On Building a Full-Text Digital Library of Land Deeds of Taiwan

In this paper we present a full-text digital library of
Taiwanese land deeds. Land deeds were the only
proof of land activities such as transaction of ownership
and leasing in Taiwan before 1900. They form a major
part of the primary documents at the grassroot level in
pre-1900 Taiwan, and are extremely valuable for studying
the evolution of the Taiwanese society.
Land deeds, on the other hand, are difficult to study because
they are hand-written and hard to read. Furthermore,
they are scattered in many different locations and,
in some cases, in the hands of families and private collectors.
In order to make the land deeds more accessible to researchers
and educators, the Council for Cultural Affairs
of Taiwan embarked on a major effort to organize available
land deeds and typed them as machine readable fulltext.
Based on this collection and collections from other
sources, the National Taiwan University built a full-text
digital library of Taiwanese land deeds. The current size
of the collection is over 23,000 which, according to one
estimation, cover about 50% of all existing land deeds.1
The collection will be expanded to around 30,000 by the
end of the year.
Our digital library is built with the goal of providing an
electronic research environment for historians to conduct
research using land deeds. Thus in addition to providing
full-text search and retrieval, we developed a concept of
regarding the query return as a sub-collection and built
tools to help the user find meaning and relationships
at the collection level. Post-processing presentation,
term frequency analysis and co-occurrence, and relation
graphs are some of the tools described in this paper. We
believe that our digital library will bring Taiwanese historical
research using land deeds to a different horizon.
Land Deeds of Taiwan
Before Taiwan was ceded to Japan by the Chinese Qing
Dynasty after the Sino-Japanese War of 1895, land deeds
were the only proof of ownership and transaction of land
in Taiwan. Land deeds were, thus, a centrally important
primary source for studying the 300 years written
history of pre-1900 Taiwanese society (Wu, Ang, Lee,
Lin, 2004). Even after the modernization of land administration
by the Japanese, many families still kept the
old land deeds either as part of the family heritage or
due to the mistrust of the government. Many, however,
were destroyed or discarded since they lost their original
significance. During the early stage of the Japanese
occupation, the government conducted research on the
old administrative systems and customs and produced
three series of books totalling 40 volumes, many of
which contained transcriptions of land deeds as examples.
During the reform of the land administration, the
Japanese government sent surveyors to systematically
transcribe land deeds so that they can be convert into
modern land administrative records. In the latter endeavor,
about 16,000 were collected. Their transcribed
versions (copied verbatim by hand) are scattered in the
13,855 volumes of the Archives of the Japanese Taiwan
Governor-Generals (Wang 1993). After Taiwan was returned
to Chinese rule in 1945 after the 2nd World War,
some research institutes and researchers recognized the
importance of the land deeds and made efforts to collect
them. The most notable, and largest scale such effort
was conducted between 1976 and 1983 by a team
lead by the historian Wang Shih-Ching who, commissioned
by the Committee for Taiwan Historical Studies,
Association for Asian Studies, U.S.A., collected about
5,600 land deeds and published a six volume catalog
Taiwanese Historical Documents in Private Holdings
(Wang, 1977). The photocopies of the land deeds were
bound into more than 100 volumes. Other notable collections
were kept at the National Taiwan University, the
Institute of Taiwanese History of the Academia Sinica,
the National Taiwan Library, and by various private collectors.
Scores of books containing the images of some
land deeds have appeared in the past ten years. Wang
estimated (1993, pp. 71) that there are 20,000 land deeds
in the hands of private collectors, libraries, museums and
research institutes that were not included in the official
collections. That makes the total number of such land
deeds about 40,000. Our experience in the past ten years
of digitization tells us there should be more, although we
cannot give a reasonable estimation. Land deeds of Taiwan are contracts about various actions
involving lands, such as the commission by the
government to cultivate previously un-owned land, the
division of family properties, the transaction of ownership,
the rental of farming right, the pawning of land, etc.
