Matthias Orlikowski, Jiaxin Pei, Paul Röttger, Philipp Cimiano, David Jurgens, Dirk Hovy. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Moritz Altemeyer, Steffen Eger, Johannes Daxenberger, Yanran Chen, Tim Altendorf, Philipp Cimiano, Benjamin Schiller. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

Since the emergence of the field of eXplainable Artificial Intelligence (XAI), a growing number of researchers have argued that XAI should consider insights from the social sciences in order to adapt explanations to the expectations and needs of human users. This has led to the emergence of a field called Social XAI, which is concerned with understanding how explanations are actively shaped in the interaction between a human user and an AI system. Recognizing this turn in XAI toward making XAI systems more “social” by providing explanations that focus on human information needs and incorporating insights from human–human explanatory interactions, in this paper we provide a formal foundation for Social XAI. We do so by proposing novel ontological accounts of the key terms used in Social XAI based on Basic Formal Ontology (BFO). Specifically, we provide novel ontological accounts for explanandum, explanans, understanding, explanation, explainer, explainee, and context. In doing so, we discuss multifaceted entities in Social XAI (having both continuant and occurrent facets; e.g., explanation) and the relationship between understanding and explanation. Additionally, we propose solutions to seemingly paradoxical views on some terms (e.g., social constructivist vs. individual constructivist perspective on explanandum).

Conversational XAI systems often follow a limited question–answering paradigm, handling user queries independently without tracking evolving understanding or adapting explanations. We present CoCoXplain, a conversational XAI agent that treats explanation as a co-constructed process shaped jointly by explainer and explainee. Built on the MAPE-K loop and implemented with interacting LLM agents, CoCoXplain monitors user comprehension, maintains an explicit user model, and adapts explanation strategies dynamically. The design draws on human–human explanation, dialogue models, and socio-cognitive factors, operationalizing co-construction through monitoring and scaffolding. In an online study with 192 participants, CoCoXplain achieved comparable objective understanding to a click-based baseline while significantly improving subjective understanding, perceived co-construction (monitoring, scaffolding, collaborative construction), and social dimensions such as sociability, trust, usability, and user–agent alliance. These results show that structured adaptive architectures enable explanations to evolve with the user, offering more effective and socially attuned XAI than static information delivery.

Link Prediction (LP) approaches based on Language Models (LMs) operate over the labels and descriptions of entities and relations in a Knowledge Graph (KG). Recent approaches have shown that incorporating a local graph neighborhood can improve the LP capabilities of LMs. These approaches usually sample a context from the neighborhood around a query triple randomly, thereby incorporating noise that might hinder the model in making correct predictions.

Federated learning (FL) has emerged as a powerful framework for privacy-preserving machine learning, especially relevant in fields like healthcare, finance, and mobile devices. Despite its success, traditional FL systems have a significant limitation: they rely on a static set of clients, forming a federation at the beginning of the training process, which remains fixed throughout the training cycle, thus limiting their scalability and adaptability in dynamic, data-rich settings. To address this, we introduce the concept of federated learning platforms (FLPs), which extend FL into a dynamic platform where client participation is continuously adapted based on their expected value and strategic incentives. In this paper, we envision FLPs as a natural extension of conventional FL that resemble dynamic, value-driven digital platforms where participants can join or leave the federation at any time. Given this dynamicity of client participation, FLPs are designed to gracefully handle changes in the client pool to uphold their value proposition. In this article, we propose a framework for implementing FLPs, outlining key components such as those for dynamic FLP governance, including client on- and offboarding as well as process monitoring. Furthermore, we demonstrate the practical viability of FLPs through a proof of concept for an exemplary use-case and discuss key challenges related to federation stability, data interoperability, as well as privacy, alongside potential solutions. Finally, we present a roadmap and future research directions, guiding the development of robust and scalable FLPs to drive innovation in FL and data interoperability. • We envision FLPs for dynamic and value-driven participation in federated learning. • We propose a framework for FLPs consisting of components for dynamic FLP governance. • We discuss key challenges like federation stability and data privacy concerns. • We provide a possible roadmap for future research directions on FLPs.

