w 012345 ya
play

}w !"#$%&'()+,-./012345<yA| Illustraons by Ji Franek. - PowerPoint PPT Presentation

Sep 24, 2022 •539 likes •711 views

Math Indexer and Searcher under the Hood: History and Development of a Winning Strategy Michal Rika, Petr Sojka, Marn Lka Masaryk University, Faculty of Informacs, Brno, Czech Republic mruzicka@mail.muni.cz, sojka@fi.muni.cz,

  1. Math Indexer and Searcher under the Hood: History and Development of a Winning Strategy Michal Růžička, Petr Sojka, Mar�n Líška Masaryk University, Faculty of Informa�cs, Brno, Czech Republic mruzicka@mail.muni.cz, sojka@fi.muni.cz, 255768@mail.muni.cz https://mir.fi.muni.cz/ }w� !"#$%&'()+,-./012345<yA| Illustra�ons by Jiří Franek.

  2. Results Comparison Approach Summary Outline 1 Results Comparison 2 Approach 3 Summary Math Indexer and Searcher under the Hood: History and Development of a Winning Strategy NTCIR 2014, Tokyo, Japan, December 11th, 2014

  3. Results Comparison Approach Summary Outline 1 Results Comparison 2 Approach 3 Summary Math Indexer and Searcher under the Hood: History and Development of a Winning Strategy NTCIR 2014, Tokyo, Japan, December 11th, 2014

  4. Results Comparison Approach Summary NTCIR-10 Math Task • The first (pilot) year of the math task event last year (i.e. 2013). • Formula search and Full-text search. • 4 runs submi�ed – differ in query language. • PMath – Run #1. • CMath – Run #2. • PCMath – Run #3. • T EX – Run #4. • Open Informa�on Retrieval. • 1 run submi�ed – T EX + text mixed queries. Math Indexer and Searcher under the Hood: History and Development of a Winning Strategy NTCIR 2014, Tokyo, Japan, December 11th, 2014

  5. Results Comparison Approach Summary NTCIR-10 Math Task Results Table 1: Result metrics for submitted runs in Formula Search with Relevance Level ≥ 3 (Relevant) Metric Run 1 Run 2 Run 4 P-10 avg 0.105 0.191 0.219 P-5 avg 0.133 0.229 0.276 MAP avg 0.060 0.112 0.127 0.109 0.185 0.123 Precision (64/589) (92/496) (96/778) Table 2: Result metrics for submitted runs in Formula Search with Relevance Level ≥ 1 (Partially Relevant) Metric Run 1 Run 2 Run 4 P-10 avg 0.143 0.214 0.267 P-5 avg 0.181 0.267 0.343 MAP avg 0.066 0.081 0.100 0.148 0.232 0.161 Precision (87/589) (115/496) (125/778) Math Indexer and Searcher under the Hood: History and Development of a Winning Strategy NTCIR 2014, Tokyo, Japan, December 11th, 2014

  6. Results Comparison Approach Summary NTCIR-11 Math-2 Task • Only one type of queries. • 50 queries, each • 1–4 formulae, • 1–4 keyphrases. • Wikipedia task in addi�on to the Main task. Math Indexer and Searcher under the Hood: History and Development of a Winning Strategy NTCIR 2014, Tokyo, Japan, December 11th, 2014

  7. Results Comparison Approach Summary NTCIR-11 Math-2 Main Task Results Table: Results of submi�ed runs with Relevance Level ≥ 3 (Relevant). Main task team rank is in [ ] for our best runs (in bold). PMath CMath PCMath T EX MAP avg 0.3073 0.3630 [1] 0.3594 0.3357 P@10 avg 0.3040 0.3520 [1] 0.3480 0.3380 P@5 avg 0.5120 0.5680 [1] 0.5560 0.5400 Table: Results of submi�ed runs with Relevance Level ≥ 1 (Par�ally Relevant). Number in [ ] is team rank of all runs. PMath CMath PCMath T EX MAP avg 0.2557 0.2807 [2] 0.2799 0.2747 P@10 avg 0.5020 0.5440 0.5520 [1] 0.5400 P@5 avg 0.8440 0.8720 [2] 0.8640 0.8480 Math Indexer and Searcher under the Hood: History and Development of a Winning Strategy NTCIR 2014, Tokyo, Japan, December 11th, 2014

