arXiv subjects

Fabian M. Suchanek

Publications and source records attributed to Fabian M. Suchanek.

17 recordsLinked to original sources

arXiv2026

LELA: an LLM-based Entity Linking Approach with Zero-Shot Domain Adaptation

Entity linking (mapping ambiguous mentions in text to entities in a knowledge base) is a foundational step in tasks such as knowledge graph construction, question-answering, and information extraction. Our method, LELA, is a modular coarse-to-fine approach that leverages the capabilities of large language models (LLMs), and works with different target domains, knowledge bases and LLMs, without any fine-tuning phase. Our experiments across various entity linking settings show that LELA is highly competitive with fine-tuned approaches, and substantially outperforms the non-fine-tuned ones.

cs.CL↗

arXiv2026

KARLA: Knowledge-base Augmented Retrieval for Language Models

We propose a new method that allows an LLM to automatically pull in factual knowledge from a knowledge base during token generation. This means that (1)~factual knowledge in the LLM output can be updated without retraining the LLM, (2)~facts in the LLM output can be traced to the knowledge base for transparency and explainability, and (3)~smaller models can achieve the same factual accuracy as larger models. Our core idea is to train the model to produce special tokens that trigger a query to the knowledge base. Our experiments show that our method improves factual grounding in both short and long-form generation, and allows factual revisions to take effect through KB edits rather than parameter updates.

cs.AI↗

arXiv2026

Query-Level Uncertainty in Large Language Models

It is important for Large Language Models (LLMs) to be aware of the boundary of their knowledge, distinguishing queries they can confidently answer from those that lie beyond their capabilities. Such awareness enables models to perform adaptive inference, such as invoking retrieval-augmented generation (RAG), engaging in slow and deep thinking, or abstaining from answering when appropriate. These mechanisms are key to developing efficient and trustworthy AI. In this work, we propose a method to detect knowledge boundaries via Query-Level Uncertainty, which estimates if a model is capable of answering a given query before generating any tokens, thus avoiding the generation cost. To this end, we propose a novel, training-free method called Internal Confidence, which leverages self-evaluations across layers and tokens to provide a reliable signal of uncertainty. Empirical studies on both factual question answering and mathematical reasoning tasks demonstrate that our Internal Confidence outperforms several baselines in quality of confidence while being computationally cheaper. Furthermore, we demonstrate its benefits in adaptive inference settings, showing that for RAG and model cascading it reduces inference costs while preserving overall performance.

cs.CL↗

arXiv2026

Detecting Greenwashing: A Natural Language Processing Literature Survey

Greenwashing refers to practices by corporations or governments that intentionally mislead the public about their environmental impact. This paper provides a comprehensive and methodologically grounded survey of natural language processing (NLP) approaches for detecting greenwashing in textual data, with a focus on corporate climate communication. Rather than treating greenwashing as a single, monolithic task, we examine the set of NLP problems, also known as climate NLP tasks, that researchers have used to approximate it, ranging from climate topic detection to the identification of deceptive communication patterns. Our focus is on the methodological foundations of these approaches: how tasks are formulated, how datasets are constructed, and how model evaluation influences reliability. Our review reveals a fragmented landscape: several subtasks now exhibit near-perfect performance under controlled settings, yet tasks involving ambiguity, subjectivity, or reasoning remain challenging. Crucially, no dataset of verified greenwashing cases currently exists. We argue that advancing automated greenwashing detection requires principled NLP methodologies that combine reliable data annotations with interpretable model design. Future work should leverage third-party judgments, such as verified media reports or regulatory records, to mitigate annotation subjectivity and legal risk, and adopt decomposed pipelines that support human oversight, traceable reasoning, and efficient model design.

cs.CL↗

arXiv2025

Large Language Models as Search Engines: Societal Challenges

Large Language Models (LLMs) may one day replace search engines as the primary portal to information on the Web. In this article, we investigate the societal challenges that such a change could bring. We focus on the roles of LLM Providers, Content Creators, and End Users, and identify 15 types of challenges. With each, we show current mitigation strategies -- both from the technical perspective and the legal perspective. We also discuss the impact of each challenge and point out future research opportunities.

cs.CY↗

arXiv2025

FLORA: Unsupervised Knowledge Graph Alignment by Fuzzy Logic

Knowledge graph alignment is the task of matching equivalent entities (that is, instances and classes) and relations across two knowledge graphs. Most existing methods focus on pure entity-level alignment, computing the similarity of entities in some embedding space. They lack interpretable reasoning and need training data to work. In this paper, we propose FLORA, a simple yet effective method that (1) is unsupervised, i.e., does not require training data, (2) provides a holistic alignment for entities and relations iteratively, (3) is based on fuzzy logic and thus delivers interpretable results, (4) provably converges, (5) allows dangling entities, i.e., entities without a counterpart in the other KG, and (6) achieves state-of-the-art results on major benchmarks.

