From the perspective of machine learning and data mining applications, expressing data in RDF rather than a domain-specific format can add complexity and obfuscate the internal structure. We investigate and illustrate this issue with an example where bio-molecular graph datasets are expressed in RDF. We use this example to inspire preprocessing techniques which reverse some of the complications...

All of the non-trivial algorithms that are necessary for building and applying a rudimentary syntax-aware statistical machine translation system are generalized parsers. This paper extends the “translation by parsing” architecture by adding two components that are invariably used by state-of-the-art statistical machine translation systems. First, the paper shows how a generic syntax-aware trans...

We de ne the notion of adding \small amounts" of nondeterminism to a deterministic function class, and give a machine model; the result is a functional AC closure of the deterministic class. We characterize, by the \safe parameters" technique, the classes of functions computable in linear and in quasilinear time on a multi-tape Turing machine. We then combine these two results by extending the ...

There is an extcnlal fitctor with vcry substantial conscquenec,,; lot the internal design o1" machine translation systems: exteno aibility. When a machine mmslation system has to allow lbr adding m'bitrary soumc m~d target languages without each time adaptint; the atmady existing pa~ts of the system, tim Reed arises for at careftflly defiv.ed interface ~;tr,ctm'e to which modules R)r addithmal ...

Pattern-based machine translation systems can be easily customized by adding new patterns. To gain full profits from this character, input of patterns should be both expressive and simple to understand. The pattern-based machine translation system we have developed simplifies the handling of features in patterns by allowing sharing constraints between non-terminal symbols, and implementing an a...

Machine learning techniques have been widely applied in many areas. In many cases, high accuracy requires training on large amount of data, adding more expressive features and/or exploring complex input and output interactions, often resulting in scalability problems. My research goal is to design practical algorithms to efficiently learn expressive models from large-scale data and massive know...

When machine learning is deployed in the real world, its performance can be significantly undermined because test data may follow a different distribution from training data. To build a reliable machine learning system in such a scenario, we propose a supervised learning framework that is explicitly robust to the uncertainty of dataset shift. Our robust learning framework is flexible in modelin...

In systems that process sensory data there is frequently a model matching stage where class hypotheses are combined to recognize a complex entity. We introduce a new model of parallelism, the Single Function Multiple Data (SFMD) model, appropriate to this stage. SFMD functionality can be added with small hardware expense to certain existing SIMD architectures, and as an incremental addition to ...

Reordering is one source of error in statistical machine translation (SMT). This paper extends the study of the statistical machine reordering (SMR) approach, which uses the powerful techniques of the SMT systems to solve reordering problems. Here, the novelties yield in: (1) using the SMR approach in a SMT phrase-based system, (2) adding a feature function in the SMR step, and (3) analyzing th...

Learning complex dependencies from time series data is an important task; dependencies can be used to make predictions and characterize a source of data. We have developed Multi-Stream Dependency Detection (msdd), a machine learning algorithm that detects complex dependencies in categorical time-series data. dmsdd attempts to balance the search for strong dependencies across a heterogeneous net...