Left Nb. | Right Nb. | Frequency |
---|---|---|
un | un | 3 |
un | ka | 25 |
un | ir | 67 |
un | par | 54 |
un | kas | 27 |
un | ar | 65 |
un | no | 52 |
un | uz | 42 |
un | arī | 145 |
ir | un | 19 |
ir | kas | 10 |
ir | ar | 30 |
ir | uz | 10 |
ir | no | 25 |
ir | par | 110 |
ir | arī | 146 |
ka | par | 41 |
ka | arī | 63 |
arī | kas | 3 |
arī | ir | 46 |
arī | no | 41 |
arī | ar | 54 |
arī | uz | 41 |
arī | par | 86 |
kas | un | 6 |
kas | kā | 3 |
kas | par | 18 |
kas | ir | 253 |
kā | un | 6 |
kā | par | 12 |
NN co-occurrences within the 10 most frequent words are presented in a table.
The graph below gives much more information. Here, the top-1000 words are plotted against each other and the dots indicate NN co-occurrences. The diameter of the dots increases with the significance of the co-occurrence. Both axis are scaled logarithmic to shift the emphasis to the top words.
The picture above is very typical for a language, therefore the name language fingerprint. Comparing these fingerprints for different languages one is able to identify determiners, prepositions etc. by its graphical properties.
Frequency of the most frequent word:
select @maxfreq:=(select freq from words where w_id=101);
Table data:
select w1.word,w2.word,c.freq from co_n c, words w1, words w2 where w1.w_id=w1_id and w2.w_id=w2_id and w1_id>100 and w2_id>100 and 110>=w1_id and 110>=w2_id and c.freq>(select count(*) from sentences)/100000 order by w1.w_id;
Picture data:
select if(12>w1_id-99,w1.word,"-"),if(12>w2_id-99,w2.word,"-"),w1_id-99,w2_id-99,1/(log(c.freq/@maxfreq)*log(c.freq/@maxfreq)/20) from co_n c, words w1, words w2 where w1.w_id=w1_id and w2.w_id=w2_id and w1_id>100 and w2_id>100 and 1100>=w1_id and 1100>=w2_id and c.freq>(select count(*) from sentences)/100000;