Florian Matthes

Phillip Schneider, Tim Schopf, Juraj Vladika et al. · deepmind

In pace with developments in the research field of artificial intelligence, knowledge graphs (KGs) have attracted a surge of interest from both academia and industry. As a representation of semantic relations between entities, KGs have proven to be particularly relevant for natural language processing (NLP), experiencing a rapid spread and wide adoption within recent years. Given the increasing amount of research work in this area, several KG-related approaches have been surveyed in the NLP research community. However, a comprehensive study that categorizes established topics and reviews the maturity of individual research streams remains absent to this day. Contributing to closing this gap, we systematically analyzed 507 papers from the literature on KGs in NLP. Our survey encompasses a multifaceted review of tasks, research types, and contributions. As a result, we present a structured overview of the research landscape, provide a taxonomy of tasks, summarize our findings, and highlight directions for future work.

Tim Schopf, Daniel Braun, Florian Matthes

Text classification of unseen classes is a challenging Natural Language Processing task and is mainly attempted using two different types of approaches. Similarity-based approaches attempt to classify instances based on similarities between text document representations and class description representations. Zero-shot text classification approaches aim to generalize knowledge gained from a training task by assigning appropriate labels of unknown classes to text documents. Although existing studies have already investigated individual approaches to these categories, the experiments in literature do not provide a consistent comparison. This paper addresses this gap by conducting a systematic evaluation of different similarity-based and zero-shot approaches for text classification of unseen classes. Different state-of-the-art approaches are benchmarked on four text classification datasets, including a new dataset from the medical domain. Additionally, novel SimCSE and SBERT-based baselines are proposed, as other baselines used in existing work yield weak classification results and are easily outperformed. Finally, the novel similarity-based Lbl2TransformerVec approach is presented, which outperforms previous state-of-the-art approaches in unsupervised text classification. Our experiments show that similarity-based approaches significantly outperform zero-shot approaches in most cases. Additionally, using SimCSE or SBERT embeddings instead of simpler text representations increases similarity-based classification results even further.

Tim Schopf, Simon Klimek, Florian Matthes

Keyphrase extraction is the process of automatically selecting a small set of most relevant phrases from a given text. Supervised keyphrase extraction approaches need large amounts of labeled training data and perform poorly outside the domain of the training data. In this paper, we present PatternRank, which leverages pretrained language models and part-of-speech for unsupervised keyphrase extraction from single documents. Our experiments show PatternRank achieves higher precision, recall and F1-scores than previous state-of-the-art approaches. In addition, we present the KeyphraseVectorizers package, which allows easy modification of part-of-speech patterns for candidate keyphrase selection, and hence adaptation of our approach to any domain.

Tim Schopf, Daniel Braun, Florian Matthes

In this paper, we consider the task of retrieving documents with predefined topics from an unlabeled document dataset using an unsupervised approach. The proposed unsupervised approach requires only a small number of keywords describing the respective topics and no labeled document. Existing approaches either heavily relied on a large amount of additionally encoded world knowledge or on term-document frequencies. Contrariwise, we introduce a method that learns jointly embedded document and word vectors solely from the unlabeled document dataset in order to find documents that are semantically similar to the topics described by the keywords. The proposed method requires almost no text preprocessing but is simultaneously effective at retrieving relevant documents with high probability. When successively retrieving documents on different predefined topics from publicly available and commonly used datasets, we achieved an average area under the receiver operating characteristic curve value of 0.95 on one dataset and 0.92 on another. Further, our method can be used for multiclass document classification, without the need to assign labels to the dataset in advance. Compared with an unsupervised classification baseline, we increased F1 scores from 76.6 to 82.7 and from 61.0 to 75.1 on the respective datasets. For easy replication of our approach, we make the developed Lbl2Vec code publicly available as a ready-to-use tool under the 3-Clause BSD license.

Oleksandra Klymenko, Stephen Meisenbacher, Florian Matthes

As the tide of Big Data continues to influence the landscape of Natural Language Processing (NLP), the utilization of modern NLP methods has grounded itself in this data, in order to tackle a variety of text-based tasks. These methods without a doubt can include private or otherwise personally identifiable information. As such, the question of privacy in NLP has gained fervor in recent years, coinciding with the development of new Privacy-Enhancing Technologies (PETs). Among these PETs, Differential Privacy boasts several desirable qualities in the conversation surrounding data privacy. Naturally, the question becomes whether Differential Privacy is applicable in the largely unstructured realm of NLP. This topic has sparked novel research, which is unified in one basic goal: how can one adapt Differential Privacy to NLP methods? This paper aims to summarize the vulnerabilities addressed by Differential Privacy, the current thinking, and above all, the crucial next steps that must be considered.

Phillip Schneider, Anum Afzal, Juraj Vladika et al.

Conversational search has evolved as a new information retrieval paradigm, marking a shift from traditional search systems towards interactive dialogues with intelligent search agents. This change especially affects exploratory information-seeking contexts, where conversational search systems can guide the discovery of unfamiliar domains. In these scenarios, users find it often difficult to express their information goals due to insufficient background knowledge. Conversational interfaces can provide assistance by eliciting information needs and narrowing down the search space. However, due to the complexity of information-seeking behavior, the design of conversational interfaces for retrieving information remains a great challenge. Although prior work has employed user studies to empirically ground the system design, most existing studies are limited to well-defined search tasks or known domains, thus being less exploratory in nature. Therefore, we conducted a laboratory study to investigate open-ended search behavior for navigation through unknown information landscapes. The study comprised of 26 participants who were restricted in their search to a text chat interface. Based on the collected dialogue transcripts, we applied statistical analyses and process mining techniques to uncover general information-seeking patterns across five different domains. We not only identify core dialogue acts and their interrelations that enable users to discover domain knowledge, but also derive design suggestions for conversational search systems.

