An Assessment of Wikidata on Premodern Writing Systems

(1) Overview of Wikidata Items related to Writing Systems

Historical writing systems are an example of a quintessentially humanistic mode of production and area of study; indeed, they are of basic importance for Wikidata and other large-scale humanities datasets precisely because they consistent the foundational matrix for the production of written information. In this article, I use this dataset as a test case to assess the Wikidata ecosystem of this interdisciplinary humanities subfield for its scope, reliability, and research utility.¹ As Wikidata and similar datasets are increasingly integrated into humanities research and teaching (Zhao, 2023), such evaluations will be useful for specialists as they consider applications of the currently existing entries, as well as identify strategies for making them more robust.

The study of writing systems is highly interdisciplinary, and goes by various names, such as grammatology (Gelb, 1952), orthography (Condorelli and Rutkowska, 2023), and grapholinguistics (Barbarić, 2023); potential sub-disciplines, such as paleography, remain largely autonomous. “Writing systems” (Q8192) are defined in Wikidata as “any conventional method of visually representing verbal or signed communication.” They have various Wikidata subclasses, represented by items such as “Alphabet” (Q9779), “Abudgida” (Q335806), “Syllabary” (Q182133), “Logographic Writing System” (Q3953107),² “Unicase Alphabet” (Q4004706), “Bicameral Alphabet” (Q65045986), “Natural Writing System” (Q29517555), and “Constructed Writing System” (Q29517555), among others.

Individual pre-modern writing systems have Wikidata labels that reflect the wide-ranging terminology of the discipline. For example, Wikidata items that are instances of alphabets regularly contain “Alphabet” in their names (i.e., “Greek Alphabet”/Q8216 or “Hebrew Alphabet”/Q33513), but items that are instances of logographic writing systems may be described as “characters” (e.g., “Chinese Characters”/Q8201), or “hieroglyphs” as in (e.g., “Egyptian Hieroglyphs”/Q132659 and “Mayan Hieroglyphs”/Q211248), or a descriptive rather than typological label (“Egyptian Hieratic”/Q208111 or “Egyptian Demotic”/Q188519). Non-alphabetic writing systems are often described as a script, including abugidas (e.g., “Ge’ez Script”/Q257634) or pictographic writing systems (e.g., “Aztec Script”/Q794182) or simply by their name (e.g., “Kharoṣṭhī”/Q1161266 and “Brahmi”/ Q185083). Sometimes alphabet and script are used interchangeably in the scholarly literature, and this can lead to some puzzling and misleading statements in Wikidata, such as “Greek Alphabet” having the “writing system” property (P282) of “Greek Script” (Q1352751300), which is actually the Graeco-Pisidian script. A few families of scripts, such as “Brahmic Scripts” (Q467037), currently have their own items, but many are not currently represented in Wikidata. Finally, writing systems themselves are rarely specified by language, such as the Japanese writing system (Q190502), which “has parts” katakana (Q82946), which is a syllabary used to write Japanese phonetically, and kanji (Q82772), namely Chinese characters used for writing Japanese; both have historically been employed.

A SPARQL search returns a total of 310 instances of writing system, of which 42 are before the 16th century.³ However, many of these do not have a date of inception listed, and multiple scripts in this category are pre-modern. Given this dataset, I offer an exploratory assessment of typical annotations in Wikidata items across a selection of representative pre-modern writing systems belonging to a variety of types: Cuneiform (a logographic system),⁴ Aramaic (an abjad system), Greek (an alphabet) and Brahmi (an abugida), and Mayan hieroglyphs (also logographic). Cuneiform (Q401), which competes with Egyptian hieroglyphics for the title of oldest writing system, is among the most extensively annotated items among pre-modern scripts, and so reviewing its statements will give a broad sense of the data categories currently in use. Cuneiform is accurately defined as an “ancient writing system used for many languages, including Akkadian and Hittite.” It is listed as an instance of “writing system,” “logographic writing system/Q3953107,” “syllabary,” “natural writing system,” and “unicase alphabet.”⁵ It provides labels for a rich variety of languages, including those of contemporary scholarship on cuneiform, on the one hand, and those for which a Wikipedia page exists, on the other. Chronological date includes “inception” (P571) and “end time” (P582). Like most writing system entries, cuneiform’s is multimodal: there are four “images” (P18), two of material artifacts, and two script charts, showing a syllabary and evolution from archaic scripts, respectively. And, like all such image statements for writing systems, the objects displayed may be helpful, but they are somewhat random and hardly comprehensive in terms of script or media features.

