combining probabilistic and translation based models for
play

Combining Probabilistic and Translation- Based Models for - PowerPoint PPT Presentation

Nov 15, 2022 •315 likes •711 views

Combining Probabilistic and Translation- Based Models for Information Retrieval based on Word Sense Annotations Elisabeth Wolf, Delphine Bernhard, Iryna Gurevych Ubiquitous Knowledge Processing (UKP) Lab Prof. Dr. Iryna Gurevych Fachbereich

  1. Combining Probabilistic and Translation- Based Models for Information Retrieval based on Word Sense Annotations Elisabeth Wolf, Delphine Bernhard, Iryna Gurevych Ubiquitous Knowledge Processing (UKP) Lab Prof. Dr. Iryna Gurevych Fachbereich Informatik Technische Universität Darmstadt

  2. UKP Motivation: monolingual task 1. Increase precision of WSD 2. Apply translation-based model + combination with probabilistic m. 1. ….. 1. ….. UBC NUS 2. ….. 2. ….. 3. ….. 3. ….. 1. ….. 2. ….. Comb 3. ….. Heuristic-based combinations Reranking of retrieved of both annotations documents I N D E X I N G R E T R I E V A L 02.10.09 | Computer Science Department | Ubiquitous Knowledge Processing Lab

  3. 1. Increase precision of WSD • Four different index types: UBC NUS <SYNSET SCORE="0.32" <SYNSET CODE="00735486-n"/> SCORE="0.82" <SYNSET CODE="00735486-n"/> SCORE="0.21" <SYNSET CODE="03857483-n"/> SCORE="0.18" <SYNSET CODE="03857483-n"/> SCORE="0.47" CODE="01252343-n"/> 02.10.09 | Computer Science Department | Ubiquitous Knowledge Processing Lab

  4. 1. Increase precision of WSD • Four different index types: UBC NUS <SYNSET SCORE="0.32" <SYNSET • UBCBest CODE="00735486-n"/> SCORE="0.82" <SYNSET CODE="00735486-n"/> SCORE="0.21" <SYNSET CODE="03857483-n"/> SCORE="0.18" <SYNSET CODE="03857483-n"/> SCORE="0.47" CODE="01252343-n"/> 02.10.09 | Computer Science Department | Ubiquitous Knowledge Processing Lab

  5. 1. Increase precision of WSD • Four different index types: UBC NUS <SYNSET SCORE="0.32" <SYNSET • UBCBest CODE="00735486-n"/> SCORE="0.82" • NUSBest <SYNSET CODE="00735486-n"/> SCORE="0.21" <SYNSET CODE="03857483-n"/> SCORE="0.18" <SYNSET CODE="03857483-n"/> SCORE="0.47" CODE="01252343-n"/> 02.10.09 | Computer Science Department | Ubiquitous Knowledge Processing Lab

  6. 1. Increase precision of WSD • Four different index types: UBC NUS <SYNSET SCORE="0.32" <SYNSET • UBCBest CODE="00735486-n"/> SCORE="0.82" • NUSBest <SYNSET CODE="00735486-n"/> • CombBest SCORE="0.21" <SYNSET CODE="03857483-n"/> SCORE="0.18" <SYNSET CODE="03857483-n"/> SCORE="0.47" CODE="01252343-n"/> 02.10.09 | Computer Science Department | Ubiquitous Knowledge Processing Lab

  7. 1. Increase precision of WSD • Four different index types: UBC NUS <SYNSET SCORE="0.32" <SYNSET • UBCBest CODE="00735486-n"/> SCORE="0.82" • NUSBest <SYNSET CODE="00735486-n"/> • CombBest SCORE="0.21" <SYNSET CODE="03857483-n"/> SCORE="0.18" <SYNSET CODE="03857483-n"/> SCORE="0.47" CODE="01252343-n"/> 02.10.09 | Computer Science Department | Ubiquitous Knowledge Processing Lab

  8. 1. Increase precision of WSD • Four different index types: UBC NUS <SYNSET SCORE="0.32" <SYNSET • UBCBest CODE="00735486-n"/> SCORE="0.82" • NUSBest <SYNSET CODE="00735486-n"/> • CombBest SCORE="0.21" <SYNSET CODE="03857483-n"/> SCORE="0.18" <SYNSET CODE="03857483-n"/> SCORE="0.47" CODE="01252343-n"/> 0.82 + 0.32 = 1.14 0.18 + 0.21 = 0.39 02.10.09 | Computer Science Department | Ubiquitous Knowledge Processing Lab

