Statistical relational learning (SRL) algorithms combine ideas from rich knowledge representations, such as first-order logic, with those from probabilistic graphical models, such as Markov networks, to address the problem of learning from multi-relational data. One challenge posed by such data is that individual instances are frequently very large and include complex relationships among the entities. Moreover, because separate instances do not follow the same structure and contain varying numbers of entities, they cannot be effectively represented as a feature-vector. SRL models and algorithms have been successfully applied to a wide variety of domains such as social network analysis, biological data analysis, and planning, among others. Markov logic networks (MLNs) are a recently-developed SRL model that consists of weighted first-order clauses. MLNs can be viewed as templates that define Markov networks when provided with the set of constants present in a domain. MLNs are therefore very powerful because they inherit the expressivity of first-order logic. At the same time, MLNs can flexibly deal with noisy or uncertain data to produce probabilistic predictions for a set of propositions. MLNs have also been shown to subsume several other popular SRL models.
The expressive power of MLNs comes at a cost: structure learning, or learning the first-order clauses of the model, is a very computationally intensive process that needs to sift through a large hypothesis space with many local maxima and plateaus. It is therefore an important research problem to develop learning algorithms that improve the speed and accuracy of this process. The main contribution of this proposal are two algorithms for learning the structure of MLNs that proceed in a more data-driven fashion, in contrast to most existing SRL algorithms. The first algorithm we present, R-TAMAR, improves learning by transferring the structure of an MLN learned in a domain related to the current one. It first diagnoses the transferred structure and then focuses its efforts only on the regions it determines to be incorrect. Our second algorithm, BUSL improves structure learning from scratch by approaching the problem in a more bottom-up fashion and first constructing a variablized Markov network template that significantly constrains the space of viable clause candidates. We demonstrate the effectiveness of our methods in three social domains.
Our proposed future work directions include testing BUSL in additional domains and extending it so that it can be used not only to learn from scratch, but also to revise a provided MLN structure. Our most ambitious long-term goal is to develop a system that transfers knowledge from multiple potential sources. An important prerequisite to such a system is a method for measuring the similarity between domains. We would also like to extend BUSL to learn other SRL models and to handle functions.
ML ID: 217
Semantic parsing involves deep semantic analysis that maps natural language sentences to their formal executable meaning representations. This is a challenging problem and is critical for developing computing systems that understand natural language input. This thesis presents a new machine learning approach for semantic parsing based on string-kernel-based classification. It takes natural language sentences paired with their formal meaning representations as training data. For every production in the formal language grammar, a Support-Vector Machine (SVM) classifier is trained using string similarity as the kernel. Meaning representations for novel natural language sentences are obtained by finding the most probable semantic parse using these classifiers. This method does not use any hard-matching rules and unlike previous and other recent methods, does not use grammar rules for natural language, probabilistic or otherwise, which makes it more robust to noisy input.
Besides being robust, this approach is also flexible and able to learn under a wide range of supervision, from extra to weaker forms of supervision. It can easily utilize extra supervision given in the form of syntactic parse trees for natural language sentences by using a syntactic tree kernel instead of a string kernel. Its learning algorithm can also take advantage of detailed supervision provided in the form of semantically augmented parse trees. A simple extension using transductive SVMs enables the system to do semi-supervised learning and improve its performance utilizing unannotated sentences which are usually easily available. Another extension involving EM-like retraining makes the system capable of learning under ambiguous supervision in which the correct meaning representation for each sentence is not explicitly given, but instead a set of possible meaning representations is given. This weaker and more general form of supervision is better representative of a natural training environment for a language-learning system requiring minimal human supervision.
For a semantic parser to work well, conformity between natural language and meaning representation grammar is necessary. However meaning representation grammars are typically designed to best suit the application which will use the meaning representations with little consideration for how well they correspond to natural language semantics. We present approaches to automatically transform meaning representation grammars to make them more compatible with natural language semantics and hence more suitable for learning semantic parsers. Finally, we also show that ensembles of different semantic parser learning systems can obtain the best overall performance.
ML ID: 215
One of the main goals of natural language processing (NLP) is to build automated systems that can understand and generate human languages. This goal has so far remained elusive. Existing hand-crafted systems can provide in-depth analysis of domain sub-languages, but are often notoriously fragile and costly to build. Existing machine-learned systems are considerably more robust, but are limited to relatively shallow NLP tasks.
