WP2:%20Evaluation%20Infrastructure%20-%20Task%202.3:%20Data%20Processing - PowerPoint PPT Presentation

About This Presentation
Title:

WP2:%20Evaluation%20Infrastructure%20-%20Task%202.3:%20Data%20Processing

Description:

LEC Power Translator. Hagen (1st monolingual German) GIRSA (GIR by semantic annotation) ... SINTRAM (Sinai Translation Module) Location index (Ling Pipe for NER ... – PowerPoint PPT presentation

Number of Views:41
Avg rating:3.0/5.0
Slides: 12
Provided by: giorgiomar
Category:

less

Transcript and Presenter's Notes

Title: WP2:%20Evaluation%20Infrastructure%20-%20Task%202.3:%20Data%20Processing


1
(No Transcript)
2
Participation
Participant Monolingual Tasks Monolingual Tasks Monolingual Tasks Bilingual Tasks Bilingual Tasks Bilingual Tasks TOTAL
Participant DE EN PT X2DE X2EN X2PT
alivale 9 9
cheshire 3 3 3 6 6 6 27
csusm 1 1 2 1 1 1 7
dfki 5 5
hagen 5 10 15
icl 9 9
inaoe 12 12
jaen 7 6 13
pittsburgh 4 4
valencia 6 6
xldb 12 12 24
TOTAL 9 68 17 17 13 7 131
3
Monolingual vs Bilingual English
Bilingual 76 wrt Monolingual
4
(No Transcript)
5
Monolingual vs BilingualGerman
Bilingual 86 wrt Monolingual
6
(No Transcript)
7
Monolingual vs Bilingual Portuguese
Bilingual 90 wrt Monolingual
8
(No Transcript)
9
Approaches
  • XLDB (2nd monolingual Portuguese)
  • REMBRANDT, language-dependent named-entity
    recognition (uses Wikipedia)
  • QueryCol, query reformulation blind relevance
    feedback, query expansion using geographic
    ontology
  • MG4J text index/retrieval, implicit and explicit
    geographical indexes
  • Jaen (2nd bilingual English)
  • Lemure
  • SINTRAM (Sinai Translation Module)
  • Location index (Ling Pipe for NER using
    GeoNames),
  • Query reformulation module
  • AliVale (2nd monolingual English)
  • Jaen Fuzzy Borda result merging
  • CSUMS (last in all the tasks!)
  • TERRIER index/retrieval (DFR)
  • CHESHIRE (1st all bilingual tasks, 1st
    monolingual Portuguese)
  • OKAPI/BM25, Logistic Regression, fusion
    approaches
  • Geographic Indexing Subsystem using World
    Gazetteer (not used this year?)
  • LEC Power Translator
  • Hagen (1st monolingual German)
  • GIRSA (GIR by semantic annotation)
  • Text and location names indexed separately
  • Results of IR and QA combined
  • DFKI (1st monolingual English)
  • Lucene
  • NER Topic parsing (event part and geographic
    part)
  • Geographic ontology (using geographic taxonomies
    such as GeoNames, World Gazetteer)
  • Event ontology (using Wikipedia)

10
Some other approaches
  • UPV
  • Lucene
  • GeoWorSE
  • Stanford NER system, Geo WordNet
  • Geo index, text index, WordNet index
  • ICL/KMI
  • Lucene
  • Geografic index, text index
  • NER Sheffields GATE
  • Pedia Crawler for geographic co-occurrence
  • Data fusion
  • Pittsburgh
  • Query pattern analysis (geo-part, topic-part)
  • Lingpipe NER
  • WikiGEO
  • INAOE

11
Conclusions
Write a Comment
User Comments (0)
About PowerShow.com