=Paper=
{{Paper
|id=Vol-1399/paper8
|storemode=property
|title=Tolstoy Digital: Mining Biographical Data in Literary Heritage Editions
|pdfUrl=https://ceur-ws.org/Vol-1399/paper8.pdf
|volume=Vol-1399
|dblpUrl=https://dblp.org/rec/conf/bd/Bonch-Osmolovskaya15
}}
==Tolstoy Digital: Mining Biographical Data in Literary Heritage Editions==
https://ceur-ws.org/Vol-1399/paper8.pdf
Tolstoy Digital: Mining Biographical Data in Literary Heritage Editions
Anastasia Bonch-Osmolovskaya, Matvey Kolbasov
National Research University Higher School of Economics
20 Myasnitskaya Ulitsa; Moscow, Russia; 101000
abonch@gmail.com, matveykolbasov@yandex.ru
Abstract
This paper presents a solution for mining the biographical information from commentaries on Leo Tolstoy’s letters. It is implemented
as a part of Tolstoy Digital Project – a semantically marked-up web publication of the 90-volume complete collection of Leo
Tolstoy’s works. Extraction of relevant biographical information will be used to create an open database for all the persons who were
somehow connected with Tolstoy or Tolstoy’s works. The paper also accounts for various subtleties of the commentary apparatus and
pays special attention to specific difficulties of biographical information extraction, such as the problem of defining the boundaries of
expressions denoting profession, or the problem of non-standardized syntactic constructions for kinship relations.
Keywords: Leo Tolstoy, commentary apparatus, biographical database, semantic edition
1. Project Description problem of automatic named entity recognition and fact
The Tolstoy Digital project aims to prepare a web-
1 extraction seems to be very well elaborated, see
published semantically marked-up version of the 90- (Grishman 2003, and Jurafsky Martin 2009) for a
volume complete collection of Leo Tolstoy’s works2. The comprehensive outline and reference list. Still,
digital version of the 90-volume edition has become easy information extraction in academic philological texts
to access thanks to the mass crowdsourcing campaign pursues slightly different aims than mining the news flow:
“All of Tolstoy in one click”3 . The next step of Tolstoy there are graphical, linguistic and conceptual problems.
digitization is devoted to the semantic tagging of Specific graphics (such as abbreviations, inner
Tolstoy’s text and the creation of a comprehensive references, font highlighting) may have non-trivial
database of all of the additional reference information that semiotic functions. Some patterns that are uncommon in
goes along with the Tolstoy oeuvres and private archive. general language may be used to render the latent
The 90-volume edition (Tolstoy 1928-1964) comprises an recurring logics of the commentary structure. Finally, the
exhaustive critical apparatus, which contains relevant notions that are relevant for characterizing historical
information on Tolstoy’s works, life, and other people events or persons may have little in common with the
connected with him. Current research, done as part of the conceptual space of data mining in the web (the domain
Tolstoy Digital project, presents an on-going work in fact- which is primarily regarded in all the works on
extraction from literary commentaries. The edition information extraction). This is the reason why the task of
contains 21 volumes of letters, dated from 1844 till 1910, biographical information extraction from academic
the year of Tolstoy’s death. Each letter is followed by a editions cannot be solved with the help of already existing
detailed commentary, where a biographical reference to solutions, but requires some specific process
the addressees and persons mentioned in the letter is modifications. This paper aims to report on the first steps
provided. Our aim is to analyze unstructured text of a that we have made in this direction. We used a rule-based
commentary, to extract person names and relevant toolbox Tomita to write and apply some basic grammar
biographical information, and to use TEI semantic rules that are used to extract relevant ontology concepts.
annotations for relevant mark-ups (Barnard, et al 1995). In part (2) we will present a short outline of the
Extracted data will be aggregated and stored in a biographical ontology we are going to elaborate and we
reference database. The database will be linked with the will briefly describe preliminary data preparation. Part (3)
text of the semantic edition. will be dedicated to the description of our grammar and
the rules for biographical fact extraction. Evaluation and
The main aim of our current experiment was to estimate analysis are presented in part (4).
the effectiveness of a rule-based approach that could be
used for data analysis and fact extraction for our type of 2. Textual material and basic ontology
texts. We started with the most basic concepts and facts: The commentary apparatus of epistolary volumes of the
names, dates, professions and family relations. The complete 90-volume edition consists of letter
commentaries, which constitute about 40% of entire text.
