Suzan Verberne

Jia-Huei Ju, Eugene Yang, Trevor Adriaanse et al.

Long-form Retrieval-Augmented Generation (RAG) brings the challenge of coverage-based ranking, because ranking methods must ensure the inclusion of comprehensive relevant nuggets (i.e., facts), which can thereby be synthesized into a comprehensive output. In this work, we propose CoveR (Our code is available at https://github.com/DylanJoo/CoveR ) a dense retrieval method optimized for coverage-aware retrieval scenarios. CoveR is a bi-encoder trained with the coverage-based contrastive and distillation objectives, which enables CoveR to capture diverse aspects of information needs. To train CoveR, we create the SCOPE dataset, (Our training data is available at https://huggingface.co/datasets/DylanJHJ/scope ) which comprises 90K training pairs from Researchy Questions with synthetic coverage signals augmented from sub-question answerability judgments generated by LLMs. Our empirical experiments show that CoveR enhances nugget coverage by 10\% over strong dense retrieval baselines without sacrificing its relevance-based retrieval capability. Further ablation studies validate the importance of our proposed learning method, showing that CoveR achieves a superior trade-off between relevance- and coverage-based ranking, which is essential for long-form RAG.

Angel Paul, Dhivin Shaji, Lifeng Han et al.

The increasing availability of sensitive textual data has created an urgent need for robust de-identification methods that enable compliant data sharing while preserving downstream utility. This paper presents DeID-Clinic, a multi-layered framework for automated pseudonymization and re-identification risk assessment of clinical free-text data. Our approach integrates domain-adapted transformer models, including BioBERT and ClinicalBERT, into the MASK de-identification framework to improve the detection and masking of protected health information (PHI). Beyond entity recognition, we introduce a novel document-level risk assessment module that quantifies residual re-identification risk using a combination of k-anonymity, l-diversity, t-closeness, contextual similarity, and entity co-occurrence analysis. Experiments conducted on the i2b2 2014 de-identification dataset demonstrate strong performance, achieving macro-level F1 scores above 0.96 for several entity categories, while enabling quantitative prioritization of high-risk documents for further review. Our results highlight the effectiveness of combining neural de-identification with explicit risk modeling, supporting privacy-preserving data sharing in sensitive domains. Although evaluated on clinical text, the proposed framework is generalizable to other privacy-critical domains such as legal and administrative documents, where reliable pseudonymization and risk-aware anonymization are essential. Keywords{Automated De-Identification, Risk Assessment, Patient Privacy, Pseudonymization, Personal Health Information}

Yanfang Hou, Peter van der Putten, Suzan Verberne

During the COVID-19 pandemic, large amounts of COVID-19 misinformation are spreading on social media. We are interested in the stance of Twitter users towards COVID-19 misinformation. However, due to the relative recent nature of the pandemic, only a few stance detection datasets fit our task. We have constructed a new stance dataset consisting of 2631 tweets annotated with the stance towards COVID-19 misinformation. In contexts with limited labeled data, we fine-tune our models by leveraging the MNLI dataset and two existing stance detection datasets (RumourEval and COVIDLies), and evaluate the model performance on our dataset. Our experimental results show that the model performs the best when fine-tuned sequentially on the MNLI dataset and the combination of the undersampled RumourEval and COVIDLies datasets. Our code and dataset are publicly available at https://github.com/yanfangh/covid-rumor-stance

Teodor-Călin Ionescu, Lifeng Han, Jan Heijdra Suasnabar et al.

This study investigates the use of neural topic modeling and LLMs to uncover meaningful themes from patient storytelling data, to offer insights that could contribute to more patient-oriented healthcare practices. We analyze a collection of transcribed interviews with cancer patients (132,722 words in 13 interviews). We first evaluate BERTopic and Top2Vec for individual interview summarization by using similar preprocessing, chunking, and clustering configurations to ensure a fair comparison on Keyword Extraction. LLMs (GPT4) are then used for the next step topic labeling. Their outputs for a single interview (I0) are rated through a small-scale human evaluation, focusing on {coherence}, {clarity}, and {relevance}. Based on the preliminary results and evaluation, BERTopic shows stronger performance and is selected for further experimentation using three {clinically oriented embedding} models. We then analyzed the full interview collection with the best model setting. Results show that domain-specific embeddings improved topic \textit{precision} and \textit{interpretability}, with BioClinicalBERT producing the most consistent results across transcripts. The global analysis of the full dataset of 13 interviews, using the BioClinicalBERT embedding model, reveals the most dominant topics throughout all 13 interviews, namely ``Coordination and Communication in Cancer Care Management" and ``Patient Decision-Making in Cancer Treatment Journey''. Although the interviews are machine translations from Dutch to English, and clinical professionals are not involved in this evaluation, the findings suggest that neural topic modeling, particularly BERTopic, can help provide useful feedback to clinicians from patient interviews. This pipeline could support more efficient document navigation and strengthen the role of patients' voices in healthcare workflows.

Junchen Fu, Xuri Ge, Alexandros Karatzoglou et al.

Generative recommendation provides a novel paradigm in which each item is represented by a discrete semantic ID (SID) learned from rich content. Most existing methods treat SIDs as predefined and train recommenders under static indexing. In practice, SIDs are typically optimized only for content reconstruction rather than recommendation accuracy. This leads to an objective mismatch: the system optimizes an indexing loss to learn the SID and a recommendation loss for interaction prediction, but because the tokenizer is trained independently, the recommendation loss cannot update it. A natural approach is to make semantic indexing differentiable so that recommendation gradients can directly influence SID learning, but this often causes codebook collapse, where only a few codes are used. We attribute this issue to early deterministic assignments that limit codebook exploration, resulting in imbalance and unstable optimization. In this paper, we propose DIGER (Differentiable Semantic ID for Generative Recommendation), a first step toward effective differentiable semantic IDs for generative recommendation. DIGER introduces Gumbel noise to explicitly encourage early-stage exploration over codes, mitigating codebook collapse and improving code utilization. To balance exploration and convergence, we further design two uncertainty decay strategies that gradually reduce the Gumbel noise, enabling a smooth transition from early exploration to exploitation of learned SIDs. Extensive experiments on multiple public datasets demonstrate consistent improvements from differentiable semantic IDs. These results confirm the effectiveness of aligning indexing and recommendation objectives through differentiable SIDs and highlight differentiable semantic indexing as a promising research direction. Our code is released under https://github.com/junchen-fu/DIGER.

Aske Plaat, Annie Wong, Suzan Verberne et al.

