Search

Topic models have shown to be one of the most effective tools in Content-Based Multimedia Retrieval (CBMR). However, the high computational learning cost together with the huge expansion of multimedia collections limit the scalability of topic-based CBMR systems in real-life multimedia applications. The present work pursues a twofold objective. On the one hand, to study the effect of using clustering-based document reduction schemes over standard topic models pLSA (probabilistic Latent Semantic Analysis) and LDA (Latent Dirichlet Allocation). On the other hand, to develop a pLSA-based extension oriented to integrate this reduction scheme within the own model in order to improve the CBMR effectiveness. The experimental part of the work includes three different multimedia databases, three ranking functions, four retrieval scenarios, three different numbers of topics and ten document reduction levels. Experiments revealed that standard topic models are highly sensitive to the document reduction level whereas the proposed model is able to provide a competitive advantage within the content-based retrieval field.

Nanotechnology is an emerging and promising field of research. Creating sufficient technological diversity among its alternatives is important for the long-term success of nanotechnologies, as well as for other emerging technologies. Diversity prevents early lock-in, facilitates recombinant innovation, increases resilience, and allows market growth. Creation of new technological alternatives usually takes place in innovation projects in which public and private partners often collaborate. Currently, there is little empirical evidence about which characteristics of innovation projects influence diversity. In this paper we study the influence of characteristics of EU-funded nanotechnology projects on the creation of technological diversity. In addition to actor diversity and the network of the project, we also include novel variables that have a plausible influence on diversity creation: the degree of multi-disciplinarity of the project and the size of the joint knowledge base of project partners. We apply topic modelling (Latent Dirichlet allocation) as a novel method to categorize technological alternatives. Using an ordinal logistic regression model, our results show that the largest contribution to diversity comes from the multi-disciplinary nature of a project. The joint knowledge base of project partners and the geographical distance between them were positively associated with technological diversity creation. In contrast, the number and diversity of actors and the degree of clustering showed a negative association with technological diversity creation. These results extend current micro-level explanations of how the diversity of an emerging technology is created. The contribution of this study could also be helpful for policy makers to influence the level of diversity in a technological field, and hence to contribute to survival of emerging technologies., The financial support of SENESCYT - Ecuador National Secretary of Higher Education, Science, Technology and Innovation (Grant Convocatoria Abierta 2013 - No. AR2Q) is gratefully acknowledged.

The growth of social media in recent years has led to an increase in online reviews that reflects consumer opinions. Firms benefit greatly from making this information available in order to respond more effectively to consumer dissatisfaction and to exploit market opportunities by observing standards that may represent unsatisfied needs. The present study aims to address this problem through a survey based on the Yelp platform. To this end, 14,000 comments related to different tourism products were used and text mining techniques and topic models were applied to find the main latent topics discussed in the online comments and their associated sentiments. The study presents 20 latent topics from online discussions and reveals that the topic that discusses “Air Travel” themes is the one with a lower sentiment connotation on average and should therefore be the subject of a deeper evaluation. info:eu-repo/semantics/acceptedVersion

In this article, we propose a framework to tackle conflict prevention, an issue which has received interest in several policy areas. A key challenge of conflict forecasting for prevention is that outbreaks of conflict in previously peaceful countries are rare events and therefore hard to predict. To make progress in this hard problem, this project summarizes more than four million newspaper articles using a topic model. The topics are then fed into a random forest to predict conflict risk, which is then integrated into a simple static framework in which a decision maker decides on the optimal number of interventions to minimize the total cost of conflict and intervention. According to the stylized model, cost savings compared to not intervening pre-conflict are over US$1 trillion even with relatively ineffective interventions and US$13 trillion with effective interventions, Mueller acknowledges financial support from the Banco de España, the Spanish Ministry of Economy and Competitiveness, through the Severo Ochoa Programme for Centres of Excellence in R&D (CEX2019-000915-S) and the Ayudas Fundación BBVA. The authors declare that they have no competing financial interests. All errors are ours. Mueller is also affiliated with MOVE and the CEPR. Rauh is affiliated with CEPR, HCEO, and Trinity College Cambridge.

A single abstract from the DHd-2019 Book of Abstracts., Sofern eine editorische Arbeit an dieser Publikation stattgefunden hat, dann bestand diese aus der Eliminierung von Bindestrichen in Überschriften, die aufgrund fehlerhafter Silbentrennung entstanden sind, der Vereinheitlichung von Namen der Autor*innen in das Schema "Nachname, Vorname" und/oder der Trennung von Überschrift und Unterüberschrift durch die Setzung eines Punktes, sofern notwendig., {"references":["https://doi.org/10.5281/zenodo.2600812","https://github.com/DHd-Verband/DHd-Abstracts-2019"]}

Medikal alanda, sağlık personeli tarafından hastalara uygulanan operasyonları kayıt altında tutmak son derece önemli ve yasal bir zorunluluk olmasına rağmen, genellikle operasyon listel-eri eksiktir. Operasyon listelerindeki ihmaller hastalar için beklenmeyen sonuçlara neden olabilmektedir. Ek olarak, operasyon listeleri faturalandırma sürecinde kullanıldığından, kabul edilemez operasyon listeleri hem sağlık kurumları hem de hastalar için finansal sorunlara neden olmaktadır. Bu nedenle, bu tez çalışmasının temel amacı, hem insan sağlığını tehdit eden hem de hastalar ve sağlık merkezleri için ekonomik sorunlara neden olan operasyon listelerindeki ihmalleri tahmin edebilen bir uzman tavsiye sistemi geliştirmektir. Bu tez çalışmasında, Latent Dirichlet Allocation yöntemini kullanarak operasyon listelerindeki ihmalleri tahmin etmeye çalışan önceki denenmiş çözümlerden farklı yeni bir model öneriyoruz, önerilen yöntem ICD-10 kodunu yeni gözlemlenen bir değişken olarak kullanıyor. İlk deneyler daha önce bu alanda başarı elde etmiş olan Lojistik Regresyon ve Latent Dirichlet Allocation yöntemleri ile gerçekleştirilmiştir. Kesinlik, hatırlama, F1 ölçümü ve Ortalama Karşılıklı Sıra değerleri, değer-lendirme ölçütleri olarak kullanılır ve önerilen modelin, Lojistik Regresyon ve klasik Latent Dirichlet Allocation yöntemiyle sonuçları değerlendirme ölçütlerine göre karşılaştırılır. Üç farklı veri setinde yapılan deney sonuçlarına göre, önerilen yöntemin LDA yönteminden %5, Lojistik Regresyon yönteminden %13 daha başarılı olduğu tespit edilmiştir. In the medical field, although it is extremely important and a legal obligation to record the pro-cedures applied to patients by the health personnel, generally the operation lists are incomplete. Omissions in the operation lists can cause unexpected results for patients. In addition, inadmis-sible operation lists on billing operations applied to patients cause financial problems for both health institutions and patients because operation lists are used for invoicing process. There-fore, the main objective of this study is to develop an expert recommender system which can predict the omissions in the operation lists with a high success rate, which both threaten human health and cause economic problems for patients and medical centers. In this thesis study, we propose a new model different from the previous attempted solutions which tried to predict omissions in the operation lists using the Latent Dirichlet Allocation method, the proposed method uses the ICD-10 code as a new observed variable. The first experiments are carried out with Logistic Regression and Latent Dirichlet Allocation methods which had previously achieved success in this field. Precision, recall, F1 measure and MRR values are used as evalu-ation metrics, and the results of the proposed model with the Logistic regression method and the classical Latent Dirichlet Allocation method are compared based on the evaluation metrics. According to the results of the experiments conducted on three different datasets, it is ob-served that the Proposed Method is 5% more successful than the LDA method and 13% more successful than the Logistic Regression method. 99