Explanations are pervasive in our lives. Mostly, they occur in dialogical form where an explainer discusses a concept or phenomenon of interest with an explainee. Leaving the explainee with a clear understanding is not straightforward due to the knowledge gap between the two participants. Previous research looked at the interaction of explanation moves, dialogue acts, and topics in successful dialogues with expert explainers. However, daily-life explanations often fail, raising the question of what makes a dialogue successful. In this work, we study explanation dialogues in terms of the interactions between the explainer and explainee and how they correlate with the quality of explanations in terms of a successful understanding on the explainee’s side. In particular, we first construct a corpus of 399 dialogues from the Reddit forum Explain Like I am Five and annotate it for interaction flows and explanation quality. We then analyze the interaction flows, comparing them to those appearing in expert dialogues. Finally, we encode the interaction flows using two language models that can handle long inputs, and we provide empirical evidence for the effectiveness boost gained through the encoding in predicting the success of explanation dialogues.

In recent years, large language models have achieved state-of-the-art performance across various NLP tasks. However, investigations have shown that these models tend to rely on shortcut features, leading to inaccurate predictions and causing the models to be unreliable at generalization to out-of-distribution (OOD) samples. For instance, in the context of relation extraction (RE), we would expect a model to identify the same relation independently of the entities involved in it. For example, consider the sentence “Leonardo da Vinci painted the Mona Lisa” expressing the created(Leonardo_da_Vinci, Mona_Lisa) relation. If we substiute “Leonardo da Vinci” with “Barack Obama”, then the sentence still expresses the created relation. A robust model is supposed to detect the same relation in both cases. In this work, we describe several semantically-motivated strategies to generate adversarial examples by replacing entity mentions and investigate how state-of-the-art RE models perform under pressure. Our analyses show that the performance of these models significantly deteriorates on the modified datasets (avg. of -48.5% in F1), which indicates that these models rely to a great extent on shortcuts, such as surface forms (or patterns therein) of entities, without making full use of the information present in the sentences.

Federated learning (FL) facilitates shared training of machine learning models while maintaining data privacy. Unfortunately, it suffers from data imbalance among participating clients, causing the performance of the shared model to drop. To diminish the negative effects of unfavorable data-specific properties, both algorithm- and data-based approaches seek to make FL more resilient against them. In this regard, data-based approaches prove to be more versatile and require less domain knowledge to be applied efficiently. Hence, they seem particularly suitable for widespread application in various FL environments. Although data-based approaches such as local data sampling have been applied to FL in the past, previous research did not provide a systematic analysis of the potential and limitations of individual data sampling strategies to improve FL. To this end, we (1) identify relevant local data sampling strategies for FL, (2) identify data-specific properties that negatively affect FL performance, and (3) provide a benchmark of local data sampling strategies regarding their effect on model performance, convergence, and training time in synthetic, real-world, and large-scale FL environments. Moreover, we propose and rigorously test a novel method for data sampling in FL that locally optimizes the choice of sampling strategy prior to FL participation. Our results show that FL can benefit from applying local data sampling in terms of performance and convergence rate, especially when data imbalance is high or the number of clients and samples is low. Furthermore, our proposed sampling strategy offers the best trade-off between model performance and training time.

The field of eXplainable Artificial Intelligence (xAI) is increasingly recognizing the need to personalize and interactively adapt the explanation to reflect better users’ explanation needs. While dialogue-based approaches to xAI have been proposed recently, the state-of-the-art in xAI is still characterized by what we call one-shot, non-personalized, and one-way explanations. In contrast, dialogue-based systems that can adapt explanations through interaction with a user promise to be superior to GUI-based or dashboard explanations as they offer a more intuitive way of requesting information. In general, while interactive xAI systems are often evaluated in terms of user satisfaction, there are limited studies that access user’s objective model understanding. This is in particular the case for dialogue-based xAI approaches. In this paper, we close this gap by carrying out controlled experiments within a dialogue framework in which we measure the understanding of users in three phases by asking them to simulate the predictions of the model they are learning about. By this, we can quantify the level of understanding of how the model works, comparing the state before and after the interaction. We further analyze the data to reveal patterns of how the interaction between groups with high vs. low understanding differs. Overall, our work thus contributes to our understanding of the effectiveness of xAI approaches.

Kenneweg S, Deigmöller J, Cimiano P, Eggert J. Benchmarking the Ability of Large Language Models to Reason About Event Sets. In: Proceedings of the 16th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management. SCITEPRESS - Science and Technology Publications; 2024: 74-82.

