Automatic Speech Recognition (ASR) enables a wide variety of practical applications. However, many applications have their own domain-specific words, creating a gap between training and test data when used in practice. Domain adaptation can be achieved through model fine-tuning, but it requires domain-specific speech data paired with transcripts, which is labor intensive to produce. Fortunately, the dependence on audio data can be mitigated to a certain extent by incorporating text-based language models during decoding. This thesis explores approaches for creating domain-specific 4-gram models for a Swedish open-source wav2vec 2.0 model. The three main approaches extend a social media corpus with domain-specific data to estimate the models. The first approach utilizes a relatively small set of in-domain text data, and the second approach utilizes machine transcripts from another ASR system. Finally, the third approach utilizes Named Entity Recognition (NER) to find words of the same entity type in a corpus to replace with in-domain words. The 4-gram models are evaluated by the error rate (ERR) of recognizing in-domain words in a custom dataset. Additionally, the models are evaluated by the Word Error Rate (WER) on the Common Voice test set to ensure good overall performance. Compared to not having a language model, the base model improves the WER on Common Voice by 2.55 percentage points and the in-domain ERR by 6.11 percentage points. Next, adding in-domain text to the base model results in a 2.61 WER improvement and a 10.38 ERR improvement over not having a language model. Finally, adding in-domain machine transcripts and using the NER approach results in the same 10.38 ERR improvement as adding in-domain text but slightly less significant WER improvements of 2.56 and 2.47, respectively. These results contribute to the exploration of state-of-the-art Swedish ASR and have the potential to enable the adoption of open-source ASR models for more use cases. / Automatisk taligenkänning (ASR) möjliggör en mängd olika praktiska tillämpningar. Men många tillämpningsområden har sin egen uppsättning domänspecifika ord vilket kan skapa problem när en taligenkänningsmodell används på data som skiljer sig från träningsdatan. Taligenkänningsmodeller kan anpassas till nya domäner genom fortsatt träning med taldata, men det kräver tillgång till domänspecifik taldata med tillhörande transkript, vilket är arbetskrävande att producera. Lyckligtvis kan beroendet av ljuddata mildras till viss del genom användande av textbaserade språkmodeller tillsammans med taligenkänningsmodellerna. Detta examensarbete utforskar tillvägagångssätt för att skapa domänspecifika 4-gram-språkmodeller för en svensk wav2vec 2.0-modell som tränats av Kungliga Biblioteket. Utöver en basmodell så används tre huvudsakliga tillvägagångssätt för att utöka en korpus med domänspecifik data att träna modellerna från. Det första tillvägagångssättet använder en relativt liten mängd domänspecifik textdata, och det andra tillvägagångssättet använder transkript från ett annat ASR-system (maskintranskript). Slutligen använder det tredje tillvägagångssättet Named Entity Recognition (NER) för att hitta ord av samma entitetstyp i en korpus som sedan ersätts med domänspecifika ord. Språkmodellerna utvärderas med ett nytt domänspecifikt evalueringsdataset samt på testdelen av Common Voice datasetet. Jämfört med att inte ha en språkmodell förbättrar basmodellen Word Error Rate (WER) på Common Voice med 2,55 procentenheter och Error Rate (ERR) inom domänen med 6,11 procentenheter. Att lägga till domänspecifik text till basmodellens korpus resulterar i en 2,61 WER-förbättringochen10,38 ERR-förbättring jämfört med att inte ha en språkmodell. Slutligen, att lägga till domänspecifika maskintranskript och att använda NER-metoden resulterar i samma 10.38 ERR-förbättringar som att lägga till domänspecifik text men något mindre WER-förbättringar på 2.56 respektive 2.47 procentenheter. Den här studien bidrar till svensk ASR och kan möjliggöra användandet av öppna taligenkänningsmodeller för fler användningsområden.