They were hand written and were usually prepared by a
scrivener. A land deed usually consists of the following
elements:
• The type of the land deed: selling, renting, pawning,
etc.
• the “seller” (or owner) of the land,
• the “buyer” (or lender) of the land,
• the location of the land and its boundaries (usually
marked on all four directions using neighboring
landmarks such as river, road, building, pond, or
even trees),
• the cost: money, maybe accompanied by other
properties such as houses, cows, storage sheds, and
farming tools,
• the names of witnesses and the scrivener, and
• the date. While each land deed may have significance only to its
owner, the collection as a whole provides a fascinating
glimpse into the pre-1900 Taiwanese grassroots society.
Through these land deeds, one can study the development
of a region, or the overall land management, society,
economy, and law of pre-1900 Taiwan. Furthermore,
since many of the deeds were contracts between indigenous
people of Taiwan and the Han immigrants from
China, they also provide clues to the intricate relationship
among the various peoples of Taiwan (Hong, 2002),
the transition of rights to land, and the gradual assimilation
of the indigenous people (in particular the Pinpu 平
埔族群) into the Han society. Our Collections of Land Deeds of Taiwan
In 2003 and 2004, the Council for Cultural Affairs (CCA)
of Taiwan commissioned the National Taichung Library
(NTL) and Professor Lee Wen-Liang of NTU to collect
and digitize (in full text) the hand-written copies of land
deeds from the Archives of the Japanese Taiwan Governor-
Generals. In this project, NTL keyed-in the full text
of 15,899 land deeds from the Archives of the Japanese
Taiwan Governor-Generals. In the meantime, National
Taiwan University (NTU) and Taichung County Cultural
Center (TCCC) also digitized their own collections of
land deeds, most notably the Anli Dashe Archive. Together,
NTL, NTU, and TCCC have collected more than
23,000 land deeds in Taiwan, all incorporated into Taiwan
History Digital Library (THDL), a full-text digital serve as a research environment for researchers in Taiwanese
history and other disciplines. All of the deeds
are available in searchable full text, with metadata and,
in some cases, images.
The building of content is an on-going effort. We project
that the size of our collection will reach 30,000 by the
end of the year.
A Research Environment for Land Deeds
We have incorporated the above-mentioned collections
of Taiwanese land deeds into Taiwan History Digital Library
(THDL) (Chen, Hsiang, Tu, and Wu, 2007), which
is built with the goal of providing an electronic research
environment for historians. Since our primary goal is to
build a digital library to be used by researchers, we spent
a great amount of time interacting with historians and
built tools that they would find useful in their research.
Full-text search is, in our view, the most basic facility.
However, what is more important is how to help the user
analyze the query results once they are retrieved.
We developed a methodology that treats query returns as
a sub-collection, instead of as individual (and independent)
documents. This seems to reflect better the need of
researchers, who usually look for significance emerged
from a set of land deeds. Under this philosophy, we have
built extensive post-query classification facilities, which
classify and present the query results according to attributes
such as year, type of deeds, origin, etc. We also
provide term frequency analysis which, using the 50,000
terms (names and locations, mostly extracted automatically)
appeared in the collection, analyzes relationships
such as geographic locations, co-occurrences, people involved.
The co-occurrence and temporal relationships are further
analyzed in the line chart of temporal distribution facility
provided, which gives a visual representation that
makes observation easier. As mentioned before, each
land deed features a list of attributes. These attributes
can, in principle at least, be extracted from a deed. This
work is quite laborious and is still under way. But we
have developed an XML format that captures the attributes
and, more importantly, makes it easy to build relation
graphs that show the relationship among land deeds.
Our preliminary experiments show that these graphs can
play a significant role in the study of land deeds.