Tim Schopf, Karim Arabi, Florian Matthes

As an efficient approach to understand, generate, and process natural language texts, research in natural language processing (NLP) has exhibited a rapid spread and wide adoption in recent years. Given the increasing research work in this area, several NLP-related approaches have been surveyed in the research community. However, a comprehensive study that categorizes established topics, identifies trends, and outlines areas for future research remains absent. Contributing to closing this gap, we have systematically classified and analyzed research papers in the ACL Anthology. As a result, we present a structured overview of the research landscape, provide a taxonomy of fields of study in NLP, analyze recent developments in NLP, summarize our findings, and highlight directions for future work.

Jonas Gebele, Florian Matthes

Collateralized prediction markets are contingent-claim markets in which economic uncertainty can disappear before winning claims become redeemable. This paper studies the pricing effect of that delay. When collateral remains locked until oracle settlement, a near-certain dollar is a delayed dollar, so prices embed a maturity-dependent settlement discount in addition to beliefs about outcomes. We recover an implied settlement-discount term structure from persistent near-certain contracts using realized settlement times and summarize it as an annualized settlement wedge (ASW). The recovered wedges are positive, maturity-dependent, and time-varying. Adjusting pricesby these curves reduces the near-certainty horizon gradient by roughly 48-88%, indicating that much of the raw maturity pattern reflects priced settlement frictions rather than forecast error alone. Market architecture changes the wedge: negRisk conversion compresses discounts by recycling part of the position into synthetic collateral, while yield-bearing collateral flattens the term structure by reducing the opportunity cost of lock-up. The results show that pricing quality in prediction markets is endogenous to settlement mechanics, collateral productivity, and capital-recycling design. Prediction-market prices therefore aggregate information through a financial infrastructure whose funding conditions are measurable and economically important.

Juraj Vladika, Florian Matthes

With the increasing number of clinical trial reports generated every day, it is becoming hard to keep up with novel discoveries that inform evidence-based healthcare recommendations. To help automate this process and assist medical experts, NLP solutions are being developed. This motivated the SemEval-2023 Task 7, where the goal was to develop an NLP system for two tasks: evidence retrieval and natural language inference from clinical trial data. In this paper, we describe our two developed systems. The first one is a pipeline system that models the two tasks separately, while the second one is a joint system that learns the two tasks simultaneously with a shared representation and a multi-task learning approach. The final system combines their outputs in an ensemble system. We formalize the models, present their characteristics and challenges, and provide an analysis of achieved results. Our system ranked 3rd out of 40 participants with a final submission.

Tim Schopf, Dennis N. Schneider, Florian Matthes

Sentence embeddings enable us to capture the semantic similarity of short texts. Most sentence embedding models are trained for general semantic textual similarity tasks. Therefore, to use sentence embeddings in a particular domain, the model must be adapted to it in order to achieve good results. Usually, this is done by fine-tuning the entire sentence embedding model for the domain of interest. While this approach yields state-of-the-art results, all of the model's weights are updated during fine-tuning, making this method resource-intensive. Therefore, instead of fine-tuning entire sentence embedding models for each target domain individually, we propose to train lightweight adapters. These domain-specific adapters do not require fine-tuning all underlying sentence embedding model parameters. Instead, we only train a small number of additional parameters while keeping the weights of the underlying sentence embedding model fixed. Training domain-specific adapters allows always using the same base model and only exchanging the domain-specific adapters to adapt sentence embeddings to a specific domain. We show that using adapters for parameter-efficient domain adaptation of sentence embeddings yields competitive performance within 1% of a domain-adapted, entirely fine-tuned sentence embedding model while only training approximately 3.6% of the parameters.

Tim Schopf, Emanuel Gerber, Malte Ostendorff et al.

Generic sentence embeddings provide a coarse-grained approximation of semantic textual similarity but ignore specific aspects that make texts similar. Conversely, aspect-based sentence embeddings provide similarities between texts based on certain predefined aspects. Thus, similarity predictions of texts are more targeted to specific requirements and more easily explainable. In this paper, we present AspectCSE, an approach for aspect-based contrastive learning of sentence embeddings. Results indicate that AspectCSE achieves an average improvement of 3.97% on information retrieval tasks across multiple aspects compared to the previous best results. We also propose using Wikidata knowledge graph properties to train models of multi-aspect sentence embeddings in which multiple specific aspects are simultaneously considered during similarity predictions. We demonstrate that multi-aspect embeddings outperform single-aspect embeddings on aspect-specific information retrieval tasks. Finally, we examine the aspect-based sentence embedding space and demonstrate that embeddings of semantically similar aspect labels are often close, even without explicit similarity training between different aspect labels.

Anum Afzal, Ribin Chalumattu, Florian Matthes et al. · eth-zurich

Despite the advances in the abstractive summarization task using Large Language Models (LLM), there is a lack of research that asses their abilities to easily adapt to different domains. We evaluate the domain adaptation abilities of a wide range of LLMs on the summarization task across various domains in both fine-tuning and in-context learning settings. We also present AdaptEval, the first domain adaptation evaluation suite. AdaptEval includes a domain benchmark and a set of metrics to facilitate the analysis of domain adaptation. Our results demonstrate that LLMs exhibit comparable performance in the in-context learning setting, regardless of their parameter scale.

Phillip Schneider, Markus Voggenreiter, Abdullah Gulraiz et al.

Over the last years, software development in domains with high security demands transitioned from traditional methodologies to uniting modern approaches from software development and operations (DevOps). Key principles of DevOps gained more importance and are now applied to security aspects of software development, resulting in the automation of security-enhancing activities. In particular, it is common practice to use automated security testing tools that generate reports after inspecting a software artifact from multiple perspectives. However, this raises the challenge of generating duplicate security findings. To identify these duplicate findings manually, a security expert has to invest resources like time, effort, and knowledge. A partial automation of this process could reduce the analysis effort, encourage DevOps principles, and diminish the chance of human error. In this study, we investigated the potential of applying Natural Language Processing for clustering semantically similar security findings to support the identification of problem-specific duplicate findings. Towards this goal, we developed a web application for annotating and assessing security testing tool reports and published a human-annotated corpus of clustered security findings. In addition, we performed a comparison of different semantic similarity techniques for automatically grouping security findings. Finally, we assess the resulting clusters using both quantitative and qualitative evaluation methods.

