Terms in Time and Times in Context: A Graph-based Term-Time Ranking Model

Transcription

1 Terms in Time and Times in Context: A Graph-based Term-Time Ranking Model Andreas Spitz, Jannik Strötgen, Thomas Bögel and Michael Gertz Heidelberg University Institute of Computer Science Database Systems Research Group spitz@informatik.uni-heidelberg.de 5th Temporal Web Analytics Workshop Florence, May 18, 2015

2 What happened on June 15, 1215? A simple question. How simple is the answer? Terms in Time and Times in Context Andreas Spitz 1 of 24

3 Motivation Co-ooccurrence Graphs With structured data: quite simple Terms in Time and Times in Context Term-Ranking Projection Application Summary Based on unstructured text data: much more challenging Andreas Spitz 2 of 24

4 Data Set and Approach A corpus of all English Wikipedia articles: Only text is considered, no info-boxes 3, 079, 620 documents with time expressions Problem statement, given such a corpus: Extract and normalize temporal expressions (dates) Find key terms that best summarize a given date Terms in Time and Times in Context Andreas Spitz 3 of 24

5 Outline Outline of the approach: Represent date-term co-occurrences efficiently Extract and normalize temporal expressions (dates) Extract content words that co-occur with dates Generate an efficient data structure Based on this representation Identify relevant terms for any given date Identify similar dates for any given date Example applications Terms in Time and Times in Context Andreas Spitz 4 of 24

6 Extraction of Temporal Expressions Normalization, e.g., May 18, Handling relative temporal expressions, e.g., in May Considering the document type Source: Strötgen, Gertz Multilingual and Cross-domain Temporal Tagging (2013) Terms in Time and Times in Context Andreas Spitz 5 of 24

7 Coverage of Dates We use a combination of dates of three granularities: YYYY-MM-DD (day) YYYY-MM (month) YYYY (year) Percentage of dates that are included in the data per year year coverage in % Terms in Time and Times in Context Andreas Spitz 6 of 24

8 Extraction of Terms and Representation For all sentences s in any Wikipedia document: Terms in Time and Times in Context Andreas Spitz 7 of 24

9 Extraction of Terms and Representation Identify/normalize dates and remove stop words Terms in Time and Times in Context Andreas Spitz 7 of 24

10 Extraction of Terms and Representation Create a bipartite graph G s = (T s D s, E s ) with weights ω s Terms in Time and Times in Context Andreas Spitz 7 of 24

11 Extraction of Terms and Representation Satisfy the inclusion condition for dates Terms in Time and Times in Context Andreas Spitz 7 of 24

12 Extraction of Terms and Representation Satisfy the inclusion condition for dates Terms in Time and Times in Context Andreas Spitz 7 of 24

13 Graph aggregation Aggregate the sentence-graphs G s : T := T s D := D s E := E s ω(e) := ω s (e) We obtain G = (T D, E, ω) with: T = 3, 748, 730 terms D = 210, 375 dates E = 110, 639, 525 edges Terms in Time and Times in Context Andreas Spitz 8 of 24

14 Formalising the Question What happened on June 15, 1215? Which terms in the graph co-occur in a significant manner with the date ? Terms in Time and Times in Context Andreas Spitz 9 of 24

15 Ranking We need a ranking-function from dates D to a list of terms T r : D R T r(d) := ranking of terms t T by their significance for d Terms in Time and Times in Context Andreas Spitz 10 of 24

16 Ranking We need a ranking-function from dates D to a list of terms T r : D R T r(d) := ranking of terms t T by their significance for d Idea: adapt tf-idf to the bipartite graph tf-idf := tf log 1 df tf : frequency of term in document df : fraction of documents that contain the term Terms in Time and Times in Context Andreas Spitz 10 of 24

17 Adapting tf-idf How does this relate to the graph? Identify dates with documents, i.e., dates contain terms Term frequency given by edge weights: tf(d, t) ω(d, t) Inverse document frequency given by number of neighbours: idf(t) D deg(t) tf-idf := tf log 1 df tf-idf(d, t) := ω(d, t) log D deg(t) Terms in Time and Times in Context Andreas Spitz 11 of 24

