Discovering Citation Relations among the Imperial Court Documents of Qing China

Discovering Citation Relations among the Imperial Court Documents of Qing China
Chen, Shih-Pei, Department of Computer Science and Information Engineering, National Taiwan University, gail@turing.csie.ntu.edu.tw
Ho, Hou-Ieong, Department of Computer Science and Information Engineering, National Taiwan University, brent@turing.csie.ntu.edu.tw
Tu, Hsieh-Chang, Department of Computer Science and Information Engineering, National Taiwan University, champiyen@turing.csie.ntu.edu.tw
Hsiang, Jieh, Department of Computer Science and Information Engineering, National Taiwan University; Research Center for Digital Humanities, National Taiwan University, jhsiang@ntu.edu.tw
1. Motivation and Introduction
The dynasties of imperial China had always had sophisticated governing systems to run the vast empire (Chien, 1952). While many of these dynasties left large quantities of imperial court documents, the last dynasty, Qing (1644-1911), produced the largest volume. These documents have been a major source of primary research material for studying Qing era China since they provided the most direct and first-hand details of how national affairs were handled. Among them, two of the most important kinds are Imperial Edicts (from the emperors to his officials) and Memorials (reports from officials to the emperor). The number of Memorials increased significantly after Emperor Kang-xi (康熙 – reigned from 1662 to 1723) allowed senior local officials to report to him directly (Chuang, 1979). The ability for the emperors to obtain first-hand information directly from local officials was among the major reasons why the Qing imperial courts did not suffer as much interference from people surrounding the emperors, such as eunuchs and family members of the empress dowagers, as in the previous Ming Dynasty.

Qing Dynasty had a systematic way to archive official documents. However, although most of the archives were organized chronologically, the court documents involved in a specific event might span several months and were often kept in different archives. For instance, if the emperor received a Memorial reporting a rebellion in some province, he might decide to issue an Imperial Edict to give instructions to relevant officials. The Memorial, depending on its character, might be kept (or had copies made) in the Archives of the Imperial Palace (宮中檔), Archives of the Grand Council (軍機處檔案), or the Grand Secretariat Archives (內閣大庫). The Edict might have records in the Imperial Decrees Archives (上諭檔), Archives of the Diary-Keepers (月摺檔, 起居注), or the Archives of the Imperial Palace (宮中檔), or the Grand Secretariat Archives mentioned above. Worse yet, these archives are now kept at different locations, notably the National Palace Museum (National Palace Museum, 2001), the Institute of History and Philology of Academia Sinica (Institute of History and Philology of Academia Sinica, 2001), both in Taipei, and the First Historical Archives of China (FHAC, 1995), in Beijing. Although digitization effort at the former two institutions made these archives easier to access than before, it remains a cumbersome task to collect documents covering the same event and rebuild their original linkage.

In this paper, we present an approach to restore an important relation: the citation links among the imperial court documents. To be more precise, a Memorial from an official often quotes earlier Imperial Edicts as the directive for the activities that he is reporting. On the other hand, an Imperial Edict may also cite earlier Memorials as the reason for issuing the decree. In an important historical event such as the Taiping Rebellion, there may be hundreds of Imperial Edicts and Memorials that form a complex web of successive citations. We call such a graph an IE-M diagram. Figure 1 is an illustration.

Fig. 1 An IE-M Diagram

Full Size Image

2. Discovering IE-M Diagrams
To discover a citation relation we need to first detect whether a document has cited previous documents. This problem is similar to plagiarism detection (or copy detection), which is to detect whether a part of a document is copied from other materials without acknowledging the source (Shivakumar and Garcia-Molina, 1995; Si, Leong and Lau, 1997; Timothy and Justin, 2003). However, unlike plagiarism detection for which an exhaustive comparison among documents might be necessary, in our case there are often specific phrases occurred around quotations, which we call syntactic anchors. In the case of a Memorial citing an Imperial Edict, the former usually contains an anchor that starts with adhering to the Imperial Edict (奉上諭) and ends with By the Emperor Himself. That is all (欽此). The text in between are quoted verbatim from the Edict (although usually not in its entirety) (see Fig. 2).

Fig. 2 In a memorial, the sentences in between the syntactic anchors are text quoted verbatim from the source edict.

Full Size Image

The anchors involved in an Imperial Edict citing a Memorial may have a number of varieties. The quotations may also be done in a more casual manner, since the emperors did not feel obliged to quote carefully. After identifying the anchor and the quoted text, our method extracts a segment of the latter (called signature) and applies a text-matching algorithm to see if it appears in any document in the database. A document of the right type that contains the signature is called a candidate.
We remark that because the quoted text may not be very precise, a certain degree of fuzziness is incorporated in our algorithm so that minor differences can be tolerated. A detailed algorithm is listed in Fig. 3.

Fig. 3 The text-matching algorithm used for Memorials citing Imperial Edicts.