Large language models (LLMs) with billions of parameters exhibit in-context learning abilities, enabling few-shot learning on tasks that the model was not specifically trained for. Traditional models achieve breakthrough performance on language tasks, but do not perform well on basic reasoning benchmarks. However, a new in-context learning approach, Chain-of-thought, has demonstrated strong multi-step reasoning abilities on these benchmarks. The research on LLM reasoning abilities started with the question whether LLMs can solve grade school math word problems, and has expanded to other tasks in the past few years. This article reviews the field of multi-step reasoning with LLMs. We propose a taxonomy that identifies different ways to generate, evaluate, and control multi-step reasoning. We provide an in-depth coverage of core approaches and open problems, and we propose a research agenda for the near future. We find that multi-step reasoning approaches have progressed beyond math word problems, and can now successfully solve challenges in logic, combinatorial games, and robotics, sometimes by first generating code that is then executed by external tools. Many studies in multi-step methods use reinforcement learning for finetuning, external optimization loops, in-context reinforcement learning, and self-reflection.

Mert Yazan, Suzan Verberne, Frederik Bungaran Ishak Situmeang

Artificial Intelligence (AI) agents personalize their responses by tailoring explanations to users' backgrounds, interests, and prior interactions, referred to as contextualization. Personalization has been identified as a persuasive strategy in politics or in marketing. However, the persuasive effect of contextualization in everyday tasks, where users often lack prior knowledge, remains unclear. We conducted a $2\times2$ between-subjects experiment ($N = 380$) examining how contextualization, combined with conversational warmth, shapes reliance and persuasiveness of an AI assistant arguing against expert recommendations. Our findings reveal that contextualization reduces the persuasive power of AI, but its combination with warmth restores persuasiveness through a crossover interaction. Reliance on AI is present across conditions and is invariant to the conversational design. Trust strongly predicts both persuasion and reliance, yet neither contextualization nor warmth operates through trust. AI literacy decouples trust from behavior: more literate users report lower trust in the assistant, yet are more persuaded and more reliant on its advice. These results suggest that users are prone to deferring to AI agents over human expert judgment; however, interface-level conversational design choices have a limited role in shaping the behavior.

Sophia Althammer, Guido Zuccon, Sebastian Hofstätter et al.

Search methods based on Pretrained Language Models (PLM) have demonstrated great effectiveness gains compared to statistical and early neural ranking models. However, fine-tuning PLM-based rankers requires a great amount of annotated training data. Annotating data involves a large manual effort and thus is expensive, especially in domain specific tasks. In this paper we investigate fine-tuning PLM-based rankers under limited training data and budget. We investigate two scenarios: fine-tuning a ranker from scratch, and domain adaptation starting with a ranker already fine-tuned on general data, and continuing fine-tuning on a target dataset. We observe a great variability in effectiveness when fine-tuning on different randomly selected subsets of training data. This suggests that it is possible to achieve effectiveness gains by actively selecting a subset of the training data that has the most positive effect on the rankers. This way, it would be possible to fine-tune effective PLM rankers at a reduced annotation budget. To investigate this, we adapt existing Active Learning (AL) strategies to the task of fine-tuning PLM rankers and investigate their effectiveness, also considering annotation and computational costs. Our extensive analysis shows that AL strategies do not significantly outperform random selection of training subsets in terms of effectiveness. We further find that gains provided by AL strategies come at the expense of more assessments (thus higher annotation costs) and AL strategies underperform random selection when comparing effectiveness given a fixed annotation cost. Our results highlight that ``optimal'' subsets of training data that provide high effectiveness at low annotation cost do exist, but current mainstream AL strategies applied to PLM rankers are not capable of identifying them.

Mert Yazan, Frederik Bungaran Ishak Situmeang, Suzan Verberne

Conversational artificial intelligence (AI) provides an efficient and convenient gateway to information access. However, it can cause overreliance when users blindly trust AI and accept its answers without fact-checking. Information search increasingly follows a hybrid interaction paradigm that combines conversational AI with web search, making fact-checking easier. In this paper, we examine whether this interaction paradigm is effective in curbing reliance. We further investigate the underlying factors (e.g., digital literacy and conversation warmth) that drive users to verify AI answers. We conduct a mixed-subjects question-answering experiment where participants interact with either a warm or a neutral chatbot. Our findings reveal that reliance persists despite users having access to both conversational and web search. The decision to verify is driven primarily by existing user perceptions (e.g., prior trust in chatbots) rather than answer properties, with some users fact-checking regardless of the context and others trusting chatbots by default. Warm conversational style has an indirect yet critical influence on reliance by increasing agreement with the chatbot when it is incorrect. Consulting additional AI sources predicts higher accuracy, while traditional web search does not. Our study extends overreliance research by: (a) demonstrating its persistence despite access to fact-checking, (b) identifying verification behavior as user-dependent, and (c) revealing conversational warmth's indirect effect on overreliance with implications for designing trustworthy conversational search systems.

Lifeng Han, Paul Rayson, Suzan Verberne et al.

This proposed tutorial focuses on Healthcare Domain Applications of NLP, what we have achieved around HealthcareNLP, and the challenges that lie ahead for the future. Existing reviews in this domain either overlook some important tasks, such as synthetic data generation for addressing privacy concerns, or explainable clinical NLP for improved integration and implementation, or fail to mention important methodologies, including retrieval augmented generation and the neural symbolic integration of LLMs and KGs. In light of this, the goal of this tutorial is to provide an introductory overview of the most important sub-areas of a patient- and resource-oriented HealthcareNLP, with three layers of hierarchy: data/resource layer: annotation guidelines, ethical approvals, governance, synthetic data; NLP-Eval layer: NLP tasks such as NER, RE, sentiment analysis, and linking/coding with categorised methods, leading to explainable HealthAI; patients layer: Patient Public Involvement and Engagement (PPIE), health literacy, translation, simplification, and summarisation (also NLP tasks), and shared decision-making support. A hands-on session will be included in the tutorial for the audience to use HealthcareNLP applications. The target audience includes NLP practitioners in the healthcare application domain, NLP researchers who are interested in domain applications, healthcare researchers, and students from NLP fields. The type of tutorial is "Introductory to CL/NLP topics (HealthcareNLP)" and the audience does not need prior knowledge to attend this. Tutorial materials: https://github.com/4dpicture/HealthNLP

Arian Askari, Chuan Meng, Mohammad Aliannejadi et al.

Existing generative retrieval (GR) methods rely on training-based indexing, which fine-tunes a model to memorise associations between queries and the document identifiers (docids) of relevant documents. Training-based indexing suffers from high training costs, under-utilisation of pre-trained knowledge in large language models (LLMs), and limited adaptability to dynamic document corpora. To address the issues, we propose a few-shot indexing-based GR framework (Few-Shot GR). It has a few-shot indexing process without any training, where we prompt an LLM to generate docids for all documents in a corpus, ultimately creating a docid bank for the entire corpus. During retrieval, we feed a query to the same LLM and constrain it to generate a docid within the docid bank created during indexing, and then map the generated docid back to its corresponding document. Moreover, we devise few-shot indexing with one-to-many mapping to further enhance Few-Shot GR. Experiments show that Few-Shot GR achieves superior performance to state-of-the-art GR methods requiring heavy training.