18 June 15, 1215 Terms in Time and Times in Context Andreas Spitz 12 of 24

19 June 15, 1215 Terms in Time and Times in Context Andreas Spitz 12 of 24

20 A Ranking for Dates Ranking dates by term works analogously: tf-idf(t, d) := ω(t, d) log T deg(d) Terms in Time and Times in Context Andreas Spitz 13 of 24

21 A Ranking for Dates Ranking dates by term works analogously: tf-idf(t, d) := ω(t, d) log T deg(d) Terms in Time and Times in Context Andreas Spitz 13 of 24

22 Ranking Nodes by Similarity Within a Set Can we create a ranking for dates by dates?... or for terms by terms? Terms in Time and Times in Context Andreas Spitz 14 of 24

23 Ranking Nodes by Similarity Within a Set Can we create a ranking for dates by dates?... or for terms by terms? Formally this is a one-mode projection of the bipartite graph: Reduce graph to a single set of nodes T or D Connect nodes that share neighbours in the bipartite graph This results in a very dense graph How can we identify relevant edges in the projection? Terms in Time and Times in Context Andreas Spitz 14 of 24

24 Cosine Similarity of Adjacency Vectors In a lesson from Collaborative Filtering : use a cosine similarity of adjacency vectors cos(t a, t b ) := tai t bi t 2 ai t 2 bi Terms in Time and Times in Context Andreas Spitz 15 of 24

25 Evaluation Ground truth: U.S. Election Days ( ) Recurs annually Always on Tuesday after the first Monday in November (Nov 2 - Nov 8) Every four years: presidential election Terms in Time and Times in Context Andreas Spitz 16 of 24

26 Evaluation Ground truth: U.S. Election Days ( ) Recurs annually Always on Tuesday after the first Monday in November (Nov 2 - Nov 8) Every four years: presidential election Expectation: For a given election day, election days in other years are ranked highly For presidential election days, other presidential election days are ranked highly Terms in Time and Times in Context Andreas Spitz 16 of 24

27 Precision at k 0.6 precision k election day presidential general prec k := Election days in top k ranks k Terms in Time and Times in Context Andreas Spitz 17 of 24

28 Area Under the ROC Curve year AUC election day general presidential Terms in Time and Times in Context Andreas Spitz 18 of 24

29 Practical Application: Hot Spots & Key Players Here: approximation of countries activity during given months For each European country c, define its name, e.g. t n (c) = italy, define the countries adjectival form, e.g. t a (c) = italian, compute individual tf-idf scores for terms and combine. act(c, d) := tf-idf(d, t n(c)) + tf-idf(d, t a (c)) max[tf-idf(d, )] Terms in Time and Times in Context Andreas Spitz 19 of 24

30 Activity by Country During World War II March 1939 September 1939 November 1939 German occupation German invasion Soviet invasion of Czechoslovakia of Poland of Finland activity 1.5 April 1940 May 1940 October 1940 German assault German invasion Battle of Britain and on Norway of France Greco Italian War Terms in Time and Times in Context Andreas Spitz 20 of 24

31 Activity by Country During World War II (2) April 1941 June 1941 September 1943 Invasion of Yugoslavia German offensive Allied invasion and Greece against the Soviets of Italy June 1944 August 1944 May 1945 Allied forces land in Liberation of Paris German Capitulation Normandy (DDay) activity Terms in Time and Times in Context Andreas Spitz 21 of 24

32 Summary Approach: Extract dates and terms from unstructured text Construct a bipartite date-term graph Allows ranking dates / terms according to co-occurrences Benefits: Simple measures already yield good results Efficient: 4GB Memory and real-time queries Flexibility of ranking methods Terms in Time and Times in Context Andreas Spitz 22 of 24

33 Ongoing Work Terms in Time and Times in Context Andreas Spitz 23 of 24

34 Thank you! Questions? Terms in Time and Times in Context Andreas Spitz 24 of 24