Cuneiform has a number of other statements that are rarely found in other writing system items. For example, “Named after” (P138) (“nail”) is accurate but refers to the Latin-based modern word, not the ancient equivalent. “Location” (P276) in the “Near East” is accurate with respect to both its origin and distribution. Under time period (P2348), cuneiform is listed with a number of specific periods, including one, Sassanian, which is rightfully flagged with a “contemporary constraint” (Q25796498) potential issue. The Sumerians are correctly identified as “discoverer or inventor” (P61) of cuneiform. A number of open-access Encyclopedias are listed under “described by source,” but mostly without page reference, a common problem discussed further below. Finally, cuneiform is “used by” (P1535) “cuneiform clay tablet” (Q66018222), one of the very few statements which link a particular writing system to the media it was written on, a lacuna which I address late in the article; in this particular case, other typical media for cuneiform, such as “inscription” (Q1640824) are not listed.

The cuneiform entry lists the various languages that were written with it, including Sumerian and Akkadian, the earliest and most extensive, through the “Language of work or name” property (P407). Whether or not the particular version of cuneiform used by each of these languages constitutes its own separate writing system, given various modifications, is an important question, with no clear answer in linguistic or philological scholarship. In Wikidata they are treated as falling under a single writing system, cuneiform; but in the case of the Latin writing system, its adaptations for specific languages are treated separately. Indeed, another option would have been to describe all these subtypes as separate writing systems that were “influenced” by cuneiform. There is also a potential inaccuracy in listing the Old Persian script here: while its modern name is Old Persian cuneiform, it is composed of entirely different symbols from the others, constituting its own independent writing system, which indeed has its own Wikidata entry (Old Persian cuneiform/Q1471822), which is again listed inaccurately as an “instance of” and “subclass of” cuneiform. Note that there are no separate Wikidata items for Luwian, Eblaite, Hurrian, Urartian, Palaic, or Ugaritic cuneiform; on the other hand, one exists for Hittite cuneiform (Q1035956) and Elamite cuneiform (Q3782237).

The Aramaic Alphabet (Q26978) is an abjad (Q185087),⁶ and is annotated as an “instance of” “abjad,” “natural writing system,” and “unicase alphabet.” Two conflicting dates are given for its “inception,” probably to distinguish between the earlier form and later “Imperial Aramaic;” indeed, the associated languages, which for this entry are listed through the “has use” property, are “Aramaic” and “Imperial Aramaic.” The “image” includes a Stele, and, strangely, a single word written with the Syriac script, which is a later, Aramaic-derived writing system. Its name is given as “Imperial Aramaic,” which is only partially correct as it refers to a particular era, which probably should have a separate entry; similarly, the “location of creation” is incorrectly given as “Achaemenid Empire,” which is in fact where Imperial Aramaic developed, not the earliest form of the script. The end time in the 6th century is similarly confusing, and is inaccurate unless it is 6th century BCE and in reference to the pre-Imperial Aramaic form. On the other hand, the “based on” “Phoenician script” statement is accurate and useful, as is the “script directionality” as “right-to-left.” In this case, unlike cuneiform, a unicode range is given; like many abjads and alphabets, individual letters are given under “has part(s)” (P527).

The Greek Alphabet (Q8216) is an “alphabet” (Q9779).⁷ Indeed, it is described as an “instance of” “alphabet,” but not “writing system.” There are three different dates for its inception, reflecting scholarly debates about this issue. It represents languages that use the script by “has use,” with “Greek” and “Ancient Greek,” but also by “facet of” (P1269) “Greek,” as well as “language of work or name,” this time with “Greek,” “Ancient Greek,” and “Pontic Greek,” in a seemingly repetitive manner. The four images are typically diverse, including a modern road sign, a palaeographical list of medieval script styles, and an example of an early printed book. Like Aramaic, it correctly notes that it is “based on” the “Phoenician script,” as well as “script directionality,” in this case “left-to-right.” On the other hand, “location of creation” is incorrectly listed as “Classical Athens.”

