En esta tesis doctoral se describen las investigaciones realizadas con el objetivo de determinar las mejores tecnicas para construir un Reconocedor de Entidades Nombradas en Arabe. Tal sistema tendria la habilidad de identificar y clasificar las entidades nombradas que se encuentran en un texto arabe de dominio abierto. La tarea de Reconocimiento de Entidades Nombradas (REN) ayuda a otras tareas de Procesamiento del Lenguaje Natural (por ejemplo, la Recuperacion de Informacion, la Busqueda de Respuestas, la Traduccion Automatica, etc.) a lograr mejores resultados gracias al enriquecimiento que a~nade al texto. En la literatura existen diversos trabajos que investigan la tarea de REN para un idioma especifico o desde una perspectiva independiente del lenguaje. Sin embargo, hasta el momento, se han publicado muy pocos trabajos que estudien dicha tarea para el arabe. El arabe tiene una ortografia especial y una morfologia compleja, estos aspectos aportan nuevos desafios para la investigacion en la tarea de REN. Una investigacion completa del REN para elarabe no solo aportaria las tecnicas necesarias para conseguir un alto rendimiento, sino que tambien proporcionara un analisis de los errores y una discusion sobre los resultados que benefician a la comunidad de investigadores del REN. El objetivo principal de esta tesis es satisfacer esa necesidad. Para ello hemos: 1. Elaborado un estudio de los diferentes aspectos del arabe relacionados con dicha tarea; 2. Analizado el estado del arte del REN; 3. Llevado a cabo una comparativa de los resultados obtenidos por diferentes tecnicas de aprendizaje automatico; 4. Desarrollado un metodo basado en la combinacion de diferentes clasificadores, donde cada clasificador trata con una sola clase de entidades nombradas y emplea el conjunto de caracteristicas y la tecnica de aprendizaje automatico mas adecuados para la clase de entidades nombradas en cuestion. Nuestros experimentos han sido evaluados sobre nueve conjuntos de test. / Benajiba, Y. (2009). Arabic named entity recognition [Tesis doctoral no publicada]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/8318 / Palancia

This study analyzes natural language processing with emphasis on morphological analysis of inflective languages and systems for named entity recognition. It analyzes effective pattern matching in dictionary by using succint structures and then analyzes practical implementation of succint structures. It describes design and implementation of named entity recognition system and morphological analyzer and compares and test their speed and effectiveness.

Der Großteil des geschäftsrelevanten Wissens liegt heute als unstrukturierte Information in Form von Textdaten auf Internetseiten, in Office-Dokumenten oder Foreneinträgen vor. Zur Extraktion und Verwertung dieser unstrukturierten Informationen wurde eine Vielzahl von Text-Mining-Lösungen entwickelt. Viele dieser Systeme wurden in der jüngeren Vergangenheit als Webdienste zugänglich gemacht, um die Verwertung und Integration zu vereinfachen. Die Kombination verschiedener solcher Text-Mining-Dienste zur Lösung konkreter Extraktionsaufgaben erscheint vielversprechend, da so bestehende Stärken ausgenutzt, Schwächen der Systeme minimiert werden können und die Nutzung von Text-Mining-Lösungen vereinfacht werden kann. Die vorliegende Arbeit adressiert die flexible Kombination von Text-Mining-Diensten in einem serviceorientierten System und erweitert den Stand der Technik um gezielte Methoden zur Auswahl der Text-Mining-Dienste, zur Aggregation der Ergebnisse und zur Abbildung der eingesetzten Klassifikationsschemata. Zunächst wird die derzeit existierende Dienstlandschaft analysiert und aufbauend darauf eine Ontologie zur funktionalen Beschreibung der Dienste bereitgestellt, so dass die funktionsgesteuerte Auswahl und Kombination der Text-Mining-Dienste ermöglicht wird. Des Weiteren werden am Beispiel entitätsextrahierender Dienste Algorithmen zur qualitätssteigernden Kombination von Extraktionsergebnissen erarbeitet und umfangreich evaluiert. Die Arbeit wird durch zusätzliche Abbildungs- und Integrationsprozesse ergänzt, die eine Anwendbarkeit auch in heterogenen Dienstlandschaften, bei denen unterschiedliche Klassifikationsschemata zum Einsatz kommen, gewährleisten. Zudem werden Möglichkeiten der Übertragbarkeit auf andere Text-Mining-Methoden erörtert.