Tobias Müller, Milena Zahn, Florian Matthes

The disruptive potential of AI systems roots in the emergence of big data. Yet, a significant portion is scattered and locked in data silos, leaving its potential untapped. Federated Machine Learning is a novel AI paradigm enabling the creation of AI models from decentralized, potentially siloed data. Hence, Federated Machine Learning could technically open data silos and therefore unlock economic potential. However, this requires collaboration between multiple parties owning data silos. Setting up collaborative business models is complex and often a reason for failure. Current literature lacks guidelines on which aspects must be considered to successfully realize collaborative AI projects. This research investigates the challenges of prevailing collaborative business models and distinct aspects of Federated Machine Learning. Through a systematic literature review, focus group, and expert interviews, we provide a systemized collection of socio-technical challenges and an extended Business Model Canvas for the initial viability assessment of collaborative AI projects.

Phillip Schneider, Nils Rehtanz, Kristiina Jokinen et al.

As the healthcare sector is facing major challenges, such as aging populations, staff shortages, and common chronic diseases, delivering high-quality care to individuals has become very difficult. Conversational agents have shown to be a promising technology to alleviate some of these issues. In the form of digital health assistants, they have the potential to improve the everyday life of the elderly and chronically ill people. This includes, for example, medication reminders, routine checks, or social chit-chat. In addition, conversational agents can satisfy the fundamental need of having access to information about daily news or local events, which enables individuals to stay informed and connected with the world around them. However, finding relevant news sources and navigating the plethora of news articles available online can be overwhelming, particularly for those who may have limited technological literacy or health-related impairments. To address this challenge, we propose an innovative solution that combines knowledge graphs and conversational agents for news search in assisted living. By leveraging graph databases to semantically structure news data and implementing an intuitive voice-based interface, our system can help care-dependent people to easily discover relevant news articles and give personalized recommendations. We explain our design choices, provide a system architecture, share insights of an initial user test, and give an outlook on planned future work.

Ibrahim Ebrar Yurt, Fabian Karl, Tejaswi Choppa et al.

Clinical question answering over electronic health records (EHRs) can help clinicians and patients access relevant medical information more efficiently. However, many recent approaches rely on large cloud-based models, which are difficult to deploy in clinical environments due to privacy constraints and computational requirements. In this work, we investigate how far grounded EHR question answering can be pushed when restricted to a single notebook. We participate in all four subtasks of the ArchEHR-QA 2026 shared task and evaluate several approaches designed to run on commodity hardware. All experiments are conducted locally without external APIs or cloud infrastructure. Our results show that such systems can achieve competitive performance on the shared task leaderboards. In particular, our submissions perform above average in two subtasks, and we observe that smaller models can approach the performance of much larger systems when properly configured. These findings suggest that privacy-preserving EHR QA systems running fully locally are feasible with current models and commodity hardware. The source code is available at https://github.com/ibrahimey/ArchEHR-QA-2026.

Tobias Müller, Maximilian Stäbler, Hugo Gascón et al.

Machine Learning (ML) has shown significant potential in various applications; however, its adoption in privacy-critical domains has been limited due to concerns about data privacy. A promising solution to this issue is Federated Machine Learning (FedML), a model-to-data approach that prioritizes data privacy. By enabling ML algorithms to be applied directly to distributed data sources without sharing raw data, FedML offers enhanced privacy protections, making it suitable for privacy-critical environments. Despite its theoretical benefits, FedML has not seen widespread practical implementation. This study aims to explore the current state of applied FedML and identify the challenges hindering its practical adoption. Through a comprehensive systematic literature review, we assess 74 relevant papers to analyze the real-world applicability of FedML. Our analysis focuses on the characteristics and emerging trends of FedML implementations, as well as the motivational drivers and application domains. We also discuss the encountered challenges in integrating FedML into real-life settings. By shedding light on the existing landscape and potential obstacles, this research contributes to the further development and implementation of FedML in privacy-critical scenarios.

Juraj Vladika, Phillip Schneider, Florian Matthes

In the digital age, seeking health advice on the Internet has become a common practice. At the same time, determining the trustworthiness of online medical content is increasingly challenging. Fact-checking has emerged as an approach to assess the veracity of factual claims using evidence from credible knowledge sources. To help advance automated Natural Language Processing (NLP) solutions for this task, in this paper we introduce a novel dataset HealthFC. It consists of 750 health-related claims in German and English, labeled for veracity by medical experts and backed with evidence from systematic reviews and clinical trials. We provide an analysis of the dataset, highlighting its characteristics and challenges. The dataset can be used for NLP tasks related to automated fact-checking, such as evidence retrieval, claim verification, or explanation generation. For testing purposes, we provide baseline systems based on different approaches, examine their performance, and discuss the findings. We show that the dataset is a challenging test bed with a high potential for future use.

Anum Afzal, Juraj Vladika, Daniel Braun et al.

Large Language Models work quite well with general-purpose data and many tasks in Natural Language Processing. However, they show several limitations when used for a task such as domain-specific abstractive text summarization. This paper identifies three of those limitations as research problems in the context of abstractive text summarization: 1) Quadratic complexity of transformer-based models with respect to the input text length; 2) Model Hallucination, which is a model's ability to generate factually incorrect text; and 3) Domain Shift, which happens when the distribution of the model's training and test corpus is not the same. Along with a discussion of the open research questions, this paper also provides an assessment of existing state-of-the-art techniques relevant to domain-specific text summarization to address the research gaps.

Anum Afzal, Yuki Saito, Hiroya Takamura et al.