Brahmi (Q18508) is an “instance of” “natural writing system,” “unicase alphabet,” and “abugida” (Q335806)⁸ but not simply “writing system.” There are four examples under “image” of inscriptions and coins. Brahmi is described as “based on” (P144) the Aramaic alphabet (Q26978), although this is controversial in scholarship, and no references are provided. One script, Gupta, is named as a “derivative work” (P4969), although there are many others. Likewise, Brahmi is also described as “replaced by” (P1366) Gupta, which is also accurate, in a limited sense: i.e., it was not replaced by Gupta in all regions. Neither of these are referenced. Six languages that use Brahmi are named under “language of work or name.” These are “major” examples (Prakrit, Sanskrit, Saka, Tamil, Kannada and Kannada; the sixth listed, Tocharian, is less well attested). The “location of creation” (P1071) is listed as “Central Asia,” although “South Asia” would no doubt be preferred by many scholars. But “location,” used for cuneiform, is not found here. Brahmi also has a statement for “end time.”

(2) Common Properties used to describe Writing Systems

In this section I will explore some relatively common properties found in writing system entries, highlighting areas for improvement as relevant. Although all the items in this study are “writing systems,” they are in fact not always marked as such in their respective entries, which often use other related but more specific terms, and often more than one. These include all of the different types of system (that is, logographic writing system, abjad, abugida, alphabet, syllabary), as well as further subcategories. Most common is “natural writing system” (Q29517555), defined as a “historically-grown writing system, in contrast to constructed writing systems,” whereas the “constructed writing system” (Q1191702) is defined as a “writing system specifically created by an individual or group.” Another distinction is between the “unicase alphabet” (Q4004706), which makes no distinction between upper- and lower-case letters; and the bicameral script (Q65045986), which does. However, these terms are not systematically used across writing system items in Wikidata, and they are sometimes recorded using “instance of” statements, and sometimes using “subclass of” statements.

More consistent is “Script directionality” (P1406), which is currently the only script-specific property. “Script directionality” can be right-to-left, left-to-right, bottom-to-top, top-to-bottom (or all, as in the case of Egyptian hieroglyphics). Note that these descriptions are limited to the orientation of characters within a single line; in other cases, additional details are given about the orientation of the lines, for example vertical left-to-right (Old Uyghur alphabet/Q1998938), specifying that characters are read vertically (which direction is not provided), while lines are read from left to right. Line orientation is only occasionally noted in Wikidata items, perhaps because of a Latin-based bias that assumes horizontal top-to-bottom.

In addition to the “Language of work or name” property (P407), scripts are sometimes related to languages with the property “facet of” (P1269), sometimes with the property “has use” (P366). Sometimes these are even both used for the same item: the ‘Phags-pa script (Q822836) includes “has use” Sinitic and “facet of” Sanskrit, while omitting the various other languages, above all Mongolian, for which it was used. As another example, the Hebrew Alphabet entry records several languages using it under “Language of work or name,” for example, Hebrew, Yiddish, and Judaeo-Spanish, but also leaves out a number, such as Babylonian Jewish Aramaic, Judeo-Arabic and Judeo-Persian. So the crucially important linkage of script and language suffers from two difficulties: languages are not consistently noted; and when they are, different properties are used, and some languages are often still left out.

Various forms of chronological data exist. “Inception” (P571) dates can be exact (i.e., ‘Phags-pa in 1249 CE) or inexact (i.e., 3200 BCE), although the inexact dates are not always marked as such. Sometimes uncertainty is registered through the use of a century, rather than a precise year, such as “13th century BCE” for Chinese Characters. In the case of the Greek Alphabet, two different numbers are given, each with its own reference. For scripts that ceased to be used, there is often an “end time” (P582; e.g., Cuneiform, Egyptian hieroglyphics, and ‘Phags-pa). Only very rarely, as in the case of cuneiform above, are specific time periods (P2348) described. And there are no cases of more granular data, such as the inception date for when the script was used by another language or languages.

