Current Search:   Glinos, Demetrios (x)

View All Items

  • CSV Spreadsheet
(1 - 2 of 2)
SYNTAX-BASED CONCEPT EXTRACTION FOR QUESTION ANSWERING
Title
SYNTAX-BASED CONCEPT EXTRACTION FOR QUESTION ANSWERING.
Creator
Glinos, Demetrios, Gomez, Fernando, University of Central Florida
Abstract / Description
Question answering (QA) stands squarely along the path from document retrieval to text understanding. As an area of research interest, it serves as a proving ground where strategies for document processing, knowledge representation, question analysis, and answer extraction may be evaluated in real world information extraction contexts. The task is to go beyond the representation of text documents as "bags of words" or data blobs that can be scanned for keyword combinations and word...
Show moreQuestion answering (QA) stands squarely along the path from document retrieval to text understanding. As an area of research interest, it serves as a proving ground where strategies for document processing, knowledge representation, question analysis, and answer extraction may be evaluated in real world information extraction contexts. The task is to go beyond the representation of text documents as "bags of words" or data blobs that can be scanned for keyword combinations and word collocations in the manner of internet search engines. Instead, the goal is to recognize and extract the semantic content of the text, and to organize it in a manner that supports reasoning about the concepts represented. The issue presented is how to obtain and query such a structure without either a predefined set of concepts or a predefined set of relationships among concepts. This research investigates a means for acquiring from text documents both the underlying concepts and their interrelationships. Specifically, a syntax-based formalism for representing atomic propositions that are extracted from text documents is presented, together with a method for constructing a network of concept nodes for indexing such logical forms based on the discourse entities they contain. It is shown that meaningful questions can be decomposed into Boolean combinations of question patterns using the same formalism, with free variables representing the desired answers. It is further shown that this formalism can be used for robust question answering using the concept network and WordNet synonym, hypernym, hyponym, and antonym relationships. This formalism was implemented in the Semantic Extractor (SEMEX) research tool and was tested against the factoid questions from the 2005 Text Retrieval Conference (TREC), which operated upon the AQUAINT corpus of newswire documents. After adjusting for the limitations of the tool and the document set, correct answers were found for approximately fifty percent of the questions analyzed, which compares favorably with other question answering systems.
Show less
Date Issued
2006
Identifier
CFE0000985, ucf:46711
Format
Document (PDF)
PURL
http://purl.flvc.org/ucf/fd/CFE0000985
An intelligent editor for natural language processing of unrestricted text
Title
An intelligent editor for natural language processing of unrestricted text.
Creator
Glinos, Demetrios George, Gomez, Fernando, Arts and Sciences
Abstract / Description
University of Central Florida College of Arts and Sciences Thesis; The understanding of natural language by computational methods has been a continuing and elusive problem in artificial intelligence. In recent years there has been a resurgence in natural language processing research. Much of this work has been on empirical or corpus-based methods which use a data-driven approach to train systems on large amounts of real language data. Using corpus-based methods, the performance of part-of...
Show moreUniversity of Central Florida College of Arts and Sciences Thesis; The understanding of natural language by computational methods has been a continuing and elusive problem in artificial intelligence. In recent years there has been a resurgence in natural language processing research. Much of this work has been on empirical or corpus-based methods which use a data-driven approach to train systems on large amounts of real language data. Using corpus-based methods, the performance of part-of-speech (POS) taggers, which assign to the individual words of a sentence their appropriate part of speech category (e.g., noun, verb, preposition), now rivals human performance levels, achieving accuracies exceeding 95%. Such taggers have proved useful as preprocessors for such tasks as parsing, speech synthesis, and information retrieval. Parsing remains, however, a difficult problem, even with the benefit of POS tagging. Moveover, as sentence length increases, there is a corresponding combinatorial explosing of alternative possible parses. Consider the following sentence from a New York Times online article: After Salinas was arrested for murder in 1995 and lawyers for the bank had begun monitoring his accounts, his personal banker in New York quietly advised Salinas' wife to move the money elsewhere, apparently without the consent of the legal department. To facilitate the parsing and other tasks, we would like to decompose this sentence into the following three shorter sentences which, taken together, convey the same meaning as the original: 1. Salinas was arrested for murder in 1995. 2. Lawyers for the bank had begun monitoring his accounts. 3. His personal banker in New York quietly advised Salinas' wife to move the money elsewhere, apparently without the consent of the legal department. This study investigates the development of heuristics for decomposing such long sentences into sets of shorter sentences without affecting the meaning of the original sentences. Without parsing or semantic analysis, heuristic rules were developed based on: (1) the output of a POS tagger (Brill's tagger); (2) the punctuation contained in the input sentences; and (3) the words themselves. The heuristic algorithms were implemented in an intelligent editor program which first augmented the POS tags and assigned tags to punctuation, and then tested the rules against a corpus of 25 New York Times online articles containing approximately 1,200 sentences and over 32,000 words, with good results. Recommendations are made for improving the algorithms and for continuing this line of research.
Show less
Date Issued
1999
Identifier
CFR0008181, ucf:53055
Format
Document (PDF)
PURL
http://purl.flvc.org/ucf/fd/CFR0008181