Causal Extraction from Text for Causal Modeling: A Case Study of Urban Resilience
碩士 === 國立成功大學 === 自然災害減災及管理國際碩士學位學程 === 105 === This paper seeks to develop a process to extract causal relationship from text, for use in causal modeling. Currently there is a gap between available the requirements for causal modeling and information resources. Causal models require causal relation...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | en_US |
Published: |
2017
|
Online Access: | http://ndltd.ncl.edu.tw/handle/7zrz94 |
id |
ndltd-TW-105NCKU5575006 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-105NCKU55750062019-05-15T23:47:01Z http://ndltd.ncl.edu.tw/handle/7zrz94 Causal Extraction from Text for Causal Modeling: A Case Study of Urban Resilience 協助因果建模之因果擷取流程: 以都市韌性文本為例 ColtBender 班德 碩士 國立成功大學 自然災害減災及管理國際碩士學位學程 105 This paper seeks to develop a process to extract causal relationship from text, for use in causal modeling. Currently there is a gap between available the requirements for causal modeling and information resources. Causal models require causal relationship information but causal relationships are difficult to produce with certainty. This usually leads to causal models being dependent on mental databases. Improved access to causal information, through text resources, would advance existing modeling and allow for greater access to causal modeling. Complex systems, like urban resilience, can be better understood through causal modeling. Depending on the relationship type, causal and non-causal relationships can be applied to models such as Directed Graphs, Bayesian Models, and System Dynamics Models. To improve understanding and better utilize relationship extraction, relationship types and their information requirements are defined. Relationship requirements are based on three elements - relation (correlation), direction (dependence), and polarity (change relative to other variables). All three elements are required to define a causal linkage within a causal model. Methods proposed to extract relationships, both causal and non-causal, include human identification, linguistic pattern recognition, and word embedding methods. To support extraction and classification efforts, linguistic patterns were adapted to fit the requirements of causal models. While the process is intended to be independent of a particular domain, the topic of urban resilience was used as a case study for this work. Causal extraction of appropriate natural text was performed, producing explicit and implicit relationships. The resulting causal relationships were translated into causal models of various depth and focus. Tai-lin Huang 黃泰霖 2017 學位論文 ; thesis 138 en_US |
collection |
NDLTD |
language |
en_US |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 國立成功大學 === 自然災害減災及管理國際碩士學位學程 === 105 === This paper seeks to develop a process to extract causal relationship from text, for use in causal modeling. Currently there is a gap between available the requirements for causal modeling and information resources. Causal models require causal relationship information but causal relationships are difficult to produce with certainty. This usually leads to causal models being dependent on mental databases. Improved access to causal information, through text resources, would advance existing modeling and allow for greater access to causal modeling. Complex systems, like urban resilience, can be better understood through causal modeling.
Depending on the relationship type, causal and non-causal relationships can be applied to models such as Directed Graphs, Bayesian Models, and System Dynamics Models. To improve understanding and better utilize relationship extraction, relationship types and their information requirements are defined. Relationship requirements are based on three elements - relation (correlation), direction (dependence), and polarity (change relative to other variables). All three elements are required to define a causal linkage within a causal model. Methods proposed to extract relationships, both causal and non-causal, include human identification, linguistic pattern recognition, and word embedding methods. To support extraction and classification efforts, linguistic patterns were adapted to fit the requirements of causal models. While the process is intended to be independent of a particular domain, the topic of urban resilience was used as a case study for this work. Causal extraction of appropriate natural text was performed, producing explicit and implicit relationships. The resulting causal relationships were translated into causal models of various depth and focus.
|
author2 |
Tai-lin Huang |
author_facet |
Tai-lin Huang ColtBender 班德 |
author |
ColtBender 班德 |
spellingShingle |
ColtBender 班德 Causal Extraction from Text for Causal Modeling: A Case Study of Urban Resilience |
author_sort |
ColtBender |
title |
Causal Extraction from Text for Causal Modeling: A Case Study of Urban Resilience |
title_short |
Causal Extraction from Text for Causal Modeling: A Case Study of Urban Resilience |
title_full |
Causal Extraction from Text for Causal Modeling: A Case Study of Urban Resilience |
title_fullStr |
Causal Extraction from Text for Causal Modeling: A Case Study of Urban Resilience |
title_full_unstemmed |
Causal Extraction from Text for Causal Modeling: A Case Study of Urban Resilience |
title_sort |
causal extraction from text for causal modeling: a case study of urban resilience |
publishDate |
2017 |
url |
http://ndltd.ncl.edu.tw/handle/7zrz94 |
work_keys_str_mv |
AT coltbender causalextractionfromtextforcausalmodelingacasestudyofurbanresilience AT bāndé causalextractionfromtextforcausalmodelingacasestudyofurbanresilience AT coltbender xiézhùyīnguǒjiànmózhīyīnguǒxiéqǔliúchéngyǐdōushìrènxìngwénběnwèilì AT bāndé xiézhùyīnguǒjiànmózhīyīnguǒxiéqǔliúchéngyǐdōushìrènxìngwénběnwèilì |
_version_ |
1719154928513974272 |