Search

Translating text into a language unknown to the text's author, dubbed outbound translation, is a modern need for which the user experience has significant room for improvement, beyond the basic machine translation facility. We demonstrate this by showing three ways in which user confidence in the outbound translation, as well as its overall final quality, can be affected: backward translation, quality estimation (with alignment) and source paraphrasing. In this paper, we describe an experiment on outbound translation from English to Czech and Estonian. We examine the effects of each proposed feedback module and further focus on how the quality of machine translation systems influence these findings and the user perception of success. We show that backward translation feedback has a mixed effect on the whole process: it increases user confidence in the produced translation, but not the objective quality., Comment: 9 pages (excluding references); to appear at NAACL-HWT 2021

Human interaction relies on a wide range of signals, including non-verbal cues. In order to develop effective Explainable Planning (XAIP) agents it is important that we understand the range and utility of these communication channels. Our starting point is existing results from joint task interaction and their study in cognitive science. Our intention is that these lessons can inform the design of interaction agents -- including those using planning techniques -- whose behaviour is conditioned on the user's response, including affective measures of the user (i.e., explicitly incorporating the user's affective state within the planning model). We have identified several concepts at the intersection of plan-based agent behaviour and joint task interaction and have used these to design two agents: one reactive and the other partially predictive. We have designed an experiment in order to examine human behaviour and response as they interact with these agents. In this paper we present the designed study and the key questions that are being investigated. We also present the results from an empirical analysis where we examined the behaviour of the two agents for simulated users.

In a review into automated and malicious activity Twitter released a list of accounts that they believed were connected to state sponsored manipulation of the 2016 American Election. This list details 2,752 accounts Twitter believed to be controlled by Russian operatives. In the absence of a similar list of operatives active within the debate on the 2016 UK referendum on membership of the European Union (Brexit) we investigated the behaviour of the same American Election focused accounts in the production of content related to the UK-EU referendum. We found that within our dataset we had Brexit-related content from 419 of these accounts, leading to 3,485 identified tweets gathered between the 29th August 2015 and 3rd October 2017. The behaviour of the accounts altered radically on the day of the referendum, shifting from generalised disruptive tweeting to retweeting each other in order to amplify content produced by other troll accounts. We also demonstrate that, while these accounts are, in general, designed to resemble American citizens, accounts created in 2016 often contained German locations and terms in the user profiles.

This work is produced by researchers at the Neuropolitics Research Lab, School of Social and Political Science and the School of Informatics at the University of Edinburgh. In this report we provide an analysis of the social media posts on the British general election 2017 over the month running up to the vote. We find that pro-Labour sentiment dominates the Twitter conversation around GE2017 and that there is also a disproportionate presence of the Scottish National Party (SNP), given the UK-wide nature of a Westminster election. Substantive issues have featured much less prominently and in a less sustained manner in the Twitter debate than pro and anti leader and political party posts. However, the issue of Brexit has provided a consistent backdrop to the GE2017 conversation and has rarely dropped out of the top three most popular hashtags in the last month. Brexit has been the issue of the GE2017 campaign, eclipsing even the NHS. We found the conversation in the GE2017 Twitter debate to be heavily influenced both by external events and by the top-down introduction of hashtags by broadcast media outlets, often associated with specific programmes and the mediatised political debates. Hashtags like these have a significant impact on the shape of the data collected from Twitter and might distort studies with short data-collection windows but are usually short-lived with little long term impact on the Twitter conversation. If the current polling is to be believed Jeremy Corbyn is unlikely to do as badly as was anticipated when the election was first called. Traditional media sources were slow to pick up on this change in public opinion whereas this trend could be seen early on in social media and throughout the month of May.