Real-time video commentary generation provides textual descriptions of ongoing events in videos. It supports accessibility and engagement in domains such as sports, esports, and livestreaming. Commentary generation involves two essential decisions: what to say and when to say it. While recent prompting-based approaches using multimodal large language models (MLLMs) have shown strong performance in content generation, they largely ignore the timing aspect. We investigate whether in-context prompting alone can support real-time commentary generation that is both semantically relevant and well-timed. We propose two prompting-based decoding strategies: 1) a fixed-interval approach, and 2) a novel dynamic interval-based decoding approach that adjusts the next prediction timing based on the estimated duration of the previous utterance. Both methods enable pause-aware generation without any fine-tuning. Experiments on Japanese and English datasets of racing and fighting games show that the dynamic interval-based decoding can generate commentary more closely aligned with human utterance timing and content using prompting alone. We release a multilingual benchmark dataset, trained models, and implementations to support future research on real-time video commentary generation.

Stephen Meisenbacher, Angelo Kleinert, Florian Matthes

The goal of differentially private text obfuscation is to obfuscate, or "perturb", input texts with Differential Privacy (DP) guarantees, such that the private output texts are quantifiably indistinguishable from the originals. While perturbation at the word level is intuitive, meaningful text privatization happens on complete documents. Recent research has laid the groundwork for reasoning about privacy budget distribution, namely, how an overall $\varepsilon$ budget can be sensibly distributed among the component pieces of a text. We perform a systematic evaluation of multiple text decomposition and budget distribution techniques in the context of DP text obfuscation, testing how different methods for chunking texts can be combined with techniques for allocating $\varepsilon$ to these chunks. Our experiments reveal that such design choices are very important, as even with comparable privacy budgets, significantly different results can occur based on which methods are chosen. In this, we provide credible evidence of the feasibility of maximizing empirical trade-offs by optimizing DP obfuscation procedures.

Phillip Schneider, Nils Rehtanz, Kristiina Jokinen et al.

Exploratory search is an open-ended information retrieval process that aims at discovering knowledge about a topic or domain rather than searching for a specific answer or piece of information. Conversational interfaces are particularly suitable for supporting exploratory search, allowing users to refine queries and examine search results through interactive dialogues. In addition to conversational search interfaces, knowledge graphs are also useful in supporting information exploration due to their rich semantic representation of data items. In this study, we demonstrate the synergistic effects of combining knowledge graphs and conversational interfaces for exploratory search, bridging the gap between structured and unstructured information retrieval. To this end, we propose a knowledge-driven dialogue system for exploring news articles by asking natural language questions and using the graph structure to navigate between related topics. Based on a user study with 54 participants, we empirically evaluate the effectiveness of the graph-based exploratory search and discuss design implications for developing such systems.

Anum Afzal, Alexander Kowsik, Rajna Fani et al.

Large Language Models have found application in various mundane and repetitive tasks including Human Resource (HR) support. We worked with the domain experts of SAP SE to develop an HR support chatbot as an efficient and effective tool for addressing employee inquiries. We inserted a human-in-the-loop in various parts of the development cycles such as dataset collection, prompt optimization, and evaluation of generated output. By enhancing the LLM-driven chatbot's response quality and exploring alternative retrieval methods, we have created an efficient, scalable, and flexible tool for HR professionals to address employee inquiries effectively. Our experiments and evaluation conclude that GPT-4 outperforms other models and can overcome inconsistencies in data through internal reasoning capabilities. Additionally, through expert analysis, we infer that reference-free evaluation metrics such as G-Eval and Prometheus demonstrate reliability closely aligned with that of human evaluation.

Stephen Meisenbacher, Nihildev Nandakumar, Alexandra Klymenko et al.

The application of Differential Privacy to Natural Language Processing techniques has emerged in relevance in recent years, with an increasing number of studies published in established NLP outlets. In particular, the adaptation of Differential Privacy for use in NLP tasks has first focused on the $\textit{word-level}$, where calibrated noise is added to word embedding vectors to achieve "noisy" representations. To this end, several implementations have appeared in the literature, each presenting an alternative method of achieving word-level Differential Privacy. Although each of these includes its own evaluation, no comparative analysis has been performed to investigate the performance of such methods relative to each other. In this work, we conduct such an analysis, comparing seven different algorithms on two NLP tasks with varying hyperparameters, including the $\textit{epsilon ($\varepsilon$)}$ parameter, or privacy budget. In addition, we provide an in-depth analysis of the results with a focus on the privacy-utility trade-off, as well as open-source our implementation code for further reproduction. As a result of our analysis, we give insight into the benefits and challenges of word-level Differential Privacy, and accordingly, we suggest concrete steps forward for the research field.

Phillip Schneider, Wessel Poelman, Michael Rovatsos et al.

Conversational search systems enable information retrieval via natural language interactions, with the goal of maximizing users' information gain over multiple dialogue turns. The increasing prevalence of conversational interfaces adopting this search paradigm challenges traditional information retrieval approaches, stressing the importance of better understanding the engineering process of developing these systems. We undertook a systematic literature review to investigate the links between theoretical studies and technical implementations of conversational search systems. Our review identifies real-world application scenarios, system architectures, and functional components. We consolidate our results by presenting a layered architecture framework and explaining the core functions of conversational search systems. Furthermore, we reflect on our findings in light of the rapid progress in large language models, discussing their capabilities, limitations, and directions for future research.

Stephen Meisenbacher, Maulik Chevli, Florian Matthes

The study of privacy-preserving Natural Language Processing (NLP) has gained rising attention in recent years. One promising avenue studies the integration of Differential Privacy in NLP, which has brought about innovative methods in a variety of application settings. Of particular note are $\textit{word-level Metric Local Differential Privacy (MLDP)}$ mechanisms, which work to obfuscate potentially sensitive input text by performing word-by-word $\textit{perturbations}$. Although these methods have shown promising results in empirical tests, there are two major drawbacks: (1) the inevitable loss of utility due to addition of noise, and (2) the computational expensiveness of running these mechanisms on high-dimensional word embeddings. In this work, we aim to address these challenges by proposing $\texttt{1-Diffractor}$, a new mechanism that boasts high speedups in comparison to previous mechanisms, while still demonstrating strong utility- and privacy-preserving capabilities. We evaluate $\texttt{1-Diffractor}$ for utility on several NLP tasks, for theoretical and task-based privacy, and for efficiency in terms of speed and memory. $\texttt{1-Diffractor}$ shows significant improvements in efficiency, while still maintaining competitive utility and privacy scores across all conducted comparative tests against previous MLDP mechanisms. Our code is made available at: https://github.com/sjmeis/Diffractor.

