Having computer systems capable of answering questions has been a goal within Natural Language Processing research for many years. Machine Learning systems have recently become increasingly proficient at this task with large language models obtaining state-of-the-art performance. Retriever-reader architectures have become a powerful approach for building systems that enable users to enter questions and get factual answers from a corpus of documents. This architecture uses a retriever component that fetches the most relevant documents and a reader which in turn extracts the answer from the documents. These systems commonly use transformer-based models for both components, which have been fine-tuned on a general domain of documents, such as Wikipedia. However, the performance of such systems on new domains, with different vocabularies, can be lacking. Furthermore, new domains of, for instance, company-specific documents often lack annotated data which makes training new models cumbersome. This thesis investigated how a retriever-reader-based architecture can be adapted to a corpus of Telecom documents by generating question-answer data using a large generative language model, GPT3.5. Also, it compared the usage of a dense retriever using BERT to a BM25-based retriever on the domain. Findings suggest that generating training data can be an effective approach for fine-tuning a dense retriever, increasing the Top-K retrieval accuracy by 20 points for k = 10, compared to a dense retriever fine-tuned on Wikipedia. Additionally, it is found that the sparse retriever outperforms the best dense retriever, although, there is reason to believe that the structure of the test dataset could influence this. Finally, the results also indicate that the performance of the reader is not improved by the generated data although future work is needed to draw better conclusions. / Datorsystem som kan svara på frågor har varit ett mål inom forskningsfältet naturlig språkbehandling i många år. System som använder sig av maskininlärning, så som stora språkmodeller har under de senaste åren uppnått hög prestanda. Att använda sig av en så kallad retriever-reader arkitektur har blivit ett kraftfullt tillvägagångssätt för att bygga system som gör det möjligt för användare att ställa frågor och få faktabaserade svar hämtade från en korpus av dokument. Denna arkitektur använder en retriever som hämtar den mest relevanta informationen och en reader som sedan extraherar ett svar från den hämtade informationen. Dessa system använder vanligtvis transformer-baserade modeller för båda komponenterna, som har tränats på en allmän domän som t.ex., Wikipedia. Dock kan prestandan hos dessa system vara bristfällig när de appliceras på mer specifika domäner med andra ordförråd. Dessutom saknas ofta annoterad data för mer specifika domäner, som exempelvis företagsdokument, vilket gör det svårt att träna modeller på dessa områden. I denna avhandling undersöktes hur en retriever-reader arkitektur kan appliceras på en korpus telekomdokument genom att generera data bestående av frågor och tillhörande svar, genom att använda en stor generativ språkmodell, GPT3.5. Rapporten jämförde även användandet av en BERT-baserad retriever med en BM25-baserad retriever för denna domän. Resultaten tyder på att generering av träningsdata kan vara ett effektivt tillvägagångssätt för att träna en BERT-baserad retriever. Den tränade modellen hade 20 poäng högre noggranhet för måttet Top-K retrieval vid k = 10 jämfört med samma model tränad på data från Wikipedia. Resultaten visade även att en BM25-baserad retriever hade högre noggranhet än den bästa BERT-baserade retrievern som tränats. Dock kan detta bero på datasetets utformning. Slutligen visade resultaten även att prestandan hos en tränad reader inte blev bättre genom att träna på genererad data men denna slutsats kräver framtida arbete för att undersökas mer noggrant.

京都大学 / 新制・課程博士 / 博士(情報学) / 甲第24259号 / 情博第803号 / 新制||情||135(附属図書館) / 京都大学大学院情報学研究科社会情報学専攻 / (主査)教授 吉川 正俊, 教授 田島 敬史, 教授 黒橋 禎夫, 特定准教授 LIN Donghui / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

