided by k-shot examples and predefined relation definitions to ensure consistent output. Extracted triples are stored in a PKG, enabling downstream recommendation and reasoning tasks [PITH_… view at source ↗

**Figure 2.** User Movie Preference Ontology for the ReDial Dataset. Note that solid edges represent object properties and dashed edges represent rdf:type (instantiation). Ontology Design: Our User Movie Preference Ontology ( [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

**Figure 3.** F1 Scores Across Model Sizes Qwen3-0.6B Qwen3-1.7B Qwen3-4B Qwen3-8B gemma-3-12b-it gemma-3-1b-it gemma-3-4b-it 0.0 0.2 0.4 0.6 0.8 F1 Score Liked Qwen3-0.6B Qwen3-1.7B Qwen3-4B Qwen3-8B gemma-3-12b-it gemma-3-1b-it gemma-3-4b-it F1 Score Seen Qwen3-0.6B Qwen3-1.7B Qwen3-4B Qwen3-8B gemma-3-12b-it gemma-3-1b-it gemma-3-4b-it 0.0 0.2 0.4 0.6 0.8 F1 Score Suggested Shot Count 0-shot 1-shot 3-shot 5-shot 10-shot [PITH… view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: F1 Scores Across Model Relation 5.1. Downstream Task Results As outlined in Section 4.3, we evaluate the utility of our extracted triples by training the FedTREK-LM recommendation framework on our augmented datasets and comparing the results against models trained on t…

**Figure 5.** Figure 5: Downstream recommendation performance evaluated on the FedTREK-LM framework. (Left) Comparison of F1-scores across Qwen3 model sizes using the original ReDial dataset versus the 5-shot augmented dataset. (Right) Downstream F1-score for the Qwen3-0.6B model across vary…

Review history (2 revisions) →

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

35 extracted references · 4 canonical work pages

[1]

Balog, T

K. Balog, T. Kenter, Personal knowledge graphs: A research agenda, in: Proceedings of the 2019 ACM SIGIR International Conference on Theory of Information Retrieval, 2019, pp. 217–220

2019
[2]

M. G. Skjæveland, K. Balog, N. Bernard, W. Łajewska, T. Linjordet, An ecosystem for personal knowledge graphs: A survey and research roadmap, AI Open 5 (2024) 55–69

2024
[3]

S. S. Shirai, O. Seneviratne, D. L. McGuinness, A survey on personal health knowledge graphs, in: Workshop on The Personal Health Knowledge Graph, Knowledge Graph Conference, 2020, 2020

2020
[4]

X. Wang, X. He, Y. Cao, M. Liu, T.-S. Chua, KGAT: Knowledge graph attention network for recommendation, in: Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining, 2019, pp. 950–958

2019
[5]

Christakopoulou, F

K. Christakopoulou, F. Radlinski, K. Hofmann, Towards conversational recommender systems, in: Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining, 2016, pp. 815–824

2016
[6]

X. Wang, X. He, Y. Cao, M. Liu, T.-S. Chua, Kgat: Knowledge graph attention network for recom- mendation, in: Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Dis- covery & Data Mining, KDD ’19, Association for Computing Machinery, New York, NY, USA, 2019, p. 950–958. URL: https://doi.org/10.1145/3292500.3330989. doi:10.1145/329250...

doi:10.1145/3292500.3330989 2019
[7]

P. Wang, X. Li, F. Du, H. Liu, S. Zhi, A personalized recommendation system based on knowledge graph embedding and neural network, in: 2019 3rd International Conference on Data Science and Business Analytics (ICDSBA), 2019, pp. 161–165. doi:10.1109/ICDSBA48748.2019.00042

work page doi:10.1109/icdsba48748.2019.00042 2019
[8]

R. Li, S. Ebrahimi Kahou, H. Schulz, V. Michalski, L. Charlin, C. Pal, Towards deep conversational recommendations, Advances in neural information processing systems 31 (2018)

2018

Show all 35 references

[9]

Hugging Face, ReDial dataset, https://huggingface.co/datasets/community-datasets/re_dial, 2024

2024
[10]

Spadea, O

F. Spadea, O. Seneviratne, Bursting the filter bubble with knowledge graph inversion, in: Compan- ion Publication of the 17th ACM Web Science Conference 2025, 2025, pp. 39–43

2025
[11]

