pln-uba-2019

Introducción al Procesamiento de Lenguaje Natural - UBA 2019

Link a las consignas: https://wiki.cs.famaf.unc.edu.ar/doku.php?id=materias:pln:uba2019:practico1

Ejercicio 1 | stats.py

Basic Statistics
================
sents: 17378
tokens: 517194
words: 46501
tags: 85

Example of word frequency
================
Frequency of "presidente": 682

Example of tag frequency
================
Frequency of "nc0s000": 63452

Most Frequent POS Tags
======================
tag	freq	%	top
sp000	79884	15.45	(de, en, a, del, con)
nc0s000	63452	12.27	(presidente, equipo, partido, país, año)
da0000	54549	10.55	(la, el, los, las, El)
aq0000	33906	6.56	(pasado, gran, mayor, nuevo, próximo)
fc	30147	5.83	(,)
np00000	29111	5.63	(Gobierno, España, PP, Barcelona, Madrid)
nc0p000	27736	5.36	(años, millones, personas, países, días)
fp	17512	3.39	(.)
rg	15336	2.97	(más, hoy, también, ayer, ya)
cc	15023	2.90	(y, pero, o, Pero, e)

Word Ambiguity Levels
=====================
n	words	%	top
1	43972	94.56	(,, con, por, su, El)
2	2318	4.98	(el, en, y, ", los)
3	180	0.39	(de, la, ., un, no)
4	23	0.05	(que, a, dos, este, fue)
5	5	0.01	(mismo, cinco, medio, ocho, vista)
6	3	0.01	(una, como, uno)
7	0	0.00	()
8	0	0.00	()
9	0	0.00	()

Tag	Meaning
sp000	Preposición
nc0s000	Sustantivo común singular
da0000	Artículo
aq0000	Adjetivo Descriptivo
fc	Coma
np00000	Nombre Propio (sustantivo)
nc0p000	Sustantivo común plural
fp	Punto
rg	Adverbio
cc	Conjunción

Ejercicio 2 | baseline.py

Resultados de dumb tagger (etiqueta todo como nc0s000)

Accuracy: 12.65% / 0.00% / 12.65% (total / known / unk)

g \ m	sp000	nc0s000	da0000	aq0000	fc	nc0p000	rg	np00000	fp	cc
sp000	-	14.39	-	-	-	-	-	-	-	-
nc0s000	-	12.65	-	-	-	-	-	-	-	-
da0000	-	9.70	-	-	-	-	-	-	-	-
aq0000	-	7.28	-	-	-	-	-	-	-	-
fc	-	5.85	-	-	-	-	-	-	-	-
nc0p000	-	5.53	-	-	-	-	-	-	-	-
rg	-	3.73	-	-	-	-	-	-	-	-
np00000	-	3.58	-	-	-	-	-	-	-	-
fp	-	3.55	-	-	-	-	-	-	-	-
cc	-	3.41	-	-	-	-	-	-	-	-

Resultados de baseline (etiqueta con la etiqueta más frecuente)

Accuracy: 87.58% / 95.27% / 18.01% (total / known / unk)

g \ m	sp000	nc0s000	da0000	aq0000	fc	nc0p000	rg	np00000	fp	cc
sp000	14.28	0.05	-	-	-	-	0.01	-	-	-
nc0s000	0.00	12.22	-	0.25	-	0.00	0.03	0.00	-	0.00
da0000	-	0.15	9.54	-	-	-	-	-	-	-
aq0000	0.01	2.05	-	4.84	-	0.13	0.00	-	-	-
fc	-	-	-	-	5.85	-	-	-	-	-
nc0p000	-	1.24	-	0.20	-	4.09	-	-	-	-
rg	0.02	0.31	-	0.04	-	-	3.27	-	-	0.02
np00000	0.00	2.05	-	0.00	-	0.00	-	1.52	-	0.00
fp	-	-	-	-	-	-	-	-	3.55	-
cc	0.00	0.01	-	-	-	-	0.05	0.00	-	3.34

