Browsing by master's degree program "Tietojenkäsittelytieteen maisteriohjelma"

Chatbotteja on kehitetty jo vuosikymmenten ajan, mutta nykyinen kiinnostus on kasvanut niihin teknologian kehityksen myötä. Chatbotit palvelevat ihmisiä eri tarkoituksissa ja niiden toiminta perustuu keskusteluun ihmisen kanssa. Chatbotit tarjoavat henkilökohtaista palvelua vuorokauden jokaisena hetkenä, jonka vuoksi niiden tarve on lisääntynyt monilla aloilla, kuten verkkomyynnissä ja terveydenhuollossa. Chatbottien kehityksessä on tärkeää pohtia niiden toteutusta. Monet käyttäjät suosivat edelleen muita informaationlähteitä heidän ongelmiensa ratkaisuun. Yksi tapa mitata chatbot-järjestelmien laatua on tutkia niiden käyttäjäkokemusta. Tässä tutkielmassa tarkastellaan empiirisesti chatbot-sovellusten käyttäjäkokemusta. Empiirisen osion muodostaa laadullinen tutkimus, jonka avulla pyritään vastaamaan seuraavaan tutkimuskysymykseen: Kuinka chatbottien käyttäjäkokemusta voitaisiin parantaa? Tutkimus järjestettiin Osaamisbotti-palvelun kanssa, joka tarjosi testiympäristön tutkimuksen suorittamiselle. Tutkimukseen osallistui kahdeksan henkilöä, jotka suorittivat heille annetun tehtävän keskustelemalla chatbotin kanssa. Tutkimuksen aineisto on saatu protokolla-analyysin ja sen jälkeisen haastattelun keinoin. Tulokset esittävät, että ihmismäiset keskustelukyvyt, pidemmät vastaukset sekä tehokas keskustelun kulku parantavat chatbottien käyttäjäkokemusta. Lisäksi riittävällä informoinnilla ohjataan keskustelua sekä vältetään virhetilanteita. Chatbottien hyvällä saatavuudella sekä helppokäyttöisyydellä kasvatetaan niiden hyväksyntää ja käyttöönottoa. Tutkielman tuloksia voidaan hyödyntää tulevissa tutkimuksissa ja chatbottien kehitystyössä.

Sequence alignment by exact or approximate string matching is one of the fundamental problems in bioinformatics. As the volume of sequenced genomes grows rapidly, pairwise sequence alignment becomes inefficient for pan-genomic analyses involving multiple sequences. The graph representation of multiple genomes has been an increasingly useful tool in pan-genomics research. Therefore, sequence-to-graph alignment becomes an important and challenging problem. For pairwise approximate sequence alignment under Levenshtein (edit) distance, subquadratic algorithms for finding an optimal solution are unknown. As a result, aligning sequences of millions of characters optimally is too challenging and impractical. Thus, many heuristics and techniques are developed for possibly suboptimal alignments. Among them, co-linear chaining (CLC) is a powerful and popular technique that approximates the alignment by finding a chain of short aligned fragments that may come from exact matching. The optimal solution to CLC on sequences can be found efficiently in subquadratic time. For sequence-to-graph alignment, the CLC problem has been solved theoretically on a special class of graphs that are narrow and have no cycles, i.e. directed acyclic graphs (DAGs) with small width, by Mäkinen et al. (ACM Transactions on Algorithms, 2019). Pan-genome graphs such as variation graphs satisfy these restrictions but allowing cycles may enable more general applications of the algorithm. In this thesis, we introduce an efficient algorithm to solve the CLC problem on general graphs with small width that may have cycles, by reducing it to a slightly modified CLC problem on DAGs. We implemented an initial version of the new algorithm on DAGs as a sequence-to-graph aligner GraphChainer. The aligner is evaluated and compared to an existing state-of-the-art aligner GraphAligner (Genome Biology, 2020) in experiments using both simulated and real genome assembly data on variation graphs. Our method improves the quality of alignments significantly in the task of aligning real human PacBio data. GraphChainer is freely available as an open source tool at https://github.com/algbio/GraphChainer.