  8. Results Comparison Approach Summary NTCIR-11 Math-2 Wikipedia Task Results • Topics with results: • 75 out of 100 (CMath run) • Average posi�on: • 64 correct results in top 100 • 58 correct results in top 20 • 56 correct results in top 10 • 53 correct results in top 5 • 52 correct results in top 4 • 50 correct results in top 3 • 48 correct results in top 2 • 46 correct results in top 1 Math Indexer and Searcher under the Hood: History and Development of a Winning Strategy NTCIR 2014, Tokyo, Japan, December 11th, 2014

  9. Results Comparison Approach Summary NTCIR-11 Math-2 Main Task Approach Math Indexer and Searcher under the Hood: History and Development of a Winning Strategy NTCIR 2014, Tokyo, Japan, December 11th, 2014

  10. Results Comparison Approach Summary NTCIR-11 Math-2 Main Task Approach: News • Query expansion & strip-merging of subresults. • Query expansion. 𝑔 𝑔 𝑙 � 𝑙 � 𝑙 � query 1 (the original query): � � 𝑔 𝑔 𝑙 � 𝑙 � query 2: � � 𝑔 𝑔 𝑙 � query 3: � � 𝑔 𝑔 query 4: � � 𝑔 𝑙 � 𝑙 � 𝑙 � query 5: � 𝑙 � 𝑙 � 𝑙 � query 6: Math Indexer and Searcher under the Hood: History and Development of a Winning Strategy NTCIR 2014, Tokyo, Japan, December 11th, 2014

  11. Results Comparison Approach Summary NTCIR-11 Math-2 Main Task Approach: News The final result list: • Strip-merging of subresults. 1: �� original 2: �� original • Example on three subqueries 3: �� original 4: �� subquery 1 (the original one and 5: �� subquery 1 6: �� subquery 2 two derived subqueries). 7: �� original 8: �� original 9: �� original Results of the subquery 1: 10: �� subquery 1 Results of the original query: 1: �� subquery 1 11: �� subquery 1 1: �� original 2: �� subquery 1 12: �� subquery 2 2: �� original 3: �� subquery 1 13: �� original 3: �� original 4: �� subquery 1 14: �� original 4: �� original 5: �� subquery 1 15: �� original 5: �� original 16: �� subquery 1 6: �� original No more results from subquery 1. 7: �� original Results of the subquery 2: 17: �� subquery 2 1: �� subquery 2 8: �� original 18: ��� original 2: �� subquery 2 9: �� original 19: ��� original 3: �� subquery 2 10: ��� original No more results from the original query. 4: �� subquery 2 11: ��� original 20: �� subquery 2 5: �� subquery 2 21: �� subquery 2 No more results from subquery 2. 22: �� random 23: �� random … 1000: ���� random Math Indexer and Searcher under the Hood: History and Development of a Winning Strategy NTCIR 2014, Tokyo, Japan, December 11th, 2014

  12. Results Comparison Approach Summary Query Expansion Results’ Insight 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0% 10% 20% 30% 40% 50% 60% 70% The percentage of results returned by individual subqueries Original Query Subquery 1 Subquery 2 Subquery 3 Subquery 4 Subquery 5 Subquery 6 Subquery 7 Figure: Rela�ve number of results found using different subqueries for every query in CMath run Math Indexer and Searcher under the Hood: History and Development of a Winning Strategy NTCIR 2014, Tokyo, Japan, December 11th, 2014

  13. Results Comparison Approach Summary NTCIR-11 Math-2 Wikipedia Task Content Topics • Completely the same fully automa�c system used for the main NTCIR Math Task and Wikipedia subtask. • Only different data. • No tuning or modifica�ons for the Wikipedia task. • Input Content MathML was transformed to the format of the main NTCIR math task. • Manually added Presenta�on MathML and TeX representa�on of the data. • Performed all the four runs (CMath, PMath, PCMath, TeX) similarly to the main task. • No query expansion & strip-merging possible as queries consist of a single formula only. Math Indexer and Searcher under the Hood: History and Development of a Winning Strategy NTCIR 2014, Tokyo, Japan, December 11th, 2014

  14. Results Comparison Approach Summary Summary • Our results significanlty improved since the last year. • Query expansion & strip-merging of subresults helps a lot . • Be�er unifica�on definitely needed. • Wikipedia task very useful. Math Indexer and Searcher under the Hood: History and Development of a Winning Strategy NTCIR 2014, Tokyo, Japan, December 11th, 2014