  9. 1. Increase precision of WSD • Four different index types: UBC NUS <SYNSET SCORE="0.32" <SYNSET • UBCBest CODE="00735486-n"/> SCORE="0.82" • NUSBest <SYNSET CODE="00735486-n"/> • CombBest SCORE="0.21" <SYNSET CODE="03857483-n"/> SCORE="0.18" <SYNSET CODE="03857483-n"/> SCORE="0.47" CODE="01252343-n"/> CombBest 0.82 + 0.32 = 1.14 0.18 + 0.21 = 0.39 02.10.09 | Computer Science Department | Ubiquitous Knowledge Processing Lab

  10. 1. Increase precision of WSD • Four different index types: UBC NUS <SYNSET SCORE="0.32" <SYNSET • UBCBest CODE="0111222-n "/> SCORE="0.82" • NUSBest <SYNSET CODE="00735486-n"/> • CombBest SCORE="0.21" <SYNSET CODE=„0333444-n"/> • CombBest+ SCORE="0.18" <SYNSET CODE="03857483-n"/> SCORE="0.47" CODE="01252343-n"/> 02.10.09 | Computer Science Department | Ubiquitous Knowledge Processing Lab

  11. 1. Increase precision of WSD • Four different index types: UBC NUS <SYNSET SCORE="0.32" <SYNSET • UBCBest CODE="0111222-n "/> SCORE="0.82" • NUSBest <SYNSET CODE="00735486-n"/> • CombBest SCORE="0.21" <SYNSET CODE=„0333444-n"/> • CombBest+ SCORE="0.18" <SYNSET CODE="03857483-n"/> SCORE="0.47" CODE="01252343-n"/> • Terrier, version 2.1 • Multi field indices: token, lemma, sense (UBCBest, NUSBest, CombBest, CombBest+) 02.10.09 | Computer Science Department | Ubiquitous Knowledge Processing Lab

  12. 2. Combination of Retrieval Models Probabilistic model + Translation-based Query expansion: model: Divergence From Randomness Monolingual translation-based model (TM): BM25 model (DFR_BM25): • Motivation: • address the lexical gap problem • learn translation probabilities between terms trained on Kullback-Leibler model (KL): parallel dataset: dictionary and encyclopedic definitions • 10 terms out of 3 top ranked docs • „the translation probability reflects the association between query term and document term” • Usage: • trained model recently successfully applied by Bernhard&Gurevych (2009) for answer finding • trained on token 02.10.09 | Computer Science Department | Ubiquitous Knowledge Processing Lab

  13. 2. Combination of Retrieval Models Probabilistic model + Translation-based Query expansion: model: Divergence From Randomness Monolingual translation-based model (TM): BM25 model (DFR_BM25): • Motivation: • address the lexical gap problem • learn translation probabilities between terms trained on Kullback-Leibler model (KL): parallel dataset: dictionary and encyclopedic definitions • 10 terms out of 3 top ranked docs • „the translation probability reflects the association between query term and document term” • Usage: • trained model recently successfully applied by Bernhard&Gurevych (2009) for answer finding • trained on token 02.10.09 | Computer Science Department | Ubiquitous Knowledge Processing Lab

  14. 2. Combination of Retrieval Models Probabilistic model + Translation-based Query expansion: model: Divergence From Randomness Monolingual translation-based model (TM): BM25 model (DFR_BM25): • Motivation: • address the lexical gap problem • learn translation probabilities between terms trained on Kullback-Leibler model (KL): parallel dataset: dictionary and encyclopedic definitions • 10 terms out of 3 top ranked docs • „the translation probability reflects the association between query term and document term” • Usage: • trained model recently successfully applied by Bernhard&Gurevych (2009) for answer finding • trained on token 02.10.09 | Computer Science Department | Ubiquitous Knowledge Processing Lab

  15. 2. Combination of Retrieval Models • Hypothesis: probabilistic and translation-based models retrieve different sets of relevant documents 1. ….. 1. ….. TM DFR_BM25 + KL 2. ….. 2. ….. 3. ….. 3. ….. token token lemma 1. ….. 2. ….. sense 3. ….. 02.10.09 | Computer Science Department | Ubiquitous Knowledge Processing Lab

