An Information%2DExtraction System for Urdu %2D A Resource%2DPoor Language.

Discovering voter preferences in blogs using mixtures of topic models.

OpinionMiner a novel machine learning system for web opinion mining and extraction.

Named Entity Recognition for Improving Retrieval and Translation of Chinese Documents.

TREC 2008 at the University at Buffalo Legal and Blog Track.

Generating hypotheses from the web.

Automatic scoring of short handwritten essays in reading comprehension tests.
comprehension tests
InfoXtract A customizable intermediate level information extraction engine.
Information extraction (IE) systems assist analysts to assimilate information from electronic documents. This paper focuses on IE tasks designed to support information discovery applications. Since information discovery implies examining large volumes of documents drawn from various sources for situations that cannot be anticipated a priori, they require IE systems to have breadth as well as depth. This implies the need for a domain-independent IE system that can easily be customized for specific domains: end users must be given tools to customize the system on their own. It also implies the need for defining new intermediate level IE tasks that are richer than the subject-verb-object (SVO) triples produced by shallow systems, yet not as complex as the domain-specific scenarios defined by the Message Understanding Conference (MUC). This paper describes a robust, scalable IE engine designed for such purposes. It describes new IE tasks such as entity profiles, and concept-based general events which represent realistic goals in terms of what can be accomplished in the near-term as well as providing useful, actionable information. These new tasks also facilitate the correlation of output from an IE engine with existing structured data. Benchmarking results for the core engine and applications utilizing the engine are presented. 1
Improving Knowledge Discovery in Document Collections through Combining Text Retrieval and Link Analysis Techniques.

A Text Mining Model for Hypothesis Generation.

On the Automatic Scoring of Handwritten Essays.
Automating the task of scoring handwritten student essays is a challenging problem of AI. The goal is to assign scores which are comparable to those of human scorers even though both human and machine handwriting recognition do not achieve perfect transcription. The research described is a first attempt based on coupling two AI technologies: optical handwriting recognition and automated essay scoring. The test-bed is that of essays written by children in statewide reading comprehension tests in schools. The process involves several image-level operations: removal of pre-printed matter, segmentation of handwritten text lines and extraction of words. Constraints provided by the reading passage, the question and the answer rubric help recognize handwritten words. The method of essay scoring is based on using a vector space model and a machine learning approach to learn scoring parameters from a set of human-scored samples. System performance is compared to scoring done by human raters. Testing on test-bed of 96 handwritten answers indicate that system performance is comparable to that of automatic scoring based on perfect manual transcription. 1
Use of ranked cross document evidence trails for hypothesis generation.

Mining Concept Associations for Knowledge Discovery Through Concept Chain Queries.



Automatically Extracting Nominal Mentions of Events with a Bootstrapped Probabilistic Classifier.
Most approaches to event extraction focus on mentions anchored in verbs. However, many mentions of events surface as noun phrases. Detecting them can increase the recall of event extraction and provide the foundation for detecting relations between events. This paper describes a weaklysupervised method for detecting nominal event mentions that combines techniques from word sense disambiguation (WSD) and lexical acquisition to create a classifier that labels noun phrases as denoting events or non-events. The classifier uses bootstrapped probabilistic generative models of the contexts of events and non-events. The contexts are the lexically-anchored semantic dependency relations that the NPs appear in. Our method dramatically improves with bootstrapping, and comfortably outperforms lexical lookup methods which are based on very much larger handcrafted resources. 1
Automated Scoring of Handwritten Essays Based on Latent Semantic Analysis.
Abstract. Handwritten essays are widely used in educational assessments, particularly in classroom instruction. This paper concerns the design of an automated system for performing the task of taking as input scanned images of handwritten student essays in reading comprehension tests and to produce as output scores for the answers which are analogous to those provided by human scorers. The system is based on integrating the two technologies of optical handwriting recognition (OHR) and automated essay scoring (AES). The OHR system performs several pre-processing steps such as forms removal, rule-line removal and segmentation of text lines and words. The final recognition step, which is tuned to the task of reading comprehension evaluation in a primary education setting, is performed using a lexicon derived from the passage to be read. The AES system is based on the approach of latent semantic analysis where a set of human-scored answers are used to determine scoring system parameters using a machine learning approach. System performance is compared to scoring done by human raters. Testing on a small set of handwritten answers indicate that system performance is comparable to that of automatic scoring based on manual transcription. 1
Knowledge Discovery across Documents through Concept Chain Queries.