To produce explanations that are more likely to be accepted by humans, Explainable Artificial Intelligence (XAI) systems need to incorporate explanation models grounded in human communication patterns. So far, little is known about how an explainee, who lacks understanding of an issue, and an explainer, who has knowledge to fill the explainee’s knowledge gap, actively shape an explanation process, and how their involvement relates to explanatory success in terms of maximizing the explainee’s level of understanding. In this paper, we characterize explanations as dialogues in which explainee and explainer take turns to advance the explanation process. We build on an existing annotation scheme of ‘explanatory moves’ to characterize such turns, and manually annotate 362 dialogical explanations from the “Explain Like I’m Five” subreddit. Building on the annotated data, we compute correlations between explanatory moves and explanatory success, measured on a five-point Likert scale, in order to identify factors that are significantly correlated with explanatory success. Based on a qualitative analysis of these factors, we develop a conceptual model of the main factors that contribute to the success of explanatory dialogues.

In recent years, we have witnessed both artificial intelligence obtaining remarkable results in clinical decision support systems (CDSSs) and explainable artificial intelligence (XAI) improving the interpretability of these models. In turn, this fosters the adoption by medical personnel and improves trustworthiness of CDSSs. Among others, counterfactual explanations prove to be one such XAI technique particularly suitable for the healthcare domain due to its ease of interpretation, even for less technically proficient staff. However, the generation of high-quality counterfactuals relies on generative models for guidance. Unfortunately, training such models requires a huge amount of data that is beyond the means of ordinary hospitals. In this paper, we therefore propose to use federated learning to allow multiple hospitals to jointly train such generative models while maintaining full data privacy. We demonstrate the superiority of our approach compared to locally generated counterfactuals. Moreover, we prove that generative models for counterfactual generation that are trained using federated learning in a suitable environment perform only marginally worse compared to centrally trained ones while offering the benefit of data privacy preservation. Finally, we integrate our method into a prototypical CDSS for treatment recommendation for sepsis patients, thus providing a proof of concept for real-world application as well as insights and sanity checks from clinical application. • Limited availability of data limits small hospitals in generating high- quality counterfactual explanations. • Integrating federated learning mitigates this limitation and maintains data privacy. • Benefit of using federated learning depends on the degree of data imbalance among hospitals. • Proof-of-concept for clinical application for sepsis treatment recommendation.

Arguments often do not make explicit how a conclusion follows from its premises.To compensate for this lack, we enrich arguments with structured background knowledge to support knowledge-intense argumentation tasks.We present a new unsupervised method for constructing Contextualized Commonsense Knowledge Graphs (CCKGs) that selects contextually relevant knowledge from large knowledge graphs (KGs) efficiently and at high quality.Our work goes beyond context-insensitive knowledge extraction heuristics by computing semantic similarity between KG triplets and textual arguments.Using these triplet similarities as weights, we extract contextualized knowledge paths that connect a conclusion to its premise, while maximizing similarity to the argument.We combine multiple paths into a CCKG that we optionally prune to reduce noise and raise precision.Intrinsic evaluation of the quality of our graphs shows that our method is effective for (re)constructing human explanation graphs.Manual evaluations in a large-scale knowledge selection setup confirm high recall and precision of implicit CSK in the CCKGs.Finally, we demonstrate the effectiveness of CCKGs in a knowledge-insensitive argument quality rating task, outperforming strong baselines and rivaling a GPT-3 based system. 1

Many annotation tasks in natural language processing are highly subjective in that there can be different valid and justified perspectives on what is a proper label for a given example.This also applies to the judgment of argument quality, where the assignment of a single ground truth is often questionable.At the same time, there are generally accepted concepts behind argumentation that form a common ground.To best represent the interplay of individual and shared perspectives, we consider a continuum of approaches ranging from models that fully aggregate perspectives into a majority label to "share nothing"-architectures in which each annotator is considered in isolation from all other annotators.In between these extremes, inspired by models used in the field of recommender systems, we investigate the extent to which architectures that include layers to model the relations between different annotators are beneficial for predicting single-annotator labels.By means of two tasks of argument quality classification (argument concreteness and validity/novelty of conclusions), we show that recommender architectures increase the averaged annotator-individual F 1 -scores up to 43% over a majority-label model.Our findings indicate that approaches to subjectivity can benefit from relating individual perspectives.