The development of scripts over time, including influences, is an important and emerging area of research, for which the “Based on” property (P144) is sometimes used. A few families of scripts, such as “Brahmic Scripts” (Q467037), currently have their own items, but many such groups are not currently represented in Wikidata. The most substantial example in Wikidata, and likely the best data collection of this type overall, is the Latin-script alphabet item (Q29575627), defined as an “alphabet that uses letters of the Latin script,” for which a SPARQL query returns 130 instances of this “Latin-script alphabet,” including the “Swedish alphabet,” the “Filipino alphabet,” the “Old English Latin alphabet,” the “Uzbek alphabet,” and the “Somali Latin alphabet.” A separate but related part of the historical context for the emergence of new scripts is the “Country of Origin” (P495), which is often omitted, perhaps because it is presumed to be obvious because of modernist assumptions neatly tying together nation, language, and script; or perhaps because the category of “Country” does not fit well within pre-modern political structures. Variations in a given script, traditionally the domain of palaeography, are treated as instances of “handwriting style” (Q33260112). In general, Wikidata on palaeographical variants of writing systems are much less complete than on scripts, despite having a plausible location in the data structure.

Sometimes Wikidata somehow collectively moves the scholarly needle, applying existing academic systems in novel ways. For example, the distinction between a “unicase alphabet” (Q4004706) as opposed to “bicameral script” (Q65045986), more at home in modern typography, is successfully applied to many pre-modern scripts, including Aramaic and Brahmi, as is the distinction between a “natural writing system” (Q29517555) and a “constructed writing system” (Q1191702). However, this creep of categories can sometimes lead to mistaken classification: the Georgian alphabet is recorded as an instance of a “constructed writing system,” although the consensus is that it was developed over time, despite several later accounts attributing it to an individual (as noted even on the Wikipedia page for the Georgian alphabet!).

In terms of quantitative data, a SPARQL result shows relative low usage of most relevant properties across the dataset of 310 writing systems described below.⁹ The most commonly used properties, that is, “Name” (287 instances), “Image” (112 instances), and “Subclass” (103 instances), are not specific to writing systems. The most common such entry, “Language of work or name,” in our dataset a description of languages using the script, is found in less than one third of the items (91 instances). The next two most common, “Inception” (82 instances) and “Based on” (82 instances), provide some historical context. Script directionality, an important typographical concept, is not well representative (68 instances), nor is Unicode range (45 instances), a key property for electronic text, or the crucial historical information about location of creation (16). In short, a large number of key properties are simply missing from the majority of writing system items in Wikidata.

(3) Form and Material of Writing Systems

Another important aspect of pre-modern writing systems for which Wikidata has great potential is their materiality, including their structure (e.g., scroll or codex), the materials out of which they are constructed (e.g., parchment or paper), and various writing implements (such as a reed brush or quill). Unfortunately, there is currently very little consistency in the way the relevant items and their statements are organized. One widespread pre-modern writing format, “tablet” (Q16744570) is very spare in its present form. It is defined as a “plane object, usually covered with some letters or pictures.” And further, it is described as a subclass of “communications media,” “artificial physical object,” “archaeological artefact,” and “writing surface,” all apt statements which denote its status as both a general category and a particular instance of object/artefact. On the other hand, there are no statements about chronological or geographical distribution of the tablet; the particular languages and scripts which used it; or the materials with which it was used.

Indeed, specific combinations of format and material often have their own specific entries, rather than placing attested materials in the statements of a given format item. For example, the item “wax tablet” (Q1428312) is defined as “two or more tablets of wood, filled with wax, for writing upon,” which does feature two helpful statements, “has characteristic” (P1552) “reusability” and “uses” “stylus” (Q1227035). I found neither of these otherwise attested among pre-modern communications media for writing systems, but both would be quite productive if more widespread—the linking of specific writing implements with particular pre-modern materials and formats is not widely documented, even in the print literature. Another item that combines material with format is “clay tablet” (Q1570005), which is a subclass of “tablet,” and, helpfully, “made from material,” “clay,” which in this case is redundant. Another rare statement is its description as “different from” (P1889) the “clay cone,” (Q62273144), which, like the clay tablet, is particularly associated with ancient Mesopotamia.