Unapparent information revelation a concept chain graph approach.

Minimal document set retrieval.
The detection of new information in a document stream is an important component of many potential applications. In this work, a new novelty detection approach based on the identification of sentence level information patterns is proposed. First, the information- pattern concept for novelty detection is presented with the emphasis on new information patterns for general topics (queries) that cannot be simply turned into specific questions whose answers are specific named entities (NEs). Then we elaborate a thorough analysis of sentence level information patterns on data from the TREC novelty tracks, including sentence lengths, named entities, sentence level opinion patterns. This analysis provides guidelines in applying those patterns in novelty detection particularly for the general topics. Finally, a unified pattern-based approach is presented to novelty detection for both general and specific topics. The new method for dealing with general topics will be the focus. Experimental results show that the proposed approach significantly improves the performance of novelty detection for general topics as well as the overall performance for all topics from the 2002-2004 TREC novelty tracks.
Weakly Supervised Learning for Cross%2Ddocument Person Name Disambiguation Supported by Information Extraction.

Document Representation for One%2DClass SVM.
Abstract. Motivated by applying Text Categorization to sorting Web search results, this paper describes an extensive experimental study of the impact of bagof-words document representations on the performance of five major classifiers – Naïve Bayes, SVM, Voted Perceptron, kNN and C4.5. The texts represent short Web-page descriptions from the dmoz Open Directory Web-page ontology. Different transformations of input data: stemming, normalization, logtf and idf, together with dimensionality reduction, are found to have a statistically significant improving or degrading effect on classification performance measured by classical metrics – accuracy, precision, recall, F1 and F2. The emphasis of the study is not on determining the best document representation which corresponds to each classifier, but rather on describing the effects of every individual transformation on classification, together with their mutual relationships. 1
Incorporating prior knowledge with weighted margin support vector machines.

UB at TREC 13 Genomics Track.
Abstract. This paper describes the experiments of the State University of New York at Buffalo in TREC 13. We participated in the Genomics track and submitted official runs to the Adhoc retrieval task. Our approach uses a language model IR system developed in house. We also present unofficial results for the triage sub-task of categorization task. 1
Orthographic case restoration using supervised learning without manual annotation.

Feature selection for text categorization on imbalanced data.
A number of feature selection metrics have been explored in text categorization, among which information gain (IG), chi-square (CHI), correlation coefficient (CC) and odds ratios (OR) are considered most effective. CC and OR are one-sided metrics while IG and CHI are two-sided. Feature selection using one-sided metrics selects the features most indicative of membership only, while feature selection using two-sided metrics implicitly combines the features most indicative of membership (e.g. positive features) and nonmembership (e.g. negative features) by ignoring the signs of features. The former never consider the negative features, which are quite valuable, while the latter cannot ensure the optimal combination of the two kinds of features especially on imbalanced data. In this work, we investigate the usefulness of explicit control of that combination within a proposed feature selection framework. Using multinomial naïve Bayes and regularized logistic regression as classifiers, our experiments show both great potential and actual merits of explicitly combining positive and negative features in a nearly optimal fashion according to the imbalanced data. 1.
Feature selection for text categorization on imbalanced data.
A number of feature selection metrics have been explored in text categorization, among which information gain (IG), chi-square (CHI), correlation coefficient (CC) and odds ratios (OR) are considered most effective. CC and OR are one-sided metrics while IG and CHI are two-sided. Feature selection using one-sided metrics selects the features most indicative of membership only, while feature selection using two-sided metrics implicitly combines the features most indicative of membership (e.g. positive features) and nonmembership (e.g. negative features) by ignoring the signs of features. The former never consider the negative features, which are quite valuable, while the latter cannot ensure the optimal combination of the two kinds of features especially on imbalanced data. In this work, we investigate the usefulness of explicit control of that combination within a proposed feature selection framework. Using multinomial naïve Bayes and regularized logistic regression as classifiers, our experiments show both great potential and actual merits of explicitly combining positive and negative features in a nearly optimal fashion according to the imbalanced data. 1.
A Bootstrapping Approach to Named Entity Classification Using Successive Learners.
approach to named entity (NE)  classification. This approach only requires  a few common noun/pronoun seeds that  correspond to the concept for the target  NE type, e.g. he/she/man/woman for  PERSON NE. The entire bootstrapping  procedure is implemented as training two  successive learners: (i) a decision list is  used to learn the parsing-based high  precision NE rules; (ii) a Hidden Markov  Model is then trained to learn string  sequence-based NE patterns. The second  learner uses the training corpus  automatically tagged by the first learner.
An Expert Lexicon Approach to Identifying English Phrasal Verbs.
Phrasal Verbs are an important feature  of the English language. Properly  identifying them provides the basis for  an English parser to decode the related  structures. Phrasal verbs have been a  challenge to Natural Language  Processing (NLP) because they sit at  the borderline between lexicon and  syntax. Traditional NLP frameworks  that separate the lexicon module from  the parser make it difficult to handle  this problem properly. This paper  presents a finite state approach that  integrates a phrasal verb expert lexicon  between shallow parsing and deep  parsing to handle morpho-syntactic  interaction. With precision/recall  combined performance benchmarked  consistently at 95.8%-97.5%, the  Phrasal Verb identification problem  has basically been solved with the  presented method.
Exploiting syntactic structure of queries in a language modeling approach to IR.