Traditional digital tools for exploring historical data mostly rely on conventional 2D visualizations, which often cannot reveal all relevant interrelationships between historical fragments (e.g., persons or events). In this paper, we present a novel interactive exploration tool for historical data in VR, which represents fragments as spheres in a 3D environment and arranges them around the user based on their temporal, geo, categorical and semantic similarity. Quantitative and qualitative results from a user study with 29 participants revealed that most participants considered the virtual space and the abstract fragment representation well-suited to explore historical data and to discover complex interrelationships. These results were particularly underlined by high usability scores in terms of attractiveness, stimulation, and novelty, while researching historical facts with our system did not impose unexpectedly high task loads. Additionally, the insights from our post-study interviews provided valuable suggestions for future developments to further expand the possibilities of our system.

Federated Learning (FL) is a learning paradigm that allows clients to profit from the data that is available across multiple clients to train a joint model. As FL allows to train such a joint model without explicitly sharing data, but only sharing model updates, it has attained popularity in healthcare settings where patient data is subject to strict privacy policies and needs to be locally stored at each hospital or healthcare provider. A particular challenge for FL settings is data imbalance across clients, as it has been found to be detrimental to model performance and impact the influence of each client on the learning process. Unfortunately, the healthcare domain is particularly prone to such imbalanced data due to regional differences in disease management, prescription behavior etc. In this paper, we introduce the two novel metrics Benefit and Contribution to quantify to which degree individual clients benefit from participation in FL and how they contribute to its success, respectively. Therefore, we measure Benefit and Contribution with respect to four types of imbalances present in data at each client side. Our results show that both client Benefit and Contribution are influenced by data imbalance in such a way that high imbalance in data quantity, label distribution and feature distribution reduces or nullifies clients’ Benefit while increasing their Contribution. Thus, the most valuable clients within a cohort benefit the least from their participation, exposing a critical thread to the success of clinical FL cohorts by withdrawing participation.

As the number of clinical trials carried out and published worldwide keeps growing, better tools for synthesizing the available knowledge become increasingly important. It still requires a significant effort and expertise to aggregate the evidence and results from different clinical trials, a task that is at the core of secondary or comparative studies, meta-analyses, and (living) systematic reviews. Our hypothesis is that the practical challenges involved in synthesizing evidence can be alleviated if the results of clinical trials would be published in a machine-readable format using a well-defined (semantic) vocabulary. Building on the C-TrO ontology that we developed in earlier work to support the aggregation of evidence from clinical trials as the main use case, in this paper we examine the question whether it is feasible for clinical researchers and medical practitioners to describe the results of clinical trials using the C-TrO ontology. For this purpose, we implemented a Web-based tool called CTrO-Editor that uses a form-based interaction paradigm to allow users to enter all the details regarding study population, arms, endpoints, observations and results of a clinical trial, and that exports the data in an RDF format. We describe the results of the evaluation of the CTrO-Editor with five medical students. Our preliminary results suggest that our paradigm for semantifying clinical trials is feasible, as the students could all successfully model a publication of their choice using our tool within a couple of hours.

There is a well-known lexical gap between content expressed in the form of natural language (NL) texts and content stored in an RDF knowledge base (KB). For tasks such as Information Extraction (IE), this gap needs to be bridged from NL to KB, so that facts extracted from text can be represented in RDF and can then be added to an RDF KB. For tasks such as Natural Language Generation, this gap needs to be bridged from KB to NL, so that facts stored in an RDF KB can be verbalized and read by humans. In this paper we propose LexExMachina, a new methodology that induces correspondences between lexical elements and KB elements by mining class-specific association rules. As an example of such an association rule, consider the rule that predicts that if the text about a person contains the token "Greek", then this person has the relation nationality to the entity Greece. Another rule predicts that if the text about a settlement contains the token "Greek", then this settlement has the relation country to the entity Greece. Such a rule can help in question answering, as it maps an adjective to the relevant KB terms, and it can help in information extraction from text. We propose and empirically investigate a set of 20 types of class-specific association rules together with different interestingness measures to rank them. We apply our method on a loosely-parallel text-data corpus that consists of data from DBpedia and texts from Wikipedia, and evaluate and provide empirical evidence for the utility of the rules for Question Answering.