Zhengliang Shi, Shen Gao, Xiuyi Chen et al. · baidu

Tool learning empowers large language models (LLMs) as agents to use external tools and extend their utility. Existing methods employ one single LLM-based agent to iteratively select and execute tools, thereafter incorporating execution results into the next action prediction. Despite their progress, these methods suffer from performance degradation when addressing practical tasks due to: (1) the pre-defined pipeline with restricted flexibility to calibrate incorrect actions, and (2) the struggle to adapt a general LLM-based agent to perform a variety of specialized actions. To mitigate these problems, we propose ConAgents, a Cooperative and interactive Agents framework, which coordinates three specialized agents for tool selection, tool execution, and action calibration separately. ConAgents introduces two communication protocols to enable the flexible cooperation of agents. To effectively generalize the ConAgents into open-source models, we also propose specialized action distillation, enhancing their ability to perform specialized actions in our framework. Our extensive experiments on three datasets show that the LLMs, when equipped with the ConAgents, outperform baselines with substantial improvement (i.e., up to 14% higher success rate).

Lifeng Han, David Lindevelt, Sander Puts et al.

Metaphors and metaphorical language (MLs) play an important role in healthcare communication between clinicians, patients, and patients' family members. In this work, we focus on Dutch language data from cancer patients. We extract metaphors used by patients using two data sources: (1) cancer patient storytelling interview data and (2) online forum data, including patients' posts, comments, and questions to professionals. We investigate how current state-of-the-art large language models (LLMs) perform on this task by exploring different prompting strategies such as chain of thought reasoning, few-shot learning, and self-prompting. With a human-in-the-loop setup, we verify the extracted metaphors and compile the outputs into a corpus named HealthQuote.NL. We believe the extracted metaphors can support better patient care, for example shared decision making, improved communication between patients and clinicians, and enhanced patient health literacy. They can also inform the design of personalized care pathways. We share prompts and related resources at https://github.com/aaronlifenghan/HealthQuote.NL

Bram M. A. van Dijk, Max J. van Duijn, Suzan Verberne et al.

In this resource paper we release ChiSCor, a new corpus containing 619 fantasy stories, told freely by 442 Dutch children aged 4-12. ChiSCor was compiled for studying how children render character perspectives, and unravelling language and cognition in development, with computational tools. Unlike existing resources, ChiSCor's stories were produced in natural contexts, in line with recent calls for more ecologically valid datasets. ChiSCor hosts text, audio, and annotations for character complexity and linguistic complexity. Additional metadata (e.g. education of caregivers) is available for one third of the Dutch children. ChiSCor also includes a small set of 62 English stories. This paper details how ChiSCor was compiled and shows its potential for future work with three brief case studies: i) we show that the syntactic complexity of stories is strikingly stable across children's ages; ii) we extend work on Zipfian distributions in free speech and show that ChiSCor obeys Zipf's law closely, reflecting its social context; iii) we show that even though ChiSCor is relatively small, the corpus is rich enough to train informative lemma vectors that allow us to analyse children's language use. We end with a reflection on the value of narrative datasets in computational linguistics.

Hugo de Vos, Suzan Verberne

In this paper, we present a transcribed corpus of the LIBE committee of the EU parliament, totalling 3.6 Million running words. The meetings of parliamentary committees of the EU are a potentially valuable source of information for political scientists but the data is not readily available because only disclosed as speech recordings together with limited metadata. The meetings are in English, partly spoken by non-native speakers, and partly spoken by interpreters. We investigated the most appropriate Automatic Speech Recognition (ASR) model to create an accurate text transcription of the audio recordings of the meetings in order to make their content available for research and analysis. We focused on the unsupervised domain adaptation of the ASR pipeline. Building on the transformer-based Wav2vec2.0 model, we experimented with multiple acoustic models, language models and the addition of domain-specific terms. We found that a domain-specific acoustic model and a domain-specific language model give substantial improvements to the ASR output, reducing the word error rate (WER) from 28.22 to 17.95. The use of domain-specific terms in the decoding stage did not have a positive effect on the quality of the ASR in terms of WER. Initial topic modelling results indicated that the corpus is useful for downstream analysis tasks. We release the resulting corpus and our analysis pipeline for future research.

Amin Abolghasemi, Zhaochun Ren, Arian Askari et al.

An important unexplored aspect in previous work on user satisfaction estimation for Task-Oriented Dialogue (TOD) systems is their evaluation in terms of robustness for the identification of user dissatisfaction: current benchmarks for user satisfaction estimation in TOD systems are highly skewed towards dialogues for which the user is satisfied. The effect of having a more balanced set of satisfaction labels on performance is unknown. However, balancing the data with more dissatisfactory dialogue samples requires further data collection and human annotation, which is costly and time-consuming. In this work, we leverage large language models (LLMs) and unlock their ability to generate satisfaction-aware counterfactual dialogues to augment the set of original dialogues of a test collection. We gather human annotations to ensure the reliability of the generated samples. We evaluate two open-source LLMs as user satisfaction estimators on our augmented collection against state-of-the-art fine-tuned models. Our experiments show that when used as few-shot user satisfaction estimators, open-source LLMs show higher robustness to the increase in the number of dissatisfaction labels in the test collection than the fine-tuned state-of-the-art models. Our results shed light on the need for data augmentation approaches for user satisfaction estimation in TOD systems. We release our aligned counterfactual dialogues, which are curated by human annotation, to facilitate further research on this topic.

Mert Yazan, Suzan Verberne, Frederik Situmeang

Personalization with retrieval-augmented generation (RAG) often fails to capture fine-grained features of authors, making it hard to identify their unique traits. To enrich the RAG context, we propose providing Large Language Models (LLMs) with author-specific features, such as average sentiment polarity and frequently used words, in addition to past samples from the author's profile. We introduce a new feature called Contrastive Examples: documents from other authors are retrieved to help LLM identify what makes an author's style unique in comparison to others. Our experiments show that adding a couple of sentences about the named entities, dependency patterns, and words a person uses frequently significantly improves personalized text generation. Combining features with contrastive examples boosts the performance further, achieving a relative 15% improvement over baseline RAG while outperforming the benchmarks. Our results show the value of fine-grained features for better personalization, while opening a new research dimension for including contrastive examples as a complement with RAG. We release our code publicly.

