Knowledge graph question answering seeks to translate natural language questions into executable queries over knowledge graphs, but existing approaches often rely on large models or full supervision in the form of gold query annotations. This study examines whether reinforcement learning with outcome-based rewards can train a small instruction-tuned language model to perform zero-shot Text-to-SPARQL generation in the scholarly domain. Group-Relative Policy Optimization (GRPO) is applied to the Qwen3-1.7B model on DBLP-QuAD, using prompts that combine natural language questions with symbolic hints about entities and relations. Training relies on execution feedback, structural constraints, and answer-level rewards, with an additional variant that incorporates gold-query-based shaping. The resulting models are compared to the unmodified zero-shot baseline and to a supervised DoRA-finetuned baseline across answer-level accuracy, execution accuracy, category-wise scores, and generalization to held-out templates. GRPO substantially improves over the zero-shot baseline and exhibits competitive generalization, while supervised DoRA finetuning achieves higher overall accuracy on the same model scale. Ablation analyses indicate that execution-based rewards account for most gains, with additional shaping yielding limited additional benefit, suggesting that outcome-based reinforcement learning is a viable training strategy when gold queries are unavailable for token-level supervision.

Generalization to different Knowledge Graphs (KGs) is a core challenge of Knowledge Graph Question Answering (KGQA). Current models struggle to adapt to unseen KGs due to their underlying heterogeneity. Most approaches depend on data-hungry methods, such as supervised fine-tuning and KG-specific few-shot demonstrations. These systems are limited in cross-KG generalization, and may exhibit hallucination problems, especially in low-resource or domain-specific scenarios. This paper presents STaRQ-Agent, an LLM-based multi-agent collaborative framework for the KGQA generalization problem. It comprises a core agent for SPARQL query generation, accompanied by three supportive agents for schema selection, template generation, and query refinement. This collaborative framework helps mitigate hallucination problems inherent in LLMs. STaRQ-Agent was evaluated on LC-QuAD 1.0, QALD-9, MetaQA, and MatKGQA datasets in few-shot and zero-shot settings. Experimental results demonstrate strong overall generalization across diverse KGs, without requiring high-quality annotated data, re-training, or fine-tuning. Specially, on LC-QuAD 1.0, QALD-9, and MetaQA, few-shot STaRQ-Agent outperforms the strongest baseline by 5.1%, 4.3%, and 0.7%, while zero-shot STaRQ-Agent shows only modest F1 drops of 8.7%, 12.3%, and 10.8%. Moreover, the state-of-the-art results on MatKGQA show its robust out-of-distribution generalization to novel KGs in low-resource specific domains.

This paper investigates whether state-of-the-art Large Language Models (LLMs) can automatically translate SPARQL between popular Knowledge Graph (KG) schemas. We focus on translations between the DBpedia and Wikidata KG, and later on DBLP and OpenAlex KG. This study addresses a notable gap in KG interoperability research by rigorously evaluating LLM performance on SPARQL-to-SPARQL translation. Two benchmarks are assembled, where the first align 100 DBpedia-Wikidata queries from QALD-9-Plus; the second contains 100 DBLP queries aligned to OpenAlex, testing generalizability beyond encyclopaedic KGs. Three open LLMs: Llama-3-8B, DeepSeek-R1-Distill-Llama-70B, and Mistral-Large-Instruct-2407 are selected based on their sizes and architectures and tested with zero-shot, few-shot, and two chain-of-thought variants. Outputs were compared with gold answers, and resulting errors were categorized. We find that the performance varies markedly across models and prompting strategies, and that translations for Wikidata to DBpedia work far better than translations for DBpedia to Wikidata.

Purpose: The heterogeneity of scholarly information in knowledge graphs (KGs) and unstructured textual sources poses challenges in building robust Scholarly Question Answering (SQA) systems. Existing datasets and models typically address a narrow spectrum, focusing exclusively on KGs or unstructured sources and limiting evaluation to simple factoid questions. This gap leaves current systems unable to answer complex, hybrid scholarly questions that require integrating evidence from multiple heterogeneous data sources. Methodology: We introduce HySQA (Hybrid Scholarly Question Answering), a large-scale benchmarking dataset containing hybrid questions over scholarly KGs and Wikipedia text. HySQA contains complex questions that need to traverse facts across structured and unstructured sources. We also develop a baseline model that adaptively decomposes each question into sub-questions, identifies their answer sources, retrieves relevant information from SKGs and Wikipedia, and generates an answer using a hybrid augmented answer generation framework. Findings: The experimental results show that integrating static and adaptive decomposition methods is more effective than static decomposition alone. Value: Introducing HySQA provides the community with resources for evaluating the advancements in scholarly QA research.