In template-based information extraction, a template is described by a predefined set of slots that need to be filled with entities mentioned in a text. Compared to traditional relation extraction that aims at classifying binary relations involving a pair of entities only, template-based slot-filling is of higher complexity as interdependencies between slot-filler values need to be taken into account. To model these dependencies, we tackle the slot-filling task as a joint inference problem and build on factorized distributions as used in conditional random fields. Dependencies are often described by textual features only, but in some cases they are of semantic nature and not directly expressed in text. To exploit this potential, we propose the incorporation of semantic dependencies extracted from knowledge graphs into an information extraction model. Dependencies are extracted from the variable bindings of queries executed over a knowledge graph and capture semantic soft constraints that are weighted as part of model training. We evaluate our approach on five distantly supervised labeled datasets extracted from Wikipedia/DBpedia and compare our approach to a most frequent entity baseline as well as a purely textual model. We show that there is an overall positive impact of integrating factual background knowledge in all datasets, yielding an improvement of up to 10 points in F1.

In this paper, we take a first step in exploring the effect of xR devices on the experiences of users in the context of human-agent interactions. We report on the design and heuristic evaluation of an embodied conversational agent integrated into either a Virtual or Mixed Reality environment. Our evaluation with four experts show that Virtual Reality may elicit a better experience from users, while Mixed Reality may evoke a better sense of social presence due to natural user embodiment and spatial references to physical objects in the real environment.

In this paper we describe the contributions made by the European H2020 project “Prêt-à-LLOD” (‘Ready-to-use Multilingual Linked Language Data for Knowledge Services across Sectors’) to the further development of the Linguistic Linked Open Data (LLOD) infrastructure. Prêt-à-LLOD aims to develop a new methodology for building data value chains applicable to a wide range of sectors and applications and based around language resources and language technologies that can be integrated by means of semantic technologies. We describe the methods implemented for increasing the number of language data sets in the LLOD. We also present the approach for ensuring interoperability and for porting LLOD data sets and services to other infrastructures, as well as the contribution of the projects to existing standards.

Matthias Hartung, Fabian Kaupmann, Soufian Jebbara, Philipp Cimiano. Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 1, Long Papers. 2017.

Sentiment analysis can be regarded as a relation extraction problem in which the sentiment of some opinion holder towards a certain aspect of a product, theme or event needs to be extracted. We present a novel neural architecture for sentiment analysis as a relation extraction problem that addresses this problem by dividing it into three subtasks: i) identification of aspect and opinion terms, ii) labeling of opinion terms with a sentiment, and iii) extraction of relations between opinion terms and aspect terms. For each subtask, we propose a neural network based component and combine all of them into a complete system for relational sentiment analysis. The component for aspect and opinion term extraction is a hybrid architecture consisting of a recurrent neural network stacked on top of a convolutional neural network. This approach outperforms a standard convolutional deep neural architecture as well as a recurrent network architecture and performs competitively compared to other methods on two datasets of annotated customer reviews. To extract sentiments for individual opinion terms, we propose a recurrent architecture in combination with word distance features and achieve promising results, outperforming a majority baseline by 18% accuracy and providing the first results for the USAGE dataset. Our relation extraction component outperforms the current state-of-the-art in aspect-opinion relation extraction by 15% F-Measure.

Scarcity of annotated corpora for many languages is a bottleneck for training finegrained sentiment analysis models that can tag aspects and subjective phrases.We propose to exploit statistical machine translation to alleviate the need for training data by projecting annotated data in a source language to a target language such that a supervised fine-grained sentiment analysis system can be trained.To avoid a negative influence of poor-quality translations, we propose a filtering approach based on machine translation quality estimation measures to select only high-quality sentence pairs for projection.We evaluate on the language pair German/English on a corpus of product reviews annotated for both languages and compare to in-target-language training.Projection without any filtering leads to 23 % F 1 in the task of detecting aspect phrases, compared to 41 % F 1 for in-target-language training.Our approach obtains up to 47 % F 1 .Further, we show that the detection of subjective phrases is competitive to in-target-language training without filtering.

Semantic grammars can be applied both as a language model for a speech recognizer and for semantic parsing, e.g. in order to map the output of a speech recognizer into formal meaning representations. Semantic speech recognition grammars are, however, typically created manually or learned in a supervised fashion, requiring extensive manual effort in both cases. Aiming to reduce this effort, in this paper we investigate the induction of semantic speech recognition grammars under weak supervision. We present empirical results, indicating that the induced grammars support semantic parsing of speech with a rather low loss in performance when compared to parsing of input without recognition errors. Further, we show improved parsing performance compared to applying n-gram models as language models and demonstrate how our semantic speech recognition grammars can be enhanced by weights based on occurrence frequencies, yielding an improvement in parsing performance over applying unweighted grammars.