The search for information about entities, such as people or movies, plays an increasingly important role on the Web. This information is still scattered across many Web pages, making it more time consuming for a user to find all relevant information about an entity. This thesis describes techniques to extract entities and information about these entities from the Web, such as facts, opinions, questions and answers, interactive multimedia objects, and events. The findings of this thesis are that it is possible to create a large knowledge base automatically using a manually-crafted ontology. The precision of the extracted information was found to be between 75–90 % (facts and entities respectively) after using assessment algorithms. The algorithms from this thesis can be used to create such a knowledge base, which can be used in various research fields, such as question answering, named entity recognition, and information retrieval.

The use cases of Information Extraction (IE) are more or less endless, often consisting of a combination of Named Entity Recognition (NER) and Relation Extraction (RE). One use case of IE is the extraction of transaction information from Norwegian insider transaction Press Releases (PRs), where a transaction consists of at most four entities: the name of the owner performing the transaction, the number of shares transferred, the transaction date, and the price of the shares bought or sold. The relationships between the entities define which entity belongs to which transaction, and whether shares were bought or sold. This report has investigated how a pair of supervised NER and RE models extract this information. Since these Norwegian PRs were not labeled, two different approaches to annotating the transaction entities and their associated relations were investigated, and it was found that it is better to annotate only entities that occur in a relation than annotating all occurrences. Furthermore, the number of PRs needed to achieve a satisfactory result in the IE pipeline was investigated. The study shows that training with about 400 PRs is sufficient for the results to converge, at around 0.85 in F1-score. Finally, the report shows that there is not much difference between a complex RE model and a simple rule-based approach, when applied on the studied corpus.

This thesis investigates automated annotation approaches to assist public authorities in Sweden in optimizing resource allocation and gaining valuable insights to enhance the preservation of high-quality welfare. The study uses pre-trained Swedish language models for the named entity recognition (NER) task of finding job titles in job advertisements from The Swedish Public Employment Service, Arbetsförmedlingen. Specifically, it evaluates the performance of the Swedish Bidirectional Encoder Representations from Transformers (BERT), developed by the National Library of Sweden (KB), referred to as KB-BERT. The thesis explores the impact of training data size on the models’ performance and examines whether active learning can enhance efficiency and accuracy compared to random sampling. The findings reveal that even with a small training dataset of 220 job advertisements, KB-BERT achieves a commendable F1-score of 0.770 in predicting job titles. The model’s performance improves further by augmenting the training data with an additional 500 annotated job advertisements, yielding an F1-score of 0.834. Notably, the highest F1-score of 0.856 is achieved by applying the active learning strategy of uncertainty sampling and the measure of mean entropy. The test data provided by Arbetsförmedlingen was re-annotated to evaluate the complexity of the task. The human annotator achieved an F1-score of 0.883. Based on these findings, it can be inferred that KB-BERT performs satisfactorily in classifying job titles from job ads.

