Learning to Interpret and Generate Instructional Recipes

Thesis (Ph.D.)--University of Washington, 2016-08

User Behavioral Modeling of Web-based Systems for Continuous User Authentication

Authentication plays an important role in how we interact with computers, mobile devices, the web, etc. The idea of authentication is to uniquely identify a user before granting access to system privileges. For example, in recent years more corporate information and applications have been accessible via the Internet and Intranet. Many employees are working from remote locations and need access to secure corporate files. During this time, it is possible for malicious or unauthorized users to gain access to the system. For this reason, it is logical to have some mechanism in place to detect whether the logged-in user is the same user in control of the user's session. Therefore, highly secure authentication methods must be used. We posit that each of us is unique in our use of computer systems. It is this uniqueness that is leveraged to "continuously authenticate users" while they use web software. To monitor user behavior, n-gram models are used to capture user interactions with web-based software. This statistical language model essentially captures sequences and sub-sequences of user actions, their orderings, and temporal relationships that make them unique by providing a model of how each user typically behaves. Users are then continuously monitored during software operations. Large deviations from "normal behavior" can possibly indicate malicious or unintended behavior. This approach is implemented in a system called Intruder Detector (ID) that models user actions as embodied in web logs generated in response to a user's actions. User identification through web logs is cost-effective and non-intrusive. We perform experiments on a large fielded system with web logs of approximately 4000 users. For these experiments, we use two classification techniques; binary and multi-class classification. We evaluate model-specific differences of user behavior based on coarse-grain (i.e., role) and fine-grain (i.e., individual) analysis. A specific set of metrics are used to provide valuable insight into how each model performs. Intruder Detector achieves accurate results when identifying legitimate users and user types. This tool is also able to detect outliers in role-based user behavior with optimal performance. In addition to web applications, this continuous monitoring technique can be used with other user-based systems such as mobile devices and the analysis of network traffic.

Generare con transformer lineare spiegazioni di concept set medicali costituiti da termini correlati estratti da social post di pazienti

Le malattie rare pongono diversi scogli ai pazienti, ai loro familiari e ai sanitari. Uno fra questi è la mancanza di informazione che deriva dall'assenza di fonti sicure e semplici da consultare su aspetti dell'esperienza del paziente. Il lavoro presentato ha lo scopo di generare da set termini correlati semanticamente, delle frasi che abbiamo la capacità di spiegare il legame fra di essi e aggiungere informazioni utili e veritiere in un linguaggio semplice e comprensibile. Il problema affrontato oggigiorno non è ben documentato in letteratura e rappresenta una sfida interessante si per complessità che per mancanza di dataset per l'addestramento. Questo tipo di task, come altri di NLP, è affrontabile solo con modelli sempre più potenti ma che richiedono risorse sempre più elevate. Per questo motivo, è stato utilizzato il meccanismo di recente pubblicazione del Performer, dimostrando di riuscire a mantenere uno stesso grado di accuratezza e di qualità delle frasi prodotte, con una parallela riduzione delle risorse utilizzate. Ciò apre la strada all'utilizzo delle reti neurali più recenti anche senza avere i centri di calcolo delle multinazionali. Il modello proposto dunque è in grado di generare frasi che illustrano le relazioni semantiche di termini estratti da un mole di documenti testuali, permettendo di generare dei riassunti dell'informazione e della conoscenza estratta da essi e renderla facilmente accessibile e comprensibile al pazienti o a persone non esperte.

Verbalizzazione di eventi biomedici espressi nella letteratura scientifica: generazione controllata di linguaggio naturale da grafi di conoscenza mediante transformer text-to-text

