ranking related news predictions
play

Ranking Related News Predictions Nattiya Kanhabua 1 , Roi Blanco 2 - PowerPoint PPT Presentation

Jan 21, 2023 •155 likes •776 views

Ranking Related News Predictions Ranking Related News Predictions Nattiya Kanhabua 1 , Roi Blanco 2 and Michael Matthews 2 1 Norwegian University of Science and Tech., Norway 2 Yahoo! Research, Barcelona, Spain SIGIR2011, Beijing Ranking

  1. Ranking Related News Predictions Ranking Related News Predictions Nattiya Kanhabua 1 , Roi Blanco 2 and Michael Matthews 2 1 Norwegian University of Science and Tech., Norway 2 Yahoo! Research, Barcelona, Spain SIGIR’2011, Beijing

  2. Ranking Related News Predictions Outline Outline Introduction Problem Statement Related Work Contributions Task Definition System Architecture Models Approach Features Ranking Method Evaluation Experiment Setting Experimental Results

  3. Ranking Related News Predictions Outline Outline Introduction Problem Statement Related Work Contributions Task Definition System Architecture Models Approach Features Ranking Method Evaluation Experiment Setting Experimental Results

  4. Ranking Related News Predictions Outline Outline Introduction Problem Statement Related Work Contributions Task Definition System Architecture Models Approach Features Ranking Method Evaluation Experiment Setting Experimental Results

  5. Ranking Related News Predictions Outline Outline Introduction Problem Statement Related Work Contributions Task Definition System Architecture Models Approach Features Ranking Method Evaluation Experiment Setting Experimental Results

  6. Ranking Related News Predictions Introduction Problem Statement Outline Introduction Problem Statement Related Work Contributions Task Definition System Architecture Models Approach Features Ranking Method Evaluation Experiment Setting Experimental Results

  7. Ranking Related News Predictions Introduction Problem Statement Problem statement People are naturally curious about the future. ◮ How long will a war in the middle east last? ◮ What is the latest health care plan? ◮ What will happen to EU economies in next 5 years? ◮ What will be potential effects of climate changes ? Over 32% of 2.5M documents from Yahoo! News (July 2009 to July 2010) contain at least one prediction. A new task called ranking related news predictions . ◮ Retrieve predictions related to a news story in news archives. ◮ Rank them according to their relevance to the news story.

  8. Ranking Related News Predictions Introduction Problem Statement Problem statement People are naturally curious about the future. ◮ How long will a war in the middle east last? ◮ What is the latest health care plan? ◮ What will happen to EU economies in next 5 years? ◮ What will be potential effects of climate changes ? Over 32% of 2.5M documents from Yahoo! News (July 2009 to July 2010) contain at least one prediction. A new task called ranking related news predictions . ◮ Retrieve predictions related to a news story in news archives. ◮ Rank them according to their relevance to the news story.

  9. Ranking Related News Predictions Introduction Problem Statement Problem statement People are naturally curious about the future. ◮ How long will a war in the middle east last? ◮ What is the latest health care plan? ◮ What will happen to EU economies in next 5 years? ◮ What will be potential effects of climate changes ? Over 32% of 2.5M documents from Yahoo! News (July 2009 to July 2010) contain at least one prediction. A new task called ranking related news predictions . ◮ Retrieve predictions related to a news story in news archives. ◮ Rank them according to their relevance to the news story.

  10. Ranking Related News Predictions Introduction Problem Statement Related News Predictions

  11. Ranking Related News Predictions Introduction Problem Statement Related News Predictions

  12. Ranking Related News Predictions Introduction Problem Statement Related News Predictions Query = <gas, emission, percent, european, global, climate>

  13. Ranking Related News Predictions Introduction Related Work Outline Introduction Problem Statement Related Work Contributions Task Definition System Architecture Models Approach Features Ranking Method Evaluation Experiment Setting Experimental Results

  14. Ranking Related News Predictions Introduction Related Work Future-related Information Analyzing Tools Recorded Future Difference: a user must specify a query in advance using “predefined” entities.