Organisations disclose their sustainability performance in corporate sustainability reports (CSRs). CSRs vary widely in structure and depth depending on the reporting framework. Such disparity, together with report complexity and volume, poses significant challenges to transparency, comparability and standardisation. To address this problem, we introduce ReportGRI, an automated system for Global Reporting Initiative (GRI) indexing and qualitative assessment of CSRs. The interactive framework leverages information retrieval techniques and zero-shot prompting to enable GRI disclosure-based report indexing and report coverage assessment by visualising well-covered topics and reporting gaps. The tool facilitates scalable and explainable benchmarking of Environmental, Social and Governance (ESG) reporting quality, enhancing report interpretation, transparency, and corporate accountability. The system is open-sourced on GitHub with an introduction video.

We present DBLP-QuAD 2.0, designed to evaluate Scholarly Knowledge Graph Question Answering (KGQA) over DBLP. Recent updates in the underlying DBLP KG, including new entities and relationships such as venues, research streams, and citation links, have necessitated a corresponding update to existing KG QA benchmarking resources. While the DBLP-QuAD dataset focused on author and publication-centered queries, DBLP-QuAD 2.0 broadens the coverage to reflect the enriched structure of the updated KG. Specifically, the questions in our dataset are formulated from SPARQL query logs that cover a wide range of entities involving authors, publications, venues, research streams, and citation relationships. DBLP-QuAD 2.0 thus provides a more comprehensive benchmark for evaluating KGQA systems with a baseline.

In this work, we release the Shortest Path subgraph Question Answering (ShortPathQA) dataset, the first dataset that provides textual questions with pre-computed relevant subgraphs retrieved from the Wikidata Knowledge Graph (KG), standardizing the evaluation framework for Knowledge Graph Question Answering (KGQA). For this purpose, we utilize the Mintaka dataset for both training and testing and additionally create a manual question-answering subset for testing. Our baseline experiments with both supervised approaches and unsupervised Large Language Model (LLM) inference indicate that even a simplified KGQA formulation with given KG subgraphs and candidate answers remains challenging. Our analysis has shown that LLMs are unable to correctly process and utilize graph data structures without detailed prompt engineering or model tuning. This limitation highlights the need for the creation of this dataset as a training ground for the development of methods that enable LLMs to work more effectively with graph data.

Human-produced emissions are growing at an alarming rate, causing already observable changes in the climate and environment in general. Each year global carbon dioxide emissions hit a new record, and it is reported that 0.5 total US greenhouse gas emissions are attributed to data centres as of 2021. The release of ChatGPT in late 2022 sparked social interest in Large Language Models (LLMs), the new generation of Language Models with a large number of parameters and trained on massive amounts of data. Currently, numerous companies are releasing products featuring various LLMs, with many more models in development and awaiting release. Deep Learning research is a competitive field, with only models that reach top performance attracting attention and being utilized. Hence, achieving better accuracy and results is often the first priority, while the model's efficiency and the environmental impact of the study are neglected. However, LLMs demand substantial computational resources and are very costly to train, both financially and environmentally. It becomes essential to raise awareness and promote conscious decisions about algorithmic and hardware choices. Providing information on training time, the approximate carbon dioxide emissions and power consumption would assist future studies in making necessary adjustments and determining the compatibility of available computational resources with model requirements. In this study, we infused T5 LLM with external knowledge and fine-tuned the model for Question-Answering task. Furthermore, we calculated and reported the approximate environmental impact for both steps. The findings demonstrate that the smaller models may not always be sustainable options, and increased training does not always imply better performance. The most optimal outcome is achieved by carefully considering both performance and efficiency factors.

Existing Scholarly Question Answering (QA) methods typically target homogeneous data sources, relying solely on either text or Knowledge Graphs (KGs). However, scholarly information often spans heterogeneous sources, necessitating the development of QA systems that integrate information from multiple heterogeneous data sources. To address this challenge, we introduce Hybrid-SQuAD (Hybrid Scholarly Question Answering Dataset), a novel large-scale QA dataset designed to facilitate answering questions incorporating both text and KG facts. The dataset consists of 10.5K question-answer pairs generated by a large language model, leveraging the KGs DBLP and SemOpenAlex alongside corresponding text from Wikipedia. In addition, we propose a RAG-based baseline hybrid QA model, achieving an exact match score of 69.65 on the Hybrid-SQuAD test set.