Extractive question answering is a popular domain in the field of natural language processing, where machine learning models are tasked with answering questions given a context. Historically the field has been centered on monolingual models, but recently more and more multilingual models have been developed, such as Google’s MT5 [1]. Because of this, machine translations of English have been used when training and evaluating these models, but machine translations can be degraded and do not always reflect their target language fairly. This report investigates if complementary information stored in other languages can improve monolingual QA performance for languages where only machine translations are available. It also investigates if exposure to more languages can improve zero-shot cross-lingual QA performance (i.e. when the question and answer do not have matching languages) by providing complementary information. We fine-tune 3 different MT5 models on QA datasets consisting of machine translations, as well as one model on the datasets together in combination with 3 other datasets that are not translations. We then evaluate the different models on the MLQA and XQuAD datasets. The results show that for 2 out of the 3 languages evaluated, complementary information stored in other languages had a positive effect on the QA performance of the MT5. For zero-shot cross-lingual QA, the complementary information offered by the fused model lead to improved performance compared to 2/3 of the MT5 models trained only on translated data, indicating that complementary information from other languages do not offer any improvement in this regard. / Frågebesvaring (QA) är en populär domän inom naturlig språkbehandling, där maskininlärningsmodeller har till uppgift att svara på frågor. Historiskt har fältet varit inriktat på enspråkiga modeller, men nyligen har fler och fler flerspråkiga modeller utvecklats, till exempel Googles MT5 [1]. På grund av detta har maskinöversättningar av engelska använts vid träning och utvärdering av dessa modeller, men maskinöversättningar kan vara försämrade och speglar inte alltid deras målspråk rättvist. Denna rapport undersöker om kompletterande information som lagras i andra språk kan förbättra enspråkig QA-prestanda för språk där endast maskinöversättningar är tillgängliga. Den undersöker också om exponering för fler språk kan förbättra QA-prestanda på zero-shot cross-lingual QA (dvs. där frågan och svaret inte har matchande språk) genom att tillhandahålla kompletterande information. Vi finjusterar 3 olika modeller på QA-datamängder som består av maskinöversättningar, samt en modell på datamängderna tillsammans i kombination med 3 andra datamängder som inte är översättningar. Vi utvärderar sedan de olika modellerna på MLQA- och XQuAD-datauppsättningarna. Resultaten visar att för 2 av de 3 utvärderade språken hade kompletterande information som lagrats i andra språk en positiv effekt på QA-prestanda. För zero-shot cross-lingual QA leder den kompletterande informationen som erbjuds av den sammansmälta modellen till förbättrad prestanda jämfört med 2/3 av modellerna som tränats endast på översättningar, vilket indikerar att kompletterande information från andra språk inte ger någon förbättring i detta avseende.

Title: Hybrid Deep Question Answering Author: Ahmad Aghaebrahimian Institute: Institute of Formal and Applied Linguistics Supervisor: RNDr. Martin Holub, Ph.D., Institute of Formal and Applied Lin- guistics Abstract: As one of the oldest tasks of Natural Language Processing, Question Answering is one of the most exciting and challenging research areas with lots of scientific and commercial applications. Question Answering as a discipline in the conjunction of computer science, statistics, linguistics, and cognitive science is concerned with building systems that automatically retrieve answers to ques- tions posed by humans in a natural language. This doctoral dissertation presents the author's research carried out in this discipline. It highlights his studies and research toward a hybrid Question Answering system consisting of two engines for Question Answering over structured and unstructured data. The structured engine comprises a state-of-the-art Question Answering system based on knowl- edge graphs. The unstructured engine consists of a state-of-the-art sentence-level Question Answering system and a word-level Question Answering system with results near to human performance. This work introduces a new Question An- swering dataset for answering word- and sentence-level questions as well. Start- ing from a...

The amount of data available and consumed by people globally is growing. To reduce mental fatigue and increase the general ability to gain insight into complex texts or documents, we have developed an application to aid in this task. The application allows users to upload documents and ask domain-specific questions about them using our web application. A summarized version of each document is presented to the user, which could further facilitate their understanding of the document and guide them towards what types of questions could be relevant to ask. Our application allows users flexibility with the types of documents that can be processed, it is publicly available, stores no user data, and uses state-of-the-art models for its summaries and answers. The result is an application that yields near human-level intuition for answering questions in certain isolated cases, such as Wikipedia and news articles, as well as some scientific texts. The application shows a decrease in reliability and its prediction as to the complexity of the subject, the number of words in the document, and grammatical inconsistency in the questions increases. These are all aspects that can be improved further if used in production. / Mängden data som är tillgänglig och konsumeras av människor växer globalt. För att minska den mentala trötthet och öka den allmänna förmågan att få insikt i komplexa, massiva texter eller dokument, har vi utvecklat en applikation för att bistå i de uppgifterna. Applikationen tillåter användare att ladda upp dokument och fråga kontextspecifika frågor via vår webbapplikation. En sammanfattad version av varje dokument presenteras till användaren, vilket kan ytterligare förenkla förståelsen av ett dokument och vägleda dem mot vad som kan vara relevanta frågor att ställa. Vår applikation ger användare möjligheten att behandla olika typer av dokument, är tillgänglig för alla, sparar ingen personlig data, och använder de senaste modellerna inom språkbehandling för dess sammanfattningar och svar. Resultatet är en applikation som når en nära mänsklig intuition för vissa domäner och frågor, som exempelvis Wikipedia- och nyhetsartiklar, samt viss vetensaplig text. Noterade undantag för tillämpningen härrör från ämnets komplexitet, grammatiska korrekthet för frågorna och dokumentets längd. Dessa är områden som kan förbättras ytterligare om den används i produktionen.