Much research work has been done to extract data from scientific papers, journals, and articles. However, Electronic Theses and Dissertations (ETDs) remain an unexplored genre of data in the research fields of natural language processing and machine learning. Moreover, much of the related research involved data that is in the English language. Arabic data such as news and tweets have begun to receive some attention in the past decade. However, Arabic ETDs remain an untapped source of data despite the vast number of benefits to students and future generations of scholars. Some ways of improving the browsability and accessibility of data include data annotation, indexing, parsing, translation, and classification. Classification is essential for the searchability and management of data, which can be manual or automated. The latter is beneficial when handling growing volumes of data. There are two main roadblocks to performing automatic subject classification on Arabic ETDs. The first is the unavailability of a public corpus of Arabic ETDs. The second is the Arabic language’s linguistic complexity, especially in academic documents. This research presents the Otrouha project, which aims at building a corpus of key metadata of Arabic ETDs as well as providing a methodology for their automatic subject classification. The first goal is aided by collecting data from the AskZad Digital Library. The second goal is achieved by exploring different machine learning and deep learning techniques. The experiments’ results show that deep learning using pretrained language models gave the highest classification performance, indicating that language models significantly contribute to natural language understanding. / M.S. / An Electronic Thesis or Dissertation (ETD) is an openly-accessible electronic version of a graduate student’s research thesis or dissertation. It documents their main research effort that has taken place and becomes available in the University Library instead of a paper copy. Over time, collections of ETDs have been gathered and made available online through different digital libraries. ETDs are a valuable source of information for scholars and researchers, as well as librarians. With the digitalization move in most Middle Eastern Universities, the need to make Arabic ETDs more accessible significantly increases as their numbers increase. One of the ways to improve their accessibility and searchability is through providing automatic classification instead of manual classification. This thesis project focuses on building a corpus of metadata of Arabic ETDs and building a framework for their automatic subject classification. This is expected to pave the way for more exploratory research on this valuable genre of data.

The loss of senior expertise in power line design poses a critical challenge to the sustainable energy transition. Current methods of knowledge transfer fail to prevent the loss of invaluable knowledge necessary for future junior power line designers. Additionally, the rise of informal deployment of generative language models may also threaten to bury hand-written knowledge documents before this knowledge can be extracted, structured, and preserved for future guidance. This thesis proposes a framework where large language models are integrated into knowledge transfer and decision-making guidance for an engineering enterprise. Using this framework, this thesis further explores how data-driven knowledge tools can assist junior design engineers by supporting information retrieval and directing to knowledge sources. The ability of a large language model to retrieve relevant knowledge from an engineering design document was validated by comparing the process of human designers manually completing a similar task. In this evaluation involving six participants and the large language model, responses to questions on mechanical dimensioning of stays for utility poles were ranked by experts. The results showed that the large language model responses were ranked similarly to the junior designers on average. Additionally, a small-scale demonstrative knowledge tool, insights from interviews, literature studies as well as the results from the validation study lead to the conclusion that large language models can assist power line designers via a knowledge tool. Beyond power line design, this thesis contributes to the understanding of how data-driven language models can assist knowledge retrieval and decision-making across other engineering design domains. This work utilizes a professional education document on the mechanical dimensioning of wooden power line poles including an analysis on the wind and weight span’s affect on the dimension of the pole, developed parallel to this work. The original design data from the document supported the tests conducted in this thesis. The professional education document on the mechanical dimensioning of wooden power line poles was developed in parallel to this thesis as a case study supporting the tests with original design data on power line design knowledge. The work also discusses risks and ethical aspects when implementing such a knowledge tool. Risks such as leakage of classified information are emphasized and need comprehensive systems and methods to be avoided. It is therefore highlighted how important it is to carry out the project with care and expertise to avoid damage to companies and society. Local language models or highly trusted AI system providers are recommended to ensure that no sensitive information is leaked to an unwanted third-party. With a high degree of caution and consideration of risks, an effective knowledge tool can contribute to increased efficiency, faster and more sustainable development of power line infrastructure, and thus an faster energy transition. / Förlusten av senior expertis inom kraftledningskonstruktion utgör en kritisk utmaning för den hållbara energiomställningen. Nuvarande metoder för kunskapsöverföring är otillräcklig för att förhindra förlusten av ovärderlig kunskap som är nödvändig för framtida juniora kraftledningsprojektörer. Dessutom kan den ökade informella användingen av generativa språkmodeller hota att begrava mänskligt skrivna kunskapsdokument. Detta arbete presenterar ett ramverk d¨ar storskaliga språkmodeller används för att underlätta kunskapsöverföring och tillhandahålla vägledning vid beslutsfattande inom ingenjörsföretag. Med hjälp av detta ramverk utforskar arbetet ytterligare hur datadrivna kunskapsverktyg kan hjälpa juniora kraftledningskonstrukt¨orer genom att stödja informationsinhämtning med hänvisning till kunskapskällorna. En storskalig språkmodells förmåga att hämta relevant kunskap från ett tekniskt designdokument validerades genom att jämföra processen för mänskliga designers som manuellt slutförde en liknande uppgift. I denna utv¨ardering, som involverade sex deltagare och den storskaliga spr˚akmodellen, rankades svaren på frågor om mekanisk dimensionering av stag för kraftledningsstolpar av experter. Resultaten visade att den storskaliga språkmodellens svar i genomsnitt rankades på liknade nivå som de juniora ingenjörerna. Tillsammans med  ett småskaligt demonstrativt kunskapsverktyg, insikter från intervjuer med kraftledningskonstruktörer, litteraturstudier samt resultat från valideringsstudien dras slutsatsen att storskaliga språkmodeller kan stödja kraftledningskonstruktörer via ett kunskapsverktyg. Utöver kraftledningskonstruktion bidrar detta arbete till förståelsen av hur datadrivna språkmodeller kan hjälpa till med kunskapsinhämtning och beslutsfattande  inom andra tekniska designområden. Arbetet använder ett professionellt utbildningsunderlag om mekanisk dimensionering av kraftledningsstolpar i träkonstruktion, inklusive en analys av vertikala- och horistontella linspannets påverkan på stolpens dimension, utvecklat parallellt med detta arbete. Orginaldesigndata från underlaget stödde de tester som genomfördes. Arbetet belyser även risker och etiska aspekter vid implementering av ett sådant kunskapsverktyg. Risker som läckage av sekretessbelagd information betonas, och omfattande system och metoder behövs för att undvika dem. Därför understryks hur viktigt det är att genomföra liknande projekt med noggrannhet, försiktighet och expertis för att undvika skador på företag och samhälle. Lokala språkmodeller eller API-leverantörer med högt förtroende rekommenderas för att minimera risken att känslig information läcker ut till en oönskad tredje part. Med stor försiktighet och hänsyn till riskerna kan ett effektivt kunskapsverktyg bidra till ökad effektivitet, snabbare och mer hållbar utveckling av kraftledningsinfrastruktur, och därmed en snabbare energiomställning.