Andreea-Elena Bodea, Stephen Meisenbacher, Florian Matthes

Despite the considerable promise of Retrieval-Augmented Generation (RAG), many real-world use cases may create privacy concerns, where the purported utility of RAG-enabled insights comes at the risk of exposing private information to either the LLM or the end user requesting the response. As a potential mitigation, using anonymization techniques to remove personally identifiable information (PII) and other sensitive markers in the underlying data represents a practical and sensible course of action for RAG administrators. Despite a wealth of literature on the topic, no works consider the placement of anonymization along the RAG pipeline, i.e., asking the question, where should anonymization happen? In this case study, we systematically and empirically measure the impact of anonymization at two important points along the RAG pipeline: the dataset and generated answer. We show that differences in privacy-utility trade-offs can be observed depending on where anonymization took place, demonstrating the significance of privacy risk mitigation placement in RAG.

Phillip Schneider, Nektarios Machner, Kristiina Jokinen et al.

Knowledge models are fundamental to dialogue systems for enabling conversational interactions, which require handling domain-specific knowledge. Ensuring effective communication in information-providing conversations entails aligning user understanding with the knowledge available to the system. However, dialogue systems often face challenges arising from semantic inconsistencies in how information is expressed in natural language compared to how it is represented within the system's internal knowledge. To address this problem, we study the potential of large language models for conversational grounding, a mechanism to bridge information gaps by establishing shared knowledge between dialogue participants. Our approach involves annotating human conversations across five knowledge domains to create a new dialogue corpus called BridgeKG. Through a series of experiments on this dataset, we empirically evaluate the capabilities of large language models in classifying grounding acts and identifying grounded information items within a knowledge graph structure. Our findings offer insights into how these models use in-context learning for conversational grounding tasks and common prediction errors, which we illustrate with examples from challenging dialogues. We discuss how the models handle knowledge graphs as a semantic layer between unstructured dialogue utterances and structured information items.

Stephen Meisenbacher, Tim Schopf, Weixin Yan et al.

The task of $\textit{keyword extraction}$ is often an important initial step in unsupervised information extraction, forming the basis for tasks such as topic modeling or document classification. While recent methods have proven to be quite effective in the extraction of keywords, the identification of $\textit{class-specific}$ keywords, or only those pertaining to a predefined class, remains challenging. In this work, we propose an improved method for class-specific keyword extraction, which builds upon the popular $\textbf{KeyBERT}$ library to identify only keywords related to a class described by $\textit{seed keywords}$. We test this method using a dataset of German business registry entries, where the goal is to classify each business according to an economic sector. Our results reveal that our method greatly improves upon previous approaches, setting a new standard for $\textit{class-specific}$ keyword extraction.

Anum Afzal, Juraj Vladika, Gentrit Fazlija et al.

Given the growing trend of many organizations integrating Retrieval Augmented Generation (RAG) into their operations, we assess RAG on domain-specific data and test state-of-the-art models across various optimization techniques. We incorporate four optimizations; Multi-Query, Child-Parent-Retriever, Ensemble Retriever, and In-Context-Learning, to enhance the functionality and performance in the academic domain. We focus on data retrieval, specifically targeting various study programs at a large technical university. We additionally introduce a novel evaluation approach, the RAG Confusion Matrix designed to assess the effectiveness of various configurations within the RAG framework. By exploring the integration of both open-source (e.g., Llama2, Mistral) and closed-source (GPT-3.5 and GPT-4) Large Language Models, we offer valuable insights into the application and optimization of RAG frameworks in domain-specific contexts. Our experiments show a significant performance increase when including multi-query in the retrieval phase.

Andreea-Elena Bodea, Stephen Meisenbacher, Alexandra Klymenko et al.

The continued promise of Large Language Models (LLMs), particularly in their natural language understanding and generation capabilities, has driven a rapidly increasing interest in identifying and developing LLM use cases. In an effort to complement the ingrained "knowledge" of LLMs, Retrieval-Augmented Generation (RAG) techniques have become widely popular. At its core, RAG involves the coupling of LLMs with domain-specific knowledge bases, whereby the generation of a response to a user question is augmented with contextual and up-to-date information. The proliferation of RAG has sparked concerns about data privacy, particularly with the inherent risks that arise when leveraging databases with potentially sensitive information. Numerous recent works have explored various aspects of privacy risks in RAG systems, from adversarial attacks to proposed mitigations. With the goal of surveying and unifying these works, we ask one simple question: What are the privacy risks in RAG, and how can they be measured and mitigated? To answer this question, we conduct a systematic literature review of RAG works addressing privacy, and we systematize our findings into a comprehensive set of privacy risks, mitigation techniques, and evaluation strategies. We supplement these findings with two primary artifacts: a Taxonomy of RAG Privacy Risks and a RAG Privacy Process Diagram. Our work contributes to the study of privacy in RAG not only by conducting the first systematization of risks and mitigations, but also by uncovering important considerations when mitigating privacy risks in RAG systems and assessing the current maturity of proposed mitigations.

Stephen Meisenbacher, Florian Matthes

Recent work in Differential Privacy with Natural Language Processing (DP NLP) has proposed numerous promising techniques in the form of text rewriting mechanisms. In the evaluation of these mechanisms, an often-ignored aspect is that of dataset size, or rather, the effect of dataset size on a mechanism's efficacy for utility and privacy preservation. In this work, we are the first to introduce this factor in the evaluation of DP text privatization, where we design utility and privacy tests on large-scale datasets with dynamic split sizes. We run these tests on datasets of varying size with up to one million texts, and we focus on quantifying the effect of increasing dataset size on the privacy-utility trade-off. Our findings reveal that dataset size plays an integral part in evaluating DP text rewriting mechanisms; additionally, these findings call for more rigorous evaluation procedures in DP NLP, as well as shed light on the future of DP NLP in practice and at scale.

