[PDF] Problems of Disambiguation of Prepositional Phrases

View - Download Problems of Disambiguation of Prepositional Phrases

Previous PDF

Next PDF

Problems of Disambiguation of Prepositional Phrases K irill Boyarskya,b, Eugeny Kanevskyb, and Anastasia Kozlovaa a IT MO University, Kronverkskiy Ave, 49-A, St. Petersburg, 197101, Russia

b Institute of Regional Economics Problems RAS, Serpukhovskaya St, 38, St. Petersburg, 190013, Russia

Abstract

This pape r describes t he features that app ear in parsing procession of multiwor d turns (phrasemes) able to act as prepos itions. These fe atures are considered in the c ontext of automatic analysis of Russian texts. Suc h ph rases have a fair ly high homonymy, which creates some difficulties in analysis and defining semantics and, consequently, reduces the accuracy of pars ing. M ore than 320 phrasemes have been classif ie d on the basis of the assumed homonymy types. In the course of the study, the phrasemes have been divided into three groups. The first group includes those phrasemes that can definitely be called prepositions, but potentially have some semantic ambiguity. The second group combines phrasemes that ar e characterized by the part-of-speech homonymy of prepos ition/adverb. The th ird group is characterized by phrasemes that determine the construction of two or three parsing options. The occurrence of multivariate parsing is based on the presence of one or two phrases related to different parts of speech, and a simple conjunction of a preposition with a noun. Within each group, lists of the most common phrasemes have been composed (according to the NCRL), indicating the probability that a certain phraseme may serve as a preposition. The paper also defines the basis on which the compilation of effectively removing homonymy rules for the SemSin parser may rely on. The examples provided in this paper prove that it is necessary to consider not only the direct encirclement of the phraseme, but also its remote context to remove homonymy.

Keywords 1

a u tomatic text analysis, disambiguation, homonymy, idiomaticity, prepositional phrases

1.I ntroduction

In the process of automatic parsing of the Russian language sentences and building a dependency tree, there is an arising problem of removing homonymy of various types - morphological, lexical, part-of-speech, etc. One of the ways to solve this problem is the broad use of standard combinations of wo rds - phrasemes. This term refers to a wide range of expressions with a varying degree of idiomaticity [1]. The common feature for phr asemes is that the value of the whol e is not a

composition of the values of the constituent parts. In general, the words that are part of phrasemes can

change, however within the scope of this study we are interested in invariable phrasemes, most of which are turns of speech that perform the functions of: boundlessly'), без устали ('tirelessly'), ...; ('for lack of' or 'failing'), на пути к ('on the way to'), ...; ('you see'), ...;

what'), разве только ('unless'), ...;IMS 2021 - International Conference "Internet and Modern Society", June 24-26, 2021, St. Petersburg, Russia

E MAIL: boyarin9@yandex.ru (A. 1); eak300@mail.ru (A. 2); stasia.kozlova@gmail.com (A. 3) ORCID: 0000-0002-0306-8276 (A. 1); 0000-0002-1498-4632 (A. 2)

© 2021 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).

CEUR Workshop Proceedings (CEUR-WS.org)

('nearly'), ...;

i predicative turns of speech - пруд пруди ('a dime a dozen'), раз плюнуть ('not a big deal').

The most complete lists of turns of speech are given in the NCRL (National Corpora of Russian Language) [2]. The dictionaries of Kuznetsov [3] and Rogozhnikova [4] have also been used. Currently, close attention is drawn to the semantics of prepositional groups, including those where more-than-one-word combinations act as a preposition [5]. As even a preliminary analysis shows, most of these phrasemes do not have homonymy and are always prepositional turns. However, it is possible that the same combination of several words can correspond to two different preposition or an adverb, depending on the context of the word on the right: a word in the genitive case, verb, or punctuation mark: 2 i iii"ii i!i!E$.

нотариально оформленное согласие... ('If the child goes unaccompanied, he must have with him,

in addition to the passport, a notarized consent'). A more complex situation arises in the event that a combination of several words, depending on the context, may or ma y not be a turn. Many wo rd combinations of this kind are considered by Rogozhnikova [4], who notes the possibility of their use as free phrases that are homonymous to

turns. So, for example, a phraseme с целью ('for the purpose of') can either perform the functions of a

preposition or remain a free word combination, depending on the presence or absence of a word on the right in the genitive case:

массовых пытках и заговоре с ц елью пыток ('Spain has dem anded extra dition, charging

i i!i i i! iii iiE$i eif i=@A IJK:i@:>B?KGBiJi <:CPaiBK@C:TPiG=J:C: GH:i BKiVIOiBigC:9 :iG=JIhGBOi9SCBi iiiii iii !iiiiiiii iii i !i 1!!i" iE$i ii!i!%i i i i"ii !i !ii i ii i ii!ii i iiiii ii%8i

eiМы ехали на концерт и по пути притормозили на Садовой, у дома Булгакова ('We were

ii"iiiii iiiiil i E$i ii i iii i i# ii i $i iii iii!i i i iii iiiiii iii i"iiii i"i ii"i iii i ii ii iiii i !i i i!i3ii%!i ii iin ,(ni i5Z7i+ii%i ii ii iii i! iii i"i3ii%!i ii ii``i i 5\7ii i iiiiii i i i ,]i5[7i !i ii ii i! i#!iii i$iii ii ii!i i5W7i"i i i i i iii i%!i',,,"i ii!iii ii56*7i +i iiiii!i i i i ii"i ,i% i !i" ii i ii i i i]i

8ii ii

!ii i iii %i iii i i i ,i%i ii ! i i"i i!ii i i iiiiiiiiiiiiiiiiii iiiiiiiiiiiiiiiiii iiii iiiiiiiiiii

2 Here and further on, all the examples are taken from the NCRL and are separated by a "" sign, and Russian-language

phrasemes that are turns of speech are highlighted in bold in the examples. The words that allow to make a particular

decision are underlined in Russian-language examples. IMS-2021. International Conference “Internet and Modern Society"99

tokens (words, phrases, punctuation marks, numbers, etc.). The token chain is then processed in the lexical analyzer using a system of production rules, the purpose of which is to transform the linear sequence of tokens into a dependency tree. The principles of building the parser dictionary are based on the ideas of Tuzov [11] The main table of the dictionary contains more than 195 thousand lexemes distributed over 1700 classes [12].

Each lexeme has morphological characteristics, as well as the number of its semantic class and actants

or va lences (for connecti ng dependent w ords) in the fo rm of cases (! Nom, !Gen, !Acc , etc. ) or

prepositions, possibly wi th the corresponding cases (!Wi thout, ! For, !inAcc, !onPrep, e tc.). Free

actants are als o used, w hich define m ore generali zed concepts (!Question , !Where, ! How,

!Fromwhere, !Why, etc. ). Often, before such an a ctant, the acc eptable classe s of wor ds tha t can

replace them are indicated. The presence of a classifier can significantly reduce ambiguity and is especially widely used when connecting a djectives and prepos itions. About 14% o f words in the dictionary have two or more lexemes.

In addition to the main table, there are auxiliary tables that provide the execution of tasks that are

of interest in this work. This is a table of word combinations (more than 5350 lines), containing stable

combinations of words w ith dif ferent types of inflect ion. These can be c ollocations (вид на

, ' residence permits'), names of organizations ( Чейз Манх еттен Банк, ' Chase

of the words can be used in different word forms. In thi s paper, w e are inte rested in immut able phrasemes that form compou nd prepositions,

adverbs, etc. If the parser decides that a certain phrase is such a phraseme, then the words included in

it are combined into a single token. The second auxiliary table is a table of prepositions (more than 2460 lines) with the cases and

semantic classes of the connected nouns. If the connection of the preposition with the dependent word

is syn tactic in nature and , as a rule, coincides with the case of t he dependent w ord, then the

connection of the prepositional group to the main word reflects the semantics more fully (Where,

When, Why, etc.).

2. Group 1. Prepositional phrases without lexical homonymy

Passing on to the analysis of prepositional phrases, we note that the largest of the three groups is the fi rst one, which contain s turns of s peech that have a lmost no homonyms and are unambiguous

1.The analysis of these prepositions does not differ from the analysis of ordinary one-

word prepositions. The group consists of two subgroups: the phrasemes of the first end with nouns (1A), the second - with prepositions (1B).

2.1. Subgroup 1A

This subgro up includes un ambiguous prepositi onal phrases, whose p hrasemes consist of two words: a preposition and a noun. There are more than 60 such phrases in our dictionary. The vast majority of them require the genitive case after them.

G5;M272;@<<в присутствии всего коллектива Центра ('At the end of their stay, students defend

i i iii iii i iii i ii1E$i 4i i&i ii i i !8iв прот ивовес ('in contra st to'), в на радость ('to smb's joy').