  15. Ranking Related News Predictions Introduction Related Work Future-related Information Analyzing Tools Yahoo’s Time Explorer Difference: No ranking or performance evaluation is done.

  16. Ranking Related News Predictions Introduction Related Work Previous Work on Future Retrieval R. Baeza-Yates. Searching the future. SIGIR’2005 Workshop on Mathematical/Formal Methods in IR. ◮ Extract temporal expressions from news articles. ◮ Retrieve future information using a probabilistic model, i.e., multiplying term similarity and a time confidence . ◮ Only a small data set and a year granularity are used.

  17. Ranking Related News Predictions Introduction Related Work Previous Work on Future Retrieval A. Jatowt et al. Supporting analysis of future-related information in news archives and the web. JCDL ’2009. ◮ Extract future mentions from news snippets obtained from search engines. ◮ Summarize and aggregate results using clustering methods. ◮ Not focus on relevance and ranking of future information.

  18. Ranking Related News Predictions Introduction Contributions Outline Introduction Problem Statement Related Work Contributions Task Definition System Architecture Models Approach Features Ranking Method Evaluation Experiment Setting Experimental Results

  19. Ranking Related News Predictions Introduction Contributions Contributions I. Formally define ranking related news predictions . II. Four classes of features: term similarity, entity-based similarity, topic similarity and temporal similarity. III. Extensive evaluation using dataset with over 6000 judgments from the NYT Annotated Corpus.

  20. Ranking Related News Predictions Task Definition System Architecture Outline Introduction Problem Statement Related Work Contributions Task Definition System Architecture Models Approach Features Ranking Method Evaluation Experiment Setting Experimental Results

  21. Ranking Related News Predictions Task Definition System Architecture System Architecture Step 1: Document annotation. ◮ Extract temporal expressions using time and event recognition. ◮ Normalize them to dates so they can be anchored on a timeline. ◮ Output: sentences annotated with named entities and dates, i.e., predictions .

  22. Ranking Related News Predictions Task Definition System Architecture System Architecture Step 2: Retrieving predictions. ◮ Automatically generate a query from a news article being read . ◮ Retrieve predictions that match the query. ◮ Rank predictions by relevance. A prediction is “relevant” if it is about the topics of the article.

  23. Ranking Related News Predictions Task Definition Models Outline Introduction Problem Statement Related Work Contributions Task Definition System Architecture Models Approach Features Ranking Method Evaluation Experiment Setting Experimental Results

  24. Ranking Related News Predictions Task Definition Models Annotated Document Model Collection C = { d 1 , . . . , d n } . Document d = {{ w 1 , . . . , w n } , time ( d ) } . ◮ time ( d ) gives the publication date of d . Annotated document ˆ d is composed of: ◮ Named entities ˆ d e = { e 1 , . . . , e n } ◮ Temporal expressions ˆ d t = { t 1 , . . . , t m } ◮ Sentences ˆ d s = { s 1 , . . . , s z }

  25. Ranking Related News Predictions Task Definition Models Annotated Document Model Collection C = { d 1 , . . . , d n } . Document d = {{ w 1 , . . . , w n } , time ( d ) } . ◮ time ( d ) gives the publication date of d . Annotated document ˆ d is composed of: ◮ Named entities ˆ d e = { e 1 , . . . , e n } ◮ Temporal expressions ˆ d t = { t 1 , . . . , t m } ◮ Sentences ˆ d s = { s 1 , . . . , s z }

  26. Ranking Related News Predictions Task Definition Models Prediction Model Let d p be the parent document of a prediction p . p is a sentence containing field/value pairs: Field Value 1136243_1 ID PARENT _ ID 1136243 Gore Pledges A Health Plan For Every Child TITLE Vice President Al Gore proposed today to guarantee access to TEXT affordable health insurance for all children by 2005, expanding on a program enacted two years ago that he conceded had had limited success so far. Mr. Gore acknowledged that the number of Americans without CONTEXT health coverage had increased steadily since he and President Clinton took office. Al Gore ENTITY FUTURE _ DATE 2005 PUB _ DATE 1999 / 09 / 08

  27. Ranking Related News Predictions Task Definition Models Query Model Query q is extracted from a news article being read d q . 1. Keywords q text 2. Time constraints q time

