We analyze the expected cost of a greedy active learning algorithm. Our analysis extends previous work to a more general setting in which different queries have different costs. Moreover, queries may have more than two possible responses and the distribution over hypotheses may be non uniform. Specific applications include active learning with label costs, active learning for multiclass and par...

We present a streaming model for large-scale classification (in the context of l2-SVM) by leveraging connections between learning and computational geometry. The streaming model imposes the constraint that only a single pass over the data is allowed. The l2-SVM is known to have an equivalent formulation in terms of the minimum enclosing ball (MEB) problem, and an efficient algorithm based on th...

This paper presents two alternative NLP architectures to analyze massive amounts of documents, using parallel processing. The two architectures focus on different processing scenarios, namely batch-processing and streaming processing. The batch-processing scenario aims at optimizing the overall throughput of the system, i.e., minimizing the overall time spent on processing all documents. The st...

We investigate two problems in word alignment for machine translation. First, we compare methods for incremental word alignment to save time for large-scale machine translation systems. Various methods of using existing word alignment models trained on a larger, general corpus for incrementally aligning smaller new corpora are compared. In addition, by training separate translation tables, we e...

We consider the batch processing of orders where either whole or part of a single order or a specific pair of different orders may be grouped in a batch with a fixed capacity. The problem can be modelled by a graph G = (V, E), where each node v ∈ V corresponds to an order, its weight w(v) corresponds to the amount of ordered quantity and a pair of orders, say u and v may be grouped in a batch i...

We argue that there are two qualitatively different modes of using a machine-readable dictionary in the context of research in computational linguistics: batch processing of the source with the purpose of collating information for subsequent use by a natural language application, and placing the dictionary on-line in an environment which supports fast interactive access to data selected on the ...

The related problems of transfer learning and multitask learning have attracted significant attention, generating a rich literature of models and algorithms. Yet most existing approaches are studied in an offline fashion, implicitly assuming that data from different domains are given as a batch. Such an assumption is not valid in many real-world applications where data samples arrive sequential...

Caused by powerful sensors, advanced digitalisation techniques, and dramatically increased storage capabilities, big data in the sense of large or streaming data sets, very high dimensionality, or complex data formats constitute one of the major challenges faced by machine learning today. In this realm, a couple of typical assumptions of machine learning can no longer be met, such as e.g. the p...