Companies across a wide range of business areas are working hard to fulfill users wishes to speak to digital voice assistants. The trend of replacing chatbots in favour for voice assistants carries a risk of companies not considering which applications will actually benefit from getting a voice user interface (VUI) resulting in poor user experience.  This thesis aims to investigate which help desk support task will benefit from being implemented in a VUI. By following the Service Design methodology, research on the topic has been conducted and a prototype has been build and tested on a target audience. The results from a user study were evaluated and conclusions have been drawn about which tasks are best suited for being handled by a digital voice assistant.  Two kinds of help desk tasks were evaluated in a user study to compared users experience of the current text based digital assistant with a prototype of a voice based assistant. The aim of the user study was to find which task would benefit from becoming voice based by looking at users acceptance level and over all experience.  The results from the user study showed that employees who use the current text based assistant for help desk tasks, will not choose to speak to a digital voice assistant because they are happy with the service available today. However, employees who don’t use the current text based assistant, will find the digital voice assistant useful. It was also found that short executing tasks such as unlocking accounts, are a better fit for the VUI compared to longer interactions providing information. Two conclusions were drawn, peoples’ preferences are different, meaning that it should be possible to interact with both a text based and voice based assistant when performing help desk tasks. Secondly, the voice based assistant should be implemented as a function in the help desk phone queue instead of being implemented in a browser. Because the users argued that they would be more comfortable speaking to a phone then to a screen. / Företag i alla branscher jobbar hårt för att uppfylla sina användaresönskan om att interagera med röstassistenter. Trenden att byta en chattbot mot en röstassistent medför en risk att företag inte tar hänsyn till huruvida en tjänst faktiskt drar nytta av att göras om till ett röstbaserat användargränssnitt, vilket kan resultera i en försämrad användarupplevelse.  Denna uppsats undersöker vilka funktioner i en kundtjänst som skulle gagnas av att implementeras i ett röstbaserat användargr ̈anssnitt. Genom att använda Service Design modellens forsknings- och idé-genererings fas har en röstbaserad prototyp tagits fram och testats på målgruppen. Resultaten från användarstudien har utvärderats och slutsatser har formulerats.  Två typer av kundtjänstfunktioner har undersökts i en användarstudie som jämfört användarnas upplevelse av den befintliga chatbotten och en röstassistentsprototyp. Målet med användarstudien var att definiera vilka kundtjänstfunktioner som skulle gynnas av att bli röstbaserade genom att titta på användarnas acceptansnivå och övergripande upplevelse.  Resultaten visar att, användare som idag använder, och är nöjda med, chatbotten förmodligen inte kommer att ersätta denna med röstassistenten. Samtidigt som användaren som idag inte använder chatbotten kan tänka sig att använda röstassistenten istället för att ringa till kundtjänsten.  En annan upptäckt från användarstudien var att funktioner som utför en uppgift, så som att låsa upp ett konto, passar bättre i ett röstbaserat sammanhang i jämförelse med när längre information ska förmedlas.  Slutligen formulerades två slutsatser. För det första, olika personer har olika preferenser, det borde alltså vara möjligt att interagera med både chatbotten och röstassistenten för kundtjänstärenden. För det andra, röstassistenten borde implementeras som en plugin som användaren kan utnyttja när denne sitter i telefonkön till kundtjänsten snarare än en egen funktion i på den befintliga hemsida. Detta på grund av att användarna uttryckte att det är mer bekväma med att prata i telefon snarare än till en skärm.

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.

As the technology develops and robots are integrating into more and more facets of our lives, the futureof human-robot interaction may take form in all kinds of arrangements and configurations. In this study, we examined the understandability of di erent conversational strategies in robot-robot communication from a human-bystander standpoint. Specifically, we examined the understandability of verbal explanations constructed under Grice's maxims of informativeness. A prediction task was employed to test the understandability of the proposed strategy among other strategies. Furthermore, participants' perception of the robots' interaction was assessed with a range of ratings and rankings. The results suggest that those robots using the proposed strategy and those using the other tested strategies were understood and perceived similarly. / I takt med att teknologin utvecklas integreras robotar mer och mer i olika delar av våra liv. Framtidens människo-robot interaktioner kan ta många olika former och konfigurationer. I den här studien undersökte vi förståelsen för olika konversationsstrategier mellan robotar ur det mänskliga perspektivet. Specifikt undersökte vi förståelsen av muntliga förklaringar konstruerade enligt Grices princip för informativitet. En uppgift för deltagarna i testet var att försöka förutsäga robotarnas agerande. Dessutom utvärderades robotarnas interaktion genom att låta deltagarna rangordna och betygsätta dem. Resultatet tyder på att de robotar som använder Grices princip och de som använder de andra testade strategierna förstås och uppfattas på ett liknande sätt.