Hashing is a one-way encryption method that can be used for data integrity verification, for example, in digital signature systems. The Ssdeep algorithm is a classic context-triggered piecewise hashing function that is commonly used for similarity file check. The input is divided into separate block segments for the signature generation so that the modification of some parts only makes a difference to certain bytes of the signature. This characteristic makes it one of the most popular fuzzy hash algorithms for detecting similar information. Nevertheless, the cryptanalysis of the Ssdeep is missing in previous research. Therefore, in this thesis, we propose collision attack methods based on such vulnerabilities to test the feasibility of using two different inputs to obtain the same signature. Specifically, our objective is to let the attacker add custom comments to a code file, so that the output signature is identical to a previously acknowledge signature by the target system using Ssdeep. In our work, we identify the vulnerabilities in the traditional hash and the rolling hash in the Ssdeep calculation. We further name three useful elements, the reset string, the matching character, and the trigger string, to control the Ssdeep process. Specifically for finding the matching character, we use brute force and modular multiplicative inverse numbers, based on which we further propose two implementation versions: coarse-grained and fine-grained differed in the number of potential solution states. Additionally, we investigate the block size, a parameter which is dependent on the file content length, so that the proposed attack methods can work under various realistic scenarios. We test our work by comprehensive experiments, and the results verify the effectiveness of our methods in collision attacks on the Ssdeep algorithm. Our work gives thoughtful insights into the breaking of data integrity in other fuzzy and context-triggered piecewise hashing algorithms.

This Master’s Thesis describes an original user study that took place at the University of Helsinki. The study compares and evaluates the usability of three different methods that are used in meeting rooms to share a private device’s screen on a big public screen in order to give a slideshow presentation: HDMI, VIA, and Ubicast. There were 18 participants. The study was conducted in a controlled environment, replicating a typical meeting room setup. The experiment consisted of screen mirroring tasks and an interview. In a screen mirroring task, the participants were asked to share their screen using each of the three technologies. They were provided with the necessary equipment and user guides if needed. Then the participants were given training on how to use the technologies, and they performed the tasks again. During the task, the time taken to complete each screen mirroring session was recorded, and any errors or difficulties encountered were noted. After completing the screen mirroring tasks, participants were interviewed to gather qualitative data on their experiences and preferences. They were asked about the ease of use, efficiency, and any difficulties they faced while using each technology. This information was used to gain insights into user preferences and potential areas for improvement in the respective technologies. To analyze the data, the System Usability Scale (SUS) scores and time taken to complete the screen mirroring tasks were calculated for each technology. Statistical analyses were conducted to determine any significant differences in SUS scores and time across the three technologies. Additionally, the interview data was analyzed using thematic analysis to identify common themes and patterns in the experiences of the users. HDMI emerged on the top, with Ubicast not far behind.

Graph drawing in two dimensions is an established area of research with many proposed drawing algorithms and drawing quality measures. Most research has remained in two dimensions because of the limitations of paper and traditional monitors. However, with the advancements in head-mounted VR and AR headsets, it is possible to visualise graphs in true three dimensions which allows the viewer to comprehend larger graphs. To provide a useful graph viewing experience, the graph needs to be laid out in a way that highlights important features and structures. In this paper, we provide a collection of measures that measure the quality of 3D graph layouts and analyse their usefulness. We also measure the quality and performance of eleven graph layout algorithms on synthetic and real-world graphs and give suggestions for choosing an algorithm.

Demand forecasts are required for optimizing multiple challenges in the retail industry, and they can be used to reduce spoilage and excess inventory sizes. The classical forecasting methods provide point forecasts and do not quantify the uncertainty of the process. We evaluate multiple predictive posterior approximation methods with a Bayesian generalized linear model that captures weekly and yearly seasonality, changing trends and promotional effects. The model uses negative binomial as the sampling distribution because of the ability to scale the variance as a quadratic function of the mean. The forecasting methods provide highest posterior density intervals in different credible levels ranging from 50% to 95%. They are evaluated with proper scoring function and calculation of hit rates. We also measure the duration of the calculations as an important result due to the scalability requirements of the retail industry. The forecasting methods are Laplace approximation, Monte Carlo Markov Chain method, Automatic Differentiation Variational Inference, and maximum a posteriori inference. Our results show that the Markov Chain Monte Carlo method is too slow for practical use, while the rest of the approximation methods can be considered for practical use. We found out that Laplace approximation and Automatic Differentiation Variational Inference have results closer to the method with best analytical quarantees, the Markov Chain Monte Carlo method, suggesting that they were better approximations of the model. The model faced difficulties with highly promotional, slow selling, and intermittent data. Best fit was provided with high selling SKUs, for which the model provided intervals with hit rates that matched the levels of the credible intervals.

