Exploring Location Indicators for Geographic Information Retrieval - PowerPoint PPT Presentation

Exploring Location Indicators for Geographic Information Retrieval Johannes Leveling and Sven Hartrumpf Intelligent Information and Communication Systems (IICS) University of Hagen (FernUniversität in Hagen) 58084 Hagen, Germany firstname.lastname@fernuni-hagen.de CLEF 2007 Workshop, Budapest, Hungary

Exploring Location Indicators for Outline Geographic Information Retrieval Johannes Leveling and 1 Introduction Sven Hartrumpf Location Indicators 2 Introduction Location Indicators 3 Location Indicator Normalization Location Indicator Normalization Semantic Analysis for GIR 4 Semantic Analysis for GIR GeoCLEF 2007 Experiments 5 GeoCLEF 2007 Experiments Conclusion Conclusion and Outlook 6 and Outlook References Johannes Leveling and Sven Hartrumpf Exploring Location Indicators for Geographic Information Retrieval 2 / 17

Exploring Location Indicators for Introduction Geographic Information Retrieval Johannes Leveling and Sven • Traditional information retrieval (IR): Hartrumpf stemming is applied to all words in a text Introduction • Geographical information retrieval (GIR): Location use named entity recognition and classification; Indicators Location avoid stemming location names (typically, proper nouns Indicator Normalization only); employ geographic knowledge Semantic • GIRSA (Geographic Information Retrieval by Semantic Analysis for GIR Annotation): GeoCLEF aims at a broader GIR approach not solely based on 2007 Experiments location names, but on location indicators Conclusion and Outlook References Johannes Leveling and Sven Hartrumpf Exploring Location Indicators for Geographic Information Retrieval 3 / 17

Exploring Location Indicators for Location Indicators Geographic Information Retrieval Johannes Leveling and Definition Sven Hartrumpf Location indicators are text segments from which the geographic scope of a document can be inferred. Introduction Location Indicators Location Indicator Normalization Semantic Analysis for GIR GeoCLEF 2007 Experiments Conclusion and Outlook References Johannes Leveling and Sven Hartrumpf Exploring Location Indicators for Geographic Information Retrieval 4 / 17

Exploring Location Indicators for Location Indicators Geographic Information Retrieval Johannes Leveling and Definition Sven Hartrumpf Location indicators are text segments from which the geographic scope of a document can be inferred. Introduction Location Indicators • Adjectives corresponding to a location. Location Indicator Example: Normalization tunesisch → Tunesien Semantic Analysis for ( Tunisian → Tunisia ) GIR GeoCLEF 2007 Experiments Conclusion and Outlook References Johannes Leveling and Sven Hartrumpf Exploring Location Indicators for Geographic Information Retrieval 4 / 17

Exploring Location Indicators for Location Indicators Geographic Information Retrieval Johannes Leveling and Definition Sven Hartrumpf Location indicators are text segments from which the geographic scope of a document can be inferred. Introduction Location Indicators • Demonyms, e.g. the name for inhabitants originating Location Indicator from a location. Normalization Semantic Example: Analysis for GIR Franzose, Französin → Frankreich GeoCLEF ( Frenchman, Frenchwoman → France ) 2007 Experiments Conclusion and Outlook References Johannes Leveling and Sven Hartrumpf Exploring Location Indicators for Geographic Information Retrieval 4 / 17

Exploring Location Indicators for Location Indicators Geographic Information Retrieval Johannes Leveling and Definition Sven Hartrumpf Location indicators are text segments from which the geographic scope of a document can be inferred. Introduction Location Indicators • Codes for a location name. Location Indicator Normalization Example: Semantic HU21 → Tolna County, Hungary (FIPS region code) Analysis for GIR GeoCLEF 2007 Experiments Conclusion and Outlook References Johannes Leveling and Sven Hartrumpf Exploring Location Indicators for Geographic Information Retrieval 4 / 17

Exploring Location Indicators for Location Indicators Geographic Information Retrieval Johannes Leveling and Definition Sven Hartrumpf Location indicators are text segments from which the geographic scope of a document can be inferred. Introduction Location Indicators • Abbreviations and acronyms for a location name, Location Indicator including adjectives. Normalization Semantic Example: Analysis for GIR franz. → französisch → Frankreich GeoCLEF ( French → France ) 2007 Experiments TX → Texas Conclusion and Outlook References Johannes Leveling and Sven Hartrumpf Exploring Location Indicators for Geographic Information Retrieval 4 / 17