cs.AI↗

arXiv2024

Reconfidencing LLMs from the Grouping Loss Perspective

Large Language Models (LLMs), including ChatGPT and LLaMA, are susceptible to generating hallucinated answers in a confident tone. While efforts to elicit and calibrate confidence scores have proven useful, recent findings show that controlling uncertainty must go beyond calibration: predicted scores may deviate significantly from the actual posterior probabilities due to the impact of grouping loss. In this work, we construct a new evaluation dataset derived from a knowledge base to assess confidence scores given to answers of Mistral and LLaMA. Experiments show that they tend to be overconfident. Further, we show that they are more overconfident on some answers than others, \emph{eg} depending on the nationality of the person in the query. In uncertainty-quantification theory, this is grouping loss. To address this, we propose a solution to reconfidence LLMs, canceling not only calibration but also grouping loss. The LLMs, after the reconfidencing process, indicate improved confidence alignment with the accuracy of their responses.

cs.CL↗

arXiv2024

Do Language Models Enjoy Their Own Stories? Prompting Large Language Models for Automatic Story Evaluation

Storytelling is an integral part of human experience and plays a crucial role in social interactions. Thus, Automatic Story Evaluation (ASE) and Generation (ASG) could benefit society in multiple ways, but they are challenging tasks which require high-level human abilities such as creativity, reasoning and deep understanding. Meanwhile, Large Language Models (LLM) now achieve state-of-the-art performance on many NLP tasks. In this paper, we study whether LLMs can be used as substitutes for human annotators for ASE. We perform an extensive analysis of the correlations between LLM ratings, other automatic measures, and human annotations, and we explore the influence of prompting on the results and the explainability of LLM behaviour. Most notably, we find that LLMs outperform current automatic measures for system-level evaluation but still struggle at providing satisfactory explanations for their answers.

cs.CL↗

arXiv2024

Learning High-Quality and General-Purpose Phrase Representations

Phrase representations play an important role in data science and natural language processing, benefiting various tasks like Entity Alignment, Record Linkage, Fuzzy Joins, and Paraphrase Classification. The current state-of-the-art method involves fine-tuning pre-trained language models for phrasal embeddings using contrastive learning. However, we have identified areas for improvement. First, these pre-trained models tend to be unnecessarily complex and require to be pre-trained on a corpus with context sentences. Second, leveraging the phrase type and morphology gives phrase representations that are both more precise and more flexible. We propose an improved framework to learn phrase representations in a context-free fashion. The framework employs phrase type classification as an auxiliary task and incorporates character-level information more effectively into the phrase representation. Furthermore, we design three granularities of data augmentation to increase the diversity of training samples. Our experiments across a wide range of tasks show that our approach generates superior phrase embeddings compared to previous methods while requiring a smaller model size. [PEARL-small]: https://huggingface.co/Lihuchen/pearl_small; [PEARL-base]: https://huggingface.co/Lihuchen/pearl_base; [Code and Dataset]: https://github.com/tigerchen52/PEARL

cs.CL↗

arXiv2023

The Locality and Symmetry of Positional Encodings

Positional Encodings (PEs) are used to inject word-order information into transformer-based language models. While they can significantly enhance the quality of sentence representations, their specific contribution to language models is not fully understood, especially given recent findings that various positional encodings are insensitive to word order. In this work, we conduct a systematic study of positional encodings in \textbf{Bidirectional Masked Language Models} (BERT-style) , which complements existing work in three aspects: (1) We uncover the core function of PEs by identifying two common properties, Locality and Symmetry; (2) We show that the two properties are closely correlated with the performances of downstream tasks; (3) We quantify the weakness of current PEs by introducing two new probing tasks, on which current PEs perform poorly. We believe that these results are the basis for developing better PEs for transformer-based language models. The code is available at \faGithub~ \url{https://github.com/tigerchen52/locality\_symmetry}

cs.CL↗

arXiv2023

GLADIS: A General and Large Acronym Disambiguation Benchmark

Acronym Disambiguation (AD) is crucial for natural language understanding on various sources, including biomedical reports, scientific papers, and search engine queries. However, existing acronym disambiguation benchmarks and tools are limited to specific domains, and the size of prior benchmarks is rather small. To accelerate the research on acronym disambiguation, we construct a new benchmark named GLADIS with three components: (1) a much larger acronym dictionary with 1.5M acronyms and 6.4M long forms; (2) a pre-training corpus with 160 million sentences; (3) three datasets that cover the general, scientific, and biomedical domains. We then pre-train a language model, \emph{AcroBERT}, on our constructed corpus for general acronym disambiguation, and show the challenges and values of our new benchmark.