In this work we create a question answering dataset over the DBLP scholarly knowledge graph (KG). DBLP is an on-line reference for bibliographic information on major computer science publications that indexes over 4.4 million publications published by more than 2.2 million authors. Our dataset consists of 10,000 question answer pairs with the corresponding SPARQL queries which can be executed over the DBLP KG to fetch the correct answer. DBLP-QuAD is the largest scholarly question answering dataset.

In this work, we present an end-to-end Knowledge Graph Question Answering (KGQA) system named GETT-QA. GETT-QA uses T5, a popular text-to-text pre-trained language model. The model takes a question in natural language as input and produces a simpler form of the intended SPARQL query. In the simpler form, the model does not directly produce entity and relation IDs. Instead, it produces corresponding entity and relation labels. The labels are grounded to KG entity and relation IDs in a subsequent step. To further improve the results, we instruct the model to produce a truncated version of the KG embedding for each entity. The truncated KG embedding enables a finer search for disambiguation purposes. We find that T5 is able to learn the truncated KG embeddings without any change of loss function, improving KGQA performance. As a result, we report strong results for LC-QuAD 2.0 and SimpleQuestions-Wikidata datasets on end-to-end KGQA over Wikidata.

In this work, we present ARDIAS, a web-based application that aims to provide researchers with a full suite of discovery and collaboration tools. ARDIAS currently allows searching for authors and articles by name and gaining insights into the research topics of a particular researcher. With the aid of AI-based tools, ARDIAS aims to recommend potential collaborators and topics to researchers. In the near future, we aim to add tools that allow researchers to communicate with each other and start new projects.

This thesis explores the topic of Knowledge Graph Question Answering with a special emphasis on semantic parsing approaches, incorporating pre-trained text-to-text language models. We use the text generation ability of these models to convert natural language questions to logical forms. We test whether correct logical forms are being generated, and if not, how to mitigate the failure cases. As a second step, we try to make the same models generate additional information to aid the process of grounding of the logical forms to entities, relations and literals in the Knowledge Graph. In experiments conducted so far, we see encouraging results on both generation of base logical forms, and grounding them to the KG elements. At the same time, we discover failure cases prompting directions in future work (The author considers himself a ‘middle-stage’ Ph.D. candidate).

In this work, we analyse the role of output vocabulary for text-to-text (T2T) models on the task of SPARQL semantic parsing. We perform experiments within the the context of knowledge graph question answering (KGQA), where the task is to convert questions in natural language to the SPARQL query language. We observe that the query vocabulary is distinct from human vocabulary. Language Models (LMs) are pre-dominantly trained for human language tasks, and hence, if the query vocabulary is replaced with a vocabulary more attuned to the LM tokenizer, the performance of models may improve. We carry out carefully selected vocabulary substitutions on the queries and find absolute gains in the range of 17% on the GrailQA dataset.

AI enabled chat bots have recently been put to use to answer customer service queries, however it is a common feedback of users that bots lack a personal touch and are often unable to understand the real intent of the user's question. To this end, it is desirable to have human involvement in the customer servicing process. In this work, we present a system where a human support agent collaborates in real-time with an AI agent to satisfactorily answer customer queries. We describe the user interaction elements of the solution, along with the machine learning techniques involved in the AI agent.

In this work, we present a web application named DBLPLink, which performs entity linking over the DBLP scholarly knowledge graph. DBLPLink uses text-to-text pre-trained language models, such as T5, to produce entity label spans from an input text question. Entity candidates are fetched from a database based on the labels, and an entity re-ranker sorts them based on entity embeddings, such as TransE, DistMult and ComplEx. The results are displayed so that users may compare and contrast the results between T5-small, T5-base and the different KG embeddings used. The demo can be accessed at https://ltdemos.informatik.uni-hamburg.de/dblplink/.

Customers desire convenient, fast, and personalized service encounters. Hence, service companies deploy self-service technology for online customer service. However, as solutions based on Artificial Intelligence cannot reliably answer the full range of requests and the demands on service employees (SEs) in live chat interaction are high, Hybrid Intelligence Systems (HIS) provide great potential to overcome current pitfalls by combining the complementary strengths of artificial and human intelligence. To ensure optimal performance of this socio-technical ensemble, human-centered design approaches are needed to realize real-time augmentation of decision-making in chat-based service encounters. Following a Design Science Research approach, we generate theory-based design principles (DPs) and implement them in a web-based HIS prototype. We contribute to Hybrid Intelligence research with results showing that the DPs enable task mastery and decision efficiency and provide avenues for future research.