K. Meng, D. Bau, A. Andonian, Y. Belinkov, Locating and editing factual associations in gpt, Advances in neural information processing systems 35 (2022) 17359–17372

2022
[12]

K. Meng, A. S. Sharma, A. J. Andonian, Y. Belinkov, D. Bau, Mass-editing memory in a transformer, in: The Eleventh International Conference on Learning Representations, 2022

2022
[13]

De Cao, W

N. De Cao, W. Aziz, I. Titov, Editing factual knowledge in language models, in: Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, 2021, pp. 6491–6506

2021
[14]

Petroni, T

F. Petroni, T. Rocktäschel, S. Riedel, P. Lewis, A. Bakhtin, Y. Wu, A. Miller, Language models as knowledge bases?, in: K. Inui, J. Jiang, V. Ng, X. Wan (Eds.), Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joi...

2019
[15]

Roberts, C

A. Roberts, C. Raffel, N. Shazeer, How much knowledge can you pack into the parameters of a language model?, in: B. Webber, T. Cohn, Y. He, Y. Liu (Eds.), Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), Association for Computatio...

2020 doi
[16]

Frey, L.-P

J. Frey, L.-P. Meyer, N. Arndt, F. Brei, K. Bulert, Benchmarking the abilities of large language models for rdf knowledge graph creation and comprehension: how well do llms speak turtle?, arXiv preprint arXiv:2309.17122 (2023)

2023 arXiv
[17]

Z. Wang, Z. Zhang, Y. Feng, Y. Zhao, Y. Feng, X. Sun, Extract, define, canonicalize: An llm-based framework for knowledge graph construction, arXiv preprint arXiv:2404.03868 (2024)

2024 arXiv
[18]

Zhang, C

T. Zhang, C. Guo, Z. Chen, Y. Fan, Beyond isolation: Multi-agent synergy for improving knowledge graph construction, arXiv preprint arXiv:2312.03022 (2023)

2023 arXiv
[19]

G. Li, Z. Xu, Z. Shang, J. Liu, K. Ji, Y. Guo, Empirical analysis of dialogue relation extraction with large language models, arXiv preprint arXiv:2404.17802 (2024)

2024 arXiv
[20]

Liang, C

T. Liang, C. Jin, L. Wang, W. Fan, C. Xia, K. Chen, Y. Yin, Llm-redial: a large-scale dataset for conversational recommender systems created from user behaviors with llms, in: Findings of the Association for Computational Linguistics ACL 2024, 2024, pp. 8926–8939

2024
[21]

W.-S. Kim, S. Lim, G.-W. Kim, S.-M. Choi, Extracting implicit user preferences in conversational recommender systems using large language models, Mathematics 13 (2025) 221

2025
[22]

H. Kook, J. Kim, S. Park, J. Lee, Empowering retrieval-based conversational recommendation with contrasting user preferences, in: Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technolo...

2025
[23]

X. Chen, S. Song, Z. Zhang, S. Liu, C. Shi, Lkpnr: Llm and kg for personalized news recommendation framework, arXiv preprint arXiv:2308.12028 (2023)

2023 arXiv
[24]

Z. Zhao, S. Wang, Y. Li, M. Wu, H. Lu, B. Zhang, H. Sun, J. Zhang, Breaking the barrier: Utilizing large language models for industrial recommendation systems through an inferential knowledge graph, arXiv preprint arXiv:2402.13750 (2024)

2024 arXiv
[25]

C.-H. Chen, D. Gruen, J. Harris, J. Hendler, D. L. McGuinness, M. Monti, N. Rastogi, O. Seneviratne, M. J. Zaki, Semantic Technologies for Clinically Relevant Personal Health Applications, Springer In- ternational Publishing, Cham, 2022, pp. 199–220. URL: https://doi.org/10.10...

2022 doi
[26]

doi:10.1007/978-3-031-07696-1_10

doi
[27]

Seneviratne, M

O. Seneviratne, M. Shukla, Personal health knowledge graph construction using internet of medical things, in: Personal Knowledge Graphs (PKGs): Methodology, tools and applications, Computing, Institution of Engineering and Technology, 2023, pp. 295–305. URL: https://digital-li...

2023 doi
[28]

Seneviratne, J

O. Seneviratne, J. Harris, C.-H. Chen, D. L. McGuinness, Personal health knowledge graph for clinically relevant diet recommendations, Workshop on Personal Knowledge Graphs Co- located with the 3rd Automatic Knowledge Base Construction Conference (AKBC’21) (2021). URL: https:/...