  28. Ranking Related News Predictions Task Definition Models Query Keywords A prediction A news article being read Field ID PARENT_ID Query keyword TITLE extraction TEXT ENTITY (1) (3) (2) CONTEXT Term query Combined query FUTURE_DATE Entity query Q Q Q E T C PUB_DATE Q E = { e 1 , . . . , e m } E.g., � Barack Obama , Iraq , America �

  29. Ranking Related News Predictions Task Definition Models Query Keywords A prediction A news article being read Field ID PARENT_ID TITLE Query keyword extraction TEXT ENTITY (1) (2) (3) CONTEXT FUTURE_DATE Entity query Term query Combined query Q Q Q PUB_DATE C E T Q E = { w 1 , . . . , w n } E.g., � troop , war , withdraw �

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

Engines Previously We talked about the motivation behind vertical search engines,

Engines Previously We talked about the motivation behind vertical search engines,

Clustering for News Search Engines Previously We talked about the motivation behind vertical search engines, especially in the case of news A learning-to-rank approach of combining relevance ranking and freshness ranking was proposed

282 views • 15 slides
Ranking Daily News Events Speaker: Shih-Han Lo Advisor: Professor Jia-Ling Koh Author: Vinay

Ranking Daily News Events Speaker: Shih-Han Lo Advisor: Professor Jia-Ling Koh Author: Vinay

Modeling Event Importance for Ranking Daily News Events Speaker: Shih-Han Lo Advisor: Professor Jia-Ling Koh Author: Vinay Setty, Abhijit Anand, Arunav Mishra, Avishek Anand Date: 2017/03/21 Source: WSDM 17 1 Outline Introduction

388 views • 22 slides
IMPROVING NEWS RANKING BY COMMUNITY TWEETS Xin Shuai, Xiaozhong Liu, Johan Bollen Sunday, April

IMPROVING NEWS RANKING BY COMMUNITY TWEETS Xin Shuai, Xiaozhong Liu, Johan Bollen Sunday, April

IMPROVING NEWS RANKING BY COMMUNITY TWEETS Xin Shuai, Xiaozhong Liu, Johan Bollen Sunday, April 15, 2012 Different information needs by the same short query Will Obama sing at the next American Idol? Alice What's Will Obama Obama's

292 views • 10 slides
TVM for Ads Ranking @ Facebook Hao Lu, Ansha Yu, Yinghai Lu, Andrew Tulloch Ads Ranking at

TVM for Ads Ranking @ Facebook Hao Lu, Ansha Yu, Yinghai Lu, Andrew Tulloch Ads Ranking at

TVM for Ads Ranking @ Facebook Hao Lu, Ansha Yu, Yinghai Lu, Andrew Tulloch Ads Ranking at Facebook . . . ad 1 ad 2 ad 3 ad n model 2 model 2 model 1 model 3 . . . model k batch 1 batch 2 predictions X 2 Ads Ranking at Facebook:

530 views • 6 slides
SO(10) - inspired Leptogenesis and its predictions Work in progress: P . Di Bari, L. Marzola,

SO(10) - inspired Leptogenesis and its predictions Work in progress: P . Di Bari, L. Marzola,

SO(10) - inspired Leptogenesis and its predictions Work in progress: P . Di Bari, L. Marzola, S. Huber, S. Peeters L. Marzola NExT PhD Workshop July 2011 SO(10) - inspired Leptogenesis and its predictions Featuring news from T2K! Work in

500 views • 46 slides
1 Predictions for 2020 Predictions for 2020 We will live in flying houses. 1966

1 Predictions for 2020 Predictions for 2020 We will live in flying houses. 1966

1 Predictions for 2020 Predictions for 2020 We will live in flying houses. 1966 Predictions for 2020 We will have personal helicopters. 1951 Predictions for 2020 C, X, and Q will not be part of the alphabet. 1900 Predictions

818 views • 49 slides
Complexity news: Also somewhat related: discrete logarithms in Im starting to analyze

Complexity news: Also somewhat related: discrete logarithms in Im starting to analyze