Ejercicio 3 | classifier.py

Linear regression

Training time: 236.8 secs

Eval time: 2.04 secs

Accuracy: 91.69% / 95.01% / 61.68% (total / known / unk)

g \ m	sp000	nc0s000	da0000	aq0000	fc	nc0p000	rg	np00000	fp	cc
sp000	14.26	0.01	-	0.05	-	-	0.01	0.00	-	-
nc0s000	0.00	11.89	0.01	0.52	-	0.02	0.02	0.09	-	-
da0000	-	0.11	9.50	0.01	-	0.01	-	0.01	-	-
aq0000	0.00	0.51	-	6.43	-	0.17	0.01	0.06	-	-
fc	-	-	-	-	5.85	-	-	-	-	-
nc0p000	-	0.35	-	0.46	-	4.61	-	0.04	-	-
rg	0.02	0.04	0.00	0.39	-	0.00	3.11	0.03	-	0.02
np00000	-	0.22	-	0.08	-	0.01	-	3.24	-	0.00
fp	-	-	-	-	-	-	-	-	3.55	-
cc	0.00	0.00	-	0.01	-	0.00	0.05	0.00	-	3.34

Support-vector machine (SVM)

Training time: 66.2 secs

Eval time: 2.05 secs

Accuracy: 94.11% / 97.57% / 62.76% (total / known / unk)

g \ m	sp000	nc0s000	da0000	aq0000	fc	nc0p000	rg	np00000	fp	cc
sp000	14.30	0.00	-	0.03	-	-	0.00	-	-	-
nc0s000	0.00	12.09	0.01	0.33	-	0.02	0.02	0.08	-	0.00
da0000	-	0.08	9.53	-	-	0.00	-	0.00	-	-
aq0000	0.00	0.34	-	6.61	-	0.15	0.01	0.05	-	-
fc	-	-	-	-	5.85	-	-	-	-	-
nc0p000	-	0.25	-	0.31	-	4.90	-	0.04	-	-
rg	0.02	0.02	0.00	0.21	-	0.01	3.35	0.01	-	0.02
np00000	0.00	0.21	-	0.08	-	0.01	-	3.26	-	0.00
fp	-	-	-	-	-	-	-	-	3.55	-
cc	0.00	-	-	0.01	-	-	0.05	0.00	-	3.34

Naive bayes classifier (NBC)

Training time: 6.5 secs

Eval time: 131.28 secs

Accuracy: 84.28% / 88.07% / 49.99% (total / known / unk)

g \ m	sp000	nc0s000	da0000	aq0000	fc	nc0p000	rg	np00000	fp	cc
sp000	14.31	-	0.01	0.00	0.00	0.00	0.00	0.00	-	-
nc0s000	0.05	12.05	0.07	0.09	0.02	0.01	0.00	0.29	-	0.00
da0000	0.00	0.14	9.52	0.00	0.00	0.00	-	0.01	-	-
aq0000	0.22	0.88	0.19	5.11	0.14	0.23	0.01	0.40	0.00	0.01
fc	0.00	-	-	-	5.85	-	-	0.00	-	-
nc0p000	0.06	0.28	0.17	0.07	0.02	4.64	0.00	0.26	-	0.00
rg	0.20	0.17	0.08	0.14	0.04	0.01	2.77	0.23	-	0.02
np00000	0.03	0.27	0.05	0.00	0.00	0.01	0.00	3.21	-	-
fp	0.00	-	-	-	-	-	-	-	3.55	-
cc	0.01	0.00	0.00	-	-	0.00	0.05	0.05	-	3.29

Ejercicio 4

Ver Ejercicio_4.ipynb

Ejericio 5

Añadimos vectores de fastText.

Training time 128.55 secs

Eval time 2.1252095699310303

Accuracy: 94.57% / 97.56% / 67.42% (total / known / unk)

g \ m	sp000	nc0s000	da0000	aq0000	fc	nc0p000	rg	np00000	fp	cc
sp000	14.30	0.01	-	0.02	-	0.00	0.00	-	-	-
nc0s000	0.00	12.14	0.01	0.30	-	0.01	0.02	0.08	-	0.00
da0000	-	0.07	9.53	0.00	-	-	-	0.01	-	-
aq0000	0.01	0.35	-	6.53	-	0.25	0.01	0.05	-	-
fc	-	-	-	-	5.85	-	-	-	-	-
nc0p000	-	0.01	-	0.21	-	5.28	-	0.02	-	-
rg	0.02	0.02	0.00	0.10	-	0.02	3.47	0.00	-	0.02
np00000	0.00	0.20	-	0.06	-	0.02	0.00	3.28	-	0.00
fp	-	-	-	-	-	-	-	-	3.55	-
cc	0.00	0.00	-	0.00	-	0.01	0.05	0.00	-	3.34