Ali Satvaty, Suzan Verberne, Fatih Turkmen

Membership inference attacks (MIA) aim to infer whether a particular data point is part of the training dataset of a model. In this paper, we propose a new task in the context of LLM privacy: entity-level discovery of membership risk focused on sensitive information (PII, credit card numbers, etc). Existing methods for MIA can detect the presence of entire prompts or documents in the LLM training data, but they fail to capture risks at a finer granularity. We propose the ``EL-MIA'' framework for auditing entity-level membership risks in LLMs. We construct a benchmark dataset for the evaluation of MIA methods on this task. Using this benchmark, we conduct a systematic comparison of existing MIA techniques as well as two newly proposed methods. We provide a comprehensive analysis of the results, trying to explain the relation of the entity level MIA susceptability with the model scale, training epochs, and other surface level factors. Our findings reveal that existing MIA methods are limited when it comes to entity-level membership inference of the sensitive attributes, while this susceptibility can be outlined with relatively straightforward methods, highlighting the need for stronger adversaries to stress test the provided threat model.

I-Fan Lin, Faegheh Hasibi, Suzan Verberne

In this paper, we propose an intuitive, training-free and label-free method for intent clustering that makes minimal assumptions using lightweight and open-source LLMs. Many current approaches rely on commercial LLMs, which are costly, and offer limited transparency. Additionally, their methods often explicitly depend on knowing the number of clusters in advance, which is often not the case in realistic settings. To address these challenges, instead of asking the LLM to match similar text directly, we first ask it to generate pseudo-labels for each text, and then perform multi-label classification in this pseudo-label set for each text. This approach is based on the hypothesis that texts belonging to the same cluster will share more labels, and will therefore be closer when encoded into embeddings. These pseudo-labels are more human-readable than direct similarity matches. Our evaluation on four benchmark sets shows that our approach achieves results comparable to and better than recent baselines, while remaining simple and computationally efficient. Our findings indicate that our method can be applied in low-resource scenarios and is stable across multiple models and datasets.

Juan Pablo Bascur, Rodrigo Costas, Suzan Verberne

Traditional science maps visualize topics by clustering documents within a network, but they are inherently biased toward clustering certain topics over others. If these topics could be chosen, then the science maps could be tailored for different needs. In this paper, we explore the extent to which the topic bias of a science map can be changed by choosing different data sources to build the document network. We analyze this by evaluating the clustering effectiveness of several topic categories over two sources that are traditionally used for the creation of science maps (citations and text similarity) and six non-traditional data sources, which we found favor different kinds of topics: Health issues for Facebook users, biotechnology topics for patent families, government and social issues for policy documents, food topics for Twitter conversations, nursing topics for Twitter users, and geographical entities for document authors (the favoring in this latter source was particularly strong). Our results show that diverse data sources can be used to control topic bias, which opens up the possibility of creating science maps tailored for different needs.

Jiabao Fang, Shen Gao, Pengjie Ren et al.

Due to strong capabilities in conducting fluent, multi-turn conversations with users, Large Language Models (LLMs) have the potential to further improve the performance of Conversational Recommender System (CRS). Unlike the aimless chit-chat that LLM excels at, CRS has a clear target. So it is imperative to control the dialogue flow in the LLM to successfully recommend appropriate items to the users. Furthermore, user feedback in CRS can assist the system in better modeling user preferences, which has been ignored by existing studies. However, simply prompting LLM to conduct conversational recommendation cannot address the above two key challenges. In this paper, we propose Multi-Agent Conversational Recommender System (MACRS) which contains two essential modules. First, we design a multi-agent act planning framework, which can control the dialogue flow based on four LLM-based agents. This cooperative multi-agent framework will generate various candidate responses based on different dialogue acts and then choose the most appropriate response as the system response, which can help MACRS plan suitable dialogue acts. Second, we propose a user feedback-aware reflection mechanism which leverages user feedback to reason errors made in previous turns to adjust the dialogue act planning, and higher-level user information from implicit semantics. We conduct extensive experiments based on user simulator to demonstrate the effectiveness of MACRS in recommendation and user preferences collection. Experimental results illustrate that MACRS demonstrates an improvement in user interaction experience compared to directly using LLMs.

Mert Yazan, Frederik Bungaran Ishak Situmeang, Suzan Verberne

Conversational search interfaces, like ChatGPT, offer an interactive, personalized, and engaging user experience compared to traditional search. On the downside, they are prone to cause overtrust issues where users rely on their responses even when they are incorrect. What aspects of the conversational interaction paradigm drive people to adopt it, and how it creates personalized experiences that lead to overtrust, is not clear. To understand the factors influencing the adoption of conversational interfaces, we conducted a survey with 173 participants. We examined user perceptions regarding trust, human-likeness (anthropomorphism), and design preferences between ChatGPT and Google. To better understand the overtrust phenomenon, we asked users about their willingness to trade off factuality for constructs like ease of use or human-likeness. Our analysis identified two distinct user groups: those who use both ChatGPT and Google daily (DUB), and those who primarily rely on Google (DUG). The DUB group exhibited higher trust in ChatGPT, perceiving it as more human-like, and expressed greater willingness to trade factual accuracy for enhanced personalization and conversational flow. Conversely, the DUG group showed lower trust toward ChatGPT but still appreciated aspects like ad-free experiences and responsive interactions. Demographic analysis further revealed nuanced patterns, with middle-aged adults using ChatGPT less frequently yet trusting it more, suggesting potential vulnerability to misinformation. Our findings contribute to understanding user segmentation, emphasizing the critical roles of personalization and human-likeness in conversational IR systems, and reveal important implications regarding users' willingness to compromise factual accuracy for more engaging interactions.

Arian Askari, Roxana Petcu, Chuan Meng et al.

Identifying user intents in information-seeking dialogs is crucial for a system to meet user's information needs. Intent prediction (IP) is challenging and demands sufficient dialogs with human-labeled intents for training. However, manually annotating intents is resource-intensive. While large language models (LLMs) have been shown to be effective in generating synthetic data, there is no study on using LLMs to generate intent-aware information-seeking dialogs. In this paper, we focus on leveraging LLMs for zero-shot generation of large-scale, open-domain, and intent-aware information-seeking dialogs. We propose SOLID, which has novel self-seeding and multi-intent self-instructing schemes. The former improves the generation quality by using the LLM's own knowledge scope to initiate dialog generation; the latter prompts the LLM to generate utterances sequentially, and mitigates the need for manual prompt design by asking the LLM to autonomously adapt its prompt instruction when generating complex multi-intent utterances. Furthermore, we propose SOLID-RL, which is further trained to generate a dialog in one step on the data generated by SOLID. We propose a length-based quality estimation mechanism to assign varying weights to SOLID-generated dialogs based on their quality during the training process of SOLID-RL. We use SOLID and SOLID-RL to generate more than 300k intent-aware dialogs, surpassing the size of existing datasets. Experiments show that IP methods trained on dialogs generated by SOLID and SOLID-RL achieve better IP quality than ones trained on human-generated dialogs.