1
See the project description here: Usually they are organized in a non-structured way as a
http://tolstoy.ru/projects/tolstoy-digital/ sequence of factual comments. Some of them are hard to
2
http://tolstoy.ru/creativity/90-volume-collection-of-the- classify, and many of them seem not to fit to any
works/ database, making them redundant. This lack of explicit
3
See, for example, review in the Guardian: structure can be explained by the fact that the
http://www.theguardian.com/books/2013/oct/16/all-leo- commentaries have been created by different authors, and
tolstoy-one-click-project-digitisation
48
each of them had used his own text template. As a result, Text Type of the Current
commentaries are represented as an accompanying text, information relevance of the
but not as an enumeration of properties and parameters of information
a common database. An example of a commentary is
provided in Table 1. Anatoly The head of YES
Ivanovich biographical
Farezov (1852- information
Russian English
1928)
Анатолий Иванович Anatoly Ivanovich writer-populist Main fact YES
Фаресов (1852—1928), Faresov (1852-1928),
публицист-народник. writer-populist. [He was He was tried for Additional NO
[Судился по «делу 193»; tried for “Case 193”; He “Case 193”; He political fact
был амнистирован в was pardoned in 1880 and was pardoned in
1880 г. и перешел в joined the camp of 1880 and joined
лагерь умеренных moderate liberals.] He the camp of
либералов.] worked in the “News,” in moderate
Сотрудничал в the “Week,” and other liberals.
«Новостях», в «Неделе» publications. He had met He worked in the Additional NO
и других изданиях. with Tolstoy February 1, “News,” in the profession fact
Познакомился с 1898 [and left unpublished “Week,” and
Толстым 1 февраля 1898 memories of him]. other
г. [и оставил Faresov’s article titled By publications.
неопубликованные yesterday incident edition
воспоминания о нем]. of “Week”" was published He had met with Tolstoy fact YES
Статья Фаресова под in the "New Times" , № Tolstoy February
заглавием «Ко 7208 of 23 March 1, 1898
вчерашнему and left Additional NO
происшествию редакции unpublished publicistic fact
в «Недели» была memories of
напечатана в «Новом him.
времени», № 7208 от 23
марта Faresov’s Article Faresov’s YES
titled "By publicistic fact
yesterday
Table 1. An example of biographical commentary in the
incident edition
90-volume Tolstoy’s edition.
of “Week”" was
published in the
This example illustrates the complexity of the information
"New Times" ,
in the comments. We have marked with square brackets
№ 7208 of 23
all information that cannot be structurally formalized.
March
That means that though we have at our disposable a
considerable corpus of short biographical texts, but it is
still a collection of unstructured texts with high degree of Table 2. The choice of facts and relationships to be
lexical variation and a mixture of relevant and excessive extracted from the commentaries.
factual statements. We aim to transfer an unstructured
commentary text into a structured database with given The corpus of letters, used for rule development and
semantic relationships between the elements. For the first testing, was based on the 63rd volume of the Complete
stage of our project, we limit ourselves to a small list of edition, which contains 281 letters by Tolstoy of the
relationships. Table 2 shows the relevant facts, which period of 1880-1886. We have created a manual markup
have been filtered out of the text, given in table 1. of 50 letters using the annotation module from GATE
framework (Cunningham, et al 2002). We annotated all
professions, kinship terms and relationships, and birth and
death dates. The observations made upon this small
annotated corpus were used to develop grammatical rules
for biographical fact extraction. For example, we have
found that the mention of a person, who has not been
commented on before in the text, corresponds to a certain
robust text structure. We called it a pattern of first
mention. It consists of a personal name, dates of life, and
additional information, which may be a profession or
kinship references to other persons (first of all, to
49
Tolstoy). Finally we have built a small ontology, presented in Figure 1. The text spans referring to
comprising the main entities and their attributes and annotation are printed in bold.
matched in with our annotated sentences. The ontology is
RUS: Николай Николаевич Страхов (1828—1896) — критик и философ.
ENG: Nikolay Nikolayevich Strahov (1828—1896) — critic and philosopher.
RUS: Анастасия Васильевна Дмоховская, урожд. Воронец. О ней см. в т. 49
ENG: Anastasia Vasil’yevna Dmohovskaya, maiden name Voronec. See vol.49.
RUS: Евдокия Александровна Новосильцева (р. 1861 г.), в замужестве Регекампф.
ENG: Evdokiya Alexandrovna Novosil’ceva (born 1861), marriage name Regekampf.
RUS: Владимир Иванович Даль ( 1801 — 1872 ) — известный лексикограф и этнограф.
ENG: Vladimir Ivanovich Dal’ ( 1801 — 1872 ) — famous lexicographer and ethnographer.
RUS: Михаил Александрович Энгельгардт (р. 1861 г. — ум. 21 июля 1915 г.)
ENG: Mihail Alexandrovich Engelgardt (born 1861 — die 21 July 1915)
RUS: Гр. Сергей Николаевич Толстой (1826—1904) — старший брат Льва Николаевича.
