Fredrik D. Johansson

Entity disambiguation in anonymized graphs using graph kernels

This paper presents a novel method for entity disambiguation in anonymized graphs using local neighborhood structure. Most existing approaches leverage node information, which might not be available in several contexts due to privacy concerns, or information about the sources of the data. We consider this problem in the supervised setting where we are provided only with a base graph and a set of nodes labelled as ambiguous or unambiguous. We characterize the similarity between two nodes based on their local neighborhood structure using graph kernels; and solve the resulting classification task using SVMs. We give empirical evidence on two real-world datasets, comparing our approach to a state-of-the-art method, highlighting the advantages of our approach. We show that using less information, our method is significantly better in terms of either speed or accuracy or both. We also present extensions of two existing graphs kernels, namely, the direct product kernel and the shortest-path kernel, with significant improvements in accuracy. For the direct product kernel, our extension also provides significant computational benefits. Moreover, we design and implement the algorithms of our method to work in a distributed fashion using the GraphLab framework, ensuring high scalability.

Neural context embeddings for automatic discovery of word senses

Word sense induction (WSI) is the problem of automatically building an inventory of senses for a set of target words using only a text corpus. We introduce a new method for embedding word instances and their context, for use in WSI. The method, Instance-context embedding (ICE), leverages neural word embeddings, and the correlation statistics they capture, to compute high quality embeddings of word contexts. In WSI, these context embeddings are clustered to find the word senses present in the text. ICE is based on a novel method for combining word embeddings using continuous Skip-gram, based on both se- mantic and a temporal aspects of context words. ICE is evaluated both in a new system, and in an extension to a previous system for WSI. In both cases, we surpass previous state-of-the-art, on the WSI task of SemEval-2013, which highlights the generality of ICE. Our proposed system achieves a 33% relative improvement.

Visions and open challenges for a knowledge-based culturomics

The concept of culturomics was born out of the availability of massive amounts of textual data and the interest to make sense of cultural and language phenomena over time. Thus far however, culturomics has only made use of, and shown the great potential of, statistical methods. In this paper, we present a vision for a knowledge-based culturomics that complements traditional culturomics. We discuss the possibilities and challenges of combining knowledge-based methods with statistical methods and address major challenges that arise due to the nature of the data; diversity of sources, changes in language over time as well as temporal dynamics of information in general. We address all layers needed for knowledge-based culturomics, from natural language processing and relations to summaries and opinions.

Classifying Large Graphs with Differential Privacy

We consider classification of graphs using graph kernels under differential privacy. We develop differentially private mechanisms for two well-known graph kernels, the random walk kernel and the graphlet kernel. We use the Laplace mechanism with restricted sensitivity to release private versions of the feature vector representations of these kernels. Further, we develop a new sampling algorithm for approximate computation of the graphlet kernel on large graphs with guarantees on sample complexity, and show that the method improves both privacy and computation speed. We also observe that the number of samples needed to obtain good accuracy in practice is much lower than the bound. Finally, we perform an extensive empirical evaluation examining the trade-off between privacy and accuracy and show that our private method is able to retain good accuracy in several classification tasks.

Intent-aware temporal query modeling for keyword suggestion

This paper presents a data-driven approach for capturing the temporal variations in user search behaviour by modeling the dynamic query relationships using query-log data. The dependence between different queries (in terms of the query words and latent user intent) is represented using hypergraphs which allows us to explore more complex relationships compared to graph-based approaches. This time-varying dependence is modeled using the framework of probabilistic graphical models. The inferred interactions are used for query keyword suggestion - a key task in web information retrieval. Preliminary experiments using query logs collected from internal search engine of a large health care organization yield promising results. In particular, our model is able to capture temporal variations between queries relationships that reflect known trends in disease occurrence. Further, hypergraph-based modeling captures relationships significantly better compared to graph-based approaches.

Generalized Shortest Path Kernel on Graphs

We consider the problem of classifying graphs using graph kernels. We define a new graph kernel, called the generalized shortest path kernel, based on the number and length of shortest paths between nodes. For our example classification problem, we consider the task of classifying random graphs from two well-known families, by the number of clusters they contain. We verify empirically that the generalized shortest path kernel outperforms the original shortest path kernel on a number of datasets. We give a theoretical analysis for explaining our experimental results. In particular, we estimate distributions of the expected feature vectors for the shortest path kernel and the generalized shortest path kernel, and we show some evidence explaining why our graph kernel outperforms the shortest path kernel for our graph classification problem.

DLOREAN: Dynamic Location-Aware Reconstruction of Multiway Networks

This paper presents a method for learning time-varying higher-order interactions based on node observations, with application to short-term traffic forecasting based on traffic flow sensor measurements. We incorporate domain knowledge into the design of a new damped periodic kernel which leverages traffic flow patterns towards better structure learning. We introduce location-based regularization for learning models with desirable geographical properties (short-range or long-range interactions). We show using experiments on synthetic and real data, that our approach performs better than static methods for reconstruction of multiway interactions, as well as time-varying methods which recover only pair-wise interactions. Further, we show on real traffic data that our model is useful for short-term traffic forecasting, improving over state-of-the-art.