Complexity news: Also somewhat related: discrete logarithms in Im starting to analyze multiplicative groups of cost of NFS + CVP small-characteristic finite fields for class groups, unit groups, the algorithm of Barbulescu, short

912 views • 80 slides
Online Submodular Set Cover, Ranking, and Repeated Active Learning Online Ranking: At each round,

Online Submodular Set Cover, Ranking, and Repeated Active Learning Online Ranking: At each round,

Online Submodular Set Cover, Ranking, and Repeated Active Learning Online Ranking: At each round, the learner produces an ordered list of items, then suffers loss or receives reward. Example: search result ranking Display ranked results Get

162 views • 4 slides
Ranking candidate genes from Ranking candidate genes from perturbation experiments Niko

Ranking candidate genes from Ranking candidate genes from perturbation experiments Niko

Ranking candidate genes from Ranking candidate genes from perturbation experiments Niko Beerenwinkel Gene ranking Goal: Identify (or prioritize) genes that affect readout, i.e., are involved in biological process of interest i l d i bi

433 views • 19 slides
Tutorial: TF-Ranking for sparse features Tutorial: TF-Ranking for sparse features This tutorial

Tutorial: TF-Ranking for sparse features Tutorial: TF-Ranking for sparse features This tutorial

Tutorial: TF-Ranking for sparse features Tutorial: TF-Ranking for sparse features This tutorial is an end-to-end walkthrough of training a TensorFlow Ranking (TF-Ranking) neural network model which incorporates sparse textual features.

257 views • 24 slides
Easy and Hard Outline Constraint Ranking in OT The Constraint Ranking problem Making fast

Easy and Hard Outline Constraint Ranking in OT The Constraint Ranking problem Making fast

Easy and Hard Outline Constraint Ranking in OT The Constraint Ranking problem Making fast ranking faster Jason Eisner Extension: Considering all competitors U. of Rochester How hard is OT generation? Making slow ranking

288 views • 6 slides
1 Similarity ranking: example Weighted scoring with linear combination A simple weighted

1 Similarity ranking: example Weighted scoring with linear combination A simple weighted

Table of Content Weighted scoring for ranking Learning to rank: A simple example Learning to ranking as classification Ranking and Learning 290N UCSB, Tao Yang, 2013 Partially based on Manning, Raghavan, and Schtzes text book.

476 views • 8 slides
Accounting for uncertainty in climate projections and the premise of decadal climate predictions

Accounting for uncertainty in climate projections and the premise of decadal climate predictions

Accounting for uncertainty in climate projections and the premise of decadal climate predictions Didier Swingedouw &amp; Giovanni Sgubin Outlines News from the global warming What is a climate projection? Decadal prediction

196 views • 15 slides
Methodology for Comparison and Ranking of SAT Solvers Mladen Nikoli c Third Workshop on

Methodology for Comparison and Ranking of SAT Solvers Mladen Nikoli c Third Workshop on

Introduction Preliminaries Methodology Evaluation Related work Conclusions Methodology for Comparison and Ranking of SAT Solvers Mladen Nikoli c Third Workshop on Formal and Automated Theorem Prooving and Applications January 29, 2010.

974 views • 27 slides
Logistics Job related news C++ Classes Co-op orientation Friday, Sept 16 th 12:00

Logistics Job related news C++ Classes Co-op orientation Friday, Sept 16 th 12:00

Logistics Job related news C++ Classes Co-op orientation Friday, Sept 16 th 12:00 1:30pm 70-1400 Friday, Oct 14 th 1:00 2:30pm Eastman Aud. Career Fair Wednesday, Sept 28 th 11am 4pm Gordon Field

555 views • 7 slides
On Human Predictions with Explanations and Predictions of Machine Learning Models: A Case Study

On Human Predictions with Explanations and Predictions of Machine Learning Models: A Case Study

On Human Predictions with Explanations and Predictions of Machine Learning Models: A Case Study on Deception Detection Vivian Lai and Chenhao Tan @vivwylai | @chenhaotan vivlai.github.io | chenhaot.com University of Colorado Boulder