Companies do not exist in isolation. They are embedded in structural relationships with each other. Mapping a given company’s relationships with other companies in terms of competitors, subsidiaries, suppliers, and customers are key to understanding a company’s major risk factors and opportunities. Conventionally, obtaining and staying up to date with this key knowledge was achieved by reading financial news and reports by highly skilled manual labor like a financial analyst. However, with the development of Natural Language Processing (NLP) and graph databases, it is now possible to systematically extract and store structured information from unstructured data sources. The current go-to method to effectively extract information uses supervised machine learning models, which require a large amount of labeled training data. The data labeling process is usually time-consuming and hard to get in a domain-specific area. This project explores an approach to construct a company domain-specific Knowledge Graph (KG) that contains company-related entities and relationships from the U.S. Securities and Exchange Commission (SEC) 10-K filings by combining a pre-trained general NLP with rule-based patterns in Named Entity Recognition (NER) and Relation Extraction (RE). This approach eliminates the time-consuming data-labeling task in the statistical approach, and by evaluating ten 10-k filings, the model has the overall Recall of 53.6%, Precision of 75.7%, and the F1-score of 62.8%. The result shows it is possible to extract company information using the hybrid methods, which does not require a large amount of labeled training data. However, the project requires the time-consuming process of finding lexical patterns from sentences to extract company-related entities and relationships. / Företag existerar inte som isolerade organisationer. De är inbäddade i strukturella relationer med varandra. Att kartlägga ett visst företags relationer med andra företag när det gäller konkurrenter, dotterbolag, leverantörer och kunder är nyckeln till att förstå företagets huvudsakliga riskfaktorer och möjligheter. Det konventionella sättet att hålla sig uppdaterad med denna viktiga kunskap var genom att läsa ekonomiska nyheter och rapporter från högkvalificerad manuell arbetskraft som till exempel en finansanalytiker. Men med utvecklingen av ”Natural Language Processing” (NLP) och grafdatabaser är det nu möjligt att systematiskt extrahera och lagra strukturerad information från ostrukturerade datakällor. Den nuvarande metoden för att effektivt extrahera information använder övervakade maskininlärningsmodeller som kräver en stor mängd märkta träningsdata. Datamärkningsprocessen är vanligtvis tidskrävande och svår att få i ett domänspecifikt område. Detta projekt utforskar ett tillvägagångssätt för att konstruera en företagsdomänspecifikt ”Knowledge Graph” (KG) som innehåller företagsrelaterade enheter och relationer från SEC 10-K-arkivering genom att kombinera en i förväg tränad allmän NLP med regelbaserade mönster i ”Named Entity Recognition” (NER) och ”Relation Extraction” (RE). Detta tillvägagångssätt eliminerar den tidskrävande datamärkningsuppgiften i det statistiska tillvägagångssättet och genom att utvärdera tio SEC 10-K arkiv har modellen den totala återkallelsen på 53,6 %, precision på 75,7 % och F1-poängen på 62,8 %. Resultatet visar att det är möjligt att extrahera företagsinformation med hybridmetoderna, vilket inte kräver en stor mängd märkta träningsdata. Projektet kräver dock en tidskrävande process för att hitta lexikala mönster från meningar för att extrahera företagsrelaterade enheter och relationer.

During the past several decades, journalists have been struggling with the ever growing amount of data on the internet. Investigating the validity of the sources or finding similar articles for a story can consume a lot of time and effort. These issues are even amplified by the declining size of the staff of news agencies. The solution is to empower the remaining professional journalists with digital tools created by computer scientists. This thesis project is inspired by an idea to provide software support for journalistic work with interactive visual interfaces and artificial intelligence. More specifically, within the scope of this thesis project, we created a backend module that supports several text mining methods such as keyword extraction, named entity recognition, sentiment analysis, fake news classification and also data collection from various data sources to help professionals in the field of journalism. To implement our system, first we gathered the requirements from several researchers and practitioners in journalism, media studies, and computer science, then acquired knowledge by reviewing literature on current approaches. Results are evaluated both with quantitative methods such as individual component benchmarks and also with qualitative methods by analyzing the outcomes of the semi-structured interviews with collaborating and external domain experts. Our results show that there is similarity between the domain experts' perceived value and the performance of the components on the individual evaluations. This shows us that there is potential in this research area and future work would be welcomed by the journalistic community.

[pt] As técnicas de processamento de linguagem natural (NLP) estão se tornando populares recentemente. A gama de aplicativos que se beneficiam de NLP é extensa, desde criar sistemas de tradução automática até ajudar no marketing de um produto. Dentro de NLP, o campo de Extração de Informações (IE) é difundido; concentra-se no processamento de textos para recuperar informações específicas sobre uma determinada entidade ou conceito. Ainda assim, a comunidade de pesquisa se concentra principalmente na construção de modelos para dados na língua inglesa. Esta tese aborda três tarefas no domínio do IE: Reconhecimento de Entidade Nomeada, Extração de Relações Semânticas e Extração Conjunta de Entidade e Relação. Primeiro, criamos um novo conjunto de dados em português no domínio biomédico, descrevemos o processo de anotação e medimos suas propriedades. Além disso, desenvolvemos um novo modelo para a tarefa de Extração Conjunta de Entidade e Relação, verificando que o mesmo é competitivo em comparação com outros modelos. Finalmente, avaliamos cuidadosamente os modelos propostos em textos de idiomas diferentes do inglês e confirmamos a dominância de modelos baseados em redes neurais. / [en] Natural language processing (NLP) techniques are becoming popular recently. The range of applications that benefit from NLP is extensive, from building machine translation systems to helping market a product. Within NLP, the Information Extraction (IE) field is widespread; it focuses on processing texts to retrieve specific information about a particular entity or concept. Still, the research community mainly focuses on building models for English data. This thesis addresses three tasks in the IE domain: Named Entity Recognition, Relation Extraction, and Joint Entity and Relation Extraction. First, we created a novel Portuguese dataset in the biomedical domain, described the annotation process, and measured its properties. Also, we developed a novel model for the Joint Entity and Relation Extraction task, verifying that it is competitive compared to other models. Finally, we carefully evaluated proposed models on non-English language datasets and confirmed the dominance of neural-based models.