In the following, we present the aforementioned features
in more detail. Historians usually do not look at a single document but,
rather, a group of documents and try to find significance
through their relationship. For example, land deeds
from the same region as a whole may reveal the gradual
change of land ownership from one ethnic group to another
which, obvious, cannot be observed from a single
document. For this purpose, we developed a concept that
regards the query returns as a sub-collection and built
tools to help the user find meaning and relationships
at the collection level. This is done in THDL mainly
through post-processing a query’s returns, presenting
and analyzing them as a whole. Figure 2 is an example of how query results are presented
in THDL. After the query results (the sub-collection)
are returned, THDL classifies the resulting land deeds
according to three predefined dimensions (year, origin,
and type) on the left of the web page, while presenting
summaries of the land deeds themselves on the right.
Each class is followed by the size of the class (Fig. 3).
By representing post-query classification, the historian
can observe the distribution and behavior of the subcollection,
and see if there is anything that contradicts
what the historian predicts. At a first glimpse, the historian
can quickly capture the outline of the sub-collection.
It’s helpful especially when the query results of full-text
search is too large to manage. Furthermore, the postquery
classification can also be used as a faceted search:
simply click on a class will refine the user’s query.
Note that the three dimensions are chosen because they
are important characteristics of land deeds. A different
corpus could define a completely different set of dimensions
to reflect the characteristics of the content. The post-query classification on year reviews the temporal
distribution of a sub-collection. To better visualize
the temporal distribution of a query’s returns, we have
built a tool to draw a line chart for any given query. It is
especially useful when comparing the temporal distributions
of two queries at a same time. For example, when a
historian suspects that there is dependency between two
concepts and wants more analysis, she can simply input
each concept as a query, and then get a line chart (Fig.
4). The line chart of Fig 4. suggests that the two concepts
are quite correlated. We have developed a term extraction method for extracting
noun phrases from old Chinese text (Chang 2006).
In the land deed corpus, we have successfully extracted
40,000 names of people and 7,000 names of locations
from metadata records and from full text (Chang, 2006).
THDL uses the names to provide term frequency analysis
by calculating the numbers of times each name appears
in the sub-collection and representing the result in
tables alongside the full text of resulting land deeds (Fig.
5). The names are listed in descending order according to
their document frequency (DF, the number of documents
in which a name appears). The user can use the tables to
observe the relevance among locations and people in the
sub-collection. Fig. 5 shows the returns of the query “Jin
Guang Cheng” (金廣成), a local reclamation cooperative
in the Guanxi (關西) area. At a closer look at the tables of names (Fig. 6) shows that the people on top
are indeed the major shareholders of Jin Guang Cheng.
Similarly, the locations on top are exactly the locations
where Jin Guang Cheng claimed lands back to 1880s.
However, the person with the highest DF only appears
in 38 documents, while the size of the sub-collection is
61, showing that none of the people appear dominantly
in the sub-collection. On the contrary, the locations with
the highest DF, “Shiliao Zhuang” (十寮庄) and “Zhubei
Er Bao” (竹北二保), appear in 57 and 56 documents accordingly,
showing the lands Jin Quang Cheng claimed
were mostly around the same area. Each land deed should, in principle, include all the attributes
we mentioned earlier in the paper. It is thus desirable
to extract them so that analysis can be done more
easily. We have developed an XML format and have
already extracted and analyzed about 13,000 of these
deeds (with the attributes of such land deed represented
as an XML file). Fig. 7 shows an example of these XML files. We have also developed a way to show the inter-relationship
of the XML files via a notion of relation graphs.
These graphs have been used to conduct role analysis, which shows how a specific person, family, or cooperative
is involved in the development of a certain region
through time. It is done by unfolding the roles they
played in land deeds. For example, Fig 8. shows the role
analysis of “Lin Benyuan” (林本源), a cooperative that
the well-known Lin family of northern Taiwan set up to
represent the family in land acquisition. We found that
most of the land deeds involving Lin Benyuan in our collection
are sales of lands or certificates of lands from thethe buyer
in all the sales and was the landowners in all the certificates.
This observation matches the general impression
of the Lin family, which has been one of the wealthiest
families in Taiwan since late 19th century till now. The
timeline of the deeds also shows that they focused on
(and systematically) acquiring lands from one geographic
location before moving on to the next.We have also built other facilities to assist research.
