Text Analysis Using Generalized Concepts and Relations

Steven Corman (Inventor), Hasan Davulcu (Inventor)

Research output: Patent

Abstract

Extremist groups use stories to frame contemporary events and persuade audiences to adopt their extremist ideology. Understanding the structure of extremist discourse allows analysts to detect frequently restated themes, creating a reliable outlet to feasibly combat radical ideology. However, a major challenge in automated text analysis systems is that word usage can differ between two texts even though the subject matter is identical. The dissimilarity in word usage results in missed narrative patterns, creating analysis errors. Scientists are now looking to extract relations and concepts from messages at the contextual level such as verb and noun phrases (contextual synonyms) to create a more reliable outlet for comparison and analysis of text samples. Researchers at ASU have developed an algorithm for extracting information from documents by using a clustering technique to categorize them into generalized concepts. The framework works by first analyzing a piece of text for narrative characteristics and, then, processes sentences line-by-line to record and extract related phrases as subject-verb-subject triplets. The algorithm compares the triplets for similar contextual usage throughout the document and merges them based on phrasing criteria, giving accurate and reliable results. Lastly, the algorithm generalizes the concepts created from the criteria and produces a set of synonyms to account for alternative phrases with similar connotation in context. Overall, the technology makes it easy for analysts to examine pieces of text for extremist propaganda by scanning a document and extracting any key phrases and their synonyms to ensure a correct analysis. Potential Applications Data/Text Mining Violence Prediction Analysis Software Defense Strategies Benefits and Advantages Effective - the algorithm smoothly analyzes text without a reference database analyzes text in triplets to search for contextual synonyms, increasing the accuracy of story detection Increases Safety - the holistic text analysis method produces more accurate results, allowing defense strategies towards extremism to be more proactive and put into play ahead of time Lower Cost - reduces the number of trained analysts needed to look over documents for story/non-story characteristics Download Original PDF For more information about the inventor(s) and their research, please see: Dr. Hasan Davulcu's directory webpage Dr. Steven R. Corman's directory webpage
Original languageEnglish (US)
StatePublished - Jun 5 2015

Fingerprint

Error analysis
Scanning
Costs
Violence

Cite this

@misc{f84b226f501442ce861f1539efd68b0d,
title = "Text Analysis Using Generalized Concepts and Relations",
abstract = "Extremist groups use stories to frame contemporary events and persuade audiences to adopt their extremist ideology. Understanding the structure of extremist discourse allows analysts to detect frequently restated themes, creating a reliable outlet to feasibly combat radical ideology. However, a major challenge in automated text analysis systems is that word usage can differ between two texts even though the subject matter is identical. The dissimilarity in word usage results in missed narrative patterns, creating analysis errors. Scientists are now looking to extract relations and concepts from messages at the contextual level such as verb and noun phrases (contextual synonyms) to create a more reliable outlet for comparison and analysis of text samples. Researchers at ASU have developed an algorithm for extracting information from documents by using a clustering technique to categorize them into generalized concepts. The framework works by first analyzing a piece of text for narrative characteristics and, then, processes sentences line-by-line to record and extract related phrases as subject-verb-subject triplets. The algorithm compares the triplets for similar contextual usage throughout the document and merges them based on phrasing criteria, giving accurate and reliable results. Lastly, the algorithm generalizes the concepts created from the criteria and produces a set of synonyms to account for alternative phrases with similar connotation in context. Overall, the technology makes it easy for analysts to examine pieces of text for extremist propaganda by scanning a document and extracting any key phrases and their synonyms to ensure a correct analysis. Potential Applications Data/Text Mining Violence Prediction Analysis Software Defense Strategies Benefits and Advantages Effective - the algorithm smoothly analyzes text without a reference database analyzes text in triplets to search for contextual synonyms, increasing the accuracy of story detection Increases Safety - the holistic text analysis method produces more accurate results, allowing defense strategies towards extremism to be more proactive and put into play ahead of time Lower Cost - reduces the number of trained analysts needed to look over documents for story/non-story characteristics Download Original PDF For more information about the inventor(s) and their research, please see: Dr. Hasan Davulcu's directory webpage Dr. Steven R. Corman's directory webpage",
author = "Steven Corman and Hasan Davulcu",
year = "2015",
month = "6",
day = "5",
language = "English (US)",
type = "Patent",