The need for interacting with machines using spoken natural language is growing, along with the expectation that synthetic speech in this context sound natural. Such interaction includes answering questions, where prosody plays an important role in producing natural English synthetic speech by communicating the information structure of utterances. CCG is a theoretical framework that exploits the notion that, in English, information structure, prosodic structure and syntactic structure are isomorphic. This provides a way to convert a semantic representation of an utterance into a prosodically natural spoken utterance. GF is a framework for writing grammars, where abstract tree structures capture the semantic structure and concrete grammars render these structures in linearised strings. This research combines these frameworks to develop a system that converts semantic representations of utterances into linearised strings of natural language that are marked up to inform the prosody-generating component of a speech synthesis system. / Computing / M. Sc. (Computing)

La recherche de réponses précises à des questions formulées en langue naturelle renouvelle le champ de la recherche d’information. De nombreux travaux ont eu lieu sur la recherche de réponses à des questions factuelles en domaine ouvert. Moins de travaux ont porté sur la recherche de réponses en domaine de spécialité, en particulier dans le domaine médical ou biomédical. Plusieurs conditions différentes sont rencontrées en domaine de spécialité comme les lexiques et terminologies spécialisés, les types particuliers de questions, entités et relations du domaine ou les caractéristiques des documents ciblés. Dans une première partie, nous étudions les méthodes permettant d’analyser sémantiquement les questions posées par l’utilisateur ainsi que les textes utilisés pour trouver les réponses. Pour ce faire nous utilisons des méthodes hybrides pour deux tâches principales : (i) la reconnaissance des entités médicales et (ii) l’extraction de relations sémantiques. Ces méthodes combinent des règles et patrons construits manuellement, des connaissances du domaine et des techniques d’apprentissage statistique utilisant différents classifieurs. Ces méthodes hybrides, expérimentées sur différents corpus, permettent de pallier les inconvénients des deux types de méthodes d’extraction d’information, à savoir le manque de couverture potentiel des méthodes à base de règles et la dépendance aux données annotées des méthodes statistiques. Dans une seconde partie, nous étudions l’apport des technologies du web sémantique pour la portabilité et l’expressivité des systèmes de questions-réponses. Dans le cadre de notre approche, nous exploitons les technologies du web sémantique pour annoter les informations extraites en premier lieu et pour interroger sémantiquement ces annotations en second lieu. Enfin, nous présentons notre système de questions-réponses, appelé MEANS, qui utilise à la fois des techniques de TAL, des connaissances du domaine et les technologies du web sémantique pour répondre automatiquement aux questions médicales. / With the dramatic growth of digital information, finding precise answers to natural language questions is more and more essential for retrieving domain knowledge in real time. Many research works tackled answer retrieval for factual questions in open domain. Less works were performed for domain-specific question answering such as the medical domain. Compared to the open domain, several different conditions are met in the medical domain such as specialized vocabularies, specific types of questions, different kinds of domain entities and relations. Document characteristics are also a matter of importance, as, for example, clinical texts may tend to use a lot of technical abbreviations while forum pages may use long “approximate” terms. We focus on finding precise answers to natural language questions in the medical field. A key process for this task is to analyze the questions and the source documents semantically and to use standard formalisms to represent the obtained annotations. We propose a medical question-answering approach based on: (i) NLP methods combing domain knowledge, rule-based methods and statistical ones to extract relevant information from questions and documents and (ii) Semantic Web technologies to represent and interrogate the extracted information.