In this thesis, we present novel statistical methods for robust natural language understanding and generation. We focus on two important sub-tasks, semantic parsing and tactical generation. The key idea is that both tasks can be treated as the translation between natural languages and formal meaning representation languages, and therefore, can be performed using state-of-the-art statistical machine translation techniques. Specifically, we use a technique called synchronous parsing, which has been extensively used in syntax-based machine translation, as the unifying framework for semantic parsing and tactical generation. The parsing and generation algorithms learn all of their linguistic knowledge from annotated corpora, and can handle natural-language sentences that are conceptually complex.
A nice feature of our algorithms is that the semantic parsers and tactical generators share the same learned synchronous grammars. Moreover, charts are used as the unifying language-processing architecture for efficient parsing and generation. Therefore, the generators are said to be the inverse of the parsers, an elegant property that has been widely advocated. Furthermore, we show that our parsers and generators can handle formal meaning representation languages containing logical variables, including predicate logic.
Our basic semantic parsing algorithm is called WASP. Most of the other parsing and generation algorithms presented in this thesis are extensions of WASP or its inverse. We demonstrate the effectiveness of our parsing and generation algorithms by performing experiments in two real-world, restricted domains. Experimental results show that our algorithms are more robust and accurate than the currently best systems that require similar supervision. Our work is also the first attempt to use the same automatically-learned grammar for both parsing and generation. Unlike previous systems that require manually-constructed grammars and lexicons, our systems require much less knowledge engineering and can be easily ported to other languages and domains.
ML ID: 214
Information Extraction, the task of locating textual mentions of specific types of entities and their relationships, aims at representing the information contained in text documents in a structured format that is more amenable to applications in data mining, question answering, or the semantic web. The goal of our research is to design information extraction models that obtain improved performance by exploiting types of evidence that have not been explored in previous approaches. Since designing an extraction system through introspection by a domain expert is a laborious and time consuming process, the focus of this thesis will be on methods that automatically induce an extraction model by training on a dataset of manually labeled examples.
Named Entity Recognition is an information extraction task that is concerned with finding textual mentions of entities that belong to a predefined set of categories. We approach this task as a phrase classification problem, in which candidate phrases from the same document are collectively classified. Global correlations between candidate entities are captured in a model built using the expressive framework of Relational Markov Networks. Additionally, we propose a novel tractable approach to phrase classification for named entity recognition based on a special Junction Tree representation.
Classifying entity mentions into a predefined set of categories achieves only a partial disambiguation of the names. This is further refined in the task of Named Entity Disambiguation, where names need to be linked to their actual denotations. In our research, we use Wikipedia as a repository of named entities and propose a ranking approach to disambiguation that exploits learned correlations between words from the name context and categories from the Wikipedia taxonomy.
Relation Extraction refers to finding relevant relationships between entities mentioned in text documents. Our approaches to this information extraction task differ in the type and the amount of supervision required. We first propose two relation extraction methods that are trained on documents in which sentences are manually annotated for the required relationships. In the first method, the extraction patterns correspond to sequences of words and word classes anchored at two entity names occurring in the same sentence. These are used as implicit features in a generalized subsequence kernel, with weights computed through training of Support Vector Machines. In the second approach, the implicit extraction features are focused on the shortest path between the two entities in the word-word dependency graph of the sentence. Finally, in a significant departure from previous learning approaches to relation extraction, we propose reducing the amount of required supervision to only a handful of pairs of entities known to exhibit or not exhibit the desired relationship. Each pair is associated with a bag of sentences extracted automatically from a very large corpus. We extend the subsequence kernel to handle this weaker form of supervision, and describe a method for weighting features in order to focus on those correlated with the target relation rather than with the individual entities. The resulting Multiple Instance Learning approach offers a competitive alternative to previous relation extraction methods, at a significantly reduced cost in human supervision.
ML ID: 213
We present a new approach to relation extraction that requires only a handful of training examples. Given a few pairs of named entities known to exhibit or not exhibit a particular relation, bags of sentences containing the pairs are extracted from the web. We extend an existing relation extraction method to handle this weaker form of supervision, and present experimental results demonstrating that our approach can reliably extract relations from web documents.
ML ID: 204
Transfer learning addresses the problem of how to leverage knowledge acquired in a source domain to improve the accuracy and speed of learning in a related target domain. This paper considers transfer learning with Markov logic networks (MLNs), a powerful formalism for learning in relational domains. We present a complete MLN transfer system that first autonomously maps the predicates in the source MLN to the target domain and then revises the mapped structure to further improve its accuracy. Our results in several real-world domains demonstrate that our approach successfully reduces the amount of time and training data needed to learn an accurate model of a target domain over learning from scratch.