}

TY - PAT

T1 - Text Analysis Using Generalized Concepts and Relations

AU - Corman, Steven

AU - Davulcu, Hasan

PY - 2015/6/5

Y1 - 2015/6/5

N2 - Extremist groups use stories to frame contemporary events and persuade audiences to adopt their extremist ideology. Understanding the structure of extremist discourse allows analysts to detect frequently restated themes, creating a reliable outlet to feasibly combat radical ideology. However, a major challenge in automated text analysis systems is that word usage can differ between two texts even though the subject matter is identical. The dissimilarity in word usage results in missed narrative patterns, creating analysis errors. Scientists are now looking to extract relations and concepts from messages at the contextual level such as verb and noun phrases (contextual synonyms) to create a more reliable outlet for comparison and analysis of text samples. Researchers at ASU have developed an algorithm for extracting information from documents by using a clustering technique to categorize them into generalized concepts. The framework works by first analyzing a piece of text for narrative characteristics and, then, processes sentences line-by-line to record and extract related phrases as subject-verb-subject triplets. The algorithm compares the triplets for similar contextual usage throughout the document and merges them based on phrasing criteria, giving accurate and reliable results. Lastly, the algorithm generalizes the concepts created from the criteria and produces a set of synonyms to account for alternative phrases with similar connotation in context. Overall, the technology makes it easy for analysts to examine pieces of text for extremist propaganda by scanning a document and extracting any key phrases and their synonyms to ensure a correct analysis. Potential Applications Data/Text Mining Violence Prediction Analysis Software Defense Strategies Benefits and Advantages Effective - the algorithm smoothly analyzes text without a reference database analyzes text in triplets to search for contextual synonyms, increasing the accuracy of story detection Increases Safety - the holistic text analysis method produces more accurate results, allowing defense strategies towards extremism to be more proactive and put into play ahead of time Lower Cost - reduces the number of trained analysts needed to look over documents for story/non-story characteristics Download Original PDF For more information about the inventor(s) and their research, please see: Dr. Hasan Davulcu's directory webpage Dr. Steven R. Corman's directory webpage

AB - Extremist groups use stories to frame contemporary events and persuade audiences to adopt their extremist ideology. Understanding the structure of extremist discourse allows analysts to detect frequently restated themes, creating a reliable outlet to feasibly combat radical ideology. However, a major challenge in automated text analysis systems is that word usage can differ between two texts even though the subject matter is identical. The dissimilarity in word usage results in missed narrative patterns, creating analysis errors. Scientists are now looking to extract relations and concepts from messages at the contextual level such as verb and noun phrases (contextual synonyms) to create a more reliable outlet for comparison and analysis of text samples. Researchers at ASU have developed an algorithm for extracting information from documents by using a clustering technique to categorize them into generalized concepts. The framework works by first analyzing a piece of text for narrative characteristics and, then, processes sentences line-by-line to record and extract related phrases as subject-verb-subject triplets. The algorithm compares the triplets for similar contextual usage throughout the document and merges them based on phrasing criteria, giving accurate and reliable results. Lastly, the algorithm generalizes the concepts created from the criteria and produces a set of synonyms to account for alternative phrases with similar connotation in context. Overall, the technology makes it easy for analysts to examine pieces of text for extremist propaganda by scanning a document and extracting any key phrases and their synonyms to ensure a correct analysis. Potential Applications Data/Text Mining Violence Prediction Analysis Software Defense Strategies Benefits and Advantages Effective - the algorithm smoothly analyzes text without a reference database analyzes text in triplets to search for contextual synonyms, increasing the accuracy of story detection Increases Safety - the holistic text analysis method produces more accurate results, allowing defense strategies towards extremism to be more proactive and put into play ahead of time Lower Cost - reduces the number of trained analysts needed to look over documents for story/non-story characteristics Download Original PDF For more information about the inventor(s) and their research, please see: Dr. Hasan Davulcu's directory webpage Dr. Steven R. Corman's directory webpage

M3 - Patent

ER -