Oliver Wardas, Florian Matthes

Legal work, characterized by its text-heavy and resource-intensive nature, presents unique challenges and opportunities for NLP research. While data-driven approaches have advanced the field, their lack of interpretability and trustworthiness limits their applicability in dynamic legal environments. To address these issues, we collaborated with legal experts to extend an existing dataset and explored the use of Large Language Models (LLMs) and in-context learning to evaluate the legality of clauses in German employment contracts. Our work evaluates the ability of different LLMs to classify clauses as "valid," "unfair," or "void" under three legal context variants: no legal context, full-text sources of laws and court rulings, and distilled versions of these (referred to as examination guidelines). Results show that full-text sources moderately improve performance, while examination guidelines significantly enhance recall for void clauses and weighted F1-Score, reaching 80\%. Despite these advancements, LLMs' performance when using full-text sources remains substantially below that of human lawyers. We contribute an extended dataset, including examination guidelines, referenced legal sources, and corresponding annotations, alongside our code and all log files. Our findings highlight the potential of LLMs to assist lawyers in contract legality review while also underscoring the limitations of the methods presented.

Stephen Meisenbacher, Maulik Chevli, Juraj Vladika et al.

The task of text privatization using Differential Privacy has recently taken the form of $\textit{text rewriting}$, in which an input text is obfuscated via the use of generative (large) language models. While these methods have shown promising results in the ability to preserve privacy, these methods rely on autoregressive models which lack a mechanism to contextualize the private rewriting process. In response to this, we propose $\textbf{DP-MLM}$, a new method for differentially private text rewriting based on leveraging masked language models (MLMs) to rewrite text in a semantically similar $\textit{and}$ obfuscated manner. We accomplish this with a simple contextualization technique, whereby we rewrite a text one token at a time. We find that utilizing encoder-only MLMs provides better utility preservation at lower $\varepsilon$ levels, as compared to previous methods relying on larger models with a decoder. In addition, MLMs allow for greater customization of the rewriting mechanism, as opposed to generative approaches. We make the code for $\textbf{DP-MLM}$ public and reusable, found at https://github.com/sjmeis/DPMLM .

Tim Schopf, Florian Matthes

Scientific literature searches are often exploratory, whereby users are not yet familiar with a particular field or concept but are interested in learning more about it. However, existing systems for scientific literature search are typically tailored to keyword-based lookup searches, limiting the possibilities for exploration. We propose NLP-KG, a feature-rich system designed to support the exploration of research literature in unfamiliar natural language processing (NLP) fields. In addition to a semantic search, NLP-KG allows users to easily find survey papers that provide a quick introduction to a field of interest. Further, a Fields of Study hierarchy graph enables users to familiarize themselves with a field and its related areas. Finally, a chat interface allows users to ask questions about unfamiliar concepts or specific articles in NLP and obtain answers grounded in knowledge retrieved from scientific publications. Our system provides users with comprehensive exploration possibilities, supporting them in investigating the relationships between different fields, understanding unfamiliar concepts in NLP, and finding relevant research literature. Demo, video, and code are available at: https://github.com/NLP-Knowledge-Graph/NLP-KG-WebApp.

Gonzalo Munilla Garrido, Joseph Near, Aitsam Muhammad et al.

An increasing number of open-source libraries promise to bring differential privacy to practice, even for non-experts. This paper studies five libraries that offer differentially private analytics: Google DP, SmartNoise, diffprivlib, diffpriv, and Chorus. We compare these libraries qualitatively (capabilities, features, and maturity) and quantitatively (utility and scalability) across four analytics queries (count, sum, mean, and variance) executed on synthetic and real-world datasets. We conclude that these libraries provide similar utility (except in some notable scenarios). However, there are significant differences in the features provided, and we find that no single library excels in all areas. Based on our results, we provide guidance for practitioners to help in choosing a suitable library, guidance for library designers to enhance their software, and guidance for researchers on open challenges in differential privacy tools for non-experts.

Ahmed Elnaggar, Wei Ding, Llion Jones et al.

Currently, a growing number of mature natural language processing applications make people's life more convenient. Such applications are built by source code - the language in software engineering. However, the applications for understanding source code language to ease the software engineering process are under-researched. Simultaneously, the transformer model, especially its combination with transfer learning, has been proven to be a powerful technique for natural language processing tasks. These breakthroughs point out a promising direction for process source code and crack software engineering tasks. This paper describes CodeTrans - an encoder-decoder transformer model for tasks in the software engineering domain, that explores the effectiveness of encoder-decoder transformer models for six software engineering tasks, including thirteen sub-tasks. Moreover, we have investigated the effect of different training strategies, including single-task learning, transfer learning, multi-task learning, and multi-task learning with fine-tuning. CodeTrans outperforms the state-of-the-art models on all the tasks. To expedite future works in the software engineering domain, we have published our pre-trained models of CodeTrans. https://github.com/agemagician/CodeTrans

Lisa C. Adams, Linus Marx, Erik Thiele Orberg et al.

Question: Does atomic fact-checking, which decomposes AI treatment recommendations into individually verifiable claims linked to source guideline documents, increase clinician trust compared to traditional explainability approaches? Findings: In this randomized trial of 356 clinicians generating 7,476 trust ratings, atomic fact-checking produced a large effect on trust (Cohen's d = 0.94), increasing the proportion of clinicians expressing trust from 26.9% to 66.5%. Traditional transparency mechanisms showed a dose-response gradient of improvement over baseline (d = 0.25 to 0.50). Meaning: Decomposing AI recommendations into individually verifiable claims linked to source guidelines produces substantially higher clinician trust than traditional explainability approaches in high-stakes clinical decisions.

