Hybrid search has emerged as an effective strategy to offset the limitations of different matching paradigms, especially in out-of-domain contexts where notable improvements in retrieval quality have been observed. However, existing research predominantly focuses on a limited set of retrieval methods, evaluated in pairs on domain-general datasets exclusively in English. In this work, we study the efficacy of hybrid search across a variety of prominent retrieval models within the unexplored field of law in the French language, assessing both zero-shot and in-domain scenarios. Our findings reveal that in a zero-shot context, fusing different domain-general models consistently enhances performance compared to using a standalone model, regardless of the fusion method. Surprisingly, when models are trained in-domain, we find that fusion generally diminishes performance relative to using the best single system, unless fusing scores with carefully tuned weights. These novel insights, among others, expand the applicability of prior findings across a new field and language, and contribute to a deeper understanding of hybrid search in non-English specialized domains.

State-of-the-art neural retrievers predominantly focus on high-resource languages like English, which impedes their adoption in retrieval scenarios involving other languages. Current approaches circumvent the lack of high-quality labeled data in non-English languages by leveraging multilingual pretrained language models capable of cross-lingual transfer. However, these models require substantial task-specific fine-tuning across multiple languages, often perform poorly in languages with minimal representation in the pretraining corpus, and struggle to incorporate new languages after the pretraining phase. In this work, we present a novel modular dense retrieval model that learns from the rich data of a single high-resource language and effectively zero-shot transfers to a wide array of languages, thereby eliminating the need for language-specific labeled data. Our model, ColBERT-XM, demonstrates competitive performance against existing state-of-the-art multilingual retrievers trained on more extensive datasets in various languages. Further analysis reveals that our modular approach is highly data-efficient, effectively adapts to out-of-distribution data, and significantly reduces energy consumption and carbon emissions. By demonstrating its proficiency in zero-shot scenarios, ColBERT-XM marks a shift towards more sustainable and inclusive retrieval systems, enabling effective information accessibility in numerous languages.

In this paper, we investigate the role of attention heads in Context-aware Machine Translation models for pronoun disambiguation in the English-to-German and English-to-French language directions. We analyze their influence by both observing and modifying the attention scores corresponding to the plausible relations that could impact a pronoun prediction. Our findings reveal that while some heads do attend the relations of interest, not all of them influence the models’ ability to disambiguate pronouns. We show that certain heads are underutilized by the models, suggesting that model performance could be improved if only the heads would attend one of the relations more strongly. Furthermore, we fine-tune the most promising heads and observe the increase in pronoun disambiguation accuracy of up to 5 percentage points which demonstrates that the improvements in performance can be solidified into the models’ parameters.

Achieving human-level translations requires leveraging context to ensure coherence and handle complex phenomena like pronoun disambiguation.Sparsity of contextually rich examples in the standard training data has been hypothesized as the reason for the difficulty of context utilization.In this work, we systematically validate this claim in both single-and multilingual settings by constructing training datasets with a controlled proportions of contextually relevant examples.We demonstrate a strong association between training data sparsity and model performance confirming sparsity as a key bottleneck.Importantly, we reveal that improvements in one contextual phenomenon do no generalize to others.While we observe some cross-lingual transfer, it is not significantly higher between languages within the same sub-family.Finally, we propose and empirically evaluate two training strategies designed to leverage the available data.These strategies improve context utilization, resulting in accuracy gains of up to 6 and 8 percentage points on the ctxPro evaluation in single-and multilingual settings respectively.1

Many individuals are likely to face a legal dispute at some point in their lives, but their lack of understanding of how to navigate these complex issues often renders them vulnerable. The advancement of natural language processing opens new avenues for bridging this legal literacy gap through the development of automated legal aid systems. However, existing legal question answering (LQA) approaches often suffer from a narrow scope, being either confined to specific legal domains or limited to brief, uninformative responses. In this work, we propose an end-to-end methodology designed to generate long-form answers to any statutory law questions, utilizing a "retrieve-then-read" pipeline. To support this approach, we introduce and release the Long-form Legal Question Answering (LLeQA) dataset, comprising 1,868 expert-annotated legal questions in the French language, complete with detailed answers rooted in pertinent legal provisions. Our experimental results demonstrate promising performance on automatic evaluation metrics, but a qualitative analysis uncovers areas for refinement. As one of the only comprehensive, expert-annotated long-form LQA dataset, LLeQA has the potential to not only accelerate research towards resolving a significant real-world issue, but also act as a rigorous benchmark for evaluating NLP models in specialized domains. We publicly release our code, data, and models.

