William W. Cohen

Beyond Independent Relevance: Methods and Evaluation Metrics for Subtopic Retrieval

We present a non-traditional retrieval problem we call subtopic retrieval. The subtopic retrieval problem is concerned with finding documents that cover many different subtopics of a query topic. In such a problem, the utility of a document in a ranking is dependent on other documents in the ranking, violating the assumption of independent relevance which is assumed in most traditional retrieval methods. Subtopic retrieval poses challenges for evaluating performance, as well as for developing effective algorithms. We propose a framework for evaluating subtopic retrieval which generalizes the traditional precision and recall metrics by accounting for intrinsic topic difficulty as well as redundancy in documents. We propose and systematically evaluate several methods for performing subtopic retrieval using statistical language models and a maximal marginal relevance (MMR) ranking strategy. A mixture model combined with query likelihood relevance ranking is shown to modestly outperform a baseline relevance ranking on a data set used in the TREC interactive track.

Adaptive name matching in information integration

Identifying approximately duplicate database records that refer to the same entity is essential for information integration. The authors compare and describe methods for combining and learning textual similarity measures for name matching.

Integration of heterogeneous databases without common domains using queries based on textual similarity

Most databases contain “name constants” like course numbers, personal names, and place names that correspond to entities in the real world. Previous work in integration of heterogeneous databases has assumed that local name constants can be mapped into an appropriate global domain by normalization. However, in many cases, this assumption does not hold; determining if two name constants should be considered identical can require detailed knowledge of the world, the purpose of the users query, or both. In this paper, we reject the assumption that global domains can be easily constructed, and assume instead that the names are given in natural language text. We then propose a logic called WHIRL which reasons explicitly about the similarity of local names, as measured using the vector-space model commonly adopted in statistical information retrieval. We describe an efficient implementation of WHIRL and evaluate it experimentally on data extracted from the World Wide Web. We show that WHIRL is much faster than naive inference methods, even for short queries. We also show that inferences made by WHIRL are surprisingly accurate, equaling the accuracy of hand-coded normalization routines on one benchmark problem, and outperforming exact matching with a plausible global domain on a second.

Joint latent topic models for text and citations

In this work, we address the problem of joint modeling of text and citations in the topic modeling framework. We present two different models called the Pairwise-Link-LDA and the Link-PLSA-LDA models. The Pairwise-Link-LDA model combines the ideas of LDA [4] and Mixed Membership Block Stochastic Models [1] and allows modeling arbitrary link structure. However, the model is computationally expensive, since it involves modeling the presence or absence of a citation (link) between every pair of documents. The second model solves this problem by assuming that the link structure is a bipartite graph. As the name indicates, Link-PLSA-LDA model combines the LDA and PLSA models into a single graphical model. Our experiments on a subset of Citeseer data show that both these models are able to predict unseen data better than the baseline model of Erosheva and Lafferty [8], by capturing the notion of topical similarity between the contents of the cited and citing documents. Our experiments on two different data sets on the link prediction task show that the Link-PLSA-LDA model performs the best on the citation prediction task, while also remaining highly scalable. In addition, we also present some interesting visualizations generated by each of the models.

Learning to match and cluster large high-dimensional data sets for data integration

Part of the process of data integration is determining which sets of identifiers refer to the same real-world entities. In integrating databases found on the Web or obtained by using information extraction methods, it is often possible to solve this problem by exploiting similarities in the textual names used for objects in different databases. In this paper we describe techniques for clustering and matching identifier names that are both scalable and adaptive, in the sense that they can be trained to obtain better performance in a particular domain. An experimental evaluation on a number of sample datasets shows that the adaptive method sometimes performs much better than either of two non-adaptive baseline systems, and is nearly always competitive with the best baseline system.

