Logical-linguistic model for multilingual Open Information Extraction

Figures & data

Table 1. The predicate variables and their values range defined in the Open IE model for the English, Russian, and Kazakh languages

vari-ables	features	values
English language
z	a certain preposition after a verb in phrases	{to, by, with, about, of, on, at, in, between, for, from, over, out}
x	the order of words in the phrase	f, l, kos
y	the usage of an apostrophe at the end of the analyzed word	ap, aps, out
m	any form of the verb “to be” in the phrase	{is, are, have been, has been, had been, was, were, am, out}
f	any form of modality in the phrase	{can, may, must, should, could, need, might, would, out}
n	the negation in the phrase	not—negative phrase, out—affirmative sentence
p	a form of the main verb in the English phrase	I, II, III, ing end ed—four the base forms of the verbs and irregular verbs
Russian language
z	six grammatical cases	{nom—nominative, gen—genitive, dat—dative, acc—accusative, ins—instrumental, loc—prepositional}
x	animacy	anim—animate noun, inan—inanimate noun
y	noun semantic characteristics	{device, hum, tool, pc:hum, space, time:moment, time:period, s:loc}
Kazakh language
x	the location of the analyzed word in the phrase	{-1, −2, −3, 0, 1, 2, 3}
f	the feature of an auxiliary verb in the phrase	aux shows the existence of any of 35 auxiliary verbs of the Kazakh language in the analyzed phrase, 0
z	the grammatical case of the Kazakh noun	Nom—nominative, Gen—genitive, Dat—dative, Acc—accusative, Ela—local, Ins—instrumental, Abl—ablative
a	the types of the Kazakh nouns declensions	NSim, NPos
n	the feature of the negative sentence	me, eme, joq, 0
c	the feature of plural suffixes	tar, ter, dar, der, lar, ler
y	the derivational suffixes for verbs, nouns, participles, adverbials	UnFu, FuCo, Psuf, Usuf, NoN, NoV, Ncom, Nder, Part, ParP, VaP, Oad, Vad, Vpas, y, 0
d	the subjunctive action of the analyzed verb	shi, 0
m	a personal predicative or possessive flexion of the analyzed verb and verbal forms	PrFl, PoF, 0
b	the supplementary semantics of the analyzed action	mic, se, 0

Figure 1. Example of the fact identification from English sentences. The predicate γ_1E defines grammatical features of the Subject, the predicate γ_2E defines grammatical features of the Object and γ_4E the predicate defines grammatical features of the location attribute of the fact

Figure 2. Example of the fact identification from the Kazakh phrase. The predicate γ₁ defines grammatical features of the Doer, the predicate γ₂ defines grammatical features of the Object and γ_VK is the predicate of the Action (the Predicate) of the fact

Figure 3. Graphical representation of Universal Dependencies for the sentence “He referred to the deaths as “the cost of doing business on that particular engagement.” Source: DependenSee

Table 2. Fragment of the table with facts extracted from the English corpus of texts

Num. of sent.	Predicate verb	Actor nsubj	Object
			Advcl	Dobj	Ccomp (object)	Xcomp
1	Consisted	War	Fighting
2	Lasted, took	War, majority		Place
3	Focused	Insurgents	Featured, ambushing
4	Featured	Fighting		Warfare
5	Killed	Iraqis				Many
6	Saw	Anbar		Fighting
7	Relinquished	Army		Command
8	Struggled	Sides				Secure
9	Secure			Valley
10	Escalated	Violence	Struggled
11	Became, turned	Qaeda		Capital		Group
12	Issued	Corps	Declaring	Report
13	Began	Tribes				Form
14	Turn			Tide
15	Maintained	Forces		Role
16	Flew	Bush	Celebrating, congratulate
17	Known	Fallujah	Secular
18	Held	System		Influence
19	Favored	Conditions		Insurgency
20	Lost	That		Power
21	Claim				Defeated

Table 3. Precision and agreement of the developed logical-linguistic model for the corpora of three different languages

The language of corpus	The size of corpus	Precision of the model	Alignment
English	200 000	87,2 %	0,91
Russian	275 000	82,4 %	0,78
Kazakh	225 000	71,0 %	0,72