  16. 2. Combination of Retrieval Models • Hypothesis: probabilistic and translation-based models retrieve different sets of relevant documents 1. ….. 1. ….. TM DFR_BM25 + KL 2. ….. 2. ….. 3. ….. 3. ….. token token lemma 1. ….. 2. ….. sense 3. ….. A) normalization: r norm (i) = (r orig (i) – r min ) / (r max – r min ) B) CombSUM by Fox&Shaw(1994): r comb (i) = SUM(Individual r norm (i)) 02.10.09 | Computer Science Department | Ubiquitous Knowledge Processing Lab

  17. Extrinsic evaluation: sense index types • Retrieval based on indexed senses (DFR_BM25 +KL): Index type MAP (training) MAP (test) UBCBest 0.2514 0.2636 NUSBest 0.2930 0.3473 CombBest 0.2921 0.3313 CombBest+ 0.3011 0.3551 • CombBest+ outperforms CombBest • Focus on „combined“ indices 02.10.09 | Computer Science Department | Ubiquitous Knowledge Processing Lab

  18. Extrinsic evaluation: sense index types • Retrieval based on indexed senses (DFR_BM25 +KL): Index type MAP (training) MAP (test) UBCBest 0.2514 0.2636 NUSBest 0.2930 0.3473 CombBest 0.2921 0.3313 CombBest+ 0.3011 0.3551 • CombBest+ outperforms CombBest • Focus on „combined“ indices 02.10.09 | Computer Science Department | Ubiquitous Knowledge Processing Lab

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend

Chapter 11 Tree-based models Statistical Machine Translation Tree-Based Models Traditional

Chapter 11 Tree-based models Statistical Machine Translation Tree-Based Models Traditional

Chapter 11 Tree-based models Statistical Machine Translation Tree-Based Models Traditional statistical models operate on sequences of words Many translation problems can be best explained by pointing to syntax reordering, e.g., verb

1.39k views • 120 slides
Chapter 4 Word-based models Statistical Machine Translation Lexical Translation How to

Chapter 4 Word-based models Statistical Machine Translation Lexical Translation How to

Chapter 4 Word-based models Statistical Machine Translation Lexical Translation How to translate a word look up in dictionary Haus house, building, home, household, shell. Multiple translations some more frequent than others

1.2k views • 61 slides
Chapter 5 Phrase-based models Statistical Machine Translation Motivation Word-Based Models

Chapter 5 Phrase-based models Statistical Machine Translation Motivation Word-Based Models

Chapter 5 Phrase-based models Statistical Machine Translation Motivation Word-Based Models translate words as atomic units Phrase-Based Models translate phrases as atomic units Advantages: many-to-many translation can handle

303 views • 27 slides
3 Language Models 1: n -gram Language Models While the final goal of a probabilistic machine

3 Language Models 1: n -gram Language Models While the final goal of a probabilistic machine

3 Language Models 1: n -gram Language Models While the final goal of a probabilistic machine translation system is to create a model of the target sentence E given the source sentence F , P ( E | F ), in this chapter we will take a step back, and

458 views • 8 slides
4CSLL5 IBM Translation Models Martin Emms October 22, 2020 4CSLL5 IBM Translation Models IBM

4CSLL5 IBM Translation Models Martin Emms October 22, 2020 4CSLL5 IBM Translation Models IBM

4CSLL5 IBM Translation Models 4CSLL5 IBM Translation Models Martin Emms October 22, 2020 4CSLL5 IBM Translation Models IBM models Probabilities and Translation Alignments IBM Model 1 definitions 4CSLL5 IBM Translation Models IBM models

1.23k views • 103 slides
4CSLL5 IBM Translation Models IBM models Probabilities and Translation Alignments Martin Emms

4CSLL5 IBM Translation Models IBM models Probabilities and Translation Alignments Martin Emms

4CSLL5 IBM Translation Models 4CSLL5 IBM Translation Models 4CSLL5 IBM Translation Models IBM models Probabilities and Translation Alignments Martin Emms IBM Model 1 definitions October 22, 2020 4CSLL5 IBM Translation Models 4CSLL5 IBM