Another pre-modern communications medium, the “scroll” (Q720106) is defined as a “roll of papyrus, parchment, or paper containing writing.” But despite this definition, the scroll is not linked to any of these materials in the statements. There is a separate item for “papyrus scroll” (Q113016548), with almost no annotation, except for the obvious “subclass of” “scroll” and “made from material” (P186) “papyrus” (Q12043767). The codex (Q213924), which largely replaced the scroll in the Mediterranean and Middle East, is given the deceptively neutral definition of a “book with handwritten content,” no doubt reflecting its default status in the Anglophone world. On the other hand, it suffers from the same lack of annotation we have seen in other communications: no mention of the different materials it was made from (in this case paper, parchment, or papyrus, not to mention binding, covers, etc.); or chronological (not even an inception date, even if this is only approximate) or geographical distribution; or languages and scripts which use it, though in the case of the codex, this would admittedly be quite long.

Beyond the scroll, which is widely attested across cultures, specifically East Asian written communications media are also represented by Wikidata items, but these too currently have only a limited number of statements. For example, “bamboo and wooden slips” (Q905725), defined as a “writing medium in ancient China,” and in turn described as a subclass of “bamboo and wooden slips and silk manuscripts” (Q10461162), itself an instance of “communications media,” a subclass of “writing surface,” with “country” (P17) as “Chinese Empire,” a rare example of geographical information associated with a writing format. The related mokkan (Q4299994) is described as “a long, narrow board that originated in China and was used in ancient Asian regions” is identified as a subclass of “wooden slip” but not a “writing surface.” Indeed, nothing beyond the image identifies it as a writing surface, though helpful geographical (country of origin=Japan) and chronological (time period=7th century CE) data is given.

No Wikidata entry exists for “pothi,” a popular format of stacked palm leaves held together by a string, which originated in South Asia and spread along the silk road. Strangely, “Eastern manuscripts” (Q7262226), defined as a “handwritten book, poem, or other document, or a collection,” is “also known as” “pothi.” As the images in this item show, pothis are indeed one of the intended manuscript formats, but not the only one. So the entire entry is misleading. Moreover, there is no statement linking pothis to palm leaf, a characteristic writing surface for them, especially in South Asia and Southeast Asia. On the other hand, it does helpfully list a number of “language of work or name” and even “writing system” (P282), although in both cases it is far from comprehensive. “Eastern manuscripts” are described as a subclass of “manuscript,” but not “communications media.”

Finally, there is spotty representation for other key material aspects of writing systems: the stylus (Q2647209) is described as a “writing utensil or small tool for marking or shaping;” subclass of “writing implement,” but almost no information. There is an image which suggests Roman stylus and tablet. The “reed pen” (Q1359162), with similar geographical and chronological distribution, has more information. It too is accurately described as a “writing implement,” but with no use of the Ink brush (Q494618) is described as a “calligraphic tool,” but in this case as a subclass of “paintbrush” and “writing brush;” no languages or regions are specified (as for “stylus,” but one of the only statements, that it is “part of” the “Four Treasures of the Study” (Q494402) suggests traditions based on Chinese calligraphy). Regarding inks, there is a separate item for “iron gall ink” (Q740353), a widespread class of medieval inks in Western Eurasia, which is a subclass of ink (Q127418), but there is no separate category for carbon ink, an important class of inks for the ancient Mediterranean world.

(4) Context and Method

Like all Wikidata datasets, those on scripts are not the result of an organized project, but grow organically from the efforts of individual contributors. The history of additions to the entry can be easily viewed in the tab, and further additions (or revisions) can be made by registered users. No particular methodology is collectively followed, beyond the provision of certain terms, such as typologies of scripts, which is not done systematically, and sometimes involves interesting extensions of usage beyond what is common in academic literature. The relative benefits and disadvantages of such a distributed process of data entry are far too big a topic for this short article, but it should be noted that they reflect, to some extent, the interdisciplinary and non-standardized nature of the academic study of scripts. And in any case it should be noted that the majority of writing system items actually have very few contributors, as is evident from the distribution of the # of statements for each item from a SPARQL query.¹⁰ Indeed, 111 items have only 0–9 statements, while 90 items have 10–19, followed by a long tail of more extensive entries. This suggests that Wikidata’s potential for multiple contributors is not being fully realized for this dataset.