cs.CL↗

arXiv2022

Of Human Criteria and Automatic Metrics: A Benchmark of the Evaluation of Story Generation

Research on Automatic Story Generation (ASG) relies heavily on human and automatic evaluation. However, there is no consensus on which human evaluation criteria to use, and no analysis of how well automatic criteria correlate with them. In this paper, we propose to re-evaluate ASG evaluation. We introduce a set of 6 orthogonal and comprehensive human criteria, carefully motivated by the social sciences literature. We also present HANNA, an annotated dataset of 1,056 stories produced by 10 different ASG systems. HANNA allows us to quantitatively evaluate the correlations of 72 automatic metrics with human criteria. Our analysis highlights the weaknesses of current metrics for ASG and allows us to formulate practical recommendations for ASG evaluation.

cs.CL↗

arXiv2022

Imputing Out-of-Vocabulary Embeddings with LOVE Makes Language Models Robust with Little Cost

State-of-the-art NLP systems represent inputs with word embeddings, but these are brittle when faced with Out-of-Vocabulary (OOV) words. To address this issue, we follow the principle of mimick-like models to generate vectors for unseen words, by learning the behavior of pre-trained embeddings using only the surface form of words. We present a simple contrastive learning framework, LOVE, which extends the word representation of an existing pre-trained language model (such as BERT), and makes it robust to OOV with few additional parameters. Extensive evaluations demonstrate that our lightweight model achieves similar or even better performances than prior competitors, both on original datasets and on corrupted variants. Moreover, it can be used in a plug-and-play fashion with FastText and BERT, where it significantly improves their robustness.

cs.CL↗

arXiv2021

A Lightweight Neural Model for Biomedical Entity Linking

Biomedical entity linking aims to map biomedical mentions, such as diseases and drugs, to standard entities in a given knowledge base. The specific challenge in this context is that the same biomedical entity can have a wide range of names, including synonyms, morphological variations, and names with different word orderings. Recently, BERT-based methods have advanced the state-of-the-art by allowing for rich representations of word sequences. However, they often have hundreds of millions of parameters and require heavy computing resources, which limits their applications in resource-limited scenarios. Here, we propose a lightweight neural method for biomedical entity linking, which needs just a fraction of the parameters of a BERT model and much less computing resources. Our method uses a simple alignment layer with attention mechanisms to capture the variations between mention and entity names. Yet, we show that our model is competitive with previous work on standard evaluation benchmarks.

cs.CL↗

arXiv2016

Predicting Completeness in Knowledge Bases

Knowledge bases such as Wikidata, DBpedia, or YAGO contain millions of entities and facts. In some knowledge bases, the correctness of these facts has been evaluated. However, much less is known about their completeness, i.e., the proportion of real facts that the knowledge bases cover. In this work, we investigate different signals to identify the areas where a knowledge base is complete. We show that we can combine these signals in a rule mining approach, which allows us to predict where facts may be missing. We also show that completeness predictions can help other applications such as fact prediction.

cs.DB↗

arXiv2015

Harvesting Entities from the Web Using Unique Identifiers -- IBEX

In this paper we study the prevalence of unique entity identifiers on the Web. These are, e.g., ISBNs (for books), GTINs (for commercial products), DOIs (for documents), email addresses, and others. We show how these identifiers can be harvested systematically from Web pages, and how they can be associated with human-readable names for the entities at large scale. Starting with a simple extraction of identifiers and names from Web pages, we show how we can use the properties of unique identifiers to filter out noise and clean up the extraction result on the entire corpus. The end result is a database of millions of uniquely identified entities of different types, with an accuracy of 73--96% and a very high coverage compared to existing knowledge bases. We use this database to compute novel statistics on the presence of products, people, and other entities on the Web.

cs.DB↗

arXiv2011

PARIS: Probabilistic Alignment of Relations, Instances, and Schema

One of the main challenges that the Semantic Web faces is the integration of a growing number of independently designed ontologies. In this work, we present PARIS, an approach for the automatic alignment of ontologies. PARIS aligns not only instances, but also relations and classes. Alignments at the instance level cross-fertilize with alignments at the schema level. Thereby, our system provides a truly holistic solution to the problem of ontology alignment. The heart of the approach is probabilistic, i.e., we measure degrees of matchings based on probability estimates. This allows PARIS to run without any parameter tuning. We demonstrate the efficiency of the algorithm and its precision through extensive experiments. In particular, we obtain a precision of around 90% in experiments with some of the world's largest ontologies.

cs.DB↗