L'émergence du Web 2.0 a changé la façon avec laquelle les gens recherchent et obtiennent des informations sur internet. Entre sites communautaires spécialisés, réseaux sociaux, l'utilisateur doit faire face à une grande quantité d'informations. Les sites communautaires de questions réponses représentent un moyen facile et rapide pour obtenir des réponses à n'importe quelle question qu'une personne se pose. Tout ce qu'il suffit de faire c'est de déposer une question sur un de ces sites et d'attendre qu'un autre utilisateur lui réponde. Dans ces sites communautaires, nous voulons identifier les personnes très compétentes. Ce sont des utilisateurs importants qui partagent leurs connaissances avec les autres membres de leurs communauté. Ainsi la détection des experts est devenue une tache très importantes, car elle permet de garantir la qualité des réponses postées sur les différents sites. Dans cette thèse, nous proposons une mesure générale d'expertise fondée sur la théorie des fonctions de croyances. Cette théorie nous permet de gérer l'incertitude présente dans toutes les données émanant du monde réel. D'abord et afin d'identifier ces experts parmi la foule d'utilisateurs présents dans la communauté, nous nous sommes intéressés à identifier des attributs qui permettent de décrire le comportement de chaque individus. Nous avons ensuite développé un modèle statistique fondé sur la théorie des fonctions de croyance pour estimer l'expertise générale des usagers de la plateforme. Cette mesure nous a permis de classifier les différents utilisateurs et de détecter les plus experts d'entre eux. Par la suite, nous proposons une analyse temporelle pour étudier l'évolution temporelle des utilisateurs pendant plusieurs mois. Pour cette partie, nous décrirons com- ment les différents usagers peuvent évoluer au cours de leur activité dans la plateforme. En outre, nous nous sommes également intéressés à la détection des experts potentiels pendant les premiers mois de leurs inscriptions dans un site. L'efficacité de ces approches a été validée par des données réelles provenant de Stack Overflow. / During the last decade, people have changed the way they seek information online. Between question answering communities, specialized websites, social networks, the Web has become one of the most widespread platforms for information exchange and retrieval. Question answering communities provide an easy and quick way to search for information needed in any topic. The user has to only ask a question and wait for the other members of the community to respond. Any person posting a question intends to have accurate and helpful answers. Within these platforms, we want to find experts. They are key users that share their knowledge with the other members of the community. Expert detection in question answering communities has become important for several reasons such as providing high quality content, getting valuable answers, etc. In this thesis, we are interested in proposing a general measure of expertise based on the theory of belief functions. Also called the mathematical theory of evidence, it is one of the most well known approaches for reasoning under uncertainty. In order to identify experts among other users in the community, we have focused on finding the most important features that describe every individual. Next, we have developed a model founded on the theory of belief functions to estimate the general expertise of the contributors. This measure will allow us to classify users and detect the most knowledgeable persons. Therefore, once this metric defined, we look at the temporal evolution of users' behavior over time. We propose an analysis of users activity for several months in community. For this temporal investigation, we will describe how do users evolve during their time spent within the platform. Besides, we are also interested on detecting potential experts during the beginning of their activity. The effectiveness of these approaches is evaluated on real data provided from Stack Overflow.

Knowledge base systems are brittle when the users of the knowledge base are unfamiliar with its content and structure. Querying a knowledge base requires users to state their questions in precise and complete formal representations that relate the facts in the question with relevant terms and relations in the underlying knowledge base. This requirement places a heavy burden on the users to become deeply familiar with the contents of the knowledge base and prevents novice users to effectively using the knowledge base for problem solving. As a result, the utility of knowledge base systems is often restricted to the developers themselves. The goal of this work is to help users, who may possess little domain expertise, to use unfamiliar knowledge bases for problem solving. Our thesis is that the difficulty in using unfamiliar knowledge bases can be addressed by an approach that funnels natural questions, expressed in English, into formal representations appropriate for automated reasoning. The approach uses a simplified English controlled language, a domain-neutral ontology, a set of mechanisms to handle a handful of well known question types, and a software component, called the Question Mediator, to identify relevant information in the knowledge base for problem solving. With our approach, a knowledge base user can use a variety of unfamiliar knowledge bases by posing their questions with simplified English to retrieve relevant information in the knowledge base for problem solving. We studied the thesis in the context of a system called ASKME. We evaluated ASKME on the task of answering exam questions for college level biology, chemistry, and physics. The evaluation consists of successive experiments to test if ASKME can help novice users employ unfamiliar knowledge bases for problem solving. The initial experiment measures ASKME's level of performance under ideal conditions, where the knowledge base is built and used by the same knowledge engineers. Subsequent experiments measure ASKME's level of performance under increasingly realistic conditions. In the final experiment, we measure ASKME's level of performance under conditions where the knowledge base is independently built by subject matter experts and the users of the knowledge base are a group of novices who are unfamiliar with the knowledge base. Results from the evaluation show that ASKME works well on different knowledge bases and answers a broad range of questions that were posed by novice users in a variety of domains. / text

The need for interacting with machines using spoken natural language is growing, along with the expectation that synthetic speech in this context sound natural. Such interaction includes answering questions, where prosody plays an important role in producing natural English synthetic speech by communicating the information structure of utterances. CCG is a theoretical framework that exploits the notion that, in English, information structure, prosodic structure and syntactic structure are isomorphic. This provides a way to convert a semantic representation of an utterance into a prosodically natural spoken utterance. GF is a framework for writing grammars, where abstract tree structures capture the semantic structure and concrete grammars render these structures in linearised strings. This research combines these frameworks to develop a system that converts semantic representations of utterances into linearised strings of natural language that are marked up to inform the prosody-generating component of a speech synthesis system. / Computing / M. Sc. (Computing)