Arranging products in stores according to planograms, optimized product arrangement maps, is important for keeping up with the highly competitive modern retail market. The planograms are realized into product arrangements by humans, a process which is prone to mistakes. Therefore, for optimal merchandising performance, the planogram compliance of the arrangements needs to be evaluated from time to time. We investigate utilizing a computer vision problem setting – retail product detection – to automate planogram compliance evaluation. We introduce the relevant problems, the state-of- the-art approaches for solving them and background information necessary for understanding them. We then propose a computer vision based planogram compliance evaluation pipeline based on the current state of the art. We build our proposed models and algorithms using PyTorch, and run tests against public datasets and an internal dataset collected from a large Nordic retailer. We find that while the retail product detection performance of our proposed approach is quite good, the planogram compliance evaluation performance of our whole pipeline leaves a lot of room for improvement. Still, our approach seems promising, and we propose multiple ways for improving the performance enough to enable possible real world utility. The code used for our experiments and the weights for our models are available at https://github.com/laitalaj/cvpce

Software testing is an important part of ensuring software quality. Studies have shown that having more tests results in a lower count of defects. Code coverage is a tool used in software testing to find parts of the software that require further testing and to learn which parts have been tested. Code coverage is generated automatically by the test suites during test execution. Many types of code coverage metrics exist, the most common being line coverage, statement coverage, function coverage, and branch coverage metrics. These four common metrics are usually enough, but there are many specific coverage types for specific purposes, such as condition coverage which tells how many boolean conditions have been evaluated as true and false. Each different metric gives hints on how the codebase is tested. A common consensus amongst practitioners is that code coverage does not correlate much with software quality. The correlation of software quality with code coverage is a historically broadly researched topic, which has importance both in academia and professional practice. This thesis investigates if code coverage correlates with software quality by performing a literature review. Surprising results are derived from the literature review, as most studies included in this thesis point towards code coverage correlating with software quality. This positive correlation comes from 22 studies conducted between 1995-2021, which include Academic and Industrial studies, with studies put into multiple categories, such as Correlation or No correlation based on the key finding, and categories such as Survey studies, Case studies, Open-source studies, based on the study type. Each category has most studies pointing towards a correlation. This finding is in contradiction with the opinions of professional practitioners.

Improving the monitoring of health and well-being of dairy cows through the use of computer vision based systems is a topic of ongoing research. A reliable and low-cost method for identifying cow individuals would enable automatic detection of stress, sickness or injury, and the daily observation of the animals would be made easier. Neural networks have been used successfully in the identification of cow individuals, but methods are needed that do not require incessant annotation work to generate training datasets when there are changes within a group. Methods for person re-identification and tracking have been researched extensively, with the aim of generalizing beyond the training set. These methods have been found suitable also for re-identifying and tracking previously unseen dairy cows in video frames. In this thesis, a metric-learning based re-identification model pre-trained on an existing cow dataset is compared to a similar model that has been trained on new video data recorded at Luke Maaninka research farm in Spring 2021, which contains 24 individually labelled cow individuals. The models are evaluated in tracking context as appearance descriptors in Kalman filter based tracking algorithm. The test data is video footage from a separate enclosure in Maaninka and a group of 24 previously unseen cow individuals. In addition, a simple procedure is proposed for the automatic labeling of cow identities in images based on RFID data collected from cow ear tags and feeding stations, and the known feeding station locations.

The research group dLearn.Helsinki has created a software for defining the work life competence skills of a person, working as a part of a group. The software is a research tool for developing the mentioned skills of users, and users can be of any age, from school children to employees in a company. As the users can be of different age groups, the data privacy of different groups has to be taken into consideration from different aspects. Children are more vulnerable than adults, and may not understand all the risks imposed to-wards them. Thus in the European Union the General Data Protection Regulation (GDPR)determines the privacy and data of children are more protected, and this has to be taken into account when designing software which uses said data. For dLearn.Helsinki this caused changes not only in the data handling of children, but also other users. To tackle this problem, existing and future use cases needed to be planned and possibly implemented. Another solution was to implement different versions of the software, where the organizations would be separate. One option would be determining organizational differences in the existing SaaS solution. The other option would be creating on-premise versions, where organizations would be locked in accordance to the customer type. This thesis introduces said use cases, as well as installation options for both SaaS and on-premise. With these, broader views of data privacy and the different approaches are investigated, and it can be concluded that no matter the approach, the data privacy of children will always prove a challenge.