Vasileios Hatzivassiloglou

Predicting the Semantic Orientation of Adjectives

We identify and validate from a large corpus constraints from conjunctions on the positive or negative semantic orientation of the conjoined adjectives. A log-linear regression model uses these constraints to predict whether conjoined adjectives are of same or different orientations, achieving 82% accuracy in this task when each conjunction is considered independently. Combining the constraints across many adjectives, a clustering algorithm separates the adjectives into groups of different orientations, and finally, adjectives are labeled positive or negative. Evaluations on real data and simulation experiments indicate high levels of performance: classification precision is more than 90% for adjectives that occur in a modest number of conjunctions in the corpus.

Towards answering opinion questions: separating facts from opinions and identifying the polarity of opinion sentences

Opinion question answering is a challenging task for natural language processing. In this paper, we discuss a necessary component for an opinion question answering system: separating opinions from fact, at both the document and sentence level. We present a Bayesian classifier for discriminating between documents with a preponderance of opinions such as editorials from regular news stories, and describe three unsupervised, statistical techniques for the significantly harder task of detecting opinions at the sentence level. We also present a first model for classifying opinion sentences as positive or negative in terms of the main perspective being expressed in the opinion. Results from a large collection of news stories and a human evaluation of 400 sentences are reported, indicating that we achieve very high performance in document classification (upwards of 97% precision and recall), and respectable performance in detecting opinions and classifying them at the sentence level as positive, negative, or neutral (up to 91% accuracy).

Effects of adjective orientation and gradability on sentence subjectivity

Subjectivity is a pragmatic, sentence-level feature that has important implications for text processing applications such as information extraction and information retrieval. We study the effects of dynamic adjectives, semantically oriented adjectives, and gradable adjectives on a simple subjectivity classifier, and establish that they are strong predictors of subjectivity. A novel trainable method that statistically combines two indicators of gradability is presented and evaluated, complementing existing automatic techniques for assigning orientation labels.

GeneWays: a system for extracting, analyzing, visualizing, and integrating molecular pathway data

The immense growth in the volume of research literature and experimental data in the field of molecular biology calls for efficient automatic methods to capture and store information. In recent years, several groups have worked on specific problems in this area, such as automated selection of articles pertinent to molecular biology, or automated extraction of information using natural-language processing, information visualization, and generation of specialized knowledge bases for molecular biology. GeneWays is an integrated system that combines several such subtasks. It analyzes interactions between molecular substances, drawing on multiple sources of information to infer a consensus view of molecular networks. GeneWays is designed as an open platform, allowing researchers to query, review, and critique stored information.

Towards multidocument summarization by reformulation: progress and prospects

By synthesizing information common to retrieved documents, multi-document summarization can help users of information retrieval systems to find relevant documents with a minimal amount of reading. We are developing a multidocument summarization system to automatically generate a concise summary by identifying and synthesizing similarities across a set of related documents. Our approach is unique in its integration of machine learning and statistical techniques to identify similar paragraphs, intersection of similar phrases within paragraphs, and language generation to reformulate the wording of the summary. Our evaluation of system components shows that learning over multiple extracted linguistic features is more effective than information retrieval approaches at identifying similar text units for summarization and that it is possible to generate a fluent summary that conveys similarities among documents even when full semantic interpretations of the input text are not available.

Categorizing web queries according to geographical locality

Web pages (and resources, in general) can be characterized according to their geographical locality. For example, a web page with general information about wildflowers could be considered a global page, likely to be of interest to a geographically broad audience. In contrast, a web page with listings on houses for sale in a specific city could be regarded as a local page, likely to be of interest only to an audience in a relatively narrow region. Similarly, some search engine queries (implicitly) target global pages, while other queries are after local pages. For example, the best results for query [wildflowers] are probably global pages about wildflowers such as the one discussed above. However, local pages that are relevant to, say, San Francisco are likely to be good matches for a query [houses for sale] that was issued by a San Francisco resident or by somebody moving to that city. Unfortunately, search engines do not analyze the geographical locality of queries and users, and hence often produce sub-optimal results. Thus query [wildflowers] might return pages that discuss wildflowers in specific U.S. states (and not general information about wildflowers), while query [houses for sale] might return pages with real estate listings for locations other than that of interest to the person who issued the query. Deciding whether an unseen query should produce mostly local or global pages---without placing this burden on the search engine users---is an important and challenging problem, because queries are often ambiguous or underspecify the information they are after. In this paper, we address this problem by first defining how to categorize queries according to their (often implicit) geographical locality. We then introduce several alternatives for automatically and efficiently categorizing queries in our scheme, using a variety of state-of-the-art machine learning tools. We report a thorough evaluation of our classifiers using a large sample of queries from a real web search engine, and conclude by discussing how our query categorization approach can help improve query result quality.

Two-Level, Many-Paths Generation

Large-scale natural language generation requires the integration of vast amounts of knowledge: lexical, grammatical, and conceptual. A robust generator must be able to operate well even when pieces of knowledge are missing. It must also be robust against incomplete or inaccurate inputs. To attack these problems, we have built a hybrid generator, in which gaps in symbolic knowledge are filled by statistical methods. We describe algorithms and show experimental results. We also discuss how the hybrid generation model can be used to simplify current generators and enhance their portability, even when perfect knowledge is in principle obtainable.

Simfinder: A flexible clustering tool for summarization

Abstract : We present a statistical similarity measuring and clustering tool, SIMFINDER, that organizes small pieces of text from one or multiple documents into tight clusters. By placing highly related text units in the same cluster, SIMFINDER enables a subsequent content selection/generation component to reduce each cluster to a single sentence, either by extraction or by reformulation. We report on improvements in the similarity and clustering components of SIMFINDER, including a quantitative evaluation, and establish the generality of the approach by interfacing SIMFINDER to two very different summarization systems.

Event-Based Extractive Summarization

Most approaches to extractive summarization define a set of features upon which selection of sentences is based, using algorithms independent of the features themselves. We propose a new set of features based on low-level, atomic events that describe relationships between important actors in a document or set of documents. We investigate the effect this new feature has on extractive summarization, compared with a baseline feature set consisting of the words in the input documents, and with state-of-the-art summarization systems. Our experimental results indicate that not only the event-based features offer an improvement in summary quality over words as features, but that this effect is more pronounced for more sophisticated summarization methods that avoid redundancy in the output.

An investigation of linguistic features and clustering algorithms for topical document clustering

We investigate four hierarchical clustering methods (single-link, complete-link, groupwise-average, and single-pass) and two linguistically motivated text features (noun phrase heads and proper names) in the context of document clustering. A statistical model for combining similarity information from multiple sources is described and applied to DARPAs Topic Detection and Tracking phase 2 (TDT2) data. This model, based on log-linear regression, alleviates the need for extensive search in order to determine optimal weights for combining input features. Through an extensive series of experiments with more than 40,000 documents from multiple news sources and modalities, we establish that both the choice of clustering algorithm and the introduction of the additional features have an impact on clustering performance. We apply our optimal combination of features to the TDT2 test data, obtaining partitions of the documents that compare favorably with the results obtained by participants in the official TDT2 competition.