Amin Abolghasemi, Leif Azzopardi, Arian Askari et al.

In most recent studies, gender bias in document ranking is evaluated with the NFaiRR metric, which measures bias in a ranked list based on an aggregation over the unbiasedness scores of each ranked document. This perspective in measuring the bias of a ranked list has a key limitation: individual documents of a ranked list might be biased while the ranked list as a whole balances the groups' representations. To address this issue, we propose a novel metric called TExFAIR (term exposure-based fairness), which is based on two new extensions to a generic fairness evaluation framework, attention-weighted ranking fairness (AWRF). TExFAIR assesses fairness based on the term-based representation of groups in a ranked list: (i) an explicit definition of associating documents to groups based on probabilistic term-level associations, and (ii) a rank-biased discounting factor (RBDF) for counting non-representative documents towards the measurement of the fairness of a ranked list. We assess TExFAIR on the task of measuring gender bias in passage ranking, and study the relationship between TExFAIR and NFaiRR. Our experiments show that there is no strong correlation between TExFAIR and NFaiRR, which indicates that TExFAIR measures a different dimension of fairness than NFaiRR. With TExFAIR, we extend the AWRF framework to allow for the evaluation of fairness in settings with term-based representations of groups in documents in a ranked list.

Rayden Tseng, Suzan Verberne, Peter van der Putten

ChatGPT, GPT-3.5, and other large language models (LLMs) have drawn significant attention since their release, and the abilities of these models have been investigated for a wide variety of tasks. In this research we investigate to what extent GPT-3.5 can generate human-like comments on Dutch news articles. We define human likeness as `not distinguishable from human comments', approximated by the difficulty of automatic classification between human and GPT comments. We analyze human likeness across multiple prompting techniques. In particular, we utilize zero-shot, few-shot and context prompts, for two generated personas. We found that our fine-tuned BERT models can easily distinguish human-written comments from GPT-3.5 generated comments, with none of the used prompting methods performing noticeably better. We further analyzed that human comments consistently showed higher lexical diversity than GPT-generated comments. This indicates that although generative LLMs can generate fluent text, their capability to create human-like opinionated comments is still limited.

I-Fan Lin, Faegheh Hasibi, Suzan Verberne

In this paper, we propose Selection and Pooling with Large Language Models (SPILL), an intuitive and domain-adaptive method for intent clustering without fine-tuning. Existing embeddings-based clustering methods rely on a few labeled examples or unsupervised fine-tuning to optimize results for each new dataset, which makes them less generalizable to multiple datasets. Our goal is to make these existing embedders more generalizable to new domain datasets without further fine-tuning. Inspired by our theoretical derivation and simulation results on the effectiveness of sampling and pooling techniques, we view the clustering task as a small-scale selection problem. A good solution to this problem is associated with better clustering performance. Accordingly, we propose a two-stage approach: First, for each utterance (referred to as the seed), we derive its embedding using an existing embedder. Then, we apply a distance metric to select a pool of candidates close to the seed. Because the embedder is not optimized for new datasets, in the second stage, we use an LLM to further select utterances from these candidates that share the same intent as the seed. Finally, we pool these selected candidates with the seed to derive a refined embedding for the seed. We found that our method generally outperforms directly using an embedder, and it achieves comparable results to other state-of-the-art studies, even those that use much larger models and require fine-tuning, showing its strength and efficiency. Our results indicate that our method enables existing embedders to be further improved without additional fine-tuning, making them more adaptable to new domain datasets. Additionally, viewing the clustering task as a small-scale selection problem gives the potential of using LLMs to customize clustering tasks according to the user's goals.

Amin Abolghasemi, Leif Azzopardi, Seyyed Hadi Hashemi et al.

Attributing answers to source documents is an approach used to enhance the verifiability of a model's output in retrieval augmented generation (RAG). Prior work has mainly focused on improving and evaluating the attribution quality of large language models (LLMs) in RAG, but this may come at the expense of inducing biases in the attribution of answers. We define and examine two aspects in the evaluation of LLMs in RAG pipelines, namely attribution sensitivity and bias with respect to authorship information. We explicitly inform an LLM about the authors of source documents, instruct it to attribute its answers, and analyze (i) how sensitive the LLM's output is to the author of source documents, and (ii) whether the LLM exhibits a bias towards human-written or AI-generated source documents. We design an experimental setup in which we use counterfactual evaluation to study three LLMs in terms of their attribution sensitivity and bias in RAG pipelines. Our results show that adding authorship information to source documents can significantly change the attribution quality of LLMs by 3% to 18%. Moreover, we show that LLMs can have an attribution bias towards explicit human authorship, which can serve as a competing hypothesis for findings of prior work that shows that LLM-generated content may be preferred over human-written contents. Our findings indicate that metadata of source documents can influence LLMs' trust, and how they attribute their answers. Furthermore, our research highlights attribution bias and sensitivity as a novel aspect of brittleness in LLMs.

Myrthe Reuver, Suzan Verberne, Antske Fokkens

For a viewpoint-diverse news recommender, identifying whether two news articles express the same viewpoint is essential. One way to determine "same or different" viewpoint is stance detection. In this paper, we investigate the robustness of operationalization choices for few-shot stance detection, with special attention to modelling stance across different topics. Our experiments test pre-registered hypotheses on stance detection. Specifically, we compare two stance task definitions (Pro/Con versus Same Side Stance), two LLM architectures (bi-encoding versus cross-encoding), and adding Natural Language Inference knowledge, with pre-trained RoBERTa models trained with shots of 100 examples from 7 different stance detection datasets. Some of our hypotheses and claims from earlier work can be confirmed, while others give more inconsistent results. The effect of the Same Side Stance definition on performance differs per dataset and is influenced by other modelling choices. We found no relationship between the number of training topics in the training shots and performance. In general, cross-encoding out-performs bi-encoding, and adding NLI training to our models gives considerable improvement, but these results are not consistent across all datasets. Our results indicate that it is essential to include multiple datasets and systematic modelling experiments when aiming to find robust modelling choices for the concept `stance'.

Ziyou Hu, Zhengliang Shi, Minghang Zhu et al.

Reward models (RMs) have become essential for aligning large language models (LLMs), serving as scalable proxies for human evaluation in both training and inference. However, existing RMs struggle on knowledge-intensive and long-form tasks, where evaluating correctness requires grounding beyond the model's internal knowledge. This limitation hinders them from reliably discriminating subtle quality differences, especially when external evidence is necessary. To address this, we introduce OpenRM, a tool-augmented long-form reward model that systematically judges open-ended responses by invoking external tools to gather relevant evidence. We train OpenRM with Group Relative Policy Optimization (GRPO) on over 27K synthesized pairwise examples generated through a controllable data synthesis framework. The training objective jointly supervises intermediate tool usage and final outcome accuracy, incentivizing our reward model to learn effective evidence-based judgment strategies. Extensive experiments on three newly-collected datasets and two widely-used benchmarks demonstrate that OpenRM substantially outperforms existing reward modeling approaches. As a further step, we integrate OpenRM into both inference-time response selection and training-time data selection. This yields consistent gains in downstream LLM alignment tasks, highlighting the potential of tool-augmented reward models for scaling reliable long-form evaluation.