Les logiciels de correction grammaticale commettent parfois des détections illégitimes (fausses alertes), que nous appelons ici surdétections. La présente étude décrit les expériences de mise au point d’un système créé pour identifier et mettre en sourdine les surdétections produites par le correcteur du français conçu par la société Druide informatique. Plusieurs classificateurs ont été entraînés de manière supervisée sur 14 types de détections faites par le correcteur, en employant des traits couvrant di-verses informations linguistiques (dépendances et catégories syntaxiques, exploration du contexte des mots, etc.) extraites de phrases avec et sans surdétections. Huit des 14 classificateurs développés sont maintenant intégrés à la nouvelle version d’un correcteur commercial très populaire. Nos expériences ont aussi montré que les modèles de langue probabilistes, les SVM et la désambiguïsation sémantique améliorent la qualité de ces classificateurs. Ce travail est un exemple réussi de déploiement d’une approche d’apprentissage machine au service d’une application langagière grand public robuste. / Grammar checking software sometimes erroneously flags a correct word sequence as an error, a problem we call overdetection in the present study. We describe the devel-opment of a system for identifying and filtering out the overdetections produced by the French grammar checker designed by the firm Druide Informatique. Various fami-lies of classifiers have been trained in a supervised way for 14 types of detections flagged by the grammar checker, using features that capture diverse linguistic phe-nomena (syntactic dependency links, POS tags, word context exploration, etc.), extracted from sentences with and without overdetections. Eight of the 14 classifiers we trained are now part of the latest version of a very popular commercial grammar checker. Moreover, our experiments have shown that statistical language models, SVMs and word sense disambiguation can all contribute to the improvement of these classifiers. This project is a striking illustration of a machine learning component suc-cessfully integrated within a robust, commercial natural language processing application.

Cette thèse fait partie du projet RAPSODIE dont l’objectif est de proposer une reconnaissance vocale spécialisée sur les besoins des personnes sourdes et malentendantes. Deux axes sont étudiées : la modélisation lexicale et l’extraction d’informations para-lexicales. Concernant la modélisation lexicale, nous avons étudié les modèles de langage hybrides combinant mots et syllabes, et nous avons proposé une nouvelle approche basée sur une notion de similarité entre mots pour l’ajout de nouveaux mots dans le modèle de langage. Concernant l’extraction d’informations para-lexicales, nous avons étudié l'utilisation des paramètres prosodiques, des paramètres linguistiques ou de leur combinaison pour la détection des questions et des affirmations. Cette détection a comme but de signaler aux personnes sourdes ou malentendantes quand une question leur est adressée / This thesis is part of the RAPSODIE project which aims at proposing a speech recognition device specialized on the needs of deaf and hearing impaired people. Two aspects are studied: optimizing the lexical models and extracting para-lexical information. Regarding the lexical modeling, we studied hybrid language models combining words and syllables, and we proposed a new approach based on a similarity measure between words to add new words in the language model. Regarding the extraction of para-lexical information, we investigated the use of prosodic features, of linguistic features and of their combination for the detection of questions and statements. This detection aims to inform the deaf and hearing impaired people when a question is addressed to them