Juraj Vladika, Florian Matthes

In today's digital world, seeking answers to health questions on the Internet is a common practice. However, existing question answering (QA) systems often rely on using pre-selected and annotated evidence documents, thus making them inadequate for addressing novel questions. Our study focuses on the open-domain QA setting, where the key challenge is to first uncover relevant evidence in large knowledge bases. By utilizing the common retrieve-then-read QA pipeline and PubMed as a trustworthy collection of medical research documents, we answer health questions from three diverse datasets. We modify different retrieval settings to observe their influence on the QA pipeline's performance, including the number of retrieved documents, sentence selection process, the publication year of articles, and their number of citations. Our results reveal that cutting down on the amount of retrieved documents and favoring more recent and highly cited documents can improve the final macro F1 score up to 10%. We discuss the results, highlight interesting examples, and outline challenges for future research, like managing evidence disagreement and crafting user-friendly explanations.

Juraj Vladika, Ivana Hacajová, Florian Matthes

Fact verification (FV) aims to assess the veracity of a claim based on relevant evidence. The traditional approach for automated FV includes a three-part pipeline relying on short evidence snippets and encoder-only inference models. More recent approaches leverage the multi-turn nature of LLMs to address FV as a step-by-step problem where questions inquiring additional context are generated and answered until there is enough information to make a decision. This iterative method makes the verification process rational and explainable. While these methods have been tested for encyclopedic claims, exploration on domain-specific and realistic claims is missing. In this work, we apply an iterative FV system on three medical fact-checking datasets and evaluate it with multiple settings, including different LLMs, external web search, and structured reasoning using logic predicates. We demonstrate improvements in the final performance over traditional approaches and the high potential of step-by-step FV systems for domain-specific claims.

Phillip Schneider, Manuel Klettner, Elena Simperl et al.

Generating natural language text from graph-structured data is essential for conversational information seeking. Semantic triples derived from knowledge graphs can serve as a valuable source for grounding responses from conversational agents by providing a factual basis for the information they communicate. This is especially relevant in the context of large language models, which offer great potential for conversational interaction but are prone to hallucinating, omitting, or producing conflicting information. In this study, we conduct an empirical analysis of conversational large language models in generating natural language text from semantic triples. We compare four large language models of varying sizes with different prompting techniques. Through a series of benchmark experiments on the WebNLG dataset, we analyze the models' performance and identify the most common issues in the generated predictions. Our findings show that the capabilities of large language models in triple verbalization can be significantly improved through few-shot prompting, post-processing, and efficient fine-tuning techniques, particularly for smaller models that exhibit lower zero-shot performance.

Juraj Vladika, Florian Matthes

Retrieval-augmented generation (RAG) has emerged as an approach to augment large language models (LLMs) by reducing their reliance on static knowledge and improving answer factuality. RAG retrieves relevant context snippets and generates an answer based on them. Despite its increasing industrial adoption, systematic exploration of RAG components is lacking, particularly regarding the ideal size of provided context, and the choice of base LLM and retrieval method. To help guide development of robust RAG systems, we evaluate various context sizes, BM25 and semantic search as retrievers, and eight base LLMs. Moving away from the usual RAG evaluation with short answers, we explore the more challenging long-form question answering in two domains, where a good answer has to utilize the entire context. Our findings indicate that final QA performance improves steadily with up to 15 snippets but stagnates or declines beyond that. Finally, we show that different general-purpose LLMs excel in the biomedical domain than the encyclopedic one, and that open-domain evidence retrieval in large corpora is challenging.

Juraj Vladika, Florian Matthes

The increasing rate at which scientific knowledge is discovered and health claims shared online has highlighted the importance of developing efficient fact-checking systems for scientific claims. The usual setting for this task in the literature assumes that the documents containing the evidence for claims are already provided and annotated or contained in a limited corpus. This renders the systems unrealistic for real-world settings where knowledge sources with potentially millions of documents need to be queried to find relevant evidence. In this paper, we perform an array of experiments to test the performance of open-domain claim verification systems. We test the final verdict prediction of systems on four datasets of biomedical and health claims in different settings. While keeping the pipeline's evidence selection and verdict prediction parts constant, document retrieval is performed over three common knowledge sources (PubMed, Wikipedia, Google) and using two different information retrieval techniques. We show that PubMed works better with specialized biomedical claims, while Wikipedia is more suited for everyday health concerns. Likewise, BM25 excels in retrieval precision, while semantic search in recall of relevant evidence. We discuss the results, outline frequent retrieval patterns and challenges, and provide promising future directions.

Phillip Schneider, Manuel Klettner, Kristiina Jokinen et al.

Conversational question answering systems often rely on semantic parsing to enable interactive information retrieval, which involves the generation of structured database queries from a natural language input. For information-seeking conversations about facts stored within a knowledge graph, dialogue utterances are transformed into graph queries in a process that is called knowledge-based conversational question answering. This paper evaluates the performance of large language models that have not been explicitly pre-trained on this task. Through a series of experiments on an extensive benchmark dataset, we compare models of varying sizes with different prompting techniques and identify common issue types in the generated output. Our results demonstrate that large language models are capable of generating graph queries from dialogues, with significant improvements achievable through few-shot prompting and fine-tuning techniques, especially for smaller models that exhibit lower zero-shot performance.

Anum Afzal, Florian Matthes, Gal Chechik et al.