The large amounts of clinical data generated by electronic health record systems are an underutilized resource, which, if tapped, has enormous potential to improve health care. Since the majority of this data is in the form of unstructured text, which is challenging to analyze computationally, there is a need for sophisticated clinical language processing methods. Unsupervised methods that exploit statistical properties of the data are particularly valuable due to the limited availability of annotated corpora in the clinical domain. Information extraction and natural language processing systems need to incorporate some knowledge of semantics. One approach exploits the distributional properties of language – more specifically, term co-occurrence information – to model the relative meaning of terms in high-dimensional vector space. Such methods have been used with success in a number of general language processing tasks; however, their application in the clinical domain has previously only been explored to a limited extent. By applying models of distributional semantics to clinical text, semantic spaces can be constructed in a completely unsupervised fashion. Semantic spaces of clinical text can then be utilized in a number of medically relevant applications. The application of distributional semantics in the clinical domain is here demonstrated in three use cases: (1) synonym extraction of medical terms, (2) assignment of diagnosis codes and (3) identification of adverse drug reactions. To apply distributional semantics effectively to a wide range of both general and, in particular, clinical language processing tasks, certain limitations or challenges need to be addressed, such as how to model the meaning of multiword terms and account for the function of negation: a simple means of incorporating paraphrasing and negation in a distributional semantic framework is here proposed and evaluated. The notion of ensembles of semantic spaces is also introduced; these are shown to outperform the use of a single semantic space on the synonym extraction task. This idea allows different models of distributional semantics, with different parameter configurations and induced from different corpora, to be combined. This is not least important in the clinical domain, as it allows potentially limited amounts of clinical data to be supplemented with data from other, more readily available sources. The importance of configuring the dimensionality of semantic spaces, particularly when – as is typically the case in the clinical domain – the vocabulary grows large, is also demonstrated. / De stora mängder kliniska data som genereras i patientjournalsystem är en underutnyttjad resurs med en enorm potential att förbättra hälso- och sjukvården. Då merparten av kliniska data är i form av ostrukturerad text, vilken är utmanande för datorer att analysera, finns det ett behov av sofistikerade metoder som kan behandla kliniskt språk. Metoder som inte kräver märkta exempel utan istället utnyttjar statistiska egenskaper i datamängden är särskilt värdefulla, med tanke på den begränsade tillgången till annoterade korpusar i den kliniska domänen. System för informationsextraktion och språkbehandling behöver innehålla viss kunskap om semantik. En metod går ut på att utnyttja de distributionella egenskaperna hos språk – mer specifikt, statistisk över hur termer samförekommer – för att modellera den relativa betydelsen av termer i ett högdimensionellt vektorrum. Metoden har använts med framgång i en rad uppgifter för behandling av allmänna språk; dess tillämpning i den kliniska domänen har dock endast utforskats i mindre utsträckning. Genom att tillämpa modeller för distributionell semantik på klinisk text kan semantiska rum konstrueras utan någon tillgång till märkta exempel. Semantiska rum av klinisk text kan sedan användas i en rad medicinskt relevanta tillämpningar. Tillämpningen av distributionell semantik i den kliniska domänen illustreras här i tre användningsområden: (1) synonymextraktion av medicinska termer, (2) tilldelning av diagnoskoder och (3) identifiering av läkemedelsbiverkningar. Det krävs dock att vissa begränsningar eller utmaningar adresseras för att möjliggöra en effektiv tillämpning av distributionell semantik på ett brett spektrum av uppgifter som behandlar språk – både allmänt och, i synnerhet, kliniskt – såsom hur man kan modellera betydelsen av flerordstermer och redogöra för funktionen av negation: ett enkelt sätt att modellera parafrasering och negation i ett distributionellt semantiskt ramverk presenteras och utvärderas. Idén om ensembler av semantisk rum introduceras också; dessa överträffer användningen av ett enda semantiskt rum för synonymextraktion. Den här metoden möjliggör en kombination av olika modeller för distributionell semantik, med olika parameterkonfigurationer samt inducerade från olika korpusar. Detta är inte minst viktigt i den kliniska domänen, då det gör det möjligt att komplettera potentiellt begränsade mängder kliniska data med data från andra, mer lättillgängliga källor. Arbetet påvisar också vikten av att konfigurera dimensionaliteten av semantiska rum, i synnerhet när vokabulären är omfattande, vilket är vanligt i den kliniska domänen. / High-Performance Data Mining for Drug Effect Detection (DADEL)