Given the international status and importance of English, English language study has attracted millions of Chinese learners. Apart from those who study abroad, more and more Chinese students are motivated to study in English-medium Transnational Higher Education (THE) programmes inside China. English is a diversifying and fragmenting language that has various functions and can be used for different purposes. Whilst, according to many scholars, English has broken free from the ownership of ‘native English’ speakers, Chinese learners of English are still worried about conforming to ‘native-speaker models’ of English and so falling victim to an English linguistic imperialism project, driven by English-medium THE programmes. Accordingly, this research sets out to investigate, the extent to which Chinese learners, in a UK affiliated THE programme in China, feel the need to orientate to or identify with ‘native English’ and its speakers, and run the risk of becoming victims of English linguistic imperialism. Results from a combination of methods: questionnaires, focus group discussions and interviews, show that students’ orientations towards and identification with English and its speakers are diverse, complex and multi-dimensional, and have gone beyond affiliation with ‘native English’ speakers. Studying in English-medium THE programmes does not necessarily lead to English linguistic imperialism, but is a process of interaction where learners may consciously mediate ‘native English’ norms and express individual, local, national or international identities, literally taking advantage of the programmes’ material benefits and deliberately learning the language for international communication. This research suggests that learners in THE programmes are conscious of the overall context individually, nationally and internationally and feel free to orientate to English in ways that are suitable for their own purposes and which represent their preferred identity.

El procesamiento del lenguaje natural es un área de aplicación de la inteligencia artificial, en particular, del reconocimiento de formas que estudia, entre otras cosas, incorporar información sintáctica (modelo de lenguaje) sobre cómo deben juntarse las palabras de una determinada lengua, para así permitir a los sistemas de reconocimiento/traducción decidir cual es la mejor hipótesis �con sentido común�. Es un área muy amplia, y este trabajo se centra únicamente en la parte relacionada con el modelado de lenguaje y su aplicación a diversas tareas: reconocimiento de secuencias mediante modelos ocultos de Markov y traducción automática estadística. Concretamente, esta tesis tiene su foco central en los denominados modelos conexionistas de lenguaje, esto es, modelos de lenguaje basados en redes neuronales. Los buenos resultados de estos modelos en diversas áreas del procesamiento del lenguaje natural han motivado el desarrollo de este estudio. Debido a determinados problemas computacionales que adolecen los modelos conexionistas de lenguaje, los sistemas que aparecen en la literatura se construyen en dos etapas totalmente desacopladas. En la primera fase se encuentra, a través de un modelo de lenguaje estándar, un conjunto de hipótesis factibles, asumiendo que dicho conjunto es representativo del espacio de búsqueda en el cual se encuentra la mejor hipótesis. En segundo lugar, sobre dicho conjunto, se aplica el modelo conexionista de lenguaje y se extrae la hipótesis con mejor puntuación. A este procedimiento se le denomina �rescoring�. Este escenario motiva los objetivos principales de esta tesis: � Proponer alguna técnica que pueda reducir drásticamente dicho coste computacional degradando lo mínimo posible la calidad de la solución encontrada. � Estudiar el efecto que tiene la integración de los modelos conexionistas de lenguaje en el proceso de búsqueda de las tareas propuestas. � Proponer algunas modificaciones del modelo original que permitan mejorar su calidad / Zamora Martínez, FJ. (2012). Aportaciones al modelado conexionista de lenguaje y su aplicación al reconocimiento de secuencias y traducción automática [Tesis doctoral no publicada]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/18066 / Palancia

When classifying texts using a linear classifier, the texts are commonly represented as feature vectors. Previous methods to represent features as vectors have been unable to capture the context of individual words in the texts, in theory leading to a poor representation of natural language. Bidirectional Encoder Representations from Transformers (BERT), uses a multi-headed self-attention mechanism to create deep bidirectional feature representations, able to model the whole context of all words in a sequence. A BERT model uses a transfer learning approach, where it is pre-trained on a large amount of data and can be further fine-tuned for several down-stream tasks. This thesis uses one multilingual, and two dedicated Swedish BERT models, for the task of classifying Swedish texts as of either easy-to-read or standard complexity in their respective domains. The performance on the text classification task using the different models is then compared both with feature representation methods used in earlier studies, as well as with the other BERT models. The results show that all models performed better on the classification task than the previous methods of feature representation. Furthermore, the dedicated Swedish models show better performance than the multilingual model, with the Swedish model pre-trained on more diverse data outperforming the other.