Example: Отдавать все сил ы организации вы боров в у щерб профессиональной

('Give all the effort to organize elections to the detriment of professional activity'). i i i i iiiiii!i"i"iiii c"ii i i i i"i ! i!! iii"i i i iii!i"iii ii i"i iii!i"i# i i i i iiii ii i $ii "i i ii i8iв честь ('in 100PART 1: Computational Linguistics

honour of') (What for, Why), из числа ('from the number of') (From, Which), на основе ('on the

basis of') (How, Which), по поводу ('concerning') (Dat, Why). The semantics of prepositional relations is examined in detail in the dictionary of Zolotova [13],

but there is a lack of formal rules that allow to correlate mainly syntactic relations developed by the

parser with the semantics of [13]. This is a rather complex task that is still under consideration in

some special cases [14]. Below are the examples of two prepositional phrases, and the connection with the main word of prepositional phrase is presented in parentheses in terms of the SemSin parser and in the semantic connections of Zolotova.

В пре ссе отмечалос ь, что это был салю т в 101 зал п вчесть возникшего в России

('In the press it was noted that it had been a salute of 101 volleys in honor of the i ii ii

E$i3#9SCi3is=T:Oi#tHGHKH@$i3i@iT:JKP$i#/i# $$i

честь приезда английского принца Чарльза ('And suddenly Oleg remembered how he once was at

iiiiii iii i ii1

E$i3i#IJK>B:GGBO - Почему

(каузатив) - в честь) (Why (causative)).

853; magnate is excluded fr om the number of co-chairs of the Republic of Latvia, the party will face

financial difficulties, influential experts believe') - (исключён - Изо (финитивно-фазисное) - из

числа) (From (phase-finitive)).

('Usually, while the ceremony of awarding the winners took place above, the execution of traitors,

cowards, losers from among the subjects of the Prince-elector was carried out below') (неудачников-

Table 1 shows the most common prepositional phrases of this subgroup that require the genitive

case after them. A questionof the validity of this table arises. Obviously, expert evaluation is very

difficult in this c ase be cause of the necessit y to view too m any sen tences . For example, for a

phraseme в глубь ('into the depth'), it would be necessary to analyse more than 27 hundred sentences

in order to identify about 70 cases of absence of a word in the geniti ve case to the right of the

phraseme. It is very likely that if we choose 300-500 sentences in any way, there will be no cases of

absence of the genitive case on the right. Therefore, such method of evaluation has been chosen. With the usage of the capabilities of the

NCRL, sentences in which there is a punctuation mark after the studied phraseme в глубь ('into the

depth') have been selected.

лучи, то одним цельным блиставшим столбом падал в глубь, перерезав всю бухту ('And the

bright moon, which now silvered the whole sea, scattering its rays on the faint ripples, then fell in one

solid shining column into the depths, cutting the entire bay'). It is obvious that in all these sentences (171 units) this phraseme does not serve as a preposition,

but is simply a combination of a noun with a preposition. Next, sentences, in which there is a verb in

the indicative or imperative mood, an infinitive or an adverbial participle after the studied phraseme,

have been selected. This set of 57 sentences requires expert analysis, since after this phraseme there

are such homonymous wo rds as души ('souls' vs 'stra ngle'), з аросли ('thickets' vs 'overgrow') ,

'sushi' vs 'dry'), etc. i iiE$i +iiZi ii !i i !ii!i iiiii i i !i i i i#ZbxZ$i! iii ii ii i !i i ii i IMS-2021. International Conference “Internet and Modern Society"101

Table 1.

Prepositions that require the genitive case

./012 $34% !5"# 6*, +7*)-

89/2:0

+7*<-

8=>0?>?

?2 :%MNO

8DMNO3

!5"# 7*+ +6*;- 9 =PK $DF4% !5"# <;<*7 9 PK :%MNO /K >9: QCK :%MNO I*6 :Q 8=/R G $34% !5"# +6*A- :Q >B? S <7*( ++*I- B>/Q TD $%&DU !5"# ;;*6quotesdbs_dbs14.pdfusesText_20

[PDF] prepositions and conjunctions worksheets

[PDF] prepositions cae exercises pdf

[PDF] prepositions examples in sentences

[PDF] prepositions exercises pdf advanced level

[PDF] prepositions exercises with answers pdf grade 4

[PDF] prepositions of direction exercises pdf

[PDF] prepositions of place exercises pdf advanced

[PDF] preposizioni di luogo esercizi italiano

[PDF] prerequisite and test score error

[PDF] prerequisite for node js

[PDF] prerequisite to learn json

[PDF] preschool alphabet worksheets pdf

[PDF] preschool hand washing curriculum

[PDF] preschool music standards

[PDF] preschool songs with lyrics