Orthographic Case Restoration Using Supervised Learning Without Manual Annotation.

Bootstrapping for Named Entity Tagging Using Concept%2Dbased Seeds.

A Case Restoration Approach to Named Entity Tagging in Degraded Documents.
This paper describes a novel approach to named entity (NE) tagging on degraded documents. NE tagging is the process of identifying salient text strings in unstructured text, corresponding to names of people, places, organizations, times/dates, etc. Although NE tagging is typically part of a larger information extraction process, it has other applications, such as improving search in an information retrieval system, and post-processing the results of an OCR system. We focus on degraded documents, i.e. case insensitive documents that lack orthographic information. Examples include output of speech recognition systems, as well as e-mail. The traditional approach involves retraining an NE tagger on degraded text, a cumbersome operation. This paper describes an approach whereby text is first “restored ” to its implicit case sensitive form, and subsequently processed by the original NE tagger. Results show that this new approach leads to far less precision loss in NE tagging of degraded documents. 1.
A Feature Selection Framework for Text Filtering.

New %ED%2DSupport Vector Machines and their Sequential Minimal Optimization.

Incorporating query term dependencies in language models for document retrieval.
Introduction  Recent advances in Information Retrieval are based on using Statistical Language Models (SLM) for representing documents and evaluating their relevance to user queries [6, 3, 4]. Language Modeling (LM) has been explored in many natural language tasks including machine translation and speech recognition [1]. In LM approach to document retrieval, each document, D, is viewed to have its own language model, MD . Given a query, Q, documents are ranked based on the probability, P (Q|MD ), of their language model generating the query. While the LM approach to information retrieval has been motivated from di#erent perspectives [3, 4], most experiments have used smoothed unigram language models that assume term independence for estimating document language models.  N-gram, specifically, bigram language models that capture context provided by the previous word(s) perform better than unigram models [7]. Biterm language models [8] that ignore the word order constraint in bigram langu
UB at TREC 12 HARD and Genomics Tracks.

Knowledge Supervised Perceptual Grouping Based Qualitative Building Detection from Monocular Aerial Images.

Location Normalization for Information Extraction.
Ambiguity is very high for location names. For  example, there are 23 cities named `Buffalo&#039; in  the U.S. Country names such as `Canada&#039;,  `Brazil&#039; and `China&#039; are also city names in the  USA. Almost every city has a Main Street or  Broadway. Such ambiguity needs to be handled  before we can refer to location names for  visualization of related extracted events. This  paper presents a hybrid approach for location  normalization which combines (i) lexical  grammar driven by local context constraints, (ii)  graph search for maximum spanning tree and  (iii) integration of semi-automatically derived  default senses. The focus is on resolving  ambiguities for the following types of location  names: island, town, city, province, and country.
Biterm language models for document retrieval.
Introduction  Statistical Language Models(LM) have been used in many natural language processing tasks including speech recognition and machine translation [5, 2]. Recently language models have been explored as a framework for information retrieval [9, 4, 7, 1, 6]. The basic idea is to view each document to have its own language model and model querying as a generative process. Documents are ranked based on the probability of their language model generating the given query. Since documents are fixed entities in information retrieval, language models for documents su#er from sparse data problem. Smoothed unigram models have been used to demonstrate better performance of language models against vector space or probabilistic retrieval models for document retrieval.  Song and Croft [10] proposed a general language model that combined bigram language models with Good-Turing estimate and corpus-based smoothing of unigram probabilities. Improved performance was observed with combined bigram l
UB at TREC 11 Batch and Adaptive Filtering.