Address parsing is the process of splitting an address string into its different address components, such as street name, street number, et cetera. Address parsing has been quite extensively researched and there exist some state-ofthe-art address parsing solutions, mostly unilingual. In more recent years research has emerged which focuses on multinational address parsing and deep architecture address parsers have been used to achieve state-of-the-art performance on multinational address data. However, training these deep architectures for address parsing requires a rather large amount of address data which is not always accessible. Generally within Natural Language Processing (NLP) data is difficult to come by and most of the NLP data available consists of data from about only 20 of the approximately 7000 languages spoken around the world, so-called high-resource languages. This also applies to address data, which can be difficult to come by for some of the so-called low-resource languages of the world for which little or no NLP data exists. To attempt to deal with the lack of address data availability for some of the less spoken languages of the world, the current project investigates the potential of FewShot Learning (FSL) for multinational address parsing. To investigate this, two few-shot transfer learning models are implemented, both implementations consist of a fine-tuned pre-trained language model (PTLM). The difference between the two models is the PTLM used, which were the multilingual language models mBERT and XLM-R, respectively. The two PTLMs are finetuned using a linear classifier layer to then be used as multinational address parsers. The two models are trained and their results are compared with a state-of-the-art multinational address parser, Deepparse, as well as with each other. Results show that the two models do not outperform Deepparse, but they do show promising results, not too far from what Deepparse achieves on holdout and zero-shot datasets. On a mix of low- and high-resource language address data, both models perform well and achieve over 96% on the overall F1-score. Out of the two models used for implementation, XLM-R achieves significantly better results than mBERT and can therefore be considered the more appropriate PTLM to use for multinational FSL address parsing. Based on these results the conclusion is that there is great potential for FSL within the field of multinational address parsing and that general FSL methods can be used and perform well on multinational address parsing tasks. / Adressavkodning är processen att dela upp en adresssträng i dess olika adresskomponenter såsom gatunamn, gatunummer, et cetera. Adressavkodning har undersökts ganska omfattande och det finns några toppmoderna adressavkodningslösningar, mestadels enspråkiga. Senaste åren har forskning fokuserad på multinationell adressavkodning börjat dyka upp och djupa arkitekturer för adressavkodning har använts för att uppnå toppmodern prestation på multinationell adressdata. Att träna dessa arkitekturer kräver dock en ganska stor mängd adressdata, vilket inte alltid är tillgängligt. Det är generellt svårt att få tag på data inom naturlig språkbehandling och majoriteten av den data som är tillgänglig består av data från endast 20 av de cirka 7000 språk som används runt om i världen, så kallade högresursspråk. Detta gäller även för adressdata, vilket kan vara svårt att få tag på för vissa av världens så kallade resurssnåla språk för vilka det finns lite eller ingen data för naturlig språkbehandling. För att försöka behandla denna brist på adressdata för några av världens mindre talade språk undersöker detta projekt om det finns någon potential för inlärning med få exempel för multinationell adressavkodning. För detta implementeras två modeller för överföringsinlärning med få exempel genom finjustering av förtränade språkmodeller. Skillnaden mellan de två modellerna är den förtränade språkmodellen som används, mBERT respektive XLM-R. Båda modellerna finjusteras med hjälp av ett linjärt klassificeringsskikt för att sedan användas som multinationella addressavkodare. De två modellerna tränas och deras resultat jämförs med en toppmodern multinationell adressavkodare, Deepparse. Resultaten visar att de två modellerna presterar båda sämre än Deepparse modellen, men de visar ändå lovande resultat, inte långt ifrån vad Deepparse uppnår för både holdout och zero-shot dataset. Vidare, så presterar båda modeller bra på en blandning av adressdata från låg- och högresursspråk och båda modeller uppnår över 96% övergripande F1-score. Av de två modellerna uppnår XLM-R betydligt bättre resultat än mBERT och kan därför anses vara en mer lämplig förtränad språkmodell att använda för multinationell inlärning med få exempel för addressavkodning. Utifrån dessa resultat dras slutsatsen att det finns stor potential för inlärning med få exempel inom området multinationall adressavkodning, samt att generella metoder för inlärning med få exempel kan användas och preseterar bra på multinationella adressavkodningsuppgifter.