I-Fan Lin, Faegheh Hasibi, Suzan Verberne

In this paper, we propose a training-free and label-free method for short text clustering that can be used on top of any existing embedder. In the context of customer-facing chatbots, companies are dealing with large amounts of user utterances that need to be clustered according to their intent. In these commercial settings, no labeled data is typically available, and the number of clusters is not known. Our method is based on iterative vector updating: it constructs sparse vectors based on representative texts, and then iteratively refines them through LLM guidance. Our method achieves comparable or superior results to state-of-the-art methods that use contrastive learning, but without assuming prior knowledge of clusters or labels. Experiments on diverse datasets and smaller LLMs show that our method is model agnostic and can be applied to any embedder, with relatively small LLMs, and different clustering methods. We also show that our method scales to large datasets, reducing the computational cost of the LLM. These low-resource, adaptable settings and the scalability of our method make it more aligned with real-world scenarios than existing clustering methods.

Rob Reijtenbach, Suzan Verberne, Gijs Wijnholds

In this paper we present and validate a new synthetic dataset for training visual entailment models. Existing datasets for visual entailment are small and sparse compared to datasets for textual entailment. Manually creating datasets is labor-intensive. We base our synthetic dataset on the SNLI dataset for textual entailment. We take the premise text from SNLI as input prompts in a generative image model, Stable Diffusion, creating an image to replace each textual premise. We evaluate our dataset both intrinsically and extrinsically. For extrinsic evaluation, we evaluate the validity of the generated images by using them as training data for a visual entailment classifier based on CLIP feature vectors. We find that synthetic training data only leads to a slight drop in quality on SNLI-VE, with an F-score 0.686 compared to 0.703 when trained on real data. We also compare the quality of our generated training data to original training data on another dataset: SICK-VTE. Again, there is only a slight drop in F-score: from 0.400 to 0.384. These results indicate that in settings with data sparsity, synthetic data can be a promising solution for training visual entailment models.

Mert Yazan, Suzan Verberne, Frederik Situmeang

Post-training quantization reduces the computational demand of Large Language Models (LLMs) but can weaken some of their capabilities. Since LLM abilities emerge with scale, smaller LLMs are more sensitive to quantization. In this paper, we explore how quantization affects smaller LLMs' ability to perform retrieval-augmented generation (RAG), specifically in longer contexts. We chose personalization for evaluation because it is a challenging domain to perform using RAG as it requires long-context reasoning over multiple documents. We compare the original FP16 and the quantized INT4 performance of multiple 7B and 8B LLMs on two tasks while progressively increasing the number of retrieved documents to test how quantized models fare against longer contexts. To better understand the effect of retrieval, we evaluate three retrieval models in our experiments. Our findings reveal that if a 7B LLM performs the task well, quantization does not impair its performance and long-context reasoning capabilities. We conclude that it is possible to utilize RAG with quantized smaller LLMs.

Fons Hartendorp, Tom Seinen, Erik van Mulligen et al.

Biomedical entity linking, a main component in automatic information extraction from health-related texts, plays a pivotal role in connecting textual entities (such as diseases, drugs and body parts mentioned by patients) to their corresponding concepts in a structured biomedical knowledge base. The task remains challenging despite recent developments in natural language processing. This paper presents the first evaluated biomedical entity linking model for the Dutch language. We use MedRoBERTa.nl as base model and perform second-phase pretraining through self-alignment on a Dutch biomedical ontology extracted from the UMLS and Dutch SNOMED. We derive a corpus from Wikipedia of ontology-linked Dutch biomedical entities in context and fine-tune our model on this dataset. We evaluate our model on the Dutch portion of the Mantra GSC-corpus and achieve 54.7% classification accuracy and 69.8% 1-distance accuracy. We then perform a case study on a collection of unlabeled, patient-support forum data and show that our model is hampered by the limited quality of the preceding entity recognition step. Manual evaluation of small sample indicates that of the correctly extracted entities, around 65% is linked to the correct concept in the ontology. Our results indicate that biomedical entity linking in a language other than English remains challenging, but our Dutch model can be used to for high-level analysis of patient-generated text.

Joost Broekens, Bernhard Hilpert, Suzan Verberne et al.

Large language models, in particular generative pre-trained transformers (GPTs), show impressive results on a wide variety of language-related tasks. In this paper, we explore ChatGPT's zero-shot ability to perform affective computing tasks using prompting alone. We show that ChatGPT a) performs meaningful sentiment analysis in the Valence, Arousal and Dominance dimensions, b) has meaningful emotion representations in terms of emotion categories and these affective dimensions, and c) can perform basic appraisal-based emotion elicitation of situations based on a prompt-based computational implementation of the OCC appraisal model. These findings are highly relevant: First, they show that the ability to solve complex affect processing tasks emerges from language-based token prediction trained on extensive data sets. Second, they show the potential of large language models for simulating, processing and analyzing human emotions, which has important implications for various applications such as sentiment analysis, socially interactive agents, and social robotics.

Amin Abolghasemi, Suzan Verberne, Leif Azzopardi

Query-by-document (QBD) retrieval is an Information Retrieval task in which a seed document acts as the query and the goal is to retrieve related documents -- it is particular common in professional search tasks. In this work we improve the retrieval effectiveness of the BERT re-ranker, proposing an extension to its fine-tuning step to better exploit the context of queries. To this end, we use an additional document-level representation learning objective besides the ranking objective when fine-tuning the BERT re-ranker. Our experiments on two QBD retrieval benchmarks show that the proposed multi-task optimization significantly improves the ranking effectiveness without changing the BERT re-ranker or using additional training samples. In future work, the generalizability of our approach to other retrieval tasks should be further investigated.

Arian Askari, Suzan Verberne, Gabriella Pasi

Expert finding has been well-studied in community question answering (QA) systems in various domains. However, none of these studies addresses expert finding in the legal domain, where the goal is for citizens to find lawyers based on their expertise. In the legal domain, there is a large knowledge gap between the experts and the searchers, and the content on the legal QA websites consist of a combination formal and informal communication. In this paper, we propose methods for generating query-dependent textual profiles for lawyers covering several aspects including sentiment, comments, and recency. We combine query-dependent profiles with existing expert finding methods. Our experiments are conducted on a novel dataset gathered from an online legal QA service. We discovered that taking into account different lawyer profile aspects improves the best baseline model. We make our dataset publicly available for future work.