Data integration using similarity joins and a word-based information representation language

The integration of distributed, heterogeneous databases, such as those available on the World Wide Web, poses many problems. Herer we consider the problem of integrating data from sources that lack common object identifiers. A solution to this problem is proposed for databases that contain informal, natural-language “names” for objects; most Web-based databases satisfy this requirement, since they usually present their information to the end-user through a veneer of text. We describe WHIRL, a “soft” database management system which supports “similarity joins,” based on certain robust, general-purpose similarity metrics for text. This enables fragments of text (e.g., informal names of objects) to be used as keys. WHIRL includes textual objects as a built-in type, similarity reasoning as a built-in predicate, and answers every query with a list of answer substitutions that are ranked according to an overall score. Experiments show that WHIRL is much faster than naive inference methods, even for short queries, and efficient on typical queries to real-world databases with tens of thousands of tuples. Inferences made by WHIRL are also surprisingly accurate, equaling the accuracy of hand-coded normalization routines on one benchmark problem, and outerperforming exact matching with a plausible global domain on a second.

Exploiting dictionaries in named entity extraction: combining semi-Markov extraction processes and data integration methods

We consider the problem of improving named entity recognition (NER) systems by using external dictionaries---more specifically, the problem of extending state-of-the-art NER systems by incorporating information about the similarity of extracted entities to entities in an external dictionary. This is difficult because most high-performance named entity recognition systems operate by sequentially classifying words as to whether or not they participate in an entity name; however, the most useful similarity measures score entire candidate names. To correct this mismatch we formalize a semi-Markov extraction process, which is based on sequentially classifying segments of several adjacent words, rather than single words. In addition to allowing a natural way of coupling high-performance NER methods and high-performance similarity functions, this formalism also allows the direct use of other useful entity-level features, and provides a more natural formulation of the NER problem than sequential word classification. Experiments in multiple domains show that the new model can substantially improve extraction performance over previous methods for using external dictionaries in NER.

Relational retrieval using a combination of path-constrained random walks

Scientific literature with rich metadata can be represented as a labeled directed graph. This graph representation enables a number of scientific tasks such as ad hoc retrieval or named entity recognition (NER) to be formulated as typed proximity queries in the graph. One popular proximity measure is called Random Walk with Restart (RWR), and much work has been done on the supervised learning of RWR measures by associating each edge label with a parameter. In this paper, we describe a novel learnable proximity measure which instead uses one weight per edge label sequence: proximity is defined by a weighted combination of simple “path experts”, each corresponding to following a particular sequence of labeled edges. Experiments on eight tasks in two subdomains of biology show that the new learning method significantly outperforms the RWR model (both trained and untrained). We also extend the method to support two additional types of experts to model intrinsic properties of entities: query-independent experts, which generalize the PageRank measure, and popular entity experts which allow rankings to be adjusted for particular entities that are especially important.

Contextual search and name disambiguation in email using graphs

Similarity measures for text have historically been an important tool for solving information retrieval problems. In many interesting settings, however, documents are often closely connected to other documents, as well as other non-textual objects: for instance, email messages are connected to other messages via header information. In this paper we consider extended similarity metrics for documents and other objects embedded in graphs, facilitated via a lazy graph walk. We provide a detailed instantiation of this framework for email data, where content, social networks and a timeline are integrated in a structural graph. The suggested framework is evaluated for two email-related problems: disambiguating names in email documents, and threading. We show that reranking schemes based on the graph-walk similarity measures often outperform baseline methods, and that further improvements can be obtained by use of appropriate learning methods.

A Comparative Study of Methods for Transductive Transfer Learning

The problem of transfer learning, where information gained in one learning task is used to improve performance in another related task, is an important new area of research. While previous work has studied the supervised version of this problem, we study the more challenging case of unsupervised transductive transfer learning, where no labeled data from the target domain are available at training. We describe some current state-of-the-art inductive and transductive approaches and then adapt these models to the problem of transfer learning for protein name extraction. In the process, we introduce a novel maximum entropy based technique, iterative feature transformation (IFT), and show that it achieves comparable performance with state-of-the-art transductive SVMs. We also show how simple relaxations, such as providing additional information like the proportion of positive examples in the test data, can significantly improve the performance of some of the transductive transfer learners.