Media material analysis of continuing article portions

1. A computer-based media material analyzer for analyzing data representative of media material having a layout and including one or more continuing articles that extend across multiple pages, comprising:

(a) a computer-based segmenter that identifies block segments associated with columnar body text in the media material pages; and

(b) a computer-based article composer that determines which of the identified block segments belong to a continuing article extending across multiple pages in the media material based on language statistics information and continuation transition information.

2. The media material analyzer of claim 1 , wherein the article composer includes a continuation layout transition analyzer that identifies a last block segment associated with a candidate continuing article portion in a first page, identifies one or more items below the last block segment, classifies the last block segment based on at least one feature of the identified one or more items, and applies a decision tree to pick out one or more continuation transition features indicative of the probability of the last block segment being in a continuing article.

3. The media material analyzer of claim 2 , wherein the article composer includes a continuation language statistics analyzer that calculates language statistics information for article portions on successive pages from the first page of the continuing article, and determines a probability that the last block segment in the candidate continuing article portion has a continuing article portion based on the calculated successive language statistics information.

4. The media material analyzer of claim 3 , wherein the continuation layout transition analyzer further identifies a first block segment associated with a candidate continuing article in a successive page, identifies one or more items above the first block segment, classifies the first block segment based on at least one feature of the one or more identified items, and applies a decision tree to pick out one or more continuation transition features indicative of the probability of the first block segment being in a continuing article, and determine probability of the first block segment being in a continuing article based on the applied decision tree.

5. The media material analyzer of claim 4 , wherein the continuation language statistics analyzer further calculates language statistics information for article portions in earlier pages than the page having the first block segment, and determines a probability that the first block segment in the candidate continuing article portion has a continuing article portion based on the calculated earlier language statistics information.

6. The media material analyzer of claim 5 , wherein the continuation layout transition analyzer further identifies a candidate pair of the last and first block segments, determines a set of continuation transition features for the pair of the last and first block segments, and applies a decision tree to determine a probability that the candidate pair of last and block segments belong to a same continuing article across pages in the media material based on the set of determined continuation transition features.

7. The media material analyzer of claim 3 , wherein the language statistics information comprises word frequency information, and the continuation language statistics analyzer calculates match scores based on word frequencies in text in the last block segment and text in the article portions on successive pages.

8. The media material analyzer of claim 5 , wherein the language statistics information comprises word frequency information, and the continuation language statistics analyzer calculates match scores based on word frequencies in text in the first block segment and text in the article portions on earlier pages.

9. A computer-implemented method for analyzing data representative of media material having a layout and including one or more continuing articles that extend across multiple pages, comprising:

(a) identifying, by a computer-based system, block segments associated with columnar body text in the media material pages; and

(b) determining, by the computer-based system, which of the identified block segments belong to a continuing article extending across multiple pages in the media material based on language statistics information and continuation transition information.

10. The method of claim 9 , wherein the article determining step includes:

analyzing continuation layout transition information for block segments in candidate continuing article portions; and

analyzing language statistics for text in the candidate continuing article portions.

11. The method of claim 10 , wherein the continuation layout transition information analyzing step comprises:

identifying a last block segment associated with a candidate continuing article portion in a first page;

identifying one or more items below the last block segment;

classifying the last block segment based on at least one feature of the identified one or more items; and

applying a decision tree to pick out one or more continuation transition features indicative of the probability of the last block segment being in a continuing article.

12. The method of claim 11 , wherein the language statistics analyzing step comprises:

calculating language statistics information for article portions on successive pages from the first page of the continuing article; and

determining a probability that the last block segment in the candidate continuing article portion has a continuing article portion based on the calculated successive language statistics information.

13. The method of claim 12 , wherein the continuation layout transition information analyzing step further comprises:

identifying a first block segment associated with a candidate continuing article in a successive page;

identifying one or more items above the first block segment;

classifying the first block segment based on at least one feature of the one or more identified items; and

applying a decision tree to pick out one or more continuation transition features indicative of the probability of the first block segment being in a continuing article, and determine probability of the first block segment being in a continuing article based on the applied decision tree.

14. The method of claim 13 , wherein the language statistics analyzing step further comprises:

calculating language statistics information for article portions in earlier pages than the page having the first block segment; and

determining a probability that the first block segment in the candidate continuing article portion has a continuing article portion based on the calculated earlier language statistics information.

15. The method of claim 12 , wherein the continuation layout transition analyzing step further includes:

identifying a candidate pair of the last and first block segments;

determining a set of continuation transition features for the pair of the last and first block segments; and

applying a decision tree to determine a probability that the candidate pair of last and block segments belong to a same continuing article across pages in the media material based on the set of determined continuation transition features.

16. The method of claim 12 , wherein the language statistics information comprises word frequency information, and the continuation language statistics analyzing step includes calculating match scores based on word frequencies in text in the last block segment and text in the article portions on successive pages.

17. The method of claim 14 , wherein the language statistics information comprises word frequency information, and the continuation language statistics analyzing step includes calculating match scores based on word frequencies in text in the first block segment and text in the article portions on earlier pages.

18. A computer-based article composer for composing continuing articles that extend across multiple pages in a media material having a layout, comprising:

a computer-based continuation layout transition analyzer; and

a computer-based continuation language statistics analyzer;

wherein the continuation layout transition analyzer applies a decision tree to pick out one or more continuation transition features indicative of the probabilities of last and first block segments of candidate article portions on different pages being in the same continuing article, and

the continuation language statistics analyzer calculates language statistics information for different article portions on different pages and determines probabilities that first and last block segments in candidate article portions have continuing article portions based on the calculated language statistics information, whereby the article composer can compose a continuing article across multiple pages based on the probabilities that the first and last block segments belong to the same continuing article in accordance with the analyzed continuation layout transition features and the calculated language statistics.