Limin Yao

Modeling relations and their mentions without labeled text

Several recent works on relation extraction have been applying the distant supervision paradigm: instead of relying on annotated text to learn how to predict relations, they employ existing knowledge bases (KBs) as source of supervision. Crucially, these approaches are trained based on the assumption that each sentence which mentions the two related entities is an expression of the given relation. Here we argue that this leads to noisy patterns that hurt precision, in particular if the knowledge base is not directly related to the text we are working with. We present a novel approach to distant supervision that can alleviate this problem based on the following two ideas: First, we use a factor graph to explicitly model the decision whether two entities are related, and the decision whether this relation is mentioned in a given sentence; second, we apply constraint-driven semi-supervision to train this model without any knowledge about which sentences express the relations in our training KB. We apply our approach to extract relations from the New York Times corpus and use Freebase as knowledge base. When compared to a state-of-the-art approach for relation extraction under distant supervision, we achieve 31% error reduction.

Efficient methods for topic model inference on streaming document collections

Topic models provide a powerful tool for analyzing large text collections by representing high dimensional data in a low dimensional subspace. Fitting a topic model given a set of training documents requires approximate inference techniques that are computationally expensive. With todays large-scale, constantly expanding document collections, it is useful to be able to infer topic distributions for new documents without retraining the model. In this paper, we empirically evaluate the performance of several methods for topic inference in previously unseen documents, including methods based on Gibbs sampling, variational inference, and a new method inspired by text classification. The classification-based inference method produces results similar to iterative inference methods, but requires only a single matrix multiplication. In addition to these inference methods, we present SparseLDA, an algorithm and data structure for evaluating Gibbs sampling distributions. Empirical results indicate that SparseLDA can be approximately 20 times faster than traditional LDA and provide twice the speedup of previously published fast sampling methods, while also using substantially less memory.

Universal schema for entity type prediction

Categorizing entities by their types is useful in many applications, including knowledge base construction, relation extraction and query intent prediction. Fine-grained entity type ontologies are especially valuable, but typically difficult to design because of unavoidable quandaries about level of detail and boundary cases. Automatically classifying entities by type is challenging as well, usually involving hand-labeling data and training a supervised predictor. This paper presents a universal schema approach to fine-grained entity type prediction. The set of types is taken as the union of textual surface patterns (e.g. appositives) and pre-defined types from available databases (e.g. Freebase)---yielding not tens or hundreds of types, but more than ten thousands of entity types, such as financier, criminologist, and musical trio. We robustly learn mutual implication among this large union by learning latent vector embeddings from probabilistic matrix factorization, thus avoiding the need for hand-labeled data. Experimental results demonstrate more than 30% reduction in error versus a traditional classification approach on predicting fine-grained entities types.