Historically author profiling has been used in forensic linguistics. However, it is not until the last decades that the analysis method has worked into computer science and machine learning. In comparison, determining author profiling characteristics in machine learning is nothing new. This paper investigates the possibility to improve upon previous results with modern frameworks using data sets that have seen limited usage. The purpose of this master thesis was to use pre-trained transformers or embeddings together with transfer learning. In addition, to examine if general author profiling characteristics of anonymous users on internet forums or conversations on social media could be determined. The data sets used to investigate the questions above were PAN15 and PANDORA, which contains various properties in text data based on authors paired with ground truth labels such as gender, age, and Big Five/OCEAN. In addition, transfer learning of BERT and GloVe was used as a starting point to decrease the learning time of a new task. PAN15, a Twitter data set, did not contain enough data when training a model and was augmented using PANDORA, a Reddit-based data set. Ultimately, BERT obtained the best performance using a stacked approach, achieving 86 − 91% accuracy for each label on unseen data.

The field of speech recognition during the last decade has left the research stage and found its way into the public market, and today, speech recognition software is ubiquitous around us. An automatic speech recognizer understands human speech and represents it as text. Most of the current speech recognition software employs variants of deep neural networks. Before the deep learning era, the hybrid of hidden Markov model and Gaussian mixture model (HMM-GMM) was a popular statistical model to solve speech recognition. In this thesis, automatic speech recognition using HMM-GMM was trained on Somali data which consisted of voice recording and its transcription. HMM-GMM is a hybrid system in which the framework is composed of an acoustic model and a language model. The acoustic model represents the time-variant aspect of the speech signal, and the language model determines how probable is the observed sequence of words. This thesis begins with background about speech recognition. Literature survey covers some of the work that has been done in this field. This thesis evaluates how different language models and discounting methods affect the performance of speech recognition systems. Also, log scores were calculated for the top 5 predicted sentences and confidence measures of pre-dicted sentences. The model was trained on 4.5 hrs of voiced data and its corresponding transcription. It was evaluated on 3 mins of testing data. The performance of the trained model on the test set was good, given that the data was devoid of any background noise and lack of variability. The performance of the model is measured using word error rate(WER) and sentence error rate (SER). The performance of the implemented model is also compared with the results of other research work. This thesis also discusses why log and confidence score of the sentence might not be a good way to measure the performance of the resulting model. It also discusses the shortcoming of the HMM-GMM model, how the existing model can be improved, and different alternatives to solve the problem.

Cross-lingual word embeddings are an increasingly important reseource in cross-lingual methods for NLP, particularly for their role in transfer learning and unsupervised machine translation, purportedly opening up the opportunity for NLP applications for low-resource languages.  However, most research in this area implicitly expects the availablility of vast monolingual corpora for training embeddings, a scenario which is not realistic for many of the world's languages.  Moreover, much of the reporting of the performance of cross-lingual word embeddings is based on a fairly narrow set of mostly European language pairs.  Our study examines the performance of cross-lingual alignment across a more diverse set of language pairs; controls for the effect of the corpus size on which the monolingual embedding spaces are trained; and studies the impact of spectral graph properties of the embedding spsace on alignment.  Through our experiments on a more diverse set of language pairs, we find that performance in bilingual lexicon induction is generally poor in heterogeneous pairs, and that even using a gold or heuristically derived dictionary has little impact on the performance on these pairs of languages.  We also find that the performance for these languages only increases slowly with corpus size.  Finally, we find a moderate correlation between the isospectral difference of the source and target embeddings and the performance of bilingual lexicon induction.  We infer that methods other than cross-lingual alignment may be more appropriate in the case of both low resource languages and heterogeneous language pairs.

Transition-based dependency parsing is known to compute the syntactic structure of a sentence efficiently, but is less accurate to predict long-distance relations between tokens as it lacks global information about the sentence. Our main contribution is the integration of attention mechanisms to replace the static token selection with a dynamic approach that takes the complete sequence into account. Though our experiments confirm that our approach fundamentally works, our models do not outperform the baseline parser. We further present a line of follow-up experiments to investigate these results. Our main conclusion is that the BiLSTM of the traditional parser is already powerful enough to encode the required global information into each token, eliminating the need for an attention-driven approach. Our secondary results indicate that the attention models require a neural network with a higher capacity to potentially extract more latent information from the word embeddings and the LSTM than the traditional parser. We further show that positional encodings are not useful for our attention models, though BERT-style positional embeddings slightly improve the results. Finally, we experiment with replacing the LSTM with a Transformer-encoder to test the impact of self-attention. The results are disappointing, though we think that more future research should be dedicated to this. For our work, we implement a UUParser-inspired dependency parser from scratch in PyTorch and extend it with, among other things, full GPU support and mini-batch processing. We publish the code under a permissive open source license at https://github.com/jgontrum/parseridge.