Sophia Althammer, Sebastian Hofstätter, Mete Sertkan et al.

Dense passage retrieval (DPR) models show great effectiveness gains in first stage retrieval for the web domain. However in the web domain we are in a setting with large amounts of training data and a query-to-passage or a query-to-document retrieval task. We investigate in this paper dense document-to-document retrieval with limited labelled target data for training, in particular legal case retrieval. In order to use DPR models for document-to-document retrieval, we propose a Paragraph Aggregation Retrieval Model (PARM) which liberates DPR models from their limited input length. PARM retrieves documents on the paragraph-level: for each query paragraph, relevant documents are retrieved based on their paragraphs. Then the relevant results per query paragraph are aggregated into one ranked list for the whole query document. For the aggregation we propose vector-based aggregation with reciprocal rank fusion (VRRF) weighting, which combines the advantages of rank-based aggregation and topical aggregation based on the dense embeddings. Experimental results show that VRRF outperforms rank-based aggregation strategies for dense document-to-document retrieval with PARM. We compare PARM to document-level retrieval and demonstrate higher retrieval effectiveness of PARM for lexical and dense first-stage retrieval on two different legal case retrieval collections. We investigate how to train the dense retrieval model for PARM on limited target data with labels on the paragraph or the document-level. In addition, we analyze the differences of the retrieved results of lexical and dense retrieval with PARM.

Myrthe Reuver, Suzan Verberne, Roser Morante et al.

Cross-topic stance detection is the task to automatically detect stances (pro, against, or neutral) on unseen topics. We successfully reproduce state-of-the-art cross-topic stance detection work (Reimers et. al., 2019), and systematically analyze its reproducibility. Our attention then turns to the cross-topic aspect of this work, and the specificity of topics in terms of vocabulary and socio-cultural context. We ask: To what extent is stance detection topic-independent and generalizable across topics? We compare the model's performance on various unseen topics, and find topic (e.g. abortion, cloning), class (e.g. pro, con), and their interaction affecting the model's performance. We conclude that investigating performance on different topics, and addressing topic-specific vocabulary and context, is a future avenue for cross-topic stance detection.

Hugo de Vos, Suzan Verberne

In an often-cited 2019 paper on the use of machine learning in political research, Anastasopoulos & Whitford (A&W) propose a text classification method for tweets related to organizational reputation. The aim of their paper was to provide a 'guide to practice' for public administration scholars and practitioners on the use of machine learning. In the current paper we follow up on that work with a replication of A&W's experiments and additional analyses on model stability and the effects of preprocessing, both in relation to the small data size. We show that (1) the small data causes the classification model to be highly sensitive to variations in the random train-test split, and that (2) the applied preprocessing causes the data to be extremely sparse, with the majority of items in the data having at most two non-zero lexical features. With additional experiments in which we vary the steps of the preprocessing pipeline, we show that the small data size keeps causing problems, irrespective of the preprocessing choices. Based on our findings, we argue that A&W's conclusions regarding the automated classification of organizational reputation tweets -- either substantive or methodological -- can not be maintained and require a larger data set for training and more careful validation.

Anne Dirkson, Suzan Verberne, Wessel Kraaij

Both generic and domain-specific BERT models are widely used for natural language processing (NLP) tasks. In this paper we investigate the vulnerability of BERT models to variation in input data for Named Entity Recognition (NER) through adversarial attack. Experimental results show that BERT models are vulnerable to variation in the entity context with 20.2 to 45.0% of entities predicted completely wrong and another 29.3 to 53.3% of entities predicted wrong partially. BERT models seem most vulnerable to changes in the local context of entities and often a single change is sufficient to fool the model. The domain-specific BERT model trained from scratch (SciBERT) is more vulnerable than the original BERT model or the domain-specific model that retains the BERT vocabulary (BioBERT). We also find that BERT models are particularly vulnerable to emergent entities. Our results chart the vulnerabilities of BERT models for NER and emphasize the importance of further research into uncovering and reducing these weaknesses.

Sophia Althammer, Arian Askari, Suzan Verberne et al.

In this paper, we present our approaches for the case law retrieval and the legal case entailment task in the Competition on Legal Information Extraction/Entailment (COLIEE) 2021. As first stage retrieval methods combined with neural re-ranking methods using contextualized language models like BERT achieved great performance improvements for information retrieval in the web and news domain, we evaluate these methods for the legal domain. A distinct characteristic of legal case retrieval is that the query case and case description in the corpus tend to be long documents and therefore exceed the input length of BERT. We address this challenge by combining lexical and dense retrieval methods on the paragraph-level of the cases for the first stage retrieval. Here we demonstrate that the retrieval on the paragraph-level outperforms the retrieval on the document-level. Furthermore the experiments suggest that dense retrieval methods outperform lexical retrieval. For re-ranking we address the problem of long documents by summarizing the cases and fine-tuning a BERT-based re-ranker with the summaries. Overall, our best results were obtained with a combination of BM25 and dense passage retrieval using domain-specific embeddings.

Alex Brandsen, Suzan Verberne, Karsten Lambers et al.

The amount of archaeological literature is growing rapidly. Until recently, these data were only accessible through metadata search. We implemented a text retrieval engine for a large archaeological text collection ($\sim 658$ Million words). In archaeological IR, domain-specific entities such as locations, time periods, and artefacts, play a central role. This motivated the development of a named entity recognition (NER) model to annotate the full collection with archaeological named entities. In this paper, we present ArcheoBERTje, a BERT model pre-trained on Dutch archaeological texts. We compare the model's quality and output on a Named Entity Recognition task to a generic multilingual model and a generic Dutch model. We also investigate ensemble methods for combining multiple BERT models, and combining the best BERT model with a domain thesaurus using Conditional Random Fields (CRF). We find that ArcheoBERTje outperforms both the multilingual and Dutch model significantly with a smaller standard deviation between runs, reaching an average F1 score of 0.735. The model also outperforms ensemble methods combining the three models. Combining ArcheoBERTje predictions and explicit domain knowledge from the thesaurus did not increase the F1 score. We quantitatively and qualitatively analyse the differences between the vocabulary and output of the BERT models on the full collection and provide some valuable insights in the effect of fine-tuning for specific domains. Our results indicate that for a highly specific text domain such as archaeology, further pre-training on domain-specific data increases the model's quality on NER by a much larger margin than shown for other domains in the literature, and that domain-specific pre-training makes the addition of domain knowledge from a thesaurus unnecessary.