Full Size Image

The comparison produces a list of document, signature, candidate tuples. We then use metadata to filter unlikely tuples, and present the findings to historians for manually validation to ensure accuracy. The overall process of our approach is shown in Fig. 4.
Fig. 4 The overall process of our approach.

Full Size Image

3. Experiments and Findings
We applied this method to the “Collection of Taiwan-related Court Documents from the Ming and Qing Dynasties” corpus in THDL, the Taiwan History Digital Library (Chen, Hsiang, Tu, and Wu, 2007; Research Center for Digital Humanities, 2009). This corpus contains 37,831 imperial court documents of mostly Qing era that are related to Taiwan. The documents, mainly Imperial Edicts and Memorials, were selected from 235 different sources, including the archives mentioned earlier in this paper. They were chosen by historians, typed as full text, punctuated, proof-read, and were supplemented with metadata records (Wu, 2004; Chiu, 2006).

Using the method described above, we discovered 5,403 pairs of citation relations from these documents, among which 3,947 pairs are Memorials citing Edicts, and 1,456 pairs are Edicts citing Memorials. By taking the transitivity on the discovered citation pairs, we produced 1,258 IE-M diagrams (see Table 1), the largest of which involves 152 documents.

Table 2 The main content of the IE-M diagrams sized over 50. All of them are about major events in the Taiwanese history.

Full Size Image

Fig. 3 is an example of such a diagram, in which the blocks at the top and bottom are Imperial Edicts while the middle three are Memorials. The arrows between the blocks indicate citations.
After examining the diagrams, we found that all of the larger ones (size over 50) are about major events in Taiwanese history (see Table 2).

Table 1 A summary on the sizes of the 1,258 IE-M diagrams produced by our method.

Full Size Image

Tracing through the citations shows the process of how the Qing Imperial Court dealt with crises occurred in Taiwan, a far-flung island of the vast empire. For example, the largest IE-M diagram illustrated how Emperor Qianlong (乾隆) handled the Lin Shuangwen (林爽文) Rebellion, the largest civil unrest in Taiwan during the Qing reign. The diagram vividly reflected how the rebellion, first dismissing as a minor local disturbance, developed into an island-wide revolt. (The rebels even overran a prefectural seat and had another under siege for more than six months). It also showed how the local officials, failed at suppressing the revolt, pointed fingers at each other or reported false victories. Qianlong finally realized the severity of the situation and sent Fukangan (福康安), one of his most trusted generals, to put down the rebellion. (Qianlong himself considered the pacification of Lin Shuangwen Rebellion one of his “Top Ten Military Achievements”.)
4. Concluding Remarks
In this paper we described an approach to discover citation relations among Imperial Edicts and Memorials of the Qing Dynasty. The transitivity closures of the relations are captured in a concept of IE-M diagram, which reveals how a historical event developed through the correspondences between the Qing imperial court and the local governments. Our method demonstrates how to use information technology to explore and identify important relations among historical documents that would be hard to find otherwise. We applied this method to THDL, Taiwan Historical Digital Library, and found 1,258 such diagrams. We should remark, however, our method can be applied to other and larger corpuses provided that the full text of the documents are available. We are currently working with historians to explore other significant relations to which our method can be applied.

Fig. 5 An example of IE-M diagram.

Full Size Image

References:
Chen, S.P., Hsiang, J., Tu, H.C. & Wu, M.C. 2007 “On Building a Full-Text Digital Library of Historical Documents, ” Asian Digital Libraries. Looking Back 10 Years and Forging New Frontiers, Lecture Notes in Computer Science no. 4822, Goh, D.H.L., ed. Springer Berlin, New York 49-60

Chien, M. 1952 Chung-kuo li tai cheng chih te shih, Hong Kong

Chiu, W.J. 2006 “The Digital Project of Taiwan-Related Archives in Ming and Qing Dynasty, ” The Library Yearbook of ROC 2006, National Central Library Taipei

Chuang, J.F. 1979 Qing dai shi liao lun shu, National Place Museum Taipei

The First Historical Archives of China (FHAC) 1995 Guangxu chao zhu pi zou zh, Zhonghua shu ju Beijing

National Palace Museum The Archives of the Grand Secretariat in Academia Sinica , 10 March 2011 (link)

Research Center for Digital Humanities of National Taiwan University Taiwan History Digital Library, 10 March 2011 (link)

Shivakumar, N. and Garcia-Molina, H. 1995 “SCAM: A Copy Detection Mechanism for Digital Documents, ” The 2nd International Conference in Theory and Practice of Digital Libraries., Austin, Texas, 1995

Si, A., Leong, H.V. and Lau, R.W.H. 1997 “CHECK: a document plagiarism detection system, ” Proceedings of the 1997 ACM symposium on Applied computing, San Jose, California, 1997

Timothy, C. H. and Justin, Z. 2003 “Methods for identifying versioned and plagiarized documents. , ” Journal of the American Society for Information Science and Technology, 54(3) 203-215.

Wu, M.C. 2004 Taiwan shi liao ji cheng ti yao, Council for Cultural Affairs Taipei