391 views • 24 slides
PRLab TUDelft NL LEARNING UNDER COVARIATE SHIFT Domain Adaptation, Transfer Learning, Data

PRLab TUDelft NL LEARNING UNDER COVARIATE SHIFT Domain Adaptation, Transfer Learning, Data

PRLab TUDelft NL LEARNING UNDER COVARIATE SHIFT Domain Adaptation, Transfer Learning, Data Shift, Concept Drift Marco Loog Pattern Recognition Laboratory Delft University of Technology PRLab TUDelft NL PRLab TUDelft NL PRLab

635 views • 15 slides
Complex Prediction Problems A novel approach to multiple Structured Output Prediction Yasemin

Complex Prediction Problems A novel approach to multiple Structured Output Prediction Yasemin

Complex Prediction Problems A novel approach to multiple Structured Output Prediction Yasemin Altun Max-Planck Institute ECML HLIE08 Yasemin Altun Complex Prediction Information Extraction Extract structured information from unstructured

826 views • 27 slides
Extreme Classification Many modern applications involve a huge number of classes . E.g., image

Extreme Classification Many modern applications involve a huge number of classes . E.g., image

Multi-class SVMs From Tighter Data-Dependent Generalization Bounds to Novel Algorithms Marius Kloft Joint work with Yunwen Lei (CU Hong Kong), Urun Dogan (Microsoft Research), and Alexander Binder (Singapore). Multi-class SVMs From Tighter

358 views • 10 slides
Multilingual Dependency Analysis with a Two-Stage Discriminative Parser R. McDonald and K. Lerman

Multilingual Dependency Analysis with a Two-Stage Discriminative Parser R. McDonald and K. Lerman

Multilingual Dependency Analysis with a Two-Stage Discriminative Parser R. McDonald and K. Lerman and F. Pereira Dept. of Computer and Information Science University of Pennsylvania Conference on Natural Language Learning 2006 Shared Task on

425 views • 28 slides
!,/&amp;&quot;012,)&quot;'34,&amp;&quot;',%.'5$.&quot;6'7/62&quot;88$%&amp;' @ANU ML Workshop,

!,/&amp;&quot;012,)&quot;'34,&amp;&quot;',%.'5$.&quot;6'7/62&quot;88$%&amp;' @ANU ML Workshop,

!,/&amp;&quot;012,)&quot;'34,&amp;&quot;',%.'5$.&quot;6'7/62&quot;88$%&amp;' @ANU ML Workshop, Sept 23, 2011 !&quot;#$%&amp;'($&quot;' )&quot;#$%&amp;*#$&quot;+,%-*&quot;.-*,-' Obama @ Texas 9&quot;,)$:;'$%':&lt;$8'6%)$%&quot;'=6/).' Mar10

662 views • 33 slides
FOSTERING CREATIVITY Letting go of the control of the visual performance and allowing your

FOSTERING CREATIVITY Letting go of the control of the visual performance and allowing your

FOSTERING CREATIVITY Letting go of the control of the visual performance and allowing your members to be creative will lead to more genuine characterization in the performance. WHY IS CREATIVITY IMPORTANT IN YOUR PERFORMANCE? Fostering

251 views • 8 slides
Group Versus Individual Use of Power-Only EPMcreate as a Creativity Enhancement Technique for

Group Versus Individual Use of Power-Only EPMcreate as a Creativity Enhancement Technique for

Group Versus Individual Use of Power-Only EPMcreate as a Creativity Enhancement Technique for Requirements Elicitation 2013 V. Sakhnini, L. Mich, and D.M. Berry EMSE Group Sizes in CETs Pg. 1 by Victoria Sakhnini 1 , Luisa Mich 2 ,

894 views • 56 slides
Creativity in the Workplace Session starts at 10am HELLO! I am Karen Maher I am an experienced

Creativity in the Workplace Session starts at 10am HELLO! I am Karen Maher I am an experienced

Creativity in the Workplace Session starts at 10am HELLO! I am Karen Maher I am an experienced HR consultant and workforce development specialist originally from the North East of England. I specialise in coaching, mentoring, mediation and

625 views • 28 slides

More recommend