Il periodo in cui viviamo rappresenta la cuspide di una forte e rapida evoluzione nella comprensione del linguaggio naturale, raggiuntasi prevalentemente grazie allo sviluppo di modelli neurali. Nell'ambito dell'information extraction, tali progressi hanno recentemente consentito di riconoscere efficacemente relazioni semantiche complesse tra entità menzionate nel testo, quali proteine, sintomi e farmaci. Tale task -- reso possibile dalla modellazione ad eventi -- è fondamentale in biomedicina, dove la crescita esponenziale del numero di pubblicazioni scientifiche accresce ulteriormente il bisogno di sistemi per l'estrazione automatica delle interazioni racchiuse nei documenti testuali. La combinazione di AI simbolica e sub-simbolica può consentire l'introduzione di conoscenza strutturata nota all'interno di language model, rendendo quest'ultimi più robusti, fattuali e interpretabili. In tale contesto, la verbalizzazione di grafi è uno dei task su cui si riversano maggiori aspettative. Nonostante l'importanza di tali contributi (dallo sviluppo di chatbot alla formulazione di nuove ipotesi di ricerca), ad oggi, risultano assenti contributi capaci di verbalizzare gli eventi biomedici espressi in letteratura, apprendendo il legame tra le interazioni espresse in forma a grafo e la loro controparte testuale. La tesi propone il primo dataset altamente comprensivo su coppie evento-testo, includendo diverse sotto-aree biomediche, quali malattie infettive, ricerca oncologica e biologia molecolare. Il dataset introdotto viene usato come base per l'addestramento di modelli generativi allo stato dell'arte sul task di verbalizzazione, adottando un approccio text-to-text e illustrando una tecnica formale per la codifica di grafi evento mediante testo aumentato. Infine, si dimostra la validità degli eventi per il miglioramento delle capacità di comprensione dei modelli neurali su altri task NLP, focalizzandosi su single-document summarization e multi-task learning.

Apprendimento non supervisionato di rappresentazioni e legami di similarità tra eventi menzionati nella letteratura biomedica

L'estrazione automatica degli eventi biomedici dalla letteratura scientifica ha catturato un forte interesse nel corso degli ultimi anni, dimostrandosi in grado di riconoscere interazioni complesse e semanticamente ricche espresse all'interno del testo. Purtroppo però, esistono davvero pochi lavori focalizzati sull'apprendimento di embedding o di metriche di similarità per i grafi evento. Questa lacuna lascia le relazioni biologiche scollegate, impedendo l'applicazione di tecniche di machine learning che potrebbero dare un importante contributo al progresso scientifico. Approfittando dei vantaggi delle recenti soluzioni di deep graph kernel e dei language model preaddestrati, proponiamo Deep Divergence Event Graph Kernels (DDEGK), un metodo non supervisionato e induttivo in grado di mappare gli eventi all'interno di uno spazio vettoriale, preservando le loro similarità semantiche e strutturali. Diversamente da molti altri sistemi, DDEGK lavora a livello di grafo e non richiede nè etichette e feature specifiche per un determinato task, nè corrispondenze note tra i nodi. A questo scopo, la nostra soluzione mette a confronto gli eventi con un piccolo gruppo di eventi prototipo, addestra delle reti di cross-graph attention per andare a individuare i legami di similarità tra le coppie di nodi (rafforzando l'interpretabilità), e impiega dei modelli basati su transformer per la codifica degli attributi continui. Sono stati fatti ampi esperimenti su dieci dataset biomedici. Mostriamo che le nostre rappresentazioni possono essere utilizzate in modo efficace in task quali la classificazione di grafi, clustering e visualizzazione e che, allo stesso tempo, sono in grado di semplificare il task di semantic textual similarity. Risultati empirici dimostrano che DDEGK supera significativamente gli altri modelli che attualmente detengono lo stato dell'arte.

Supporting Scientific Research Through Machine and Deep Learning: Fluorescence Microscopy and Operational Intelligence Use Cases

Although the debate of what data science is has a long history and has not reached a complete consensus yet, Data Science can be summarized as the process of learning from data. Guided by the above vision, this thesis presents two independent data science projects developed in the scope of multidisciplinary applied research. The first part analyzes fluorescence microscopy images typically produced in life science experiments, where the objective is to count how many marked neuronal cells are present in each image. Aiming to automate the task for supporting research in the area, we propose a neural network architecture tuned specifically for this use case, cell ResUnet (c-ResUnet), and discuss the impact of alternative training strategies in overcoming particular challenges of our data. The approach provides good results in terms of both detection and counting, showing performance comparable to the interpretation of human operators. As a meaningful addition, we release the pre-trained model and the Fluorescent Neuronal Cells dataset collecting pixel-level annotations of where neuronal cells are located. In this way, we hope to help future research in the area and foster innovative methodologies for tackling similar problems. The second part deals with the problem of distributed data management in the context of LHC experiments, with a focus on supporting ATLAS operations concerning data transfer failures. In particular, we analyze error messages produced by failed transfers and propose a Machine Learning pipeline that leverages the word2vec language model and K-means clustering. This provides groups of similar errors that are presented to human operators as suggestions of potential issues to investigate. The approach is demonstrated on one full day of data, showing promising ability in understanding the message content and providing meaningful groupings, in line with previously reported incidents by human operators.