THDL allows users to bookmark documents and thus
create their own sub-collections. Moreover, users can
manipulate sub-collections by applying three set operations
– union, intersection, and complement – on subcollections
and save the result as a sub-collection (Tu,
1998). This facility enables the user to adjust a query’s
returns by adding or removing designated documents, or
documents from another sub-collection. THDL automatically
calculates the similarity between documents and
reports to user when there are documents similar to the
present one. The user can check if it’s a duplicate land
deed or a copy. For reading assistance, THDL provides
the facility highlights query terms and user-designated
keywords. This facility can help the user to quickly identify
keywords of interests when reading, and thus can
help quick evaluation of relevance.
Concluding Remarks
In this paper we described the land deed portion of THDL
(Taiwan History Digital Library) that we build with the
goal of providing a research environment with primary
documents in full-text for research in Taiwanese history.
We have developed a concept of regarding query results
as a sub-collection, and have built tools that help users
observe the relationship and collective meaning of a set
of documents. On the aspect of land deeds, we have noticed
that most of the existing research in Taiwan on this
subject had used no more than a few old deeds (often
within a hundred). It would be interesting to see, with
over 20,000 land deeds available in searchable full-text
and with tools to help discovering and analyzing their relationship,
what kind of research issues can emerge and
what kind of observations can be made.
Notes 1This estimation, however, could be significantly lower
than the real number.
Acknowledgements
The authors gratefully acknowledge the National Taichung
Library for their effort in producing the full text
of many of the land deeds mentioned in this article; the
Council of Cultural Affairs of Taiwan for authorizing
the inclusion of those deeds into our system; the staff of
the Special Collections Division of the National Taiwan
University Library in producing the full text of their collection
and for further checking the typing correctness of
all the deeds; Professors Wu Micha and Lee Wen-Liang
of the Department of History of NTU for their continuing
support; the members of the Lab of Digital Archives
and Automated Reasoning of the Department of Computer
Science of NTU for producing many of the tools
mentioned here; and the staff of the Research Center for
Digital Humanities for their general support.References
Chang, S.P. (2006). A Word-Clip Algorithm for Named
Entity Recognition: by Example of Historical Documents.
Master Thesis. National Taiwan University.
Chen, S.P., Hsiang, J., Tu, H.C., and Wu, M.C. (2007).
On Building a Full-Text Digital Library of Historical
Documents. In: Dion Hoe Lian Goh, Tru Hoang Cao,
Ingeborg Sølvberg, Edie Rasmussen (Eds.), 10th International
Conference on Asian Digital Libraries (ICADL
2007): Asian Digital Libraries. Looking Back 10 Years
and Forging New Frontiers. Hanoi, Vietnam, December
10-13, 2007. New York: Springer Berlin, pp. 49-60.
Hong, L.W. (2005). A Study of Aboriginal Contractual
Behavior and the Relationship between Aborigines and
Han Immigrants in West-Central Taiwan. Vol. 1. Taiwan:
Taichung County Cultural Center, pp. 5.
Tu, H.C. (1998). Interactive Web IR: Focalization Model,
Effectiveness Measures, and Experiments. Ph. D. Dissertation.
National Taiwan University.
Wu, M.C., Ang K.I., Lee, W.L., and Lin, H.Y. (2004).
A Brief Introduction to the Integrated Collections of
Taiwan-related Historical Records. Taiwan: CCA and
Yuan-Liou Publishing, pp. 101.
Wang, S.C. (1977). Taiwanese Historical Documents in
Private Holdings. Taipei: Huanqiu Publishing.
Wang, S.C. (1993). Introducing Historical Documents of
Taiwan: Government Archives, Old Documents, and Genealogies.
In: Yan-xian Zhang and Mei-rong Chen eds.
Taiwan History and Historical Documents of Taiwan.
Taipei: Zili Evening News Publishing,
Wang, S.C. (2004). Papers of Taiwan Historical Documents,
Volume 1. Taipei: Daw Shiang Publishing, pp.
375-376.