Exploring Location Indicators for Location Indicators Geographic Information Retrieval Johannes Leveling and Definition Sven Hartrumpf Location indicators are text segments from which the geographic scope of a document can be inferred. Introduction Location Indicators • Orthographic variants, exonyms, historic names. Location Indicator Normalization Example: Semantic Lower Saxony → Niedersachsen Analysis for GIR GeoCLEF 2007 Experiments Conclusion and Outlook References Johannes Leveling and Sven Hartrumpf Exploring Location Indicators for Geographic Information Retrieval 4 / 17

Exploring Location Indicators for Location Indicators Geographic Information Retrieval Johannes Leveling and Definition Sven Hartrumpf Location indicators are text segments from which the geographic scope of a document can be inferred. Introduction Location Indicators • Unique entities associated with a geographic Location Indicator location, e.g. headquarters of an organization, Normalization persons, buildings. Semantic Analysis for GIR Example: GeoCLEF 2007 Eiffel Tower → Paris Experiments Moliére → France (?) Conclusion and Outlook VW → Wolfsburg (?) References Johannes Leveling and Sven Hartrumpf Exploring Location Indicators for Geographic Information Retrieval 4 / 17

Exploring Location Indicators for Location Indicators Geographic Information Retrieval Johannes Leveling and Definition Sven Hartrumpf Location indicators are text segments from which the geographic scope of a document can be inferred. Introduction Location Indicators • The location names itself (full names and short Location Indicator forms). Normalization Semantic Example: Analysis for GIR Republik Korea → Südkorea GeoCLEF ( Republic of Korea → South Korea ) 2007 Experiments Conclusion and Outlook References Johannes Leveling and Sven Hartrumpf Exploring Location Indicators for Geographic Information Retrieval 4 / 17

Exploring Location Indicators for Location Indicator Normalization Geographic Information Retrieval Johannes Leveling and Normalization on surface (character), morphologic, Sven Hartrumpf syntactic, semantic, and lexical level. Introduction Location Character level Indicators Location • Diacritical marks replaced with non-accented Indicator Normalization characters Semantic • Orthographic variants normalized by selecting a Analysis for GIR representative GeoCLEF 2007 Example: Experiments Conclusion Québec → Quebec and Outlook References Johannes Leveling and Sven Hartrumpf Exploring Location Indicators for Geographic Information Retrieval 5 / 17

Exploring Location Indicators for Location Indicator Normalization Geographic Information Retrieval Johannes Leveling and Normalization on surface (character), morphologic, Sven Hartrumpf syntactic, semantic, and lexical level. Introduction Location Morphologic level Indicators Location • Inflectional endings are identified and removed Indicator Normalization • Morphologic variations of location names are reduced Semantic Analysis for to their base form GIR • Derivational morphology: adjective → location name GeoCLEF 2007 Experiments Examples: Conclusion des Roten Meer(e)s → Rote Meer and Outlook bayrisch → Bayern References dänisch → Dänemark Johannes Leveling and Sven Hartrumpf Exploring Location Indicators for Geographic Information Retrieval 5 / 17

Exploring Location Indicators for Location Indicator Normalization Geographic Information Retrieval Johannes Leveling and Normalization on surface (character), morphologic, Sven Hartrumpf syntactic, semantic, and lexical level. Introduction Location Semantic level Indicators Location • Prefixes are separated from the name Indicator Normalization • Location indicators are mapped to location names Semantic Analysis for Examples: GIR Norddeutschland → Nord-Deutschland GeoCLEF 2007 exception: Experiments Conclusion Südafrika → Südafrika and Outlook References Johannes Leveling and Sven Hartrumpf Exploring Location Indicators for Geographic Information Retrieval 5 / 17

Exploring Location Indicators for Location Indicator Normalization Geographic Information Retrieval Johannes Leveling and Normalization on surface (character), morphologic, Sven Hartrumpf syntactic, semantic, and lexical level. Introduction Location Lexical level Indicators Location • Name variations are normalized using synset Indicator Normalization representatives Semantic Analysis for Example: GIR Burma → Myanmar GeoCLEF 2007 Birma → Myanmar Experiments Conclusion and Outlook References Johannes Leveling and Sven Hartrumpf Exploring Location Indicators for Geographic Information Retrieval 5 / 17