571 views • 7 slides
Combining Crowd and AI to scale professional-quality translation Joo Graa Joo Graa CTO

Combining Crowd and AI to scale professional-quality translation Joo Graa Joo Graa CTO

Building universal understanding Combining Crowd and AI to scale professional-quality translation Joo Graa Joo Graa CTO CTO Proceedings for AMTA 2018 Workshop: Translation Quality Estimation and Automatic Post-Editing Boston, March

499 views • 45 slides
III.3 Probabilistic Retrieval Models 1. Probabilistic Ranking Principle 2. Binary Independence

III.3 Probabilistic Retrieval Models 1. Probabilistic Ranking Principle 2. Binary Independence

III.3 Probabilistic Retrieval Models 1. Probabilistic Ranking Principle 2. Binary Independence Model 3. Okapi BM25 4. Tree Dependence Model 5. Bayesian Networks for IR ! ! Based on MRS Chapter 11 IR&amp;DM 13/14 ! 48

628 views • 30 slides
Language models Chapter 3 in Martin/Jurafsky Probabilistic Language Models Goal: assign a

Language models Chapter 3 in Martin/Jurafsky Probabilistic Language Models Goal: assign a

10/17/19 Language models Chapter 3 in Martin/Jurafsky Probabilistic Language Models Goal: assign a probability to a sentence Machine Translation: Why? P(hi high h winds tonite) &gt; P(la large winds tonite) Spell Correction

321 views • 11 slides
Vector Space Models for Phrase-based Machine Translation Tamer Alkhouli, Andreas Guta, and

Vector Space Models for Phrase-based Machine Translation Tamer Alkhouli, Andreas Guta, and

Vector Space Models for Phrase-based Machine Translation Tamer Alkhouli, Andreas Guta, and Hermann Ney &lt;surname&gt;@cs.rwth-aachen.de Eighth Workshop on Syntax, Semantics and Structure in Statistical Translation Doha, Qatar October 25, 2014

275 views • 16 slides
Table of Contents I Probabilistic Reasoning Classical Probabilistic Models Basic Probabilistic

Table of Contents I Probabilistic Reasoning Classical Probabilistic Models Basic Probabilistic

Table of Contents I Probabilistic Reasoning Classical Probabilistic Models Basic Probabilistic Reasoning: The Jungle Story Multiple Random Selection Rules: Dice New Constructs Causal Probability Observations and Intentions Dynamic Range

1.08k views • 73 slides
Probabilistic Graphical Models Probabilistic Graphical Models Undirected Models Fall 2019

Probabilistic Graphical Models Probabilistic Graphical Models Undirected Models Fall 2019

Probabilistic Graphical Models Probabilistic Graphical Models Undirected Models Fall 2019 Siamak Ravanbakhsh Learning objectives Learning objectives Markov networks: how it represents a prob. dist. independence assumptions

941 views • 57 slides
Combining Interval and Towards a Better . . . How General Can We . . . Probabilistic Uncertainty

Combining Interval and Towards a Better . . . How General Can We . . . Probabilistic Uncertainty

Introduction How to Get More . . . How to Speed Up . . . Combining Interval and Towards a Better . . . How General Can We . . . Probabilistic Uncertainty in Proofs Engineering Applications Home Page Title Page Andrew Pownuk

872 views • 72 slides
Extended Translation Models in Phrase-based Decoding Andreas Guta, Joern Wuebker, Miguel Graa,

Extended Translation Models in Phrase-based Decoding Andreas Guta, Joern Wuebker, Miguel Graa,

Extended Translation Models in Phrase-based Decoding Andreas Guta, Joern Wuebker, Miguel Graa, Yunsu Kim and Hermann Ney surname@cs.rwth-aachen.de Tenth Workshop on Statistical Machine Translation (WMT) Lisbon, Portugal 18.09.2015 Human

456 views • 22 slides
Computationally efficient probabilistic inference with noisy threshold models based on a CP

Computationally efficient probabilistic inference with noisy threshold models based on a CP

Computationally efficient probabilistic inference with noisy threshold models based on a CP tensor decomposition Jirka Vomlel and Petr Tichavsk y Institute of Information Theory and Automation ( UTIA) Academy of Sciences of the Czech

1.06k views • 53 slides
4CSLL5 IBM Translation Models Martin Emms October 23, 2020 4CSLL5 IBM Translation Models