A method for measuring the complexity of image databases.

Subspace morphing theory for appearance based object identification.

Keyblock an approach for content%2Dbased image retrieval.

A Question Answering System Supported by Information Extraction.

A Question Answering System Supported by Information Extraction.

A Hybrid Approach for Named Entity and Sub%2DType Tagging.
This paper presents a hybrid approach for named entity (NE) tagging which combines Maximum Entropy Model (MaxEnt), Hidden Markov Model (HMM) and handcrafted grammatical rules. Each has innate strengths and weaknesses; the combination results in a very high precision tagger. MaxEnt includes external gazetteers in the system. Sub-category generation is also discussed.
A Model for Multimodal Information Retrieval.
Finding useful information from large multimodal document  collections such as the WWW without encountering numerous false positives poses a challenge to multimodal information retrieval systems (MMIR). A general model for multimodal information retrieval is proposed by which a user&#039;s information need is expressed through composite, multimodal queries, and the most appropriate weighted combination of indexing techniques is determined by a machine learning approach in order to best satisfy the information need. The focus is on improving precision and recall in a MMIR system by optimally combining text and image similarity. Experiments are presented which demonstrate the utility of individual indexing systems in improving overall average precision.  Keywords: Multimodal Information Retrieval, Contentbased Retrieval, Image Retrieval, Multimodal Query, Machine Learning  1. 
Applications of Image Understanding in Semantics%2DOriented Multimedia Information Retrieval.

Show%26Tell A Semi%2DAutomated Image Annotation System.

Intelligent Indexing and Semantic Retrieval of Multimodal Documents.
Finding useful information from large multimodal document collections such as the WWW without encountering numerous false positives poses a challenge to multimedia information retrieval systems (MMIR). This research addresses the problem of finding pictures. The fact that images do not appear in isolation, but rather with accompanying, collateral text is exploited. Taken independently, existing techniques for picture retrieval using (i) text-based and (ii) image-based methods have several limitations. This research presents a general model for multimodal information retrieval that addresses the follow- ing issues: (i) users&#039; information need, (ii) expressing information need through composite, multimodal queries, and (iii) determining the most appropriate weighted combination of indexing techniques in order to best satisfy information need. A machine learning approach is proposed for the latter. The focus is on improving precision and recall in a MMIR system by optimally combining text and image similarity. Experiments are presented which demonstrate the utility of individual indexing systems in improving overall average precision.
Face Detection and Its Applications in Intelligent and Focused Image Retrieval.
This paper presents a face detection technique and its applications in image retrieval. Even though this face detection method has relatively high false positives and low detection rate (as compared with the dedicated face detection systems in the literature of image understanding) , due to its simple and fast nature, it has been shown that this system may be well applied in image retrieval in certain focused application domains. Two application examples are given: one combining face detection with indexed collateral text for image retrieval regarding human beings, and the other combining face detection with conventional similarity matching techniques for image retrieval with similar background. Experimental results show that our proposed approaches have significantly improved the image retrieval precisions than the existing search engines in these focused application domains.  Key Words: Face Detection, Image Retrieval, Collateral Text, Similarity Matching, Focused Domain.  1 
Spatial Color Histograms for Content%2DBased Image Retrieval.
Color histogram is an important technique for color image database indexing and retrieving. In this paper, traditional color histogram is modified to capture spatial layout information of each color and three types spatial color histograms are introduced: annular, angular and hybrid color histograms. Experiments show that with a proper trade-off between the granularity in color dimensions and that in spatial dimensions, the performance of these histograms outperforms the traditional color histogram and some existing histogram refinement such as color coherent vector.  Keywords: Color Histogram, Spatial Color Histogram, Content-Based Image Retrieval, Annular Color Histogram, Angular Color Histogram, Hybrid Color Histogram.  1 
Information Extraction Supported Question Answering.
This paper discusses the use of our information extraction (IE) system, Textract, in the questionanswering (QA) track of the recently held TREC-8 tests. One of our major objectives is to examine how IE can help IR (Information Retrieval) in applications like QA. Our study shows: (i) IE can provide solid support for QA; (ii) low-level IE like Named Entity tagging is often a necessary component in handling most types of questions; (iii) a robust natural language shallow parser provides a structural basis for handling questions; (iv) high-level domain independent IE, i.e. extraction of multiple relationships and general events, is expected to bring about a breakthrough in QA. 1 Introduction  Natural language QA (Question Answering) is an ideal test bed for demonstrating the power of IE (Information Extraction). In our vision, there is a natural co-operation between IE and IR (Information Retrieval). An important question then is, what type of IE can support IR in QA and how well does it s...
TREC%2D8 Experiments at SUNY Buffalo.
this document ranking.
Exploiting Multimodal Context in Image Retrieval.