ML ID: 203
Markov logic networks (MLNs) are a statistical relational model that consists of weighted first-order clauses and generalizes first-order logic and Markov networks. The current state-of-the-art algorithm for learning MLN structure follows a top-down paradigm where many potential candidate structures are systematically generated without considering the data and then evaluated using a statistical measure of their fit to the data. Even though this existing algorithm outperforms an impressive array of benchmarks, its greedy search is susceptible to local maxima or plateaus. We present a novel algorithm for learning MLN structure that follows a more bottom-up approach to address this problem. Our algorithm uses a ``propositional'' Markov network learning method to construct ``template'' networks that guide the construction of candidate clauses. Our algorithm significantly improves accuracy and learning time over the existing top-down approach in three real-world domains.
ML ID: 202
We present a new approach to multiple instance learning (MIL) that is particularly effective when the positive bags are sparse (i.e. contain few positive instances). Unlike other SVM-based MIL methods, our approach more directly enforces the desired constraint that at least one of the instances in a positive bag is positive. Using both artificial and real-world data, we experimentally demonstrate that our approach achieves greater accuracy than state-of-the-art MIL methods when positive bags are sparse, and performs competitively when they are not. In particular, our approach is the best performing method for image region classification.
ML ID: 201
This paper presents a method for learning a semantic parser from ambiguous supervision. Training data consists of natural language sentences annotated with multiple potential meaning representations, only one of which is correct. Such ambiguous supervision models the type of supervision that can be more naturally available to language-learning systems. Given such weak supervision, our approach produces a semantic parser that maps sentences into meaning representations. An existing semantic parsing learning system that can only learn from unambiguous supervision is augmented to handle ambiguous supervision. Experimental results show that the resulting system is able to cope up with ambiguities and learn accurate semantic parsers.
ML ID: 200
This paper presents the first empirical results to our knowledge on learning synchronous grammars that generate logical forms. Using statistical machine translation techniques, a semantic parser based on a synchronous context-free grammar augmented with lambda-operators is learned given a set of training sentences and their correct logical forms. The resulting parser is shown to be the best-performing system so far in a database query domain.
ML ID: 199
We present a method for utilizing unannotated sentences to improve a semantic parser which maps natural language (NL) sentences into their formal meaning representations (MRs). Given NL sentences annotated with their MRs, the initial supervised semantic parser learns the mapping by training Support Vector Machine (SVM) classifiers for every production in the MR grammar. Our new method applies the learned semantic parser to the unannotated sentences and collects unlabeled examples which are then used to retrain the classifiers using a variant of transductive SVMs. Experimental results show the improvements obtained over the purely supervised parser, particularly when the annotated training set is small.
ML ID: 198
This paper explores the use of statistical machine translation (SMT) methods for tactical natural language generation. We present results on using phrase-based SMT for learning to map meaning representations to natural language. Improved results are obtained by inverting a semantic parser that uses SMT methods to map sentences into meaning representations. Finally, we show that hybridizing these two approaches results in still more accurate generation systems. Automatic and human evaluation of generated sentences are presented across two domains and four languages.
ML ID: 197
Semantic parsing is the task of mapping a natural language sentence into a complete, formal meaning representation. Over the past decade, we have developed a number of machine learning methods for inducing semantic parsers by training on a corpus of sentences paired with their meaning representations in a specified formal language. We have demonstrated these methods on the automated construction of natural-language interfaces to databases and robot command languages. This paper reviews our prior work on this topic and discusses directions for future research.
ML ID: 196
ML ID: 186
Understanding natural language presents many challenging problems that lend themselves to statistical relational learning (SRL). Historically, both logical and probabilistic methods have found wide application in natural language processing (NLP). NLP inevitably involves reasoning about an arbitrary number of entities (people, places, and things) that have an unbounded set of complex relationships between them. Representing and reasoning about unbounded sets of entities and relations has generally been considered a strength of predicate logic. However, NLP also requires integrating uncertain evidence from a variety of sources in order to resolve numerous syntactic and semantic ambiguities. Effectively integrating multiple sources of uncertain evidence has generally been considered a strength of Bayesian probabilistic methods and graphical models. Consequently, NLP problems are particularly suited for SRL methods that combine the strengths of first-order predicate logic and probabilistic graphical models. In this article, we review our recent work on using Relational Markov Networks (RMNs) for information extraction, the problem of identifying phrases in natural language text that refer to specific types of entities. We use the expressive power of RMNs to represent and reason about several specific relationships between candidate entities and thereby collectively identify the appropriate set of phrases to extract. We present experiments on learning to extract protein names from biomedical text, which demonstrate the advantage of this approach over existing IE methods.
ML ID: 165