Aiming at the design of adaptive artificial agents which are able to learn autonomously from experience and human tutoring, in this paper we present a system for learning syntactic constructions grounded in perception. These constructions are learned from examples of natural language utterances and parallel performances of actions, i.e. their trajectories and involved objects. From the input, the system learns linguistic structures and qualitative action models. Action models are represented as Hidden Markov Models over sequences of qualitative relations between a trajector and a landmark and abstract away from concrete action trajectories. Learning of action models is driven by linguistic observations, and linguistic patterns are, in turn, grounded in learned action models. The proposed system is applicable for both language understanding and language generation. We present empirical results, showing that the learned action models generalize well over concrete instances of the same action and also to novel performers, while allowing accurate discrimination between different actions. Further, we show that the system is able to describe novel dynamic scenes and to understand novel utterances describing such scenes.

Semantic parsers map natural language input into semantic representations. In this paper, we present an approach that learns a semantic parser in the form of a lexicon and an inventory of syntactic patterns from ambiguous training data which is applicable to spoken utterances. We only assume the availability of a task-independent phoneme recognizer, making it easy to adapt to other tasks and yielding no a priori restriction concerning the vocabulary that the parser can process. In spite of these low requirements, we show that our approach can be successfully applied to both spoken and written data.

Using different sources of information for grammar induction results in grammars that vary in coverage and precision. Fusing such grammars with a strategy that exploits their strengths while minimizing their weaknesses is expected to produce grammars with superior performance. We focus on the fusion of grammars produced using a knowledge-based approach using lexicalized ontologies and a data-driven approach using semantic similarity clustering. We propose various algorithms for finding the map-ping between the (non-terminal) rules generated by each gram-mar induction algorithm, followed by rule fusion. Three fusion approaches are investigated: early, mid and late fusion. Results show that late fusion provides the best relative F-measure per-formance improvement by 20%. Index Terms: spoken dialogue systems, corpus-based grammar induction, ontology-based grammar induction, grammar fusion

Nowadays, digital cameras are very popular among people and quite every mobile phone has a build-in camera. Social events have a prominent role in people's life. Thus, people take pictures of events they take part in and more and more of them upload these to well-known online photo community sites like Flickr. The number of pictures uploaded to these sites is still proliferating and there is a great interest in automatizing the process of event clustering so that every incoming (picture) document can be assigned to the corresponding event without the need of human interaction. These social events are defined as events that are planned by people, attended by people and for which the social multimedia are also captured by people. There is an urgent need to develop algorithms which are capable of grouping media by the social events they depict or are related to. In order to train, test, and evaluate such algorithms and frameworks, we present a dataset that consists of about 430,000 photos from Flickr together with the underlying ground truth consisting of about 21,000 social events. All the photos are accompanied by their textual metadata. The ground truth for the event groupings has been derived from event calendars on the Web that have been created collaboratively by people. The dataset has been used in the Social Event Detection (SED) task that was part of the MediaEval Benchmark for Multimedia Evaluation 2013. This task required participants to discover social events and organize the related media items in event-specific clusters within a collection of Web multimedia documents. In this paper we describe how the dataset has been collected and the creation of the ground truth together with a proposed evaluation methodology and a brief description of the corresponding task challenge as applied in the context of the Social Event Detection task.

Cross-lingual topic modelling has applications in machine translation, word sense disambiguation and terminology alignment.Multilingual extensions of approaches based on latent (LSI), generative (LDA, PLSI) as well as explicit (ESA) topic modelling can induce an interlingual topic space allowing documents in different languages to be mapped into the same space and thus to be compared across languages.In this paper, we present a novel approach that combines latent and explicit topic modelling approaches in the sense that it builds on a set of explicitly defined topics, but then computes latent relations between these.Thus, the method combines the benefits of both explicit and latent topic modelling approaches.We show that on a crosslingual mate retrieval task, our model significantly outperforms LDA, LSI, and ESA, as well as a baseline that translates every word in a document into the target language.

Events play a prominent role in our lives, such that many social media documents describe or are related to some event. Organizing social media documents with respect to events thus seems a promising approach to better manage and organize the ever-increasing amount of content in social media applications. A challenge is to automatize this process so that incoming documents can be assigned to their corresponding event without any user intervention. We present a system that is able to classify a stream of social media data into a growing and evolving set of events. By doing this, we successfully address two key problems that arise in this context: i) scaling to the data sizes and rates encountered in social media applications, and ii) tackling the new event detection problem, i.e. the problem of determining whether an incoming data item belongs to a new or a known event. We successfully address these problems by i) including a candidate retrieval step that retrieves a set of event candidates that the incoming data point is likely to belong to and ii) by including a function trained using machine learning techniques to determine whether the incoming data item belongs to the top scoring candidate or rather to a new event. We show that our system addresses the above mentioned challenging issues successfully and that it outperforms other state-of-the-art approaches in terms of quality and scalability.