Indexing and Retrieval%2C SIGIR%2799 Workshop Summary.

Finding Pictures in Context.

Multimedia Indexing and Retrieval%2C Summary Report.

Using Speech Input for Image Interpretation%2C Annotation%2C and Retrieval.

Large Vocabulary Recognition of On%2DLine Handwritten Cursive Words.
A critical feature of any computer system is its interface with the user. This has led to the development of user interface technologies such as mouse, touchscreen and penbased input devices. Since handwriting is one of the most familiar communication media, pen-based interfaces combined with automatic handwriting recognition offers a very easy and natural input method. Pen-based interfaces are also essential in mobile computing because they are scalable. Recent advances in pen-based hardware and wireless communication have been influential factors in the renewed interest in on-line recognition systems. On-line handwriting recognition is fundamentally a pattern classification task; the objective is to take an input pattern, the handwritten signal collected on-line via a digitizing device, and classify it as one of a pre-specified set of words (i.e., the system&#039;s lexicon). Because exact recognition is very difficult, a lexicon is used to constrain the recognition output to a known vocab...
Generalizing edit distance to incorporate domain information Handwritten text recognition as a case study.
In this paper the Damerau-Levenshtein string difference metric is generalized in two ways to more accurately compensate for the types of errors that are present in the script recognition domain. First, the basic dynamic programming method for computing such a measure is extended to allow for merges, splits and two-letter substitutions. Second, edit operations are refined into categories according to the effect they have on the visual &#034;appearance&#034; of words. A set of recognizer-independent constraints is developed to reflect the severity of the information lost due to each operation. These constraints are solved to assign specific costs to the operations. Experimental results on 2,335 corrupted strings and a lexicon of 21,299 words show higher correcting rates than with the original form.  Keywords: string distance, string matching, spelling error correction, word recognition and correction, text editing, script recognition and post-processing 1 INTRODUCTION  Since the goal of text recog...
Automatic indexing and content%2Dbased retrieval of captioned photographs.

Use of Multimedia Input in Automated Image Annotation and Content%2DBased Retrieval.
This research explores the interaction of linguistic and photographic information in an integrated text/image database. By utilizing linguistic descriptions of a picture (speech and text input) coordinated with pointing references to the picture, we extract information useful in two aspects: image interpretation and image retrieval. In the image interpretation phase, objects and regions mentioned in the text are identified; the annotated image is stored in a database for future use. We incorporate techniques from our previous research on photo understanding using accompanying text: a system, PICTION, which identifies human faces in a newspaper photograph based on the caption. In the image retrieval phase, images matching natural language queries are presented to a user in a ranked order. This phase combines the output of (i) the image interpretation/annotation phase, (ii) statistical text retrieval methods, and (iii) image retrieval methods (e.g., colour indexing). The system allows bo...
Combining text and image information in content%2Dbased retrieval.