We investigate whether the success of a zero-shot Chain-of-Thought (CoT) process can be predicted before completion. We discover that a probing classifier, based on LLM representations, performs well \emph{even before a single token is generated}, suggesting that crucial information about the reasoning process is already present in the initial steps representations. In contrast, a strong BERT-based baseline, which relies solely on the generated tokens, performs worse, likely because it depends on shallow linguistic cues rather than deeper reasoning dynamics. Surprisingly, using later reasoning steps does not always improve classification. When additional context is unhelpful, earlier representations resemble later ones more, suggesting LLMs encode key information early. This implies reasoning can often stop early without loss. To test this, we conduct early stopping experiments, showing that truncating CoT reasoning still improves performance over not using CoT at all, though a gap remains compared to full reasoning. However, approaches like supervised learning or reinforcement learning designed to shorten CoT chains could leverage our classifier's guidance to identify when early stopping is effective. Our findings provide insights that may support such methods, helping to optimize CoT's efficiency while preserving its benefits.

Stephen Meisenbacher, Maulik Chevli, Florian Matthes

The field of text privatization often leverages the notion of $\textit{Differential Privacy}$ (DP) to provide formal guarantees in the rewriting or obfuscation of sensitive textual data. A common and nearly ubiquitous form of DP application necessitates the addition of calibrated noise to vector representations of text, either at the data- or model-level, which is governed by the privacy parameter $\varepsilon$. However, noise addition almost undoubtedly leads to considerable utility loss, thereby highlighting one major drawback of DP in NLP. In this work, we introduce a new sentence infilling privatization technique, and we use this method to explore the effect of noise in DP text rewriting. We empirically demonstrate that non-DP privatization techniques excel in utility preservation and can find an acceptable empirical privacy-utility trade-off, yet cannot outperform DP methods in empirical privacy protections. Our results highlight the significant impact of noise in current DP rewriting mechanisms, leading to a discussion of the merits and challenges of DP in NLP, as well as the opportunities that non-DP methods present.

Juraj Vladika, Stephen Meisenbacher, Martina Preis et al.

In recent years, the field of Legal Tech has risen in prevalence, as the Natural Language Processing (NLP) and legal disciplines have combined forces to digitalize legal processes. Amidst the steady flow of research solutions stemming from the NLP domain, the study of use cases has fallen behind, leading to a number of innovative technical methods without a place in practice. In this work, we aim to build a structured overview of Legal Tech use cases, grounded in NLP literature, but also supplemented by voices from legal practice in Germany. Based upon a Systematic Literature Review, we identify seven categories of NLP technologies for the legal domain, which are then studied in juxtaposition to 22 legal use cases. In the investigation of these use cases, we identify 15 ethical, legal, and social aspects (ELSA), shedding light on the potential concerns of digitally transforming the legal domain.

Johannes Kirmayr, Lukas Stappen, Phillip Schneider et al.

In today's assistant landscape, personalisation enhances interactions, fosters long-term relationships, and deepens engagement. However, many systems struggle with retaining user preferences, leading to repetitive user requests and disengagement. Furthermore, the unregulated and opaque extraction of user preferences in industry applications raises significant concerns about privacy and trust, especially in regions with stringent regulations like Europe. In response to these challenges, we propose a long-term memory system for voice assistants, structured around predefined categories. This approach leverages Large Language Models to efficiently extract, store, and retrieve preferences within these categories, ensuring both personalisation and transparency. We also introduce a synthetic multi-turn, multi-session conversation dataset (CarMem), grounded in real industry data, tailored to an in-car voice assistant setting. Benchmarked on the dataset, our system achieves an F1-score of .78 to .95 in preference extraction, depending on category granularity. Our maintenance strategy reduces redundant preferences by 95% and contradictory ones by 92%, while the accuracy of optimal retrieval is at .87. Collectively, the results demonstrate the system's suitability for industrial applications.

Anum Afzal, Mehul Kumawat, Florian Matthes

Large Language Models (LLMs), being generic task solvers, are versatile. However, despite the vast amount of data they are trained on, there are speculations about their adaptation capabilities to a new domain. Additionally, the simple fine-tuning of the model to incorporate knowledge of a new domain is computationally expensive and time-consuming. This becomes more challenging when the domain in question is also low-resource, and labeled data is unavailable. We leverage parameter-efficient fine-tuning techniques (PEFTs) on high-resource datasets to address these challenges to improve performance on unseen low-resource domains. Throughout our experiments, we evaluate whether intrinsic linguistic commonalities between datasets can be leveraged for efficient domain adaptation. We benchmark six PEFTs with \texttt{Llama-3-8B-Instruct} on 14 training datasets from the Scientific, Medical, Legal, and News domains for a Text Summarization task. Our experiments show that for low-resource domains, inference using Within-Domain Adapters can achieve better performance than Few-Shot as well as a much larger \texttt{Llama-3-70B-Instruct}. Lastly, in the absence of Within-Domain Adapters, we explore the concept of using Cross-Domain Adapters as well as the strategic combinations of adapters to leverage intrinsic language similarities across domains, facilitating better adaptability and performance in low-resource settings.

Stephen Meisenbacher, Alexandra Klymenko, Florian Matthes

Despite advances in the field of privacy-preserving Natural Language Processing (NLP), a significant challenge remains the accurate evaluation of privacy. As a potential solution, using LLMs as a privacy evaluator presents a promising approach $\unicode{x2013}$ a strategy inspired by its success in other subfields of NLP. In particular, the so-called $\textit{LLM-as-a-Judge}$ paradigm has achieved impressive results on a variety of natural language evaluation tasks, demonstrating high agreement rates with human annotators. Recognizing that privacy is both subjective and difficult to define, we investigate whether LLM-as-a-Judge can also be leveraged to evaluate the privacy sensitivity of textual data. Furthermore, we measure how closely LLM evaluations align with human perceptions of privacy in text. Resulting from a study involving 10 datasets, 13 LLMs, and 677 human survey participants, we confirm that privacy is indeed a difficult concept to measure empirically, exhibited by generally low inter-human agreement rates. Nevertheless, we find that LLMs can accurately model a global human privacy perspective, and through an analysis of human and LLM reasoning patterns, we discuss the merits and limitations of LLM-as-a-Judge for privacy evaluation in textual data. Our findings pave the way for exploring the feasibility of LLMs as privacy evaluators, addressing a core challenge in solving pressing privacy issues with innovative technical solutions.