Scientific literature search engines typically index abstracts instead of the full-text of publications. The expectation is that the abstract provides a comprehensive summary of the article, enumerating key points for the reader to assess whether their information needs could be satisfied by reading the full-text. Furthermore, from a practical standpoint, obtaining the full-text is more complicated due to licensing issues, in the case of commercial publishers, and resource limitations of public repositories and pre-print servers. In this article, we use topic modelling to represent content in abstracts and full-text articles. Using Computer Science as a case study, we demonstrate that how well the abstract summarises the full-text is subfield-dependent. Indeed, we show that abstract representativeness has a direct impact on retrieval performance, with poorer abstracts leading to degraded performance. Finally, we present evidence that how well an abstract represents the full-text of an article is not random, but is a consequence of style and writing conventions in different subdisciplines and can be used to infer an "evolutionary" tree of subfields within Computer Science.

Rapid growth over the past two decades in digitized textual information represents untapped potential for methodological innovations in the adaptation governance literature that draw on machine learning approaches already being applied in other areas of computational social sciences. This Focus Article explores the potential for text mining techniques, specifically topic modeling, to leverage this data for large‐scale analysis of the content of adaptation policy documents. We provide an overview of the assumptions and procedures that underlie the use of topic modeling, and discuss key areas in the adaptation governance literature where topic modeling could provide valuable insights. We demonstrate the diversity of potential applications for topic modeling with two examples that examine: (a) how adaptation is being talked about by political leaders in United Nations Framework Convention on Climate Change; and (b) how adaptation is being discussed by decision‐makers and public administrators in Canadian municipalities using documents collected from 25 city council archives. This article is categorized under: Vulnerability and Adaptation to Climate Change > Institutions for Adaptation

Political competition, party strategy and communication in the era of social media are growing issues. Due to the increasing social media presence of parties and voters alike, direct communication is more important for party competition. This paper aims to improve the methodological approach used to analyze political competition and communication. The dataset includes over 30,000 Facebook status messages posted by seven German parties from January 2014 until February 2018. Topic modeling, which is commonly used in other fields, allows for evaluating party communication on a daily basis. The results show the high accuracy of calculating party-relevant issues. To determine the tone of the debate, a sentiment analysis was conducted. The prevalence of topics and sentiments over time allows for precise monitoring of the political debate.