Advancing digital healthcare research and machine learning-aided knowledge extraction amidst the COVID-19 pandemic

The rapid progression of biomedical research coupled with the explosion of scientific literature has generated an exigent need for efficient and reliable systems of knowledge extraction. This dissertation contends with this challenge through a concentrated investigation of digital health, Artificial Intelligence, and specifically Machine Learning and Natural Language Processing's (NLP) potential to expedite systematic literature reviews and refine the knowledge extraction process. The surge of COVID-19 complicated the efforts of scientists, policymakers, and medical professionals in identifying pertinent articles and assessing their scientific validity. This thesis presents a substantial solution in the form of the COKE Project, an initiative that interlaces machine reading with the rigorous protocols of Evidence-Based Medicine to streamline knowledge extraction. In the framework of the COKE (“COVID-19 Knowledge Extraction framework for next-generation discovery science”) Project, this thesis aims to underscore the capacity of machine reading to create knowledge graphs from scientific texts. The project is remarkable for its innovative use of NLP techniques such as a BERT + bi-LSTM language model. This combination is employed to detect and categorize elements within medical abstracts, thereby enhancing the systematic literature review process. The COKE project's outcomes show that NLP, when used in a judiciously structured manner, can significantly reduce the time and effort required to produce medical guidelines. These findings are particularly salient during times of medical emergency, like the COVID-19 pandemic, when quick and accurate research results are critical.

prism: rinfrangere distribuzioni di probabilità in riassunti astrattivi mediante strategie di decoding

Negli ultimi quattro anni la summarization astrattiva è stata protagonista di una evoluzione senza precedenti dettata da nuovi language model neurali, architetture transformer-based, elevati spazi dimensionali, ampi dataset e innovativi task di pre-training. In questo contesto, le strategie di decoding convertono le distribuzioni di probabilità predette da un modello in un testo artificiale, il quale viene composto in modo auto regressivo. Nonostante il loro cruciale impatto sulla qualità dei riassunti inferiti, il ruolo delle strategie di decoding è frequentemente trascurato e sottovalutato. Di fronte all'elevato numero di tecniche e iperparametri, i ricercatori necessitano di operare scelte consapevoli per ottenere risultati più affini agli obiettivi di generazione. Questa tesi propone il primo studio altamente comprensivo sull'efficacia ed efficienza delle strategie di decoding in task di short, long e multi-document abstractive summarization. Diversamente dalle pubblicazioni disponibili in letteratura, la valutazione quantitativa comprende 5 metriche automatiche, analisi temporali e carbon footprint. I risultati ottenuti dimostrano come non vi sia una strategia di decoding dominante, ma come ciascuna possieda delle caratteristiche adatte a task e dataset specifici. I contributi proposti hanno l'obiettivo di neutralizzare il gap di conoscenza attuale e stimolare lo sviluppo di nuove tecniche di decoding.

Model morphisms (MoMo) to enable language independent information models and interoperable business networks

MSc. Dissertation presented at Faculdade de Ciências e Tecnologia of Universidade Nova de Lisboa to obtain the Master degree in Electrical and Computer Engineering

The effects of language complexity in a counterattitudinal appeal to laypeople : an application of the elaboration likelihood model of persuasion to a legal issue