As an increasing amount of RDF data is published as Linked Data, intuitive ways of accessing this data become more and more important. Question answering approaches have been proposed as a good compromise between intuitiveness and expressivity. Most question answering systems translate questions into triples which are matched against the RDF data to retrieve an answer, typically relying on some similarity metric. However, in many cases, triples do not represent a faithful representation of the semantic structure of the natural language question, with the result that more expressive queries can not be answered. To circumvent this problem, we present a novel approach that relies on a parse of the question to produce a SPARQL template that directly mirrors the internal structure of the question. This template is then instantiated using statistical entity identification and predicate detection. We show that this approach is competitive and discuss cases of questions that can be answered with our approach but not with competing approaches.

In this paper we introduce online semi-supervised growing neural gas (OSSGNG), a novel online semi-supervised classification approach based on growing neural gas (GNG). Existing semi-supervised classification approaches based on GNG require that the training data is explicitly stored as the labeling is performed a posteriori after the training phase. As main contribution, we present an approach that relies on online labeling and prediction functions to process labeled and unlabeled data uniformly and in an online fashion, without the need to store any of the training examples explicitly. We show that using on-the-fly labeling strategies does not significantly deteriorate the performance of classifiers based on GNG, while circumventing the need to explicitly store training examples. Armed with this result, we then present a semi-supervised extension of GNG (OSSGNG) that relies on the above mentioned online labeling functions to label unlabeled examples and incorporate them into the model on-the-fly. As an important result, we show that OSSGNG performs as good as previous semi-supervised extensions of GNG which rely on offline labeling strategies. We also show that OSSGNG compares favorably to other state-of-the-art semi-supervised learning approaches on standard benchmarking datasets.

In order to support police officers in their daily activities, we have designed a rule-based system which can deliver contextualized information to police officers, thus supporting decision making. In particular, we present a framework that has been designed on the basis of requirements elicited in a previous study, focusing on the rule language and the engine that essentially defines and allows to configure the behaviour of the system. The rules consist of a body which specifies conditions that need to be fulfilled in a certain context. The head of the rules specifies how the relevance ratings of certain information items for specific users need to be updated given that the conditions in the body are met. On the basis of cumulated ratings, the system generates a user- and context specific ranking of information items. Quantitative evaluations in terms of precision and recall with respect to a gold standard determined in cooperation with police officers show that the system can cater for the requirements of our end users and yields reasonable precision and recall values.

Keyword queries enjoy widespread usage as they represent an intuitive way of specifying information needs. Recently, answering keyword queries on graph-structured data has emerged as an important research topic. The prevalent approaches build on dedicated indexing techniques as well as search algorithms aiming at finding substructures that connect the data elements matching the keywords. In this paper, we introduce a novel keyword search paradigm for graph-structured data, focusing in particular on the RDF data model. Instead of computing answers directly as in previous approaches, we first compute queries from the keywords, allowing the user to choose the appropriate query, and finally, process the query using the underlying database engine. Thereby, the full range of database optimization techniques can be leveraged for query processing. For the computation of queries, we propose a novel algorithm for the exploration of top-k matching subgraphs. While related techniques search the best answer trees, our algorithm is guaranteed to compute all k subgraphs with lowest costs, including cyclic graphs. By performing exploration only on a summary data structure derived from the data graph, we achieve promising performance improvements compared to other approaches.

We present a user-centered model for porting natural language interfaces (NLIs) between domains efficiently. The model assumes that domain experts without any background knowledge about computational linguistics will perform the customization of the NLI to a specific domain. In fact, it merely requires familiarity with the underlying knowledge base as well as with a few basic subcategorization types. Our model is iterative in the sense that the adaption of the NLI is performed in several cycles on the basis of the questions which the NLI failed to answer, thus iteratively increasing the coverage of the system. We provide experimental evidence in form of a user study as well as a case study involving a real-world application corroborating that our model is indeed a feasible way of customizing the interface to a certain domain.