Rad se bavi računalnim postupcima analize medijske agende (engl. Media Agenda) temeljenima na tematskim modelima (engl. Topic Models) te metodama vrednovanja tematskih modela. Analiza medijske agende provodi se radi stjecanja uvida u strukturu i zastupljenost medijskih tema, što je od interesa za društvenoznanstvena istraživanja te za medijsku industriju i druge komercijalne i političke aktere. Računalni postupci analize medijske agende omogućuju auto- matsko otkrivanje tema u velikim skupovima tekstova i mjerenje njihove zastupljenosti. Ovi postupci pružaju analitičaru uvid u teme prisutne u medijima i uvid u zastupljenost tema u poje- dinim medijima i vremenskim razdobljima te omogućuju analizu korelacije zastupljenosti tema sa podacima poput ljudske percepcije njihove važnosti. Cilj istraživanja bio je razvoj računalnih postupaka za eksplorativnu analizu i mjerenje me- dijske agende temeljenih na tematskim modelima, klasi modela strojnog učenja pogodnih za analizu tematske strukture teksta. Istraživanje obuhvaća razvoj postupaka primjene tematskih modela na otkrivanje medijskih tema i mjerenje njihove zastupljenosti te razvoj računalnih alata za unaprijed̄enje i provedbu tih postupaka. Ti alati obuhvaćaju metode vrednovanja tematskih modela te programsku potporu za implementaciju postupaka analize agende i vrednovanja mo- dela. Primjena postupaka na analizu medijskih tekstova brzo je pokazala potrebu za razvojem novih metoda vrednovanja tematskih modela radi povećanja efikasnosti na modelima temelje- nih postupaka. Iz tog je razloga poseban naglasak istraživanja bio na razvoju i analizi metoda vrednovanja tematskih modela. Prvo je provedeno istraživanje postupaka primjene tematskih modela na analizu medijske agende. Na temelju istraživanja postojećih postupaka predložen je poboljšani postupak koji se sastoji od tri koraka: koraka otkrivanja tema, koraka definicije tema i koraka mjerenja tema. Predloženi postupak otklanja uočene nedostatke ranijih metoda: upotrebu samo jednog modela za otkrivanje tema, nemogućnost prilagodbe i definicije novih tema te izostanak kvantitativnog vrednovanja metoda mjerenja. Postupak je primijenjen u dvije analize medijske agende prove- dene na zbirkama američkih i hrvatskih političkih vijesti. Na temelju opažanja i podataka iz tih analiza uočena je potreba za mjerom interpretabilnosti tema modela te za metodom mjerenja pokrivenosti skupa koncepata od strane modela. Drugi istraženi problem bio je problem mjerenja interpretabilnosti tema modela. Standardni pristup ovom problemu je mjerenje semantičke koherentosti tema, a postojeće mjere koherent- nosti temelje se na računanju koherentosti skupa uz temu vezanih riječi. Ove mjere pokazale su se nepogodnima u slučaju prolaznih medijskih tema karakteriziranih semantički nepovezanim riječima. Predložena je nova klasa mjera koherentosti medijskih tema temeljenih na uz teme vezanim dokumentima. Vrednovanje niza predloženih mjera na skupovima engleskih i hrvat- skih medijskih tema otkrilo je najbolju mjeru koja računa koherentnost agregacijom lokalne povezanosti grafa dokumenata. Provedena je kvantitativna i kvalitativna usporedba razvijenih mjera dokumentne koherentosti s postojećim mjerama koherentnosti riječi koja je otkrila kom- plementarnost ova dva tipa mjera. Treći istraženi problem je problem pokrivenosti tema, motiviran podacima iz primjene pos- tupka analize medijske agende, koji su pokazali da jedan tematski model pokriva samo dio svih otkrivenih koncepata. Problem pokrivenosti nadilazi domenu medijskih tekstova i unatoč važnosti ovog problema dosadašnja istraživanja na tu temu su rudimentarana. Problem pokri- venosti razmotren je u općenitosti i definiran kao problem mjerenja poklapanja izmed̄u skupa automatski naučenih tema modela i skupa referentnih tema koji sadrži od ljudi uočene koncepte. Predložena je metoda izrade skupa referentnih tema i dvije metode mjerenja pokrivenosti teme- ljene na računanju poklapanja tema. Predložene mjere vrednovane su na dva raznorodna skupa podataka, medijskom i biološkom, te primijenjene na analizu četiri različite klase standardnih tematskih modela. Završni korak istraživanja postupka analize medijske agende bio je poboljšanje postupka na temelju predloženih metoda vrednovanja tematskih modela i iskustava iz primjena postupka na analizu hrvatskih i američkih medija. Glavna poboljšanja odnose se na korak eksplorativne analize odnosno otkrivanja tema i temelje se na razvijenim mjerama pokrivenosti i dokumentne koherentosti tema. Ova poboljšanja imaju za cilj brže otkrivanje većeg broja koncepata. Ostala poboljšanja odnose se na povećanje efikasnosti postupka interpretacije tema modela. Tijekom istraživanja postupka analize medijske agende i metoda vrednovanja tematskih mo- dela uočen je niz problema vezanih uz upotrebu, izgradnju, pohranu i dohvat tematskih modela i vezanih resursa. Ovi problemi javljaju se kod implementacije grafičkog korisničkog sučelja za provedbu postupka i kod provedbe eksperimenata vrednovanja. Rješavanju ovih problema pristupilo se sustavno i oblikovan je radni okvir za izgradnju i upravljanje resursima u temat- skom modeliranju. Arhitektura okvira temelji se na četiri načela koja u kombinaciji definiraju općenitu i fleksibilnu metodu izrade programske potpore za primjenu i vrednovanje tematskih modela. Razvijeni su i grafičko korisničko sučelje za eksplorativnu analizu i potporu mjere- nju zastupljenosti tema te aplikacija namijenjena izradi zbirki medijskih tekstova koja tijekom duljeg vremenskog razdoblja sakuplja tekstove iz niza web-izvora. This thesis focuses on computational methods for media agenda analysis based on topic mo- dels and methods of topic model evaluation. The goal of a media agenda analysis is gaining insights into the structure and frequency of media topics. Such analyses are of interest for social scientists studying news media, journalists, media analysts, and other commercial and political actors. Computational methods for media agenda analysis enable automatic discovery of topics in large corpora of news text and measuring of topics’ frequency. Data obtained by such analyses provides insights into the type and structure of topics occurring in the media, enables the analysis of topic cooccurrence, and analysis of correlation between topics and other variables such as text metadata and human perception of topic significance. The goal of the research presented in the thesis is development of efficient computational methods for the discovery of topics that constitute the media agenda and methods for measu- ring frequencies of these topics. The proposed methods are based on topic models – a class of unsupervised machine learning models widely used for exploratory analysis of topical text structure. The research encompasses the development of applications of topic models for dis- covery of media topics and for measuring topics’ frequency, as well as development of methods for improvement and facilitation of these applications. The improvement and facilitation met- hods encompass methods of topic model evaluation and software tools for working with topic models. Methods of topic model evaluation can be used for selection of high-quality models and for accelerating the process of topic discovery. Namely, topic models are a useful tool, but due to the stohasticity of the model learning algorithms the quality of learned topics varies. For this reason the methods of topic model evaluation have the potential to increase the efficiency of the methods based on topic models. Media agenda consists of a set of topics discussed in the media, and the problem of media agenda analysis consists of two sub-tasks: discovery of the topics on the agenda and measu- ring the frequencies of these topics. The first contribution of the thesis is a method for media agenda analysis based on topic models that builds upon previous approaches to the problem and addresses their deficiencies. Three notable deficiencies are: usage of a single topic model for topic discovery, lack of possibility to define new topics that match the analyst’s interests, and the lack of precise evaluation of methods for measuring topics’ frequency. In addition to addressing the identified deficiencies, the method also systematizes the previous approaches to the problem and is evaluated in two case studies of media agenda analysis. The proposed experimental method for media agenda analysis consists of three steps: topic discovery, topic definition, and topic measuring steps. In order to achieve better topic coverage, the discovery step is based not on a single model but on a set of topic models. The type and number of topic models used depends on available model implementations and the time available for topic annotation, while the hyperparameter defining the number of model topics depends on the desired generality of learned topics. Reaso- nable default settings for model construction are proposed based on the existing agenda analysis studies and an iterative procedure for tuning the number of topics is described. After the topic models are constructed, topic discovery is performed by human inspection and interpretation of the topics. Topic interpretation produces semantic topics (concepts) that are recorded in a reference table of semantic topics that serves as a record of topics and as a tool for synchroni- zation of human annotators. After all the model topics are inspected, annotators can optionally perform the error correcting step of revising the semantic topics, as well as the step of building a taxonomy of semantic topics. Topic discovery is supported with a graphical user interface developed for topic inspection and annotation. The step of topic definition is based on semantic topics obtained during topic discovery. The purpose of topic definition is to define new semantic topics that closely match the analyst’s exact interests. The possibility of defining new semantic topics is an important difference between the proposed and the existing media agenda analysis approaches. Namely, the existing approaches base the analysis only on model-produced topics, although there is no guarantee that these topics will match the concepts of interest to the analyst. During topic definition, the analysts infers definitions of new semantic topics based on previously discovered topics and describes these topics with word lists. Discovered semantic topics that already closely match the concepts of interest are used without modification. During the step of topic measuring the frequencies of semantic topics obtained during the discovery and definition steps are measured. Topic frequency is defined as the number of news articles in which a topic occurrs, and the measuring problem is cast as the problem of multi-label classification in which each news article is being tagged with one or more semantic topics. This formulation allows for precise quantitative evaluation of methods for measuring topic frequency. Two measuring methods are considered. The baseline is a supervised method using the method of binary relevance in combination with a linear kernel SVM model. The second method is a newly proposed weakly supervised approach, in which the measured semantic topics are first described by sets of highly discriminative words, after which a new LDA model is constructed in such a way that the topics of the model correspond to measured topics, which is achieved via prior probabilities of model topics. The method for selecting words highly discriminative for a semantic topic represents the main difference between the proposed and the previous weakly supervised approaches. This method consists of inspecting, for each measured semantic topic, closely related model topics, and selecting words highly discriminative for the topic by means of inspecting word-related documents and assessing their correspondence with the topic. The proposed three-step method for media agenda analysis is applied to two media agenda analyses: the analysis of mainstream US political news and the analysis of mainstream Croatian political news in the election period. The applications of the proposed method show that the topic discovery step gives a good overview of the media agenda and leads to the discovery of useful topics, and that the usage of more than one topic model leads to a more comprehensive set of topics. The two analyses also demonstrate the necessity of the proposed topic definition step – in the case of US news new sensible topics corresponding to issues are pinpointed during this step, while in the case of Croatian election-related news the analysis is based entirely on newly defined semantic topics that describe the pre- and post-election processes. Quantitative evaluation of topic frequency measuring shows that the proposed weakly supervised approach works better than the supervised SVM-based method since it achieves better or comparable performance with less labeling effort. In contrast to the supervised method, weakly supervi- sed models have a higher recall and work well for smaller topics. Qualitative evaluation of measuring models confirms the quality of the proposed approach – measured topic frequency correlates well with real-world events and the election-related conclusions based on measuring models are in line with conclusions drawn from social-scientific studies. Observations from two media agenda analysis studies and the analysis of collected topic data underlined two problems related to methods of topic model evaluation. The first is the problem of measuring topic quality – the studies both confirmed variations in topic quality and indicated the inadequacy of existing word-based measures of topic coherence. The second is the problem of topic coverage – while the data confirms the limited ability of a single topic model to cover all the semantic topics, no available methods for measuring topic coverage exist, so it is not possible to identify the high-coverage models. These observations