Modern natural language processing methods requires big textual datasets to function well. A common method is to scrape the internet to acquire the needed data. This does, however, come with the issue that some of the data may be unwanted – for instance, spam websites. As a consequence, the datasets become larger and thus increasing training cost. This thesis defines text as written by humans as running text, and automatically generated texts as non-running text. The goal of the thesis was then to fine-tune the KB-BERT model, BERT pre-trained on Swedish textual data, to classify tokens as either running or non-running text. To do this, texts from the Swedish C4 corpus were manually annotated. In total, 1000 texts were annotated and used for the fine-tuning phase. As the annotated data was a bit skewed in favour of running-text, it was also tested how using class weights to balance the training data affected the end results. When using the BERT-based method with no class weights, the method got a precision and recall for non-running text of 95.13% and 78.84%, and for running text the precision and recall was 83.87% and 96.46%. When using class weights to balance the data, the precision and recall for non-running text were 90.08% and 87.4%, and for running text 89.36% and 92.40%. From these results, one can see that it is possible to alter how strict the model is depending on one’s needs, for instance, purpose and amount of available textual data by using class weights. The number of samples in the manually annotated dataset is too small to make a definite conclusion from, but this thesis shows that using a BERT-based method has the potential to handle problems such as these, as it produced much better results when compared to a more simple baseline-method. Therefore, further research in this area of natural language processing is encouraged. / Moderna språkteknologi-metoder behöver i regel en stor mängd data i form av text för att fungera väl. En vanlig metod för att samla ihop tillräckliga datamängder är att använda tekniker såsom webbskrapning. Detta leder dock i regel till problemet att man även får med oönskad data – till exempel spamhemsidor. Detta leder till att datamängden blir större, vilket innebär en ökad kostnad att träna modellen. Denna avhandling definierar text som löpande ifall den är skriven av människor, och automatiskt genererad text som icke-löpande. Målet med denna avhandling var sedan att finjustera KB-BERT, en BERT-modell som tidigare tränats med svensk text-data, för att klassificera tokens som antingen löpande eller icke-löpande text. För att genomföra detta så annoterades 1000 texter från den svenska delen av C4-korpuset manuellt som sedan användes för att finjustera KB-BERT. Då den annoterade datan innehöll mer löpande än icke-löpande text testades det också hur resultatet påverkades av att använda vikter för att jämna ut förhållandet. När den BERT-baserade metoden utan vikter användes så uppnåddes ett precision och recall för icke-löpande text till 95.13% respektive 78.84%, och för löpande text var precision och recall 83.87% respektive 96.46%. När vikter användes för att balansera datan, så var precision och recall för icke-löpande text 90.08% respektive 87.4%, och för löpande text 89.36% respektive 92.40%. Från dessa resultat kan man tydligt se att det är möjligt att påverka hur strikt modellen är. Hur strikt man vill att modellen ska vara kan variera beroende på, till exempel, ens syfte och hur mycket data man har tillgång till. Dock, det är viktigt att notera att mängden manuellt annoterad data är för liten för att kunna nå en definitiv slutsats. Däremot så visar denna avhandling att BERT-baserade metoder har potentialen att kunna användas för problem likt denna avhandlings frågeställning då den uppnådde mycket bättre resultat än den simplare metod de jämfördes med. Således uppmuntras fortsatt forskning inom detta område av språkteknologi.

Since recent years, the use of non-traditional data sources by hedge funds in order to support investment decisions has increased. One of the data sources which has increased most is social media and it has become popular to analyze the public opinion with help of sentiment analysis in order to predict the performance of a company. In order to evaluate the public opinion one need big sets of Twitter data. The Twitter data was collected by streaming the Twitter feed and the stock data was collected from a Bloomberg Terminal. The aim of this study was to examine if there is a correlation between the public opinion of a stock and the stock price, and also what affects this relationship. While such a relationship cannot be established in general, we are able to show that if the data quality is good, there is a high correlation between the public opinion and stock price, and that significant events surrounding the company results in a higher correlation during that period. / De senaste åren har användandet av icke-traditionella datakällor ökat av hedgefonder för att ta investeringsbeslut. En av datakällorna som blivit populära är sociala medier och det har blivit vanligt att analysera folkopinionen med hjälp av sentimentanalys för att kunna förutspå ett företags resultat. För att analysera folkopinionen krävdes stora mängder Twitterdata. Twitter-datan hämtades genom att strömma Twitter-flödet och aktiedatan hämtades från en Bloomberg Terminal. Målet med studien var att undersöka ifall det finns en korrelation mellan folkopinionen av en aktie och aktiens prisutveckling, och även vad som påverkar denna relationen. Även om en sådan relation inte kan fastställas i allmänhet så kan vi visa att om datakvaliten är god, så finns det en hög korrelation mellan folkopinionen och aktiepriset, samt att vid betydande händelser som rör företaget, så resultar det i en hög korrelation under den perioden.