| 1 |
<?xml version="1.0" encoding="utf-8"?> |
| 2 |
<!DOCTYPE chapter SYSTEM "docbook-dtd-45/docbookx.dtd"> |
| 3 |
|
| 4 |
<chapter id="probabilistic"> |
| 5 |
<title>Making use of Probabilistic / Statistic Input</title> |
| 6 |
|
| 7 |
<para> |
| 8 |
If your input contains confidence values or similar, you can make use of those in the grammar. |
| 9 |
<emphasis> |
| 10 |
See <link linkend="numerical-matches">numeric tags</link> for the specific feature. |
| 11 |
</emphasis> |
| 12 |
</para> |
| 13 |
|
| 14 |
<para> |
| 15 |
For example, given the input sentence "<emphasis>Bear left at zoo.</emphasis>" a statistical tagger |
| 16 |
may assign confidence and frequency values to the readings: |
| 17 |
<screen> |
| 18 |
"<Bear>" |
| 19 |
"bear" N NOM SG <Noun:784> <Conf:80> @SUBJ |
| 20 |
"bear" V INF <Verb:140> <Conf:20> @IMV @#ICL-AUX< |
| 21 |
"<left>" |
| 22 |
"leave" PED @IMV @#ICL-N< |
| 23 |
"leave" V PAST VFIN @FMV |
| 24 |
"<at>" |
| 25 |
"at" PRP @ADVL |
| 26 |
"<zoo>" |
| 27 |
"zoo" N NOM SG @P< |
| 28 |
"<$.>" |
| 29 |
</screen> |
| 30 |
which you could query with e.g. |
| 31 |
<screen> |
| 32 |
# Remove any reading with Confidence below 5% |
| 33 |
REMOVE (<Conf<5>) ; |
| 34 |
# Select N NOM SG if Confidence is above 60% |
| 35 |
SELECT (N NOM SG <Conf>60>) ; |
| 36 |
# Remove the Verb reading if the frequency is under 150 |
| 37 |
# and Noun's frequency is above 700 |
| 38 |
REMOVE (<Verb<150>) (0 (<Noun>700>)) ; |
| 39 |
</screen> |
| 40 |
</para> |
| 41 |
|
| 42 |
<para> |
| 43 |
These are just examples of what <link linkend="numerical-matches">numeric tags</link> could be used for. |
| 44 |
There is no reason Confidence values are in % and there is no requirement that they must add up to 100%. |
| 45 |
The only requirement of a numerical tag is an alphanumeric identifier and a double-precision floating point value that fits |
| 46 |
in the range -281474976710656.0 to +281474976710655.0. |
| 47 |
</para> |
| 48 |
|
| 49 |
</chapter> |