motivated the development of new methods of topic model evaluation – document-based coherence measures and methods for topic coverage analysis. As described, the analysis of topics produced during the applications of topic discovery confirmed variations in topics’ quality and underlined the need for better measures of topic quality. The analysis also indicated that existing word-based measures of topic coherence are inadequate for evaluating quality of media topics often characterized by semantically unrelated word sets. Based on the observation that media topics can be successfully interpreted using topic-related documents, a new class of document-based topic coherence measures is proposed. The proposed measures calculate topic coherence in three steps: selection of topic-related documents, document vectorization, and computation of the coherence score from document vectors. Topic-related documents are selected using a simple model-independent strategy – a fixed number of documents with top document-topic weights is selected. Two families of docu- ment vectorization methods are considered. The first family consists of two standard methods based on calculation of word and document frequencies: probabilistic bag-of-words vectoriza- tion and tf-idf vectorization. Methods in the second family vectorize documents by aggregating either CBOW or GloVe word embeddings. Three types of methods are considered for cohe- rence score computation: distance-based methods that model coherence via mutual document distance, probability-based methods that model coherence as probabilistic compactness of do- cument vectors, and graph-based methods that model coherence via connectivity of the docu- ment graph. The space of all the coherence measures is parametrized and sensible parameter values are defined to obtain a smaller set of several thousand measures. Then the selection and evaluation of the coherence measures is performed, using model topics manually labeled with document-based coherence scores and using the area under the ROC curve (AUC) as the per- formance criterion. The measures are partitioned in structural categories and the best measure from each category is selected using AUC on the development set as a criterion. These best measures are then evaluated on two test sets containing English and Croatian news topics. The evaluation of document-based coherence measures shows that the graph-based measu- res achieve best results. Namely, best approximators of human coherence scores are the graph- based measures that use frequency-based document vectorization, build sparse graphs of locally connected documents and calculate coherence by aggregating a local connectivity score such as closeness centrality. Quantitative evaluation of word-based measures confirms the observations that word-based measures fail to approximate document-based coherence scores well and qu- alitative evaluation of coherence measures indicates that document- and word-based coherence measures complement each other and should be used in combination to obtain a more complete model of topic coherence. Motivated by the data from the topic discovery steps performed in two media agenda analyses and by the obvious need to increase the number of topics discovered by a single topic model, the problem of topic coverage is defined and solutions are proposed. This problem occurrs in ap- plication of topic models to any text domain, i.e., it is domain-independent and extends beyond applications to media text. The problem of topic coverage consists of measuring how automati- cally learned model topics cover a set of reference topics – topical concepts defined by humans. Two basic aspects of the problem are the reference topics that represent the concepts topic mo- dels are expected to cover and the measures of topic coverage that calculate a score measuring overlap between the model topics and reference topics. Finally, the third aspect encompasses evaluation of a set of topic models using a reference set and coverage measures. The coverage experiments are conducted using two datasets that correspond to two separate text domains – news media texts and biological texts. Each dataset contains a text corpus, a set of reference topics, and a set of topic models. Reference topics consist of topics that standard topic models are expected to be able to cover. These topics are constructed by human inspection, selection, and modification of model-learned topics. Both sets of reference topics are representative of useful topics discovered during the process of exploratory text analysis. Two approaches to measuring topic coverage are developed – an approach based on super- vised approximation of topic matching and an unsupervised approach based on integrating co- verage across a range of topic-matching criteria. The supervised approach is based on building a classification model that approximates human intuition of topic matching. A binary classifier is learned from a set of topic pairs annotated with matching scores. Four standard classification models are considered: logistic regression, support vector machine, random forest, and multi- layer perceptron. Topic pairs are represented as distances of topic-related word and document vectors using four distinct distance measures: cosine, hellinger, L1 , and L2. Model selection and evaluation shows that the proposed method approximates human scores very well, and that logistic regression is the best-performing model. The second proposed method for measuring coverage uses a measure of topic distance and a distance threshold to approximate the equality of a reference topic and a model topic. The threshold value is varied and for each threshold coverage is calculated as a proportion of reference topics that are matched by at least one model topic at a distance below the threshold. Varying the threshold results in a curve with threshold values on the x-axis and coverage scores on the y-axis. The final coverage score is calculated as the area under this curve. This unsupervised measure of coverage, dubbed area under the coverage-distance curve, correlates very well with the supervised measures of coverage, while the curve itself is a useful tool for visual analysis of topic coverage. This measure enables the users to quickly perform coverage measurements on new domains, without the need to annotate topic pairs in order to construct a supervised coverage measure. Using the proposed coverage measures and two sets of reference topics, coverage experi- ments in two distinct text domains are performed. Experiments consist of measuring coverages obtained by a set of topic models of distinct types constructed using different hyperparameters. In addition to demonstrating application of coverage methods, the experiments show that the NMF model has high coverage scores, is robust to domain change and able to discover topics on a high level of precision. Nonparametric model based on Pitman-Yor priors achieves the best coverage for news topics. Two proposed methods of topic model evaluation – document-based coherence measures and methods devised for solving the coverage problem – are applied in order to improve the previously proposed topic-model-based method of media agenda analysis. The improvements refer to the step of topic discovery and lead to quicker discovery of a larger number of concepts. This is achieved by using more interpretable models with higher coverage, and by ordering mo- del topics, before human inspection, in the descending order of their coherence. These impro- vements conclude the contribution of the thesis related to the methods of computational media agenda analysis. The first improvement is based on the analysis of the coverage and document- based coherence scores measured for a large number of different topic models. The main result is the recommendation for using the NMF model as the default model for topic discovery, due to the fact that NMF proved as a robust, interpretable, and a high-coverage model with the ad- ditional advantage of being fast to train. In addition, the nonparametric topic model based on Pitman-Yor priors also proved as a good choice for exploratory analysis of news texts since it achieves a very high coverage. The second improvement is model-agnostic and consists of or- dering model topics, inspected during the topic discovery step, by descending topic coherence. This results in low-quality topics being pushed towards the end of the topic inspection queue. The experiments show that applying the best graph-based coherence measure in the described way significantly improves the discovery rate of semantic topics. Several other improvement recommendations are given based on the experience gained in the course of application of the media agenda analysis methods. These improvements include: improving the topic inspection and interpretation process by way of discarding the shared reference table of semantic topics, improving the step of measuring topics’ frequency, and using tools that lead to quick guided discovery of topics of interest. Research of the media agenda analysis methods and the methods of topic model evaluation revealed a number of technical problems related to usage, construction, storage, and retrieval of topic models. Namely, in topic modelling experiments it is often necessary to construct a large number of models by varying model types and various parameters of the construction process such as hyperparameters, low-level resources, and preprocessing components. A systematic solution to these problems is proposed – a framework for resource building and management in topic modeling. The framework’s architecture is based on four principles, which in com- bination define a general and flexible method for designing and building code for evaluation and application of topic models. In addition, an application for building corpora of media text by collecting texts from a set of web news feeds was developed, as well as a graphical user interface that supports the topic discovery and topic frequency measurement. The topic modelling framework, dubbed pytopia, is an object-oriented Python framework that can be viewed as a middleware framework located between application-level code and the algorithm-level frameworks such as TensorFlow. The framework’s architecture is based on four design principles: the principle of standard interfaces and the adaptation of various components to these interfaces, the principle of component identifiability, the principle of using the abs- traction dubbed Context to organize and retrieve components, and the principle of hierarchical compositionality that reflects the structure of text-mining components and facilitates their de- sign and implementation. The framework contains core functionality that supports the four de- sign principles, and functionality for component building, saving, and loading. The framework also contains a set of components related to topic modeling, ranging form basic resources such as dictionaries and corpora to complex components such as sets of vectorized texts and topic models. Finally, the framework includes several tools for topic model evaluation as well as logging and testing functionality. A corpus of media texts is a basis for media agenda analysis based on topic models. Mo- tivated by the need for a tool that enables maximum flexibility in defining and building such a corpus, an application for collecting texts from a set of news feeds is developed. This applica- tion, dubbed feedsucker, enables the user to build a corpus of news texts containing texts from a set of sources corresponding to the exact user interest. The user specifies a set of news feeds in a text format and runs the application that continuously collects new texts and stores them in a database. The application is Java-based, object-oriented, and extensible. This thesis describes the research of computational methods for media agenda analysis, which enable discovery and measuring of topics in large news corpora and find applications in a range of scientific and commercial analyses of media text. The researched methods are based on topic models, standard machine learning models for analysis of topical text structure. In the first phase of the research, an analysis of existing media agenda methods is performed and a new method that improves and systematizes the existing ones is proposed. The application of the proposed method in two use-cases underlined the need for new methods of topic model evaluation that would improve the efficiency of topic-model based tools. Consequently, two new methods of topic model evaluation are proposed – document-based measures of topic coherence and methods for analysis of topic coverage. These evaluation methods are then applied to improve the initially proposed method for media agenda analysis. In addition, research of topic model applications and methods of topic model evaluation led to a framework for resource building and management in topic modelling. The four main contributions of the thesis are: a method for computational analysis of the media agenda based on topic models, document- based measures of topic coherence, methods for analysis of topic coverage, and the framework for resource building and management in topic modelling. The research described in the thesis led to an improved method for media agenda analysis and new methods of topic model evaluation. The evaluation methods find applications more general than the media agenda analysis – the measures of document-based coherence are appli- cable to any topic-model-based analysis of news text while the methods related to the problem of topic coverage are domain-independent. These evaluation methods represent new approac- hes with a potential to provide new insights about topic models, a class of widely-used machine learning models of text. The topic modelling framework could serve the same purpose since it facilitates complex experiments.