2021 arXiv
[29]

Z. Yang, I. Azimi, M. J. Zaki, M. Gaur, O. Seneviratne, D. L. McGuinness, S. M. Rashid, A. M. Rahmani, Transforming Personal Health AI: Integrating Knowledge and Causal Graphs with Large Language Models, in: Proceedings of the ISCAP Conference, 2024

2024
[30]

Zhang, M

P. Zhang, M. Fnu, Y. Song, O. Seneviratne, Z. Yang, I. Azimi, A. M. Rahmani, Adaptive Constraint Relaxation in Personalized Nutrition Recommendations: An LLM-Driven Knowledge Graph Retrieval Approach, in: AMIA Annual Symposium Proceedings, American Medical Informatics Associat...

2025
[31]

Shirai, O

S. Shirai, O. Seneviratne, D. L. McGuinness, Applying personal knowledge graphs to health, in: Personal Health Knowledge Graph Workshop at Knowledge Graph Conference, 2021. URL: https://arxiv.org/abs/2104.07587

2021 arXiv
[32]

Experiments used Qwen3-0.6B, Qwen3-1.7B, Qwen3-4B, and Qwen3-8B

Hugging Face, Qwen3-0.6b, https://huggingface.co/Qwen/Qwen3-0.6B, 2024. Experiments used Qwen3-0.6B, Qwen3-1.7B, Qwen3-4B, and Qwen3-8B

2024
[33]

Includes Gemma-3-1B-IT, Gemma-3-4B-IT, and Gemma-3-12B-IT

Google, Gemma 3: Open instruction-tuned language models, https://huggingface.co/google/ gemma-3-12b-it, 2025. Includes Gemma-3-1B-IT, Gemma-3-4B-IT, and Gemma-3-12B-IT

2025
[34]

Spadea, O

F. Spadea, O. Seneviratne, Federated personal knowledge graph completion with lightweight large language models for personalized recommendations, in: Proceedings of the European Semantic Web Conference, 2026

2026
[35]

Spadea, O

F. Spadea, O. Seneviratne, TREK-LM, https://github.com/brains-group/TREK-LM, 2025

2025

Pith tools

Reviewed July 12, 2026 · model on record in the stance chip above.

[1] [1]

Balog, T

K. Balog, T. Kenter, Personal knowledge graphs: A research agenda, in: Proceedings of the 2019 ACM SIGIR International Conference on Theory of Information Retrieval, 2019, pp. 217–220

2019

[2] [2]

M. G. Skjæveland, K. Balog, N. Bernard, W. Łajewska, T. Linjordet, An ecosystem for personal knowledge graphs: A survey and research roadmap, AI Open 5 (2024) 55–69

2024

[3] [3]

S. S. Shirai, O. Seneviratne, D. L. McGuinness, A survey on personal health knowledge graphs, in: Workshop on The Personal Health Knowledge Graph, Knowledge Graph Conference, 2020, 2020

2020

[4] [4]

X. Wang, X. He, Y. Cao, M. Liu, T.-S. Chua, KGAT: Knowledge graph attention network for recommendation, in: Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining, 2019, pp. 950–958

2019

[5] [5]

Christakopoulou, F

K. Christakopoulou, F. Radlinski, K. Hofmann, Towards conversational recommender systems, in: Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining, 2016, pp. 815–824

2016

[6] [6]

X. Wang, X. He, Y. Cao, M. Liu, T.-S. Chua, Kgat: Knowledge graph attention network for recom- mendation, in: Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Dis- covery & Data Mining, KDD ’19, Association for Computing Machinery, New York, NY, USA, 2019, p. 950–958. URL: https://doi.org/10.1145/3292500.3330989. doi:10.1145/329250...

doi:10.1145/3292500.3330989 2019

[7] [7]

P. Wang, X. Li, F. Du, H. Liu, S. Zhi, A personalized recommendation system based on knowledge graph embedding and neural network, in: 2019 3rd International Conference on Data Science and Business Analytics (ICDSBA), 2019, pp. 161–165. doi:10.1109/ICDSBA48748.2019.00042

work page doi:10.1109/icdsba48748.2019.00042 2019

[8] [8]

R. Li, S. Ebrahimi Kahou, H. Schulz, V. Michalski, L. Charlin, C. Pal, Towards deep conversational recommendations, Advances in neural information processing systems 31 (2018)

2018