On the other hand, the distributed data entry system invariably leads to different properties being used in entries on different writing systems to describe the same concept. For example, we have seen how the use of a script to write a particular language is described with several different properties, including “language work or name” and “has use.” This feature, which we can call annotation misalignment, prevents the full mobilization of Wikidata for research on this key relationship.

The most obscure aspect of a Wikidata entry is the standards of citation of references and sources. While there is a discussion about this,¹¹ the practical implementation, at least in the writing systems entries, is spotty. Many statements in this group go unreferenced; indeed, the primary means of citation is the use of the “described by source” property (P1343), which are usually open-source (and thus old) encyclopedias of surprising variety, such as the Armenian Soviet Encyclopedia (Q2657718) for Aramaic, or Otto’s Encyclopedia (Q2041543). These almost invariably have no modern standing in the field and almost certainly do not provide relevant information for all of the statements in their entries which are missing references. Especially in cases where there is scholarly disagreement, addressed through a qualifier such as rank, these sources are simply not authoritative enough to represent current debate about the status of the data; specialized studies, whether monographs or articles, should be referenced instead.

(5) Conclusion

Wikidata on writing systems constitutes the best open-source linked dataset on the topic, and is the most promising basis for future data collection and analysis in this field of study. It is already being used in a number of research workflows, as is evident in this collected volume, for example through integration with external projects’ own developing datasets (Berti, 2025), or as a means to align two different existing authoritative datasets (Ripoll-Alberola, Le Bris, and Fischer, 2026). But if it is to be used as a major or primary source for cultural, historical, and literary analysis, there must be an accounting of the scope and quality of the data subset to be used by the researcher. Some studies have already pointed to existing bias in Wikidata datasets.

This chapter has been a more general, preliminary evaluation of the Wikidata related to pre-modern writing systems. The demonstrated incompleteness and inconsistency of the data in its current manifestation discourages its use, at least in isolation, as a primary source for hypothesis formation or new argumentation related to the comparative study of writing systems. On the other hand, building the dataset up should be seen as a valuable research exercise in itself, whether by academics or citizen scientists. Indeed many more contributors are needed both to add entries, expand existing ones, and review them for accuracy. Given that Wikidata is constantly evolving, and is open to annotation from registered contributors of diverse backgrounds, there is indeed much promise for substantial expansion of the dataset, by both individuals and organized projects or teams.

Most fundamentally, every Wikidata entry on a historical writing system should be enriched with references, including conflicting data in cases of scholarly debate, to ensure its usability.¹² Perhaps the most straightforward way to do so would be to through related Wikipedia pages, which often have relevant bibliography; but a fresh consultation of print sources would be preferable. In addition, more consistent annotation of key features of writing systems, including chronological, geographical, grammatological, and linguistic information, would greatly increase the value of this dataset for research. Of course, because Wikidata grows organically, from multiple contributors, it is not possible or even desirable to enforce conformity of annotation. But suggested guidelines could be offered through the creation of an EntitySchema, described as “living blueprints by which we model classes of items on Wikidata.”¹³ These describe expected properties to be included in various subsets of items, such as the EntitySchema for humans, which specifies, e.g., place of birth (P19) and family name (P 734).¹⁴ No EntitySchema for writing systems currently exists, but one could be developed, with a standard set of properties, based both on existing coverage within Wikidata and the recommendations of disciplinary experts. An EntitySchema for writing systems could function not only as a set of guidelines for annotators, but a way for researchers to assess the current structuring of individual items or the entire dataset.

Additional File

The additional file for this article can be found as follows:

Competing Interests

The author has no competing interests to declare.

Author Contributions

Paul C. Dilley is the sole author.