  15. Results Comparison Approach Summary Ques�ons? Math Indexer and Searcher under the Hood: History and Development of a Winning Strategy NTCIR 2014, Tokyo, Japan, December 11th, 2014

Recommend

}w !&quot;#$%&amp;'()+,-./012345&lt;yA| Illustraons by Ji Franek. Movaon Math

}w !&quot;#$%&amp;'()+,-./012345&lt;yA| Illustraons by Ji Franek. Movaon Math

Maths Informaon Retrieval for Digital Libraries Michal Rika Masaryk University, Faculty of Informacs, Brno, Czech Republic mruzicka@mail.muni.cz https://mir.fi.muni.cz/ }w !&quot;#$%&amp;'()+,-./012345&lt;yA| Illustraons by

720 views • 55 slides
!&quot;##$%&amp;'()*'++*,-./' 012345'1'0678*'9:8'48;'6&lt;&lt;3=8'!&gt;1=8' 0%?&quot;@.A%@*'

!&quot;##$%&amp;'()*'++*,-./' 012345'1'0678*'9:8'48;'6&lt;&lt;3=8'!&gt;1=8' 0%?&quot;@.A%@*'

!&quot;##$%&amp;'()*'++*,-./' 012345'1'0678*'9:8'48;'6&lt;&lt;3=8'!&gt;1=8' 0%?&quot;@.A%@*' !&quot;#&quot;$%&amp;'$%$()*&amp;+,%&quot;-.&amp;/.&quot;%)&quot;0$1*&amp;'$%$()&amp;2($$)&amp;&amp; &gt;.&amp;&quot;B*'

581 views • 17 slides
}w !&quot;#$%&amp;'()+,-./012345&lt;yA| Illustraons by Ji Franek. Semanc Indexing

}w !&quot;#$%&amp;'()+,-./012345&lt;yA| Illustraons by Ji Franek. Semanc Indexing

Flexible Similarity Search of Semanc Vectors Using Fulltext Search Engines Michal Rika, Vt Novotn, Petr Sojka; Jan Pomiklek, Radim ehek Masaryk University, Faculty of Informacs, Brno, Czech Republic mruzicka@mail.muni.cz ,

459 views • 32 slides
}w !&quot;#$%&amp;'()+,-./012345&lt;yA| Motivation MIaS at NTCIR Meaning search Formulae

}w !&quot;#$%&amp;'()+,-./012345&lt;yA| Motivation MIaS at NTCIR Meaning search Formulae

Towards a Meaning-Aware Math Information Retrieval Petr Sojka et al Masaryk University, Faculty of Informatics, Brno, Czech Republic &lt;https://mir.fi.muni.cz/&gt; Wikipedia Subtask Pre-Conference Meeting, NTCIR-11 2014, Tokyo, Japan December

630 views • 45 slides
}w !&quot;#$%&amp;'()+,-./012345&lt;yA| Motivation Semantics Semantics search Formulae

}w !&quot;#$%&amp;'()+,-./012345&lt;yA| Motivation Semantics Semantics search Formulae

Towards Structure-Aware Information Retrieval Petr Sojka et al Masaryk University, Faculty of Informatics, Brno, Czech Republic &lt;https://mir.fi.muni.cz/&gt; Informatics Colloquium, Faculty of Informatics, MU, Brno, Czech Republic October

833 views • 55 slides
!&quot;#$%&amp;$'()*()+&amp;),+&amp;-(./'#()0&amp;'&amp;1-&quot;'$2&amp;3/% ) ) 4/%()5650789:;5

!&quot;#$%&amp;$'()*()+&amp;),+&amp;-(./'#()0&amp;'&amp;1-&quot;'$2&amp;3/% ) ) 4/%()5650789:;5

!&quot;#$%&amp;$'()*()+&amp;),+&amp;-(./'#()0&amp;'&amp;1-&quot;'$2&amp;3/% ) ) 4/%()5650789:;5 ) !&quot;#$%&amp;#!'()# (# :&lt;(+=&lt;(2)0&gt;$?'(2) ! ! 6/1&amp;+$2&amp;3/%)@) ! ! *+,#-.#/012345#6#7#6&quot;(# &quot;('#2R# ! ! ! 823

411 views • 18 slides

More recommend