In a recent paper [arXiv:1612.04918] there are new results concerning the polytopic set of possible states of a linear discrete-time SISO system subject to bounded disturbances from measurements corrupted by bounded noise. Using these results we construct an algorithm which, for the special case of a plant with a lag, recursively updates these polytopic sets when new measurements arrive., Comment: 10 pages, 5 figures. Paper submitted to 56th CDC, Melbourne 2017

This paper addresses the classical problem of determining the set of possible states of a linear discrete-time system subject to bounded disturbances from measurements corrupted by bounded noise. These so-called uncertainty sets evolve with time as new measurements become available. We present two theorems which describe completely how they evolve with time, and this yields an efficient algorithm for recursively updating uncertainty sets. Numerical simulations demonstrate performance improvements over existing exact methods., Comment: 16 pages, 9 figures. Two theorems describing the propagation of uncertainty set for discrete-time, time-invariant, single-input-single-output plants are presented. A mistake in the labeling of subscripts in equations (4), (5) and (12) has been corrected

Fault tolerance is achieved through multiply redundant hardware systems in large civil aircraft. This means of achieving fault tolerance is infeasible for small compact unmanned aerial vehicles. In this paper we apply a fault tolerant control system which exploits analytical redundancy rather than hardware redundancy to an actual UAV model currently in operation via model-in-the-loop simulation. The fault tolerant control system comprises a nonlinear model predictive controller integrated with an unscented Kalman filter for fault detection and identification. The results show that our fault tolerant control system design is able to identify engine failure within seconds of fault occurrence and distribute control authority to the healthy actuators to maintain safe flight.

In this paper we investigate the design of an active fault tolerant control system applicable to autonomous flight. The system comprises a nonlinear model predictive based controller integrated with an unscented Kalman filter for fault detection and identification. We apply the fault tolerant control system design to a generic aircraft model, and simulate a failed engine scenario. The results show that the system correctly identifies the fault within seconds of occurrence and updates the nonlinear model predictive controller which is then able to reallocate control authority to the healthy actuators based upon up to date fault information.

This paper develops a new active fault tolerant control system based on the concept of analytical redundancy. The novel design consists of an observation filter based fault detection and identification system integrated with a nonlinear model predictive controller. A number of observation filters were designed, integrated with the nonlinear controller and tested before reaching the final design which comprises an unscented Kalman filter for fault detection and identification together with a nonlinear model predictive controller to form an active fault tolerant control system design.

This paper describes the design process for developing a nonlinear model predictive controller for fault tolerant flight control. After examining and implementing a number of numerical techniques, this paper identifies pseudospectral discretisation as the most suitable for this design. Applying the controller to a 2D robot model shows that the nonlinear controller performs much better than the linear controller, especially in the closed loop scenario. Assuming fault detection information, applying the technique to the longitudinal motion of a generic aircraft model shows the design to be eminently suitable for flight control.

In this paper we describe some properties of companion matrices and demonstrate some special patterns that arise when a Toeplitz or a Hankel matrix is multiplied by a related companion matrix. We present a new condition, generalizing known results, for a Toeplitz or a Hankel matrix to be the transforming matrix for a similarity between a pair of companion matrices. A special case of our main result shows that a Toeplitz or a Hankel matrix can be extended using associated companion matrices, preserving the Toeplitz or Hankel structure respectively.

This paper addresses the classical problem of determining the sets of possible states of a linear discrete-time system subject to bounded disturbances from measurements corrupted by bounded noise. These so-called uncertainty sets evolve with time as new measurements become available. We present an exact, computationally simple procedure that propagates a point on the boundary of the uncertainty set at some time instant to a set of points on the boundary of the uncertainty set at the next time instant., Comment: 21 pages, 6 figures

In this note we discover and prove some interesting and important relations among sub-matrices of Sylvester matrices and triangular toeplitz matrices. The main result is Hill's identity discovered by R. D. Hill which has an important application in optimal control problems.

A dual formulation for the problem of determining absolute performance limitations on overshoot, undershoot, maximum amplitude and fluctuation minimization for continuous-time feedback systems is constructed. Determining, for example, the minimum possible overshoot attainable by all possible stabilizing controllers is an optimization task that cannot be expressed as a minimum-norm problem. It is this fact, coupled with the continuous-time rather than discrete-time formulation, that makes these problems challenging. We extend previous results to include more general reference functions, and derive new results (in continuous time) on the influence of pole/zero locations on achievable time-domain performance.