4CSLL5 IBM Translation Models Martin Emms October 23, 2020 4CSLL5 IBM Translation Models

4CSLL5 IBM Translation Models 4CSLL5 IBM Translation Models Martin Emms October 23, 2020 4CSLL5 IBM Translation Models Parameter learning (brute force) Introduction The brute force EM algorithm defined A formula for p ( a | o , s ) Examples

759 views • 50 slides
Experiments on Active Learning for Croatian Word Sense Disambiguation c and Jan Domagoj

Experiments on Active Learning for Croatian Word Sense Disambiguation c and Jan Domagoj

Experiments on Active Learning for Croatian Word Sense Disambiguation c and Jan Domagoj Alagi Snajder TakeLab UNIZG BSNLP 2015 @ RANLP, Hissar 10 Sep 2015 Problem Many words are polysemous: The flight was delayed due to trouble with

528 views • 35 slides
Online: Unit Testjng Michael Meeks &lt;michael.meeks@collabora.com&gt; mmeeks / irc.freenode.net

Online: Unit Testjng Michael Meeks &lt;michael.meeks@collabora.com&gt; mmeeks / irc.freenode.net

Collabora Productivity Online: Unit Testjng Michael Meeks &lt;michael.meeks@collabora.com&gt; mmeeks / irc.freenode.net @mmeeks +mejmeeks In his great mercy he has given us new birth into a living hope through the resurrectjon of Jesus

327 views • 14 slides
Lexical Semantics &amp; WSD Ling571 Deep Processing Techniques for NLP February 24, 2016

Lexical Semantics &amp; WSD Ling571 Deep Processing Techniques for NLP February 24, 2016

Lexical Semantics &amp; WSD Ling571 Deep Processing Techniques for NLP February 24, 2016 Roadmap Distributional models Compression Integration Dictionary-based models Thesaurus-based similarity models WordNet

472 views • 35 slides
A Sublinear Bipartiteness Tester for Bounded Degree Graphs Oded Goldreich Dana Ron

A Sublinear Bipartiteness Tester for Bounded Degree Graphs Oded Goldreich Dana Ron

A Sublinear Bipartiteness Tester for Bounded Degree Graphs Oded Goldreich Dana Ron February 5, 1998 Abstract We present a sublinear-time algorithm for testing whether a bounded degree graph is bipartite or far from being bipartite.

562 views • 26 slides
Topic Models for Word Sense Disambiguation and Token-based Idiom Detection Linlin Li, Benjamin

Topic Models for Word Sense Disambiguation and Token-based Idiom Detection Linlin Li, Benjamin

Introduction The Sense Disambiguation Model Experimental Setup Experiments Conclusion Topic Models for Word Sense Disambiguation and Token-based Idiom Detection Linlin Li, Benjamin Roth and Caroline Sporleder Cluster of Excellence, MMCI

543 views • 35 slides
INF4080 2020 FALL NATURAL LANGUAGE PROCESSING Jan Tore Lnning 2 (Mostly Text)

INF4080 2020 FALL NATURAL LANGUAGE PROCESSING Jan Tore Lnning 2 (Mostly Text)

1 INF4080 2020 FALL NATURAL LANGUAGE PROCESSING Jan Tore Lnning 2 (Mostly Text) Classification, Naive Bayes Lecture 3, 31 Aug Today - Classification 3 Motivation Classification Naive Bayes classification NB for text

909 views • 62 slides
Computational Semantics and Pragmatics Autumn 2011 Raquel Fernndez Institute for Logic,

Computational Semantics and Pragmatics Autumn 2011 Raquel Fernndez Institute for Logic,

Computational Semantics and Pragmatics Autumn 2011 Raquel Fernndez Institute for Logic, Language &amp; Computation University of Amsterdam Raquel Fernndez COSP 2011 1 / 24 Plan for Today Well discuss the main features and differences

400 views • 24 slides
Identifying Generic Expressions Nils Reiter and Anette Frank Department of Computational

Identifying Generic Expressions Nils Reiter and Anette Frank Department of Computational

Identifying Generic Expressions Nils Reiter and Anette Frank Department of Computational Linguistics Heidelberg University Germany Elephants [Elephants] can crush and kill any other land animal [...] In Africa, groups of young teenage

439 views • 43 slides

More recommend