ENG: Earl Sergey Nikolayevich Tolstoy (1826—1904) — the elder brother of Leo Tolstoy.
RUS: Иван Васильевич Сютаев (р. 1856), крестьянин дер. Шевелино
ENG: Ivan Vasil’evich Sutaev (born 1856), peasant of Shevelino village
RUS: Афанасий Афанасьевич Фет (Шеншин) (1820—1892) — поэт.
ENG: Afanasiy Afanas’evich Fet (Shenshin) (1820—1892) — poet.
RUS: Михаил Матвеевич Стасюлевич (1826—1911) — общественный деятель, историк и
публицист, с 1865 года редактор-издатель журнала «Вестник Европы».
ENG: Mihail Matveevich Stasulevich (1826—1911) — social activist, historian and publicist, since 1865
editor and publisher of the magazine "Herald of Europe".
RUS: Лев Львович (р. в 1869 г.), третий сын Толстого.
ENG: Lev Lvovich (b. 1869), the third son of Tolstoy.
< friend of: name>
RUS: Дмитрий Алексеевич Дьяков (1823—1891) — сын Алексея Николаевича и Ирины Дмитриевны,
рожд. Полторацкой, друг Толстого.
ENG: Dmitri Alekseevich Dyakov (1823—1891) — the son of Aleksey Nikolayevich and Irina Dmitrievna, maiden
Figure 1. The ontology of biographical facts and attributes in commentaries to Tolstoy’s letters.
For the first stage of our research, we decided to consider Yandex.ru4 and is available for free download. Tomita
the most basic categories, such as name, dates of life, provides modules of morphological analysis, as well as
profession, and kinship. We have created several ready-made rules for extracting names and numbers.
grammatical rules and have evaluated those rules with the Tomita grammars consist of rules. The user may create
help of our annotated commentary corpora. his or her own grammars and dictionaries for a certain
language. To construct a grammar, the user should write a
3. Methods number of transducer rules, for which one can use regular
We used Tomita parser to create rules for fact extraction expressions, words lists, and other rules, built-in or
(Tomita 1984, 1985). Tomita parser is a free NLP composed. Each rule has a left and a right side; the
platform customized for creating small and light transducer operation is denoted by arrow, separating left
information extraction modules. It can be used as a tool and right sides. The left side can be only represented by a
for extracting structured data (facts) from texts by terminal, while the right side can be both terminals or
context-free grammars and dictionaries of keywords. The nonterminals. An example of a rule sequence can be
API for Tomita parser has been developed by the team of found in Figures 2-3.
4
https://tech.yandex.ru/tomita/
50
INPUT
RUS:
Figure 2. Rule for date extraction. Александр Андреевич Иванов (Шешин)
(1806 — 1858) — художник .
This rule says that a date is a number, consisting of four ENG:
characters from 0 to 9. Then, we can define date chains Alexandr Andreevich Ivanov (Sheshin)
(for example dates of birth and death) using previously set (1806 — 1858) — а painter.
rules. The operator interp in Figure 3 assigns the extracted ► PATTERN RECOGNITION
text span to a specific fact. Here Dt1 stands for the birth
date, while Dt2 corresponds to the date of death : ► KINSHIP DICTIONARY CHECK
► CHOICE OF RULE SET
Figure 3. Rule for date chains extraction. ► FACT EXTRACTION
OUTPUT
Using Tomita parser we have written a grammar with
specific rules and dictionaries aimed to extract the
following biographic facts from letter commentaries: RUS: Александр
ProfessionFact (person’s name, dates of life and ENG: Aleksandr
professions) and FamilyFact (kinship). We quickly
realized that the built-in date-extraction and name-
extraction rules are not well-sutied to philological RUS: Андреевич
commentaries, so we also made some specific ENG: Andreevich
modifications. For description of kinship relations of a
person, we have created several dictionaries containing
different kinship types. The kinship dictionary is applied RUS: Иванов
after the pattern is recognized. If no kinship term is ENG: Ivanov
detected in the pattern, then the rules that extract
profession are applied. Overall there have been created 31
rules and 3 dictionaries. The processing scheme is RUS: Shenshin
presented by Figure 4. ENG: Шеньшин
1806
4. Results and analysis 1858
The performance rank of these rules has been measured
with the help of testing commentary corpus of 560,000
RUS: художник
tokens. The results are considerably acceptable for
precision, but are much lower for recall, which is quite ENG: painter
typical for rule-based approaches. The overall score for
both facts is demonstrated in Table 3
Figure 4. Extracting information from text patterns.