The effects of a complexly worded counterattitudinal appeal on laypeople's attitudes toward a legal issue were examined, using the Elaboration Likelihood Model (ELM) of persuasion as a theoretical framework. This model states that persuasion can result from the elaboration and scrutiny of the message arguments (i.e., central route processing), or can result from less cognitively effortful strategies, such as relying on source characteristics as a cue to message validity (i.e., peripheral route processing). One hundred and sixty-seven undergraduates (85 men and 81 women) listened to eitller a low status or high status source deliver a counterattitudinal speech on a legal issue. The speech was designed to contain strong or weak arguments. These arguments were 'worded in a simple and, therefore, easy to comprehend manner, or in a complex and, therefore, difficult to comprehend manner. Thus, there were three experimental manipulations: argument comprehensibility (easy to comprehend vs. difficult to comprehend), argumel11 strength (weak vs. strong), and source status (low vs. high). After listening to tIle speec.J] participants completed a measure 'of their attitude toward the legal issue, a thought listil1g task, an argument recall task,manipulation checks, measures of motivation to process the message, and measures of mood. As a result of the failure of the argument strength manipulation, only the effects of the comprehel1sibility and source status manipulations were tested. There was, however, some evidence of more central route processing in the easy comprehension condition than in the difficult comprehension condition, as predicted. Significant correlations were found between attitude and favourable and unfavourable thoughts about the legal issue with easy to comprehend arguments; whereas, there was a correlation only between attitude and favourable thoughts 11 toward the issue with difficult to comprehend arguments, suggesting, perhaps, that central route processing, \vhich involves argument scrutiny and elaboration, occurred under conditions of easy comprehension to a greater extent than under conditions of difficult comprehension. The results also revealed, among other findings, several significant effects of gender. Men had more favourable attitudes toward the legal issue than did women, men recalled more arguments from the speech than did women, men were less frustrated while listening to the speech than were ,vomen, and men put more effort into thinking about the message arguments than did women. When the arguments were difficult to comprehend, men had more favourable thoughts and fewer unfavourable thoughts about the legal issue than did women. Men and women may have had different affective responses to the issue of plea bargaining (with women responding more negatively than men), especially in light of a local and controversial plea bargain that occurred around the time of this study. Such pre-existing gender differences may have led to tIle lower frustration, the greater effort, the greater recall, and more positive attitudes for men than for WOlnen. Results· from this study suggest that current cognitive models of persuasion may not be very applicable to controversial issues which elicit strong emotional responses. Finally, these data indicate that affective responses, the controversial and emotional nature ofthe issue, gender and other individual differences are important considerations when experts are attempting to persuade laypeople toward a counterattitudinal position.

A Dynamical Systems Model for Language Change

Formalizing linguists' intuitions of language change as a dynamical system, we quantify the time course of language change including sudden vs. gradual changes in languages. We apply the computer model to the historical loss of Verb Second from Old French to modern French, showing that otherwise adequate grammatical theories can fail our new evolutionary criterion.

Linguistic and cognitive motivations for the typological primacy model of third language (L3) transfer: considering the role of timing of acquisition and proficiency in the previous languages

This article elucidates the Typological Primacy Model (TPM; Rothman, 2010, 2011, 2013) for the initial stages of adult third language (L3) morphosyntactic transfer, addressing questions that stem from the model and its application. The TPM maintains that structural proximity between the L3 and the L1 and/or the L2 determines L3 transfer. In addition to demonstrating empirical support for the TPM, this article articulates a proposal for how the mind unconsciously determines typological (structural) proximity based on linguistic cues from the L3 input stream used by the parser early on to determine holistic transfer of one previous (the L1 or the L2) system. This articulated version of the TPM is motivated by argumentation appealing to cognitive and linguistic factors. Finally, in line with the general tenets of the TPM, I ponder if and why L3 transfer might obtain differently depending on the type of bilingual (e.g. early vs. late) and proficiency level of bilingualism involved in the L3 process.

Linguistic and cognitive motivations for the typological primacy model of third language (L3) transfer: timing of acquisition and proficiency considered

This article elucidates the Typological Primacy Model (TPM; Rothman, 2010, 2011, 2013) for the initial stages of adult third language (L3) morphosyntactic transfer, addressing questions that stem from the model and its application. The TPM maintains that structural proximity between the L3 and the L1 and/or the L2 determines L3 transfer. In addition to demonstrating empirical support for the TPM, this article articulates a proposal for how the mind unconsciously determines typological (structural) proximity based on linguistic cues from the L3 input stream used by the parser early on to determine holistic transfer of one previous (the L1 or the L2) system. This articulated version of the TPM is motivated by argumentation appealing to cognitive and linguistic factors. Finally, in line with the general tenets of the TPM, I ponder if and why L3 transfer might obtain differently depending on the type of bilingual (e.g. early vs. late) and proficiency level of bilingualism involved in the L3 process.

Evaluation of the Computer-based Battery for Oral Language (BILOv3) through the Rasch model for a Brazilian Sample

Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)