Den här studien utforskar den textmässiga likheten mellan AI-genererade texter av ChatGPT och ursprungliga akademiska texter, jämför prestandan hos AI-detekteringsverktyg och maskininlärningsklassificerare, inklusive SVM, Logistic Regression och Random Forest, vid detektering av AI-genererat innehåll, och undersöker hur användarinstruktioner påverkar textkvaliteten. En rad mätvärden som stilometri, sentiment, textlikhet, läsbarhet och relevans användes för att analysera textegenskaper. Resultaten visar att även om AI-genererade texter uppvisar textegenskaper som originaltexter i viss utsträckning, finns det tydliga skillnader. Maskinlärande klassificerare, tränade på DistilBERT-inbäddningar, uppnådde ett F1 Score på 99 % för SVM och Logistic Regression och 96 % för Random Forest, vilket överträffade prestandan för AI-detektionsverktyget, som fick mellan 64– 83 % i F1 Score. Detaljerade instruktioner till ChatGPT visade sig förbättra likheten med originaltexter och minska effektiviteten hos detektionsverktyg. Denna studie bidrar till förståelsen av AI-genererat innehåll och hjälper till att utveckla mer effektiva identifieringsmetoder. / This study explores the textual resemblance between AI-generated texts by ChatGPT and original academic texts, compares the performance of AI-detection tools and machine-learning classifiers, including SVM, Logistic Regression, and Random Forest, in detecting AI-generated content, and investigates the influence of user instructions on text quality. A range of metrics such as stylometry, sentiment, text similarity, readability, and relevance were utilized to analyze text characteristics. Findings reveal that while AI-generated texts do exhibit textual characteristics like original texts to some extent, there are clear differences. Machine-learning classifiers, trained on DistilBERT embeddings, achieved an F1 score of 99% for SVM and Logistic Regression, and 96% for Random Forest, surpassing the performance of the AI detection tool, which scored between 64-83% in F1 measure. Detailed instructions to ChatGPT were found to improve the resemblance to original texts and reduce the effectiveness of detection tools. This study contributes to the understanding of AI-generated content and aids the development of more efficient identification methods.

This thesis explores how large language models can be used to generate configuration files for Terraform from natural language descriptions. Few-shot and fine-tuning paradigms are evaluated on decoder-only models of varying size, including the state-of-the-art Codex model. The generated configuration files are evaluated with regard to functional correctness on a custom dataset using Terraform, to account for the large space of functionally equivalent configuration files. Results show that the largest model Codex is very capable at generating configuration files given an English description of network infrastructure even without fine-tuning. The result could be a useful tool for engineers who know Terraform fundamentals and have experience with the cloud platforms: AWS, GCP, or Azure. A future study could fine-tune Codex for Terraform using OpenAI's API or create an open source Codex-replication by fine-tuning the GPT-3 replication OPT, which in turn can be \hbox{fine-tuned}. / Denna avhandling undersöker hur stora språkmodeller kan användas till att generera konfigurationsfiler för Terraform med hjälp av språkbeskrivningar. Både few-shot och fine-tuning paradigm utvärderas på decoder-only modeller i olika storlekar, inklusive Codex. För att ta hänsyn till konfigurationsfiler som i utseende ser olika ut men som är funktionellt ekvivalenta utvärderas konfigurationsfilerna utifrån deras funktion. Resultaten visar att Codex, som är den största modellen, har förmågan att generera konfigurationsfiler givet en engelsk beskrivning av nätverksinfrastruktur, trots att Codex inte har undergått fine-tuning. Resultatet kan vara ett användbart verktyg för ingenjörer som har grundläggande kunskap om Terraform och erfarenhet av molnplattformarna: AWS, GCP eller Azure. En framtida studie skulle kunna träna Codex för Terraform med OpenAI:s API eller skapa en Codex-kopia genom att träna GPT-3 kopian OPT som i sin tur kan bli tränad för Terraform.