Robust Bayesian models are appealing alternatives to standard models, providing protection from data that contains outliers or other departures from the model assumptions. Historically, robust models were mostly developed on a case-by-case basis; examples include robust linear regression, robust mixture models, and bursty topic models. In this paper we develop a general approach to robust Bayesian modeling. We show how to turn an existing Bayesian model into a robust model, and then develop a generic strategy for computing with it. We use our method to study robust variants of several models, including linear regression, Poisson regression, logistic regression, and probabilistic topic models. We discuss the connections between our methods and existing approaches, especially empirical Bayes and James-Stein estimation.

Patients in hospitals, particularly in critical care, are susceptible to many complications affecting morbidity and mortality. Digitized clinical data in electronic medical records can be effectively used to develop machine learning models to identify patients at risk of complications early and provide prioritized care to prevent complications. However, clinical data from heterogeneous sources within hospitals pose signi ficant modeling challenges. In particular, unstructured clinical notes are a valuable source of information containing regular assessments of the patient's condition but contain inconsistent abbreviations and lack the structure of formal documents. Our contributions in this paper are twofold. First, we present a new preprocessing technique for extracting features from informal clinical notes that can be used in a classifi cation model to identify patients at risk of developing complications. Second, we explore the use of collective matrix factorization, a multi-view learning technique, to model heterogeneous clinical data text-based features in combination with other measurements, such as clinical investigations, comorbidities, and demographic data. We present a detailed case study on postoperative respiratory failure using more than 700 patient records from the MIMIC II database. Our experiments demonstrate the ef ficacy of our preprocessing technique in extracting discriminatory features from clinical notes as well as the bene fits of multi-view learning to combine clinical measurements with text data for predicting complications.

Non-convex optimization has an important role in machine learning. However, the theoretical understanding of non-convex optimization remained rather limited. Studying efficient algorithms for non-convex optimization has attracted a great deal of attention from many researchers around the world but these problems are usually NP-hard to solve. In this paper, we have proposed a new algorithm namely GS-OPT (General Stochastic OPTimization) which is effective for solving the non-convex problems. Our idea is to combine two stochastic bounds of the objective function where they are made by a commonly discrete probability distribution namely Bernoulli. We consider GS-OPT carefully on both the theoretical and experimental aspects. We also apply GS-OPT for solving the posterior inference problem in the latent Dirichlet allocation. Empirical results show that our approach is often more efficient than previous ones.

This paper introduces a novel probabilistic activity modeling approach that mines recurrent sequential patterns called motifs from documents given as word $$\times $$ time count matrices (e.g., videos). In this model, documents are represented as a mixture of sequential activity patterns (our motifs) where the mixing weights are defined by the motif starting time occurrences. The novelties are multi fold. First, unlike previous approaches where topics modeled only the co-occurrence of words at a given time instant, our motifs model the co-occurrence and temporal order in which the words occur within a temporal window. Second, unlike traditional Dynamic Bayesian networks (DBN), our model accounts for the important case where activities occur concurrently in the video (but not necessarily in synchrony), i.e., the advent of activity motifs can overlap. The learning of the motifs in these difficult situations is made possible thanks to the introduction of latent variables representing the activity starting times, enabling us to implicitly align the occurrences of the same pattern during the joint inference of the motifs and their starting times. As a third novelty, we propose a general method that favors the recovery of sparse distributions, a highly desirable property in many topic model applications, by adding simple regularization constraints on the searched distributions to the data likelihood optimization criteria. We substantiate our claims with experiments on synthetic data to demonstrate the algorithm behavior, and on four video datasets with significant variations in their activity content obtained from static cameras. We observe that using low-level motion features from videos, our algorithm is able to capture sequential patterns that implicitly represent typical trajectories of scene objects

This paper presents a novel approach for spectral unmixing of remotely sensed hyperspectral data. It exploits probabilistic latent topics in order to take advantage of the semantics pervading the latent topic space when identifying spectral signatures and estimating fractional abundances from hyperspectral images. Despite the contrasted potential of topic models to uncover image semantics, they have been merely used in hyperspectral unmixing as a straightforward data decomposition process. This limits their actual capabilities to provide semantic representations of the spectral data. The proposed model, called dual-depth sparse probabilistic latent semantic analysis (DEpLSA), makes use of two different levels of topics to exploit the semantic patterns extracted from the initial spectral space in order to relieve the ill-posed nature of the unmixing problem. In other words, DEpLSA defines a first level of deep topics to capture the semantic representations of the spectra, and a second level of restricted topics to estimate endmembers and abundances over this semantic space. An experimental comparison in conducted using the two standard topic models and the seven state-of-the-art unmixing methods available in the literature. Our experiments, conducted using four different hyperspectral images, reveal that the proposed approach is able to provide competitive advantages over available unmixing approaches.