Ioannis Chios, Suzan Verberne

In this paper we address the explainability of web search engines. We propose two explainable elements on the search engine result page: a visualization of query term weights and a visualization of passage relevance. The idea is that search engines that indicate to the user why results are retrieved are valued higher by users and gain user trust. We deduce the query term weights from the term gating network in the Deep Relevance Matching Model (DRMM) and visualize them as a doughnut chart. In addition, we train a passage-level ranker with DRMM that selects the most relevant passage from each document and shows it as snippet on the result page. Next to the snippet we show a document thumbnail with this passage highlighted. We evaluate the proposed interface in an online user study, asking users to judge the explainability and assessability of the interface. We found that users judge our proposed interface significantly more explainable and easier to assess than a regular search engine result page. However, they are not significantly better in selecting the relevant documents from the top-5. This indicates that the explainability of the search engine result page leads to a better user experience. Thus, we conclude that the proposed explainable elements are promising as visualization for search engine users.

Alex Brandsen, Suzan Verberne, Karsten Lambers et al.

This paper presents AGNES, the first information retrieval system for archaeological grey literature, allowing full-text search of these long archaeological documents. This search system has a web interface that allows archaeology professionals and scholars to search through a collection of over 60,000 Dutch excavation reports, totalling 361 million words. We conducted a user study for the evaluation of AGNES's search interface, with a small but diverse user group. The evaluation was done by screen capturing and a think aloud protocol, combined with a user interface feedback questionnaire. The evaluation covered both controlled use (completion of a pre-defined task) as well as free use (completion of a freely chosen task). The free use allows us to study the information needs of archaeologists, as well as their interactions with the search system. We conclude that: (1) the information needs of archaeologists are typically recall-oriented, often requiring a list of items as answer; (2) the users prefer the use of free-text queries over metadata filters, confirming the value of a free-text search system; (3) the compilation of a diverse user group contributed to the collection of diverse issues as feedback for improving the system. We are currently refining AGNES's user interface and improving its precision for archaeological entities, so that AGNES will help archaeologists to answer their research questions more effectively and efficiently, leading to a more coherent narrative of the past.

Ken Voskuil, Suzan Verberne

In this paper we address the challenge of extracting scientific references from patents. We approach the problem as a sequence labelling task and investigate the merits of BERT models to the extraction of these long sequences. References in patents to scientific literature are relevant to study the connection between science and industry. Most prior work only uses the front-page citations for this analysis, which are provided in the metadata of patent archives. In this paper we build on prior work using Conditional Random Fields (CRF) and Flair for reference extraction. We improve the quality of the training data and train three BERT-based models on the labelled data (BERT, bioBERT, sciBERT). We find that the improved training data leads to a large improvement in the quality of the trained models. In addition, the BERT models beat CRF and Flair, with recall scores around 97% obtained with cross validation. With the best model we label a large collection of 33 thousand patents, extract the citations, and match them to publications in the Web of Science database. We extract 50% more references than with the old training data and methods: 735 thousand references in total. With these patent-publication links, follow-up research will further analyze which types of scientific work lead to inventions.

Jeroen Offerijns, Suzan Verberne, Tessa Verhoef

For the field of education, being able to generate semantically correct and educationally relevant multiple choice questions (MCQs) could have a large impact. While question generation itself is an active research topic, generating distractors (the incorrect multiple choice options) receives much less attention. A missed opportunity, since there is still a lot of room for improvement in this area. In this work, we train a GPT-2 language model to generate three distractors for a given question and text context, using the RACE dataset. Next, we train a BERT language model to answer MCQs, and use this model as a filter, to select only questions that can be answered and therefore presumably make sense. To evaluate our work, we start by using text generation metrics, which show that our model outperforms earlier work on distractor generation (DG) and achieves state-of-the-art performance. Also, by calculating the question answering ability, we show that larger base models lead to better performance. Moreover, we conducted a human evaluation study, which confirmed the quality of the generated questions, but showed no statistically significant effect of the QA filter.

Benjamin van der Burgh, Suzan Verberne

We evaluated the effectiveness of using language models, that were pre-trained in one domain, as the basis for a classification model in another domain: Dutch book reviews. Pre-trained language models have opened up new possibilities for classification tasks with limited labelled data, because representation can be learned in an unsupervised fashion. In our experiments we have studied the effects of training set size (100-1600 items) on the prediction accuracy of a ULMFiT classifier, based on a language models that we pre-trained on the Dutch Wikipedia. We also compared ULMFiT to Support Vector Machines, which is traditionally considered suitable for small collections. We found that ULMFiT outperforms SVM for all training set sizes and that satisfactory results (~90%) can be achieved using training sets that can be manually annotated within a few hours. We deliver both our new benchmark collection of Dutch book reviews for sentiment classification as well as the pre-trained Dutch language model to the community.

Suzan Verberne, Jiyin He, Gineke Wiggers et al.

Search conducted in a work context is an everyday activity that has been around since long before the Web was invented, yet we still seem to understand little about its general characteristics. With this paper we aim to contribute to a better understanding of this large but rather multi-faceted area of `professional search'. Unlike task-based studies that aim at measuring the effectiveness of search methods, we chose to take a step back by conducting a survey among professional searchers to understand their typical search tasks. By doing so we offer complementary insights into the subject area. We asked our respondents to provide actual search tasks they have worked on, information about how these were conducted and details on how successful they eventually were. We then manually coded the collection of 56 search tasks with task characteristics and relevance criteria, and used the coded dataset for exploration purposes. Despite the relatively small scale of this study, our data provides enough evidence that professional search is indeed very different from Web search in many key respects and that this is a field that offers many avenues for future research.

Alex Brandsen, Anne Dirkson, Wessel Kraaij et al.

This volume contains the papers presented at DIR 2018: 17th Dutch-Belgian Information Retrieval Workshop (DIR) held on November 23, 2018 in Leiden. DIR aims to serve as an international platform (with a special focus on the Netherlands and Belgium) for exchange and discussions on research & applications in the field of information retrieval and related fields. The committee accepted 4 short papers presenting novel work, 3 demo proposals, and 8 compressed contributions (summaries of papers recently published in international journals and conferences). Each submission was reviewed by at least 3 programme committee members.

Suzan Verberne, Arjen P. de Vries, Wessel Kraaij

We implemented and evaluated a two-stage retrieval method for personalized academic search in which the initial search results are re-ranked using an author-topic profile. In academic search tasks, the user's own data can help optimizing the ranking of search results to match the searcher's specific individual needs. The author-topic profile consists of topic-specific terms, stored in a graph. We re-rank the top-1000 retrieved documents using ten features that represent the similarity between the document and the author-topic graph. We found that the re-ranking gives a small but significant improvement over the reproduced best method from the literature. Storing the profile as a graph has a number of advantages: it is flexible with respect to node and relation types; it is a visualization of knowledge that is interpretable by the user, and it offers the possibility to view relational characteristics of individual nodes.