4.6 Sentence Rankv

4.6 Sentence Rankv

After previous section’s good results, the linguistic information plays a key role in LS query finding. According to Rada Mihalcea’s study on the academic paper’s keywords extraction and paragraph summarization by graph-based ranking algorithms, undirected graph sentence rank, forward graph sentence rank and backward graph sentence rank which have achieved great effect in plain text retrieval, now are all practiced and tested in the project on all 225 pages.

SentenceRank	Google		Yahoo
3	87.00	38.67%	93.00	41.33%
4	113.00	50.22%	122.00	54.22%
5	133.00	59.11%	136.00	60.44%
6	146.00	64.89%	151.00	67.11%
7	152.00	67.56%	155.00	68.89%
8	157.00	69.78%	157.00	69.78%
9	159.00	70.67%	161.00	71.56%
10	162.00	72.00%	157.00	69.78%
11	166.00	73.78%	158.00	70.22%
12	165.00	73.33%	160.00	71.11%
13	165.00	73.33%	161.00	71.56%
14	164.00	72.89%	166.00	73.78%
15	167.00	74.22%	167.00	74.22%
Average	148.92	66.19%	149.54	66.46%

Table4. 25

 

(a)                                                                                        (b)

Figure4. 27 success retrieved pages’ counts per 225 pages and corresponding percentage value by undirected graph sentence rank.

Forward	Google		Yahoo
3	85.00	37.78%	93.00	41.33%
4	105.00	46.67%	127.00	56.44%
5	139.00	61.78%	144.00	64.00%
6	155.00	68.89%	155.00	68.89%
7	161.00	71.56%	160.00	71.11%
8	165.00	73.33%	159.00	70.67%
9	167.00	74.22%	163.00	72.44%
10	169.00	75.11%	168.00	74.67%
11	172.00	76.44%	167.00	74.22%
12	172.00	76.44%	168.00	74.67%
13	170.00	75.56%	169.00	75.11%
14	169.00	75.11%	171.00	76.00%
15	173.00	76.89%	174.00	77.33%
Average	154.00	68.44%	155.23	68.99%

Table4. 26

(a)                                                                                        (b)

Figure4. 28 success retrieved pages’ counts per 225 pages and corresponding percentage value by forwarded graph sentence rank.

Backward	Google		Yahoo
3	79.00	35.11%	87.00	38.67%
4	109.00	48.44%	105.00	46.67%
5	128.00	56.89%	117.00	52.00%
6	147.00	65.33%	136.00	60.44%
7	154.00	68.44%	135.00	60.00%
8	158.00	70.22%	153.00	68.00%
9	159.00	70.67%	151.00	67.11%
10	164.00	72.89%	152.00	67.56%
11	165.00	73.33%	162.00	72.00%
12	164.00	72.89%	166.00	73.78%
13	164.00	72.89%	155.00	68.89%
14	168.00	74.67%	153.00	68.00%
15	170.00	75.56%	160.00	71.11%
Average	148.38	65.95%	140.92	62.63%

Table4. 27

 

(a)                                                                                        (b)

Figure4. 29 success retrieved pages’ counts per 225 pages and corresponding percentage value by backward graph sentence rank.

Besides a higher success retrieve rate by sentence rank, which is reaching up to 75%, it is worth to mention that the results from Google and Yahoo are very closed and similar to each other, rather than all the previous sections’ results. The average results are shown in Figure4.30 , for easy comparison, the Title method is also included in Figure4.30 . The benefits from sentence rank cannot be disregarded.

Figure4. 30 all sentence rank related methods comparison along with title method

After all, the comprehensive chart which includes all the average success retrieve rates is shown in Figure4.31 .

Figure4. 31 all methods comparison

x-axis	Method	Google	Yahoo
1	Title	50.05%	44.00%
2	TF	60.24%	51.52%
3	DF	71.35%	58.36%
4	TFIDF	66.36%	55.62%
5	PW	60.55%	49.94%
6	TF3DF2	71.38%	61.09%
7	TF4DF1	66.84%	55.90%
8	TF5DF5	71.25%	61.23%
9	TFIDF3DF2	71.62%	63.08%
10	TFIDF4DF1	69.16%	58.63%
11	TFIDF5DF5	71.69%	62.12%
12	Word Rank	55.79%	49.57%
13	Nouns & Verbs Rank	47.32%	44.55%
14	WordRank3DF2	71.69%	59.90%
15	WordRank4DF1	65.95%	56.79%
16	WordRank5DF5	71.52%	60.72%
17	WordRank3TFIDF2	53.64%	44.79%
18	WordRank4TFIDF1	54.87%	47.90%
19	WordRank5TFIDF5	51.38%	43.01%
20	Random Sentence Pick	67.97%	64.14%
21	Sentence Rank	66.19%	66.46%
22	Forward Sentence Rank	68.44%	68.99%
23	Backward Sentence Rank	65.95%	62.63%

Table4. 28