A single abstract from the DHd-2018 Book of Abstracts., Sofern eine editorische Arbeit an dieser Publikation stattgefunden hat, dann bestand diese aus der Eliminierung von Bindestrichen in Überschriften, die aufgrund fehlerhafter Silbentrennung entstanden sind, der Vereinheitlichung von Namen der Autor*innen in das Schema "Nachname, Vorname" und/oder der Trennung von Überschrift und Unterüberschrift durch die Setzung eines Punktes, sofern notwendig., {"references":["https://doi.org/10.5281/zenodo.3684897","https://github.com/DHd-Verband/DHd-Abstracts-2018"]}

Conspiracy theories play a troubling role in political discourse. Online forums provide a valuable window into everyday conspiracy theorizing, and can give a clue to the motivations and interests of those who post in such forums. Yet this online activity can be difficult to quantify and study. We describe a unique approach to studying online conspiracy theorists which used non-negative matrix factorization to create a topic model of authors' contributions to the main conspiracy forum on Reddit.com. This subreddit provides a large corpus of comments which spans many years and numerous authors. We show that within the forum, there are multiple sub-populations distinguishable by their loadings on different topics in the model. Further, we argue, these differences are interpretable as differences in background beliefs and motivations. The diversity of the distinct subgroups places constraints on theories of what generates conspiracy theorizing. We argue that traditional “monological” believers are only the tip of an iceberg of commenters. Neither simple irrationality nor common preoccupations can account for the observed diversity. Instead, we suggest, those who endorse conspiracies seem to be primarily brought together by epistemological concerns, and that these central concerns link an otherwise heterogenous group of individuals.

Source at https://doi.org/10.1080/23308249.2017.1416331. Modeling has become the most commonly used method in fisheries science, with numerous types of models and approaches available today. The large variety of models and the overwhelming amount of scientific literature published yearly can make it difficult to effectively access and use the output of fisheries modeling publications. In particular, the underlying topic of an article cannot always be detected using keyword searches. As a consequence, identifying the developments and trends within fisheries modeling research can be challenging and time-consuming. This paper utilizes a machine learning algorithm to uncover hidden topics and subtopics from peer-reviewed fisheries modeling publications and identifies temporal trends using 22,236 full-text articles extracted from 13 top-tier fisheries journals from 1990 to 2016. Two modeling topics were discovered: estimation models (a topic that contains the idea of catch, effort, and abundance estimation) and stock assessment models (a topic on the assessment of the current state of a fishery and future projections of fish stock responses and management effects). The underlying modeling subtopics show a change in the research focus of modeling publications over the last 26 years.

There is a rising need for automated analysis of news text, and topic models have proven to be useful tools for this task. However, as the quality of the topics induced by topic models greatly varies, much research effort has been devoted to their automated evaluation. Recent research has focused on topic coherence as a measure of a topic’s quality. Existing topic coherence measures work by considering the semantic similarity of topic words. This makes them unfit to detect the coherence of transient topics with semantically unrelated topic words, which abound in news media texts. In this paper, we introduce the notion of document-based topic coherence and propose novel topic coherence measures that estimate topic coherence based on topic documents rather than topic words. We evaluate the proposed measures on two datasets containing topics manually labeled for document-based coherence, on which the proposed measures outperform a strong baseline as well as word-based coherence measures. We also demonstrate the usefulness of document-based coherence measures for automated topic discovery from news media texts.

Probabilistic topic models have proven to be an extremely versatile class of mixed-membership models for discovering the thematic structure of text collections. There are many possible applications, covering a broad range of areas of study: technology, natural science, social science and the humanities. In this thesis, a new efficient parallel Markov Chain Monte Carlo inference algorithm is proposed for Bayesian inference in large topic models. The proposed methods scale well with the corpus size and can be used for other probabilistic topic models and other natural language processing applications. The proposed methods are fast, efficient, scalable, and will converge to the true posterior distribution. In addition, in this thesis a supervised topic model for high-dimensional text classification is also proposed, with emphasis on interpretable document prediction using the horseshoe shrinkage prior in supervised topic models. Finally, we develop a model and inference algorithm that can model agenda and framing of political speeches over time with a priori defined topics. We apply the approach to analyze the evolution of immigration discourse in the Swedish parliament by combining theory from political science and communication science with a probabilistic topic model. Probabilistiska ämnesmodeller (topic models) är en mångsidig klass av modeller för att estimera ämnessammansättningar i större corpusar. Applikationer finns i ett flertal vetenskapsområden som teknik, naturvetenskap, samhällsvetenskap och humaniora. I denna avhandling föreslås nya effektiva och parallella Markov Chain Monte Carlo algoritmer för Bayesianska ämnesmodeller. De föreslagna metoderna skalar väl med storleken på corpuset och kan användas för flera olika ämnesmodeller och liknande modeller inom språkteknologi. De föreslagna metoderna är snabba, effektiva, skalbara och konvergerar till den sanna posteriorfördelningen. Dessutom föreslås en ämnesmodell för högdimensionell textklassificering, med tonvikt på tolkningsbar dokumentklassificering genom att använda en kraftigt regulariserande priorifördelningar. Slutligen utvecklas en ämnesmodell för att analyzera "agenda" och "framing" för ett förutbestämt ämne. Med denna metod analyserar vi invandringsdiskursen i Sveriges Riksdag över tid, genom att kombinera teori från statsvetenskap, kommunikationsvetenskap och probabilistiska ämnesmodeller.