Control Structures for Incorporating Picture%2DSpecific Context in Image Interpretation.
This paper describes an efficient control mechanism  for incorporating picture-specific context  in the task of image interpretation. Although  other knowledge-based vision systems use general  domain context in reducing the computational  burden of image interpretation, to our  knowledge, this is the first effort in exploring  picture-specific collateral information. We assume  that constraints on the picture are generated  from a natural language understanding  module which processes descriptive text accompanying  the pictures. We have developed a unified  framework for exploiting these constraints  both in the object location and identification  (labeling) stage. In particular, we describe a  technique for incorporating constrained search  in context-based vision. Finally, we demonstrate  the effectiveness of this approach in PICTION,  a system that uses captions to label human  faces in newspaper photographs.  1 Introduction  To solve the inherently under-constrained task of image ...
Use of Multimedia Input in Automated Image Annotation and Content%2DBased Retrieval.
This research explores the interaction of linguistic and photographic information in an integrated text/image database. By utilizing linguistic descriptions of a picture (speech and text input) coordinated with pointing references to the picture, we extract information useful in two aspects: image interpretation and image retrieval. In the image interpretation phase, objects and regions mentioned in the text are identified; the annotated image is stored in a database for future use. We incorporate techniques from our previous research on photo understanding using accompanying text: a system, PICTION, which identifies human faces in a newspaper photograph based on the caption. In the image retrieval phase, images matching natural language queries are presented to a user in a ranked order. This phase combines the output of (i) the image interpretation/annotation phase, (ii) statistical text retrieval methods, and (iii) image retrieval methods (e.g., colour indexing). The system allows bo...


Use of Lexical and Syntactic Techniques in Recognizing Handwritten Text.
The output of handwritten word recognizers (Wit) tends to be very noisy due to various factors. In order to compensate for this behaviour, several choices of the WR must be ini-tially considered. In the case of handwritten sentence/phrase recognition, linguistic constraints may be applied in order to improve the results of the Wit. This paper discusses two statistical methods of applying linguistic constraints to the output of an Wit on input consisting of sentences/phrases. The first is based on collocations and can be used to prOmote lower ranked word choices or to propose new words. The second is a Markov model of syntax and is based on syn-tactic categories (tags) associated with words. In each case, we show the improvement in the word recognition rate as a result of applying these constraints. 1.
Computational Models for Integrating Linguistic and Visual Information A Survey.
This paper surveys research in developing computational models for integrating linguistic and visual information. It begins with a discussion of systems which have been actually implemented and continues with computationally motivated theories of human cognition. Since existing research spans several disciplines (e.g., natural language understanding, computer vision, knowledge representation), as well as several application areas, an important contribution of this paper is to categorize existing research based on inputs and objectives. Finally, some key issues related to integrating information from two such diverse sources are outlined and related to existing research. Throughout, the key issue addressed is the correspondence problem, namely how to associate visual events with words and vice versa.  1 Introduction  Much has been said about the necessity of linking language and vision in order for a system to exhibit intelligent behaviour [Win73, Wal81]. A complete natural-language und...
Use of Captions and Other Collateral Text in Understanding Photographs.

Incorporating Syntactic Constraints in Recognizing Handwritten Sentences.
The output of handwritten word recognizers (HWR) tends to be very noisy due to various factors. In order to compensate for this behaviour, several choices of the HWR must be initially considered. In the case of handwritten sentence/phrase recognition, linguistic constraints may be applied in order to improve the results of the HWR. This paper discusses two statistical methods of applying syntactic constraints to the output of an HWR on input consisting of sentences/phrases. Both methods are based on syntactic categories (tags) associated with words. The first is a purely statistical method, the second is a hybrid method which combines higher-level syntactic information (hypertags) with statistical information regarding transitions between hypertags.  1 Introduction  Recently, there has been a great deal of interest in handwriting recognition systems. Such systems are referred to as on-line or off-line systems. In the former, a light pen and a special tablet are used by the writer to en...
Piction A System That Uses Captions to Label Human Faces in Newspaper Photographs.
This paper describes an efficient control mechanism  for incorporating picture-specific context  in the task of image interpretation. Although  other knowledge-based vision systems use general  domain context in reducing the computational  burden of image interpretation, to our  knowledge, this is the first effort in exploring  picture-specific collateral information. We assume  that constraints on the picture are generated  from a natural language understanding  module which processes descriptive text accompanying  the pictures. We have developed a unified  framework for exploiting these constraints  both in the object location and identification  (labeling) stage. In particular, we describe a  technique for incorporating constrained search  in context-based vision. Finally, we demonstrate  the effectiveness of this approach in PICTION,  a system that uses captions to label human  faces in newspaper photographs.  1 Introduction  To solve the inherently under-constrained task of image ...
Newspaper Image Understanding.