Increased availability of mobile computing, such as personal digital assistants (PDAs), creates the potential for constant and intelligent access to up-to-date, integrated and detailed information from the Web, regardless of one's actual geographical position.Intelligent question-answering requires the representation of knowledge from various domains, such as the navigational and discourse context of the user, potential user questions, the information provided by Web services and so on, for example in the form of ontologies.Within the context of the SmartWeb project, we have developed a number of domain-specific ontologies that are relevant for mobile and intelligent user interfaces to open-domain question-answering and information services on the Web.To integrate the various domain-specific ontologies, we have developed a foundational ontology, the SmartSUMO ontology, on the basis of the DOLCE and SUMO ontologies.This allows us to combine all the developed ontologies into a single SmartWeb Integrated Ontology 1 (SWIntO) having a common modeling basis with conceptual clarity and the provision of ontology design patterns for modeling consistency.In this paper, we present SWIntO, describe the design choices we made in its construction, illustrate the use of the ontology through a number of applications, and discuss some of the lessons learned from our experiences.1 http://www.smartweb-project.org 2 SmartWeb is a large international

In this poster we present an approach to query answering over knowledge sources that makes use of different ontology management components within an application scenario of the BT Digital Library. The novelty of the approach lies in the combination of different semantic technologies providing a clear benefit for the application scenario considered.

The realization of the Semantic Web is constrained by a knowledge acquisition bottleneck, i.e. the problem of how to add RDF mark-up to the millions of ordinary web pages that already exist. Information Extraction (IE) has been proposed as a solution to the annotation bottleneck. In the task based evaluation reported here, we compared the performance of users without access to annotation, users working with annotations which had been produced from manually constructed knowledge bases, and users working with annotations augmented using IE. We looked at retrieval performance, overlap between retrieved items and the two sets of annotations, and usage of annotation options. Automatically generated annotations were found to add value to the browsing experience in the scenario investigated.

Without the proliferation of formal semantic annotations, the Semantic Web is certainly doomed to failure. In earlier work we presented a new paradigm to avoid this: the 'Self Annotating Web', in which globally available knowledge is used to annotate resources such as web pages. In particular, we presented a concrete method instantiating this paradigm, called PANKOW (Pattern-based ANnotation through Knowledge On the Web). In PANKOW, a named entity to be annotated is put into several linguistic patterns that convey competing semantic meanings. The patterns that are matched most often on the Web indicate the meaning of the named entity --- leading to automatic or semi-automatic annotation.In this paper we present C-PANKOW (Context-driven PANKOW), which alleviates several shortcomings of PANKOW. First, by downloading abstracts and processing them off-line, we avoid the generation of large number of linguistic patterns and correspondingly large number of Google queries.Second, by linguistically analyzing and normalizing the downloaded abstracts, we increase the coverage of our pattern matching mechanism and overcome several limitations of the earlier pattern generation process. Third, we use the annotation context in order to distinguish the significance of a pattern match for the given annotation task. Our experiments show that C-PANKOW inherits all the advantages of PANKOW (no training required etc.), but in addition it is far more efficient and effective.

We present a novel approach to the automatic acquisition of taxonomies or concept hierarchies from a text corpus. The approach is based on Formal Concept Analysis (FCA), a method mainly used for the analysis of data, i.e. for investigating and processing explicitly given information. We follow Harris' distributional hypothesis and model the context of a certain term as a vector representing syntactic dependencies which are automatically acquired from the text corpus with a linguistic parser. On the basis of this context information, FCA produces a lattice that we convert into a special kind of partial order constituting a concept hierarchy. The approach is evaluated by comparing the resulting concept hierarchies with hand-crafted taxonomies for two domains: tourism and finance. We also directly compare our approach with hierarchical agglomerative clustering as well as with Bi-Section-KMeans as an instance of a divisive clustering algorithm. Furthermore, we investigate the impact of using different measures weighting the contribution of each attribute as well as of applying a particular smoothing technique to cope with data sparseness.

The success of the Semantic Web depends on the availability of ontologies as well as on the proliferation of web pages annotated with metadata conforming to these ontologies. Thus, a crucial question is where to acquire these metadata from. In this paper wepropose PANKOW (Pattern-based Annotation through Knowledge on theWeb), a method which employs an unsupervised, pattern-based approach to categorize instances with regard to an ontology. The approach is evaluated against the manual annotations of two human subjects. The approach is implemented in OntoMat, an annotation tool for the Semantic Web and shows very promising results.