This explorative article contributes to research on models of “good parenting” by analyzing whether current demands of politics, science and advice literature are discussed in a large German parenting online forum. The data consists of 58,240 user submissions to a digital discussion board. Using this digital data, it can not only be analyzed which topics users discuss but also how they negotiate them. With “topic modeling”, an innovative approach from the computational social sciences (CSS), is combined with qualitative content analysis. The article shows that expert knowledge is picked up by users to justify and reflect on their childrearing practices. By and large, parents refer to expert knowledge in a positive manner, however, some authors are viewed rather critically. Zusammenfassung Der explorative Artikel leistet einen Beitrag zur Forschung um Leitbilder „guter Erziehung“, indem untersucht wird, ob gegenwärtige Ansprüche aus Politik, Wissenschaft und Ratgeberliteratur innerhalb eines großen deutschen Elternforums diskutiert werden. Als Datenmaterial dienen 58.240 Nutzerbeiträge zu einer digitalen Diskussionsplattform. Anhand dieser digitalen Daten kann nicht nur analysiert werden, welche Themen die Nutzer besprechen, sondern auch wie sie diese verhandeln. Hierzu wird mit „topic modeling“ ein innovatives Verfahren der „computational social sciences“ (CSS) mit der qualitativen Inhaltsanalyse kombiniert. Der Beitrag zeigt, dass insbesondere Expertenwissen von Nutzern aufgegriffen wird, um die eigene Erziehungspraxis zu rechtfertigen und zu reflektieren. Die Bezugnahmen sind in weiten Teilen positiv auch wenn einzelne Autoren kritisch betrachtet werden., JFR - Journal of Family Research, Vol 30 No 2 (2018)

�� ���������������� ���������������������� ������ ������������������ ������������ �������� ���������� ������ ������������������ �������������� ������ ������ ������������������������ �������� ���������������� ������ �������������� ������ ���������������� ������������������ ���������������� ������ ������ ���������������� ��������������. ���� ���������� ���������������� ������ ������������������������������ ���������� ������ �������������� �������� �������������� ������ ������������ ������������������ ������ ���� ���������� �������������������� ������ �������� ���������������� ������������ ���� �������������������� ������ ������������������������ ������ �������� ���������������������������� ��������������, �� ������ ���������������� ������ ����������. �� ������������ ������ ���� �������� ������ ������������������, ���������� ���� ���������������������������� ������ �������������� ���������� ������ �������� ���������������� ������ �������������� ������������������. �������� ������������������ ������ �������������� ������ ���������������� ������������������ ����������������, ���� ���������������� ������������ �������� ������������������������ ������������: �������������� ������ ������������ ������������������ ���������������������� ������ ������ �������������� ������������������ �������������� ������ �� ������������ ���������� ���� ���������������������� ������ �������������� ������������������ ���������������� ������ ���� ������������������ ���������������� �������������� ���� ������ ����������������. ���� �������������� �������������� �������������� ������ �������������� ���� �������������������������� ��������������, ���������������������� �������� ������������������������ �������� ������ �������������������� ������ ��������������, ������������ ������ �������������� ������ ����������������, ������ ������ ������ �������������� ������ ���������������������� �������� ���������������������� ������������������, ������������ ������ ������������������ ������������������������-��������������. ���� ������ ���� ���������� ������������������ ������ ������������������ ���������� ���� ��������������, ������������������ ���� ���������������������� ������ �������������� ������������ ���� �������������������� ������ ���� ������������ ������������ ������������������. ���� �������� ������ ��������������, ������ �������������������� ���� �������� ��������������������: ��) �� ������������������������������ �������������������� ���������������� �������������� ������������������ ����������������, ��) �� ���������������� ������ ���������������������� ������ �������������������� ������ ������������������������ ������ ���� ������������������������ �������������� �������������� ��) �� ���������������������������� ������������������������ ������ ���������������������� ������ ������������������������ �������������� ������ ������ �������������� ���������������� ���������������������� �������������� ������ ���������������� ������������������ ���������������� �������� ������ �������������������� ������ ���������� ���������������� ��) �� ���������������� �������������������������� ���������������� ��������������, �������� ���� ���������������������������� ���������������� �������������������� �������������� ������ ���������������� ������������������ ���������������� ������ ���������� �������� �������� �������������� �������������������� ��������������. ������������, �������������������������� ������ ������������ �������������� ������ ���������������� �������������� ������ ���������������� ������������������ ����������������, ������ ������������������ ���������������������� ������������������ ���������������� ������ ������������������ ������ �������� ���������������������� ���������������������������� �������� ���� ������������������������ ������ ���������������� ������. ������ ����������������, ������������������ ������ ������������������ �������������� ���������������������� ������ �������������������� ������ �������������� ��������������. �� �������������� ���������� ���������� ������ �������������� ������������������ ���������������������� ������ ���������������������� ������������������ Dirichlet, ������������������������ ������������������������ ������������������ �������������������� �������� ������ �������������� ���������������� ������ ������ ���������������� ����������������������������. �� �������������� �������������� ���������� ������ ���������������� �������� �������������� ������ �������������������� ������ �������� ������ ���������������� ������ ������������������ ������ �������������� �������������� ������ ���������������� ������ ������ ���������������������������� ������������������������ �������������� ���������������� ����������������������, ���� ������������ ���������������������� ��������������. ����������, ���������������������� ������ ���������������� ������ �������������� �������������� �������������������������� ��������������, ������ ���������������������� ���� ���������������������������� ���������������� �������������������� ���� ���������������������� ���������������� ��������������. ���� ���������������� �������������� �������������������������� ���� ������������ ������������������ ���� ������ ���������� ������ ���������������������� ���� �������� ����������, ���� ���� ������������������������ ���� �������������������� ������������������ �������������������������� ������ ������ ���������������� ������, ���� �������� ������ ����������������������., This thesis presents an original research in the field of machine learning and more specifically in the areas of multi-label learning and topic models. The main task considered in this thesis involves a setting in which there exists a collection of data instances and we assume that each of these instances can be described with multiple concepts, or a distribution over them. Our goal then, is to determine these concepts for each instance. In the case of multi-label learning, the task is supervised: there exist a training data set and a given concepts ontology and the goal is to train a multi-label learning model that will effectively predict concepts on new data. Topic models entail an unsupervised context, aiming at specifying both the ontology of concepts, or, in other words, the topics of the collection, as well as the concepts describing each data instance, i.e., the instance-topics distribution. Throughout the dissertation the application domain is text, nevertheless the contributions of the thesis can readily be extended to other data domains. In this thesis, we are interested in the following problems: a) effective combination of multi-label learning methods, b) improving parameter estimation for supervised and unsupervised topic models c) partitioning of the output space to effectively solve large-scale multi-label tasks and d) adaptation of supervised topic modelling algorithms, to deal successfully with extremely large multi-label learning tasks. First, we present a multi-label ensemble method, that combines multi-label classifiers and employs a statistical significance test in order to validate its choices. Second, a novel method for parameter estimation for topic models is introduced. This method, employs a full probability distribution to compute improved estimators for the topic model parameters, both in terms of performance and time complexity. Third, we present a method to deal with large-scale multi-label classification, which proceeds by learning a hierarchy out of the label set and then solves the resulting easier and simpler sub-tasks. Finally, we propose an extension to supervised topic models, that effectively scales up to tasks with hundreds of thousands of concepts. All of the above methods are evaluated and compared against the relevant state-of-the-art, in multiple experimental settings and for diverse evaluation procedures, the results demonstrating significant improvements in all cases.

International audience; Mining medical data has significantly gained interest in the recent years thanks to the advances in data mining and machine learning fields. In this work, we focus on a challenging issue in medical data mining: automatic diagnosis code assignment to discharge summaries, i.e., characterizing patient's hospital stay (diseases, symptoms, treatments, etc.) with a set of codes usually derived from the International Classification of Diseases (ICD). We cast the problem as a machine learning task and we experiment some recent approaches based on the probabilis-tic topic models. We demonstrate the efficiency of these models in terms of high predictive scores and ease of result interpretation. As such, we show how topic models enable gaining insights into this field and provide new research opportunities for possible improvements.

