Computational Linguistics Techniques for the Study of Ancient Languages

. This paper presents a grammatical comparison of the Minoan language with the proto-Ugric and proto-Hungarian languages. Recent research showed that these languages are closely related, but this paper presents a novel grammatical comparison. The grammatical comparison shows the Minoan language to have an agglutinative type of grammar, with a CVCV type root word structure. The Minoan language also features assimilation between the word roots and the suffixes and a possessive phrase structure that is similar to that in Hungarian.


Introduction
The Minoan Cretan Hieroglyph writing [19,20,21] and Linear A [7,8] have been recently deciphered and shown to be closely related to the Ugric branch of the Finno-Ugric languages [12,13,14,15]. The corpus of Cretan Hieroglyph texts includes the Phaistos Disk [1], the Arkalochori Axe and the Malia Altar Stone inscriptions [14]. The Phaistos Disk and the Arkalochori Axe are sometimes described as separate scripts, but the Malia Altar Stone is listed as item number 328 in the Corpus Hieroglyphicarum Inscriptionum Cretae (CHIC) [8].
The key to the decipherments came from the application of phylogenetic algorithms [10,16] applied to ancient scripts. Using these clustering algorithms, Revesz [11] analyzed the evolutionary relationships within the Cretan script family, which includes the following scripts: Cretan Hieroglyph, Linear A, Linear B [3], Cypriot, Greek, Phoenician, South Arabic, Tifinagh, and Old Hungarian [4,5], which is also called rovásírás in Hungarian and also written sometimes as Rovas in English language publications. To these scripts, Revesz [15] added the Carian alphabets [18], which shows a particularly strong relationship with the Old Hungarian alphabet. In fact, the Carian alphabet is either an intermediary or very close to some intermediary between the Linear A and the Old Hungarian scripts.
The Carian and Old Hungarian phonetic values were used to find the syllabic phonetic values of the Linear A symbols [15] under the assumption that CV type Linear A phonetic values later simplified to C phonetic values during the gradual evolution of writing within the Cretan Script Family. (Here, as usual, C denotes a consonant and V denotes a vowel.) Moreover, there was enough circumstantial evidence of the relationship between the Minoan and the Ugric languages, which includes Hungarian, Khanty, and Mansi. Therefore, it seemed logical to assume that each Linear A sign represented some syllabic value CV and originally depicted a picture of an object or action that in the common proto-Ugric language had a name that started with the same CV syllable. In other words, searching for the phonetic values is a constraint problem solving [6,9] application. In particular, the syllabic value CV of any Linear A symbol σ has to satisfy the following constraints: 1. σ depicts the picture of an object or action, which is described by some word ω.
2. ω starts with the same syllable CV.

C is the phonetic value of the Carian and Old
Hungarian letters that are visually closest to σ according to an objective mathematical measure of visual similarity.
Applying the constrained acrophonic principle, it was possible to establish the syllabic values of close to fifty Linear A symbols. The substitution of the syllabic values yielded texts that were similar to some early medieval Hungarian texts. The above description underestimates the role that grammatical considerations played in an early stage of decipherment. This paper illustrates some of the grammatical considerations.
This rest of this paper is organized as follows. Sections 2 and 3 describe a set of Minoan and Hungarian language and script similarities, respectively. Finally Section 4 gives some conclusions and directions for future work.

Minoan and Hungarian grammar similarities
In this section, we list some grammatical similarities between the Minoan and the Hungarian languages. Our grammatical comparisons could be applied to any writing, not just the Minoan writing.

Minoan is also an agglutinative language
Agglutinative languages append suffixes to word roots without changing those roots. Table 1 displays some blocks of the Phaistos Disk and the Arkalochori Axe. Such a table can be built either manually or by a computer program, which would be preferable when the number of words in the corpus of a script is large. We built such a table manually because the Phaistos Disk and the Arkalochori Axe were previously considered as a separate corpus of texts from those of the Linear A and the Cretan Hieroglyph corpuses, which later were shown to be actually closely related variants for writing the same Minoan language [12,14]. We notice in Table 1  4. Whenever the endings are attached to a root, the root does not change. Table 1 indicates by red some of the apparent roots.
The above items imply that the Minoan language is an agglutinative language.

CVCV root structure
Linear A has been suspected to be a mostly syllabic writing with CV type syllables, where C is a consonant and V is a vowel, similar to Linear B, which is its immediate descendant.
Words with a CVCV structure can be represented by two symbols that stand for CV syllables. This fact may have influenced the Linear A script to develop as a syllabic script. Table 1 already shows several roots that contain two Linear A symbols. These include for example A M and Z e and G s.

Suffixes ending in -k
A problem with a pure syllabic script is that many suffixes do not fit into a CV structure. For example, consider the following Hungarian suffixes that end with a /k/ phoneme. We also give some examples, as they appear in the earliest Hungarian language documents. One of the frequently consulted documents is the Halotti Beszéd (Funeral Sermon) [2], which will be referenced as HB below.
1. /-ak, -ek, -ok/ is the plural of words that end in a consonant. The vowel that is chosen according to vowel harmony rules. For example, hal-ak (fishes) and kez-ek (hands) and ablak-ok (windows).
2. /-k/ is the plural of words that end in a vowel. For example, falu-k (villages) or kapu-k (gates).
3. /-k/ is the 1st person singular present tense verbal suffix in the indeterminate case.
4. /-juk, -jük/ is the 1st person plural present tense verbal suffix in the determinate case. For example, tümet-jük (we bury) appears in HB. As another example, present Hungarian uses számol-juk a pénzt (we count the money).
6. /-muk/ is the 1st person plural present tense verbal suffix in the indeterminate case. This appears as vogy-muk in HB. This suffix appears to be simply the composition of the /-om, em/ first person singular verbal suffix in the 8. /-nak, -nek/ is also a marker of the possessor of an object. For example, a ló-nak a lába (the horse's leg).
The Minoan B symbol represents not a syllable but some single phoneme because it used only at the end of the words with one exception. According to Table 1, in the Minoan language about half of the suffixes end with a B symbol. Remarkably, about half of the suffixes end in /k/ in Hungarian. Therefore, it is tempting to associate Minoan B with Hungarian /k/. Moreover, the above Hungarian suffixes could be grouped into three groups: (1-3), which have the form /-Vk