Justus-Jonas Erker, Florian Mai, Nils Reimers, Gerasimos Spanakis, Iryna Gurevych. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

The anonymity on the Darknet allows vendors to stay undetected by using multiple vendor aliases or frequently migrating between markets.Consequently, illegal markets and their connections are challenging to uncover on the Darknet.To identify relationships between illegal markets and their vendors, we propose Ven-dorLink, an NLP-based approach that examines writing patterns to verify, identify, and link unique vendor accounts across text advertisements (ads) on seven public Darknet markets.In contrast to existing literature, VendorLink utilizes the strength of supervised pre-training to perform closed-set vendor verification, openset vendor identification, and low-resource market adaption tasks.Through VendorLink, we uncover (i) 15 migrants and 71 potential aliases in the Alphabay-Dreams-Silk dataset, (ii) 17 migrants and 3 potential aliases in the Valhalla-Berlusconi dataset, and (iii) 75 migrants and 10 potential aliases in the Traderoute-Agora dataset.Altogether, our approach can help Law Enforcement Agencies (LEA) make more informed decisions by verifying and identifying migrating vendors and their potential aliases on existing and Low-Resource (LR) emerging Darknet markets.

Statutory article retrieval (SAR), the task of retrieving statute law articles relevant to a legal question, is a promising application of legal text processing. In particular, high-quality SAR systems can improve the work efficiency of legal professionals and provide basic legal assistance to citizens in need at no cost. Unlike traditional ad-hoc information retrieval, where each document is considered a complete source of information, SAR deals with texts whose full sense depends on complementary information from the topological organization of statute law. While existing works ignore these domain-specific dependencies, we propose a novel graph-augmented dense statute retriever (G-DSR) model that incorporates the structure of legislation via a graph neural network to improve dense retrieval performance. Experimental results show that our approach outperforms strong retrieval baselines on a real-world expert-annotated SAR dataset.

The scientific innovation in Natural Language Processing (NLP) and more broadly in artificial intelligence (AI) is at its fastest pace to date.As large language models (LLMs) unleash a new era of automation, important debates emerge regarding the benefits and risks of their development, deployment and use.Currently, these debates have been dominated by often polarized narratives mainly led by the AI Safety and AI Ethics movements.This polarization, often amplified by social media, is swaying political agendas on AI regulation and governance and posing issues of regulatory capture.Capture occurs when the regulator advances the interests of the industry it is supposed to regulate, or of special interest groups rather than pursuing the general public interest.Meanwhile in NLP research, attention has been increasingly paid to the discussion of regulating risks and harms.This often happens without systematic methodologies or sufficient rooting in the disciplines that inspire an extended scope of NLP research, jeopardizing the scientific integrity of these endeavors.Regulation studies are a rich source of knowledge on how to systematically deal with risk and uncertainty, as well as with scientific evidence, to evaluate and compare regulatory options.This resource has largely remained untapped so far.In this paper, we argue how NLP research on these topics can benefit from proximity to regulatory studies and adjacent fields.We do so by discussing basic tenets of regulation, and risk and uncertainty, and by highlighting the shortcomings of current NLP discussions dealing with risk assessment.Finally, we advocate for the development of a new multidisciplinary research space on regulation and NLP (RegNLP), focused on connecting scientific knowledge to regulatory processes based on systematic methodologies.

Human trafficking (HT) is a pervasive global issue affecting vulnerable individuals, violating their fundamental human rights. Investigations reveal that many HT cases are associated with online advertisements (ads), particularly in escort markets. Consequently, identifying and connecting HT vendors has become increasingly challenging for Law Enforcement Agencies (LEAs). To address this issue, we introduce IDTraffickers, an extensive dataset consisting of 87,595 text ads and 5,244 vendor labels to enable the verification and identification of potential HT vendors on online escort markets. To establish a benchmark for authorship identification, we train a DeCLUTR-small model, achieving a macro-F1 score of 0.8656 in a closed-set classification environment. Next, we leverage the style representations extracted from the trained classifier to conduct authorship verification, resulting in a mean r-precision score of 0.8852 in an open-set ranking environment. Finally, to encourage further research and ensure responsible data sharing, we plan to release IDTraffickers for the authorship attribution task to researchers under specific conditions, considering the sensitive nature of the data. We believe that the availability of our dataset and benchmarks will empower future researchers to utilize our findings, thereby facilitating the effective linkage of escort ads and the development of more robust approaches for identifying HT indicators (1).