This commentary reflects on how the themes in the Journal of Organization Design have evolved along with the field of organization design itself. Using topic modeling of abstracts, I assess the key topics or themes in the Journal of Organization Design and compare them to themes underlying design studies in mainstream management journals. I propose some mechanisms behind the differences and similarities and offer some implications for the future of the journal.

What is and was the sociology of sports all about? Through the method of topic modelling, I will investigate the content of all articles (full text, N¼1923) from what has historically been the three leading sociology of sport journals – Sociology of Sport Journal, International Review for the Sociology of Sport and Journal of Sport and Social Issues – from 1984 to 2014. The study extracts 20 dominant topics: from the vague but central topic of Culture via Organization & Politics, Gender, Race, and Body to the less central, but more specific, topic of Football, Nationalism and Globalization. Additionally, I look at how the three journals have their strongholds in these topics, and how the topics have fluctuated over time. The results are discussed against the background of previous reviews and studies in the sociology of sports.

International audience; We have relied on the abstracts of the scientific articles that have been published in the journal Le Travail humain since its first publication in 1933 in anattempt to grasp the evolution of the topics addressed.We analyzed these scientific publications in order to extract their essence in terms of research topics and to understand the scientific orientation of the journal and its evolution. First, we analysed the common features and the specificities of the publications in the different editions of the journal and the main differences between the consecutive editions. Then we were interested in the way the publications were organized around the extracted topics and we tried to visualize a semantic approximation. Finally, we were interested in the evolution of the topics from the creation of the journal to the 2016 edition.; Les résumés des articles scientifiques publiés dans la revue Le Travail humain depuis 1933, date de sa création, constituent un matériau sur lequel nous nous sommes appuyés pour tenter de saisir l’évolution des thématiques abordées. Nous nous sommes penchés sur l’analyse des 1665 publications scientifiques afin d’en extraire l’essence en termes de thématiques de recherches abordées et comprendreles dynamiques scientifiques et les orientations thématiques de la revue. Premièrement, nous avons analysé les points communs et les spécificités des publications dans les différentes éditions de la revue ainsi que les principales différences entre les éditions consécutives. Puis nous nous sommes intéressés à la façon dont les publications s’articulent autour des thématiques extraites et sur lesquelles nous avons essayé de visualiser une approximation sémantique. Enfin, nous nous sommes intéressés à l’évolution des thématiques depuis la création de cette revue jusqu’àl’édition 2016.

We study the role of global media during the Greek debt crisis and relate it to the transmission of events on financial actors' expectations. To identify news coverage about the Greek debt crisis, we apply topic modeling to a newly compiled dataset of over 430,000 articles from The International New York Times and Financial Times from 2009 to 2015. We identify a Greek debt crisis topic and relate it to events concerning Greece during this time period. Our finding is that events are only relevant for financial markets when they are covered in the media, whereas events without media coverage have no effect. News coverage without immediate events is equally irrelevant for financial markets.

Our search for new sources of analysis has led us to build a quantitative inductive method based on the analysis of lexical fields (topic models) to study teachers’ blogs. This approach generates new insights about teachers’ concerns in expected areas, such as their discipline or their own use of digital technology, but also in unexpected areas, such as dress code issues, or terrorist attacks. This article presents our method, and explains how and why it provides us with new opportunities for the analysis of a relatively new type of written source, namely blogs., Notre recherche de nouvelles sources d'analyse nous a conduits à construire une méthode quantitative inductive, basée sur l'analyse de champs lexicaux (topic model) pour étudier les blogs des enseignants. Cette approche permet d'acquérir de nouvelles connaissances sur les préoccupations des enseignants dans des domaines attendus, tels que leur discipline ou leur usage de la technologie numérique, mais aussi dans des domaines inattendus, tels que les questions de code vestimentaire ou les attentats terroristes. L'article présente la méthode, comment et pourquoi celle-ci nous offre de nouvelles opportunités pour l'analyse d'un type relativement nouveau de source écrite, les blogs.

The advent of online platforms such as weblogs and social networking sites provided Internet users with an unprecedented means to express their opinions on a wide range of topics, including policy and commercial products. This large volume of opinionated data can be explored and exploited through text mining techniques known as opinion mining or sentiment analysis. Contrarily to traditional opinion mining work which mostly focuses on positive and negative opinions (or an intermediate in-between), we study a more challenging type of opinions: viewpoints. Viewpoint mining reaches beyond polarity-based opinions (positive/negative) and enables the analysis of more subtle opinions such as political opinions. In this thesis, we proposed unsupervised approaches – i.e., approaches which do not require any labeled data – based on probabilistic topic models to jointly discover topics and viewpoints expressed in opinionated data. In our first contribution, we explored the idea of separating opinion words (specific to both viewpoints and topics) from topical, neutral words based on parts of speech, inspired by similar practices in the litterature of non viewpoint-related opinion mining. Our second contribution tackles viewpoints expressed by social network users. We aimed to study to what extent social interactions between users – in addition to text content – can be beneficial to identify users' viewpoints. Our different contributions were evaluated and benchmarked against state-of-the-art baselines on real-world datasets; Les plateformes en ligne telles que les blogs et les réseaux sociaux permettent aux internautes de s'exprimer sur des sujets d'une grande variété (produits commerciaux, politique, services, etc.). Cet important volume de données d'opinions peut être exploré et exploité grâce à des techniques de fouille de texte connues sous le nom de fouille d'opinions ou analyse de sentiments. Contrairement à la majorité des travaux actuels en fouille d'opinions, qui se focalisent sur les opinions simplement positives ou négatives (ou un intermédiaire entre ces deux extrêmes), nous nous intéressons dans cette thèse aux points de vue. La fouille de point de vue généralise l'opinion au delà de son acception usuelle liée à la polarité (positive ou négative) et permet l'étude d'opinions exprimées plus subtilement, telles que les opinions politiques. Nous proposons dans cette thèse des approches non supervisées - ne nécessitant aucune annotation préalable - basées sur des modèles thématiques probabilistes afin de découvrir simultanément les thèmes et les points de vue exprimés dans des corpus de textes d'opinion. Dans notre première contribution, nous avons exploré l'idée de différencier mots d'opinions (spécifiques à la fois à un point de vue et à un thème) et mots thématiques (dépendants du thème mais neutres vis-à-vis des différents points de vue) en nous basant sur les parties de discours, inspirée par des pratiques similaires dans la littérature de fouille d'opinions classique - restreinte aux opinions positives et négatives. Notre seconde contribution se focalise quant à elle sur les points de vue exprimés sur les réseaux sociaux. Notre objectif est ici d'analyser dans quelle mesure l'utilisation des interactions entre utilisateurs, en outre de leur contenu textuel généré, est bénéfique à l'identification de leurs points de vue. Nos différentes contributions ont été évaluées et comparées à l'état de l'art sur des collections de documents réels.