Fact name Precision Recall F-measure
ProfessionFact 1 0,82 0,9 Another problem is the boundaries of the text span which
FamilyFact 0,78 0,3 0,43 refers to the profession itself. Thus, in example 2, we see
a description of Shamil’s social activities. So the question
Table 3. Evaluation of biographical fact extraction. is what part of the complicated NP should be extracted as
Shamil’s profession. Should it be the leader and
The analysis of results reveals that some important consolidator or the full NP (leader and consolidator of
problems lie in the conceptual part. First of all, it is hillmen of Daghestan and Chechnya)?
problematic to define a strict classification of what may
(or may not) be considered as a profession. Thus in 1) Василий Кириллович Сютаев (1819—1892) — ...
example 1 the profession extracted is a peasant. But in крестьянин дер. Шевелино.
context of biographical commentaries this person’s origin
(village Shevelevo) becomes even more important than Vasiliy Kirillovich Syutaev (1819—1892) — ... peasant of
the profession of peasant itself. It’s obvious that Tolstoy village Shevelino.
had met a lot of peasants during his life, so any special
2) Шамиль (1797—1874) — знаменитый вождь и
attributes are of much value in the context. There remains
объединитель горцев Дагестана и Чечни в их борьбе
a problem of distinguishing those cases, which are
с русскими.
meaningless without additional information.
51
Shamil (1797-1874) – a popular leader and consolidator extract them using algorithms of machine learning. In
of hillmen of Daghestan and Chechnya in their struggle general, during the next stage of the research we intend to
with Russians. develop the ontology (i.e. add important locations and
relations between person and location), to process all the
To resume, if such types of professions as writer, 31 volumes of letters, to make an open database with all
musician, and philosopher are good categorizers, that the persons connected to Tolstoy, and to provide every
allow to define a group of people in one way or another input referring to a person with a short semantically
connected with Tolstoy or his work, such status as a marked-up biography. The database is to be used as an
peasant or consolidator are meaningless without their interlinked reference base to Leo Tolstoy’s 90 volume
attributes (genitive groups), in this case – “village edition, and also as an aggregator of exterior information
Shevelino” and “hillmen of Daghestan and Chechnya in from other sources. The specific syntactic constructions,
their struggle with Russians.” Accordingly, the that are intrinsic for the commentary genre (such as
significant problem is with the boundary of the nominal abundance of names and titles in a sentence, special
group that determines the professional status. The second means of reference inside text, etc.) will be certainly
problem is concerned with intricate chains of kinships taken into account.
that have a very specific syntax, as shown in Example 3.
In this phrase there is an inversion (see Figure 5), References
probably made especially for publication, so that Sophia
Tolstaya, wife of Leo Tolstoy, would take a more Barnard, David T., et al. “Hierarchical encoding of text:
prominent position. Technical problems and SGML solutions.” Text
Encoding Initiative. Springer Netherlands, 1995. 211-
3) Александр Михайлович Кузминский ... муж сестры
231.Computers and the Humanities 29.3 (1995): 211-
Софьи Андреевны, Татьяны Андреевны (1846—
231.
1925).
Cunningham, Hamish, et al. “GATE: an architecture for
Alexandr Mihaylovich Kuzminskiy ... the husband of the
sister of Sofiya Andreevna, Tatiana Andreevna (1846— development of robust HLT applications.” Proceedings
1925). of the 40th annual meeting on association for
computational linguistics. Association for
Computational Linguistics, 2002.
Grishman, Ralph. ”Information extraction: Techniques
and challenges.” Information extraction a
the husband of the sister of SA, TA multidisciplinary approach to an emerging information
technology. Springer Berlin Heidelberg, 1997. 10-27.
Figure 5. Complicated syntactic structures in kinship Tolstoy, Lev. Polnoe sobranie sochinenij v 90 tomah,
patterns. Moskva, 1928-1954. (Tolstoy, Leo. Complete collected
works in 90 volumes, Moscow 1928-1954)
Another problem is the discrepancy between singular and
plural forms when it comes to the descriptions of relations Tomita, Masaru. LR parsers for natural languages.
between a person and a family group, as shown in COLING. 10th International Conference on
Example 4. Computational Linguistics. 1984. P. 354—357.
4) Николай Михайлович Нагорнов (1845—1896) — сын Tomita, Masaru. An efficient context-free parsing
Михаила Михайловича и Надежды Ивановны algorithm for natural languages. IJCAI. International
Нагорновых. Joint Conference on Artificial Intelligence. 1985. P.
756—764.
Nikolay Mihaylovich Nagornov (1845—1896) — the son
of Mihail Mihaylovich and Nadejda Ivanovna
Nagornovs. Acknowledgements
This study (research grant No 15-06-99523 А) was sup-
Unlike profession extraction which shows good quality
ported by the Russian Foundation for Basic Research in
by rule-based approach, kinship relation patterns seem to
2014-2015.
be much less regular, so perhaps it is worth trying to
52