With the widespread use of e-commerce, proper oversight and regulatory compliance become increasingly difficult, if not impossible, resulting in a heightened risk of harm to consumers from unsafe products. In this paper, we explore how online consumer reviews can be utilized to identify hazardous products that have previously been flagged in the European Union Safety Gate reports. Our research presents a general framework that can be beneficial for regulatory authorities, as well as a specific application to consumer electronics. We contribute a dataset of 3000 reviews of electronic products, 755 of which reference hazardous products, and conduct classification baselines, achieving an AUC of up to 80% with room for improvement. Furthermore, we discuss the legal basis for annotation and potential issues that may arise. Our proposed methodology and dataset are valuable resources for regulatory authorities in the European Union and provide evidence of the effectiveness of digital surveillance in protecting consumers.

Statutory article retrieval is the task of automatically retrieving law articles relevant to a legal question.While recent advances in natural language processing have sparked considerable interest in many legal tasks, statutory article retrieval remains primarily untouched due to the scarcity of large-scale and high-quality annotated datasets.To address this bottleneck, we introduce the Belgian Statutory Article Retrieval Dataset (BSARD), which consists of 1,100+ French native legal questions labeled by experienced jurists with relevant articles from a corpus of 22,600+ Belgian law articles.Using BSARD, we benchmark several state-of-theart retrieval approaches, including lexical and dense architectures, both in zero-shot and supervised setups.We find that fine-tuned dense retrieval models significantly outperform other systems.Our best performing baseline achieves 74.8% R@100, which is promising for the feasibility of the task and indicates there is still room for improvement.By the specificity of the domain and addressed task, BSARD presents a unique challenge problem for future research on legal information retrieval.Our dataset and source code are publicly available."Should physicians, surgeons, health officers, pharmacists, midwives, and all others who, through their status or profession, be in possession of information confided to them reveal such secrets, they shall be punished with imprisonment of one to three years and a fine of 100 to 1000 euros or one of these penalties only -unless called to testify as a witness in a court of law (or before a parliamentary commission of inquiry) or compelled by a decree or order to divulge the secret."

Detecting false information in the form of fake news has become a bigger challenge than anticipated. There are multiple promising ways of approaching such a problem, ranging from source-based detection, linguistic feature extraction, and sentiment analysis of articles. While analyzing the sentiment of text has produced some promising results, this paper explores a rather more fine-grained strategy of classifying news as fake or real, based solely on the emotion profile of an article's title. A RoBERTa model was first trained to perform Emotion Classification, achieving test accuracy of about 90%. Six basic emotions were used for the task, based on the prominent psychologist Paul Ekman - fear, joy, anger, sadness, disgust and surprise. A seventh emotional category was also added to represent neutral text. Model performance was also validated by comparing classification results to other state-of-the-art models, developed by other groups. The model was then used to make inference on the emotion profile of news titles, returning a probability vector, which describes the emotion that the title conveys. Having the emotion probability vectors for each article's title, another Binary Random Forest classifier model was trained to evaluate news as either fake or real, based solely on their emotion profile. The model achieved up to 88% accuracy on the Kaggle Fake and Real News Dataset, showing there is a connection present between the emotion profile of news titles and if the article is fake or real.

Harmony in visual compositions is a concept that cannot be defined or easily expressed mathematically, even by humans. The goal of the research described in this paper was to find a numerical representation of artistic compositions with different levels of harmony. We ask humans to rate a collection of grayscale images based on the harmony they convey. To represent the images, a set of special features were designed and extracted. By doing so, it became possible to assign objective measures to subjectively judged compositions. Given the ratings and the extracted features, we utilized machine learning algorithms to evaluate the efficiency of such representations in a harmony classification problem. The best performing model (SVM) achieved 80% accuracy in distinguishing between harmonic and disharmonic images, which reinforces the assumption that concept of harmony can be expressed in a mathematical way that can be assessed by humans.

The tiled convolutional neural network (TCNN) has been applied only to computer vision for learning invariances. We adjust its architecture to NLP to improve the extraction of the most salient features for sentiment analysis. Knowing that the major drawback of the TCNN in the NLP field is its inflexible filter structure, we propose a novel architecture called hybrid tiled convolutional neural network (HTCNN) that applies a filter only on the words that appear in similar contexts and on their neighbouring words (a necessary step for preventing the loss of some n-grams). The experiments on the IMDB movie reviews dataset demonstrate the effectiveness of the HTCNN that has a higher level of performance of more than 3% and 1% respectively than both the convolutional neural network (CNN) and the TCNN. These results are confirmed by the SemEval-2017 dataset where the recall of the HTCNN model exceeds by more than six percentage points the recall of its simple variant, CNN.

Encoder-decoder models provide a generic architecture for sequence-to-sequence tasks such as speech recognition and translation. While offline systems are often evaluated on quality metrics like word error rates (WER) and BLEU, latency is also a crucial factor in many practical use-cases. We propose three latency reduction techniques for chunk-based incremental inference and evaluate their efficiency in terms of accuracy-latency trade-off. On the 300-hour How2 dataset, we reduce latency by 83% to 0.8 second by sacrificing 1% WER (6% rel.) compared to offline transcription. Although our experiments use the Transformer, the hypothesis selection strategies are applicable to other encoder-decoder models. To avoid expensive re-computation, we use a unidirectionally-attending encoder. After an adaptation procedure to partial sequences, the unidirectional model performs on-par with the original model. We further show that our approach is also applicable to low-latency speech translation. On How2 English-Portuguese speech translation, we reduce latency to 0.7 second (-84% rel.) while incurring a loss of 2.4 BLEU points (5% rel.) compared to the offline system.

An obstacle to the development of many natural language processing products is the vast amount of training examples necessary to get satisfactory results. The generation of these examples is often a tedious and time-consuming task. This paper this paper proposes a method to transform the sentiment of sentences in order to limit the work necessary to generate more training data. This means that one sentence can be transformed to an opposite sentiment sentence and should reduce by half the work required in the generation of text. The proposed pipeline consists of a sentiment classifier with an attention mechanism to highlight the short phrases that determine the sentiment of a sentence. Then, these phrases are changed to phrases of the opposite sentiment using a baseline model and an autoencoder approach. Experiments are run on both the separate parts of the pipeline as well as on the end-to-end model. The sentiment classifier is tested on its accuracy and is found to perform adequately. The autoencoder is tested on how well it is able to change the sentiment of an encoded phrase and it was found that such a task is possible. We use human evaluation to judge the performance of the full (end-to-end) pipeline and that reveals that a model using word vectors outperforms the encoder model. Numerical evaluation shows that a success rate of 54.7% is achieved on the sentiment change.

Conversational agents have begun to rise both in the academic (in terms of research) and commercial (in terms of applications) world. This paper investigates the task of building a non-goal driven conversational agent, using neural network generative models and analyzes how the conversation context is handled. It compares a simpler Encoder-Decoder with a Hierarchical Recurrent Encoder-Decoder architecture, which includes an additional module to model the context of the conversation using previous utterances information. We found that the hierarchical model was able to extract relevant context information and include them in the generation of the output. However, it performed worse (35-40%) than the simple Encoder-Decoder model regarding both grammatically correct output and meaningful response. Despite these results, experiments demonstrate how conversations about similar topics appear close to each other in the context space due to the increased frequency of specific topic-related words, thus leaving promising directions for future research and how the context of a conversation can be exploited.

Domains such as logo synthesis, in which the data has a high degree of multi-modality, still pose a challenge for generative adversarial networks (GANs). Recent research shows that progressive training (ProGAN) and mapping network extensions (StyleGAN) enable both increased training stability for higher dimensional problems and better feature separation within the embedded latent space. However, these architectures leave limited control over shaping the output of the network. This paper explores a conditional extension to the StyleGAN architecture with the aim of firstly, improving on the low resolution results of previous research and, secondly, increasing the controllability of the output through the use of synthetic class-conditions. Furthermore, methods of extracting such class conditions are explored, where the challenge lies in the fact that, visual logo characteristics are hard to define. The introduced conditional style-based generator architecture is trained on the extracted class-conditions in two experiments and studied relative to the performance of an unconditional model. Results show that, whilst the unconditional model more closely matches the training distribution, high quality conditions enabled the embedding of finer details onto the latent space, leading to more diverse output.

As of February 2016 Facebook allows users to express their experienced emotions about a post by using five so-called `reactions'. This research paper proposes and evaluates alternative methods for predicting these reactions to user posts on public pages of firms/companies (like supermarket chains). For this purpose, we collected posts (and their reactions) from Facebook pages of large supermarket chains and constructed a dataset which is available for other researches. In order to predict the distribution of reactions of a new post, neural network architectures (convolutional and recurrent neural networks) were tested using pretrained word embeddings. Results of the neural networks were improved by introducing a bootstrapping approach for sentiment and emotion mining on the comments for each post. The final model (a combination of neural network and a baseline emotion miner) is able to predict the reaction distribution on Facebook posts with a mean squared error (or misclassification rate) of 0.135.

Designing a logo is a long, complicated, and expensive process for any designer. However, recent advancements in generative algorithms provide models that could offer a possible solution. Logos are multi-modal, have very few categorical properties, and do not have a continuous latent space. Yet, conditional generative adversarial networks can be used to generate logos that could help designers in their creative process. We propose LoGAN: an improved auxiliary classifier Wasserstein generative adversarial neural network (with gradient penalty) that is able to generate logos conditioned on twelve different colors. In 768 generated instances (12 classes and 64 logos per class), when looking at the most prominent color, the conditional generation part of the model has an overall precision and recall of 0.8 and 0.7 respectively. LoGAN's results offer a first glance at how artificial intelligence can be used to assist designers in their creative process and open promising future directions, such as including more descriptive labels which will provide a more exhaustive and easy-to-use system.

This work addresses the instability in asynchronous data parallel optimization. It does so by introducing a novel distributed optimizer which is able to efficiently optimize a centralized model under communication constraints. The optimizer achieves this by pushing a normalized sequence of first-order gradients to a parameter server. This implies that the magnitude of a worker delta is smaller compared to an accumulated gradient, and provides a better direction towards a minimum compared to first-order gradients, which in turn also forces possible implicit momentum fluctuations to be more aligned since we make the assumption that all workers contribute towards a single minima. As a result, our approach mitigates the parameter staleness problem more effectively since staleness in asynchrony induces (implicit) momentum, and achieves a better convergence rate compared to other optimizers such as asynchronous \textsceasgd and \textscdynsgd, which we show empirically.

Assessing learning outcomes for students in higher education institutes is an interesting task with many potential applications for all involved stakeholders (students, administrators, potential employers, etc.). In this paper, we propose a course recommendation system for students based on the assessment of their "graduate attributes" (i.e. attributes that describe the developing values of students). Students rate the improvement in their graduating attributes after a course is finished and a collaborative filtering algorithm is utilized in order to suggest courses that were taken by fellow students and rated in a similar way. An extension to weigh the most recent ratings as more important is included in the algorithm which is shown to have better accuracy than the baseline approach. Experimental results using correlation thresholding and the nearest neighbors approach show that such a recommendation system can be effective when an active neighborhood of 10-15 students is used and show that the numbers of users used can be decreased effectively to one fourth of the whole population for improving the performance of the algorithm.

Finding semantically rich and computer-understandable representations for textual dialogues, utterances and words is crucial for dialogue systems (or conversational agents), as their performance mostly depends on understanding the context of conversations. In recent research approaches, responses have been generated utilizing a decoder architecture, given the distributed vector representation (embedding) of the current conversation. In this paper, the utilization of embeddings for answer retrieval is explored by using Locality-Sensitive Hashing Forest (LSH Forest), an Approximate Nearest Neighbor (ANN) model, to find similar conversations in a corpus and rank possible candidates. Experimental results on the well-known Ubuntu Corpus (in English) and a customer service chat dataset (in Dutch) show that, in combination with a candidate selection method, retrieval-based approaches outperform generative ones and reveal promising future research directions towards the usability of such a system.

Massive Open Online Courses (MOOCs) are attracting the attention of people all over the world. Regardless the platform, numbers of registrants for online courses are impressive but in the same time, completion rates are disappointing. Understanding the mechanisms of dropping out based on the learner profile arises as a crucial task in MOOCs, since it will allow intervening at the right moment in order to assist the learner in completing the course. In this paper, the dropout behaviour of learners in a MOOC is thoroughly studied by first extracting features that describe the behavior of learners within the course and then by comparing three classifiers (Logistic Regression, Random Forest and AdaBoost) in two tasks: predicting which users will have dropped out by a certain week and predicting which users will drop out on a specific week. The former has showed to be considerably easier, with all three classifiers performing equally well. However, the accuracy for the second task is lower, and Logistic Regression tends to perform slightly better than the other two algorithms. We found that features that reflect an active attitude of the user towards the MOOC, such as submitting their assignment, posting on the Forum and filling their Profile, are strong indicators of persistence.

The rise of internet and mobile technologies (such as smartphones) provide a harness of data and an opportunity to learn about peoples’ states, behavior, and context in regard to several application areas such as health. Eating behavior is an area that can benefit from the development of effective e-coaching applications which utilize psychological theories and data science techniques. In this paper, we propose a framework of how machine learning techniques can effectively be used in order to fully exploit data collected from a mobile application (“Think Slim”) which is designed to assess eating behavior using experience sampling methods. The overall goal is to analyze individual states of a person status (emotions, location, activity, etc.) and assess their impact on unhealthy eating. Building on data collected from different participants, a classification algorithm (decision tree tailored to longitudinal data) is used to warn people prior to a possible unhealthy eating event and a clustering algorithm (hierarchical agglomerative clustering) is used for profiling the participants and generalize for new users of the application. Finally, a framework to offer feedback via adaptive messages (intervention) and recommendations prior to possible unhealthy eating events is presented. Results from applying our methods reveal that participants can be clustered to six robust groups based on their eating behavior and that there are specific rules that discriminate which conditions lead to healthy versus unhealthy eating. Consequently, these rules can be utilized to provide adaptive semi-tailored feedback to users who, through this method, are assisted in learning under which conditions are more prone to unhealthy eating. Effectiveness of the approach is confirmed by observing a decreasing trend in rule activation towards the end of intervention period.

Ecological Momentary Assessment (EMA) data is organized in multiple levels (per-subject, per-day, etc.) and this particular structure should be taken into account in machine learning algorithms used in EMA like decision trees and its variants. We propose a new algorithm called BBT (standing for Bagged Boosted Trees) that is enhanced by a over/under sampling method and can provide better estimates for the conditional class probability function. Experimental results on a real-world dataset show that BBT can benefit EMA data classification and performance.

Self-Organizing Map (SOM) is a neural network model which is used to obtain a topology-preserving mapping from the (usually high dimensional) input/feature space to an output/map space of fewer dimensions (usually two or three in order to facilitate visualization). Neurons in the output space are connected with each other but this structure remains fixed throughout training and learning is achieved through the updating of neuron reference vectors in feature space. Despite the fact that growing variants of SOM overcome the fixed structure limitation they increase computational cost and also do not allow the removal of a neuron after its introduction. In this paper, a variant of SOM is proposed called AMSOM (Adaptive Moving Self-Organizing Map) that on the one hand creates a more flexible structure where neuron positions are dynamically altered during training and on the other hand tackles the drawback of having a predefined grid by allowing neuron addition and/or removal during training. Experiments using multiple literature datasets show that the proposed method improves training performance of SOM, leads to a better visualization of the input dataset and provides a framework for determining the optimal number and structure of neurons.

Ecological Momentary Assessment (EMA) techniques gain more ground in studies and data collection among different disciplines. Decision tree algorithms and their ensemble variants are widely used for classifying this type of data, since they are easy to use and provide satisfactory results. However, most of these algorithms do not take into account the multiple levels (per-subject, per-day, etc.) in which EMA data are organized. In this paper we explore how the EMA data organization can be taken into account when dealing with decision trees and specifically how a combination of bagging and boosting can be utilized in a classification task. A new algorithm called BBT (standing for Bagged Boosted Trees) is proposed which is enhanced by an over/under sampling method leading to better estimates of the conditional class probability function. BBT's necessity and effects are demonstrated using both simulated datasets and real-world EMA data collected using a mobile application following the eating behavior of 100 people. Experimental analysis shows that BBT leads to clear improvements with respect to prediction error reduction and conditional class probability estimation.

In this paper, we propose a novel method for conceptual hierarchical clustering of documents using knowledge extracted from Wikipedia. The proposed method overcomes the classic bag-of-words models disadvantages through the exploitation of Wikipedia textual content and link structure. A robust and compact document representation is built in real-time using the Wikipedia application programmer's interface, without the need to store locally any Wikipedia information. The clustering process is hierarchical and extends the idea of frequent items by using Wikipedia article titles for selecting cluster labels that are descriptive and important for the examined corpus. Experiments show that the proposed technique greatly improves over the baseline approach, both in terms of F-measure and entropy on the one hand and computational cost on the other.

In this paper, we build a hybrid Web-based metric for computing semantic relatedness between words. The method exploits page counts, titles, snippets and URLs returned by a Web search engine. Our technique uses traditional information retrieval methods and is enhanced by page-count-based similarity scores which are integrated with automatically extracted lexico-synantic patterns from titles, snippets and URLs for all kinds of semantically related words provided by WordNet (synonyms, hypernyms, meronyms, antonyms). A support vector machine is used to solve the arising regression problem of word relatedness and the proposed method is evaluated on standard benchmark datasets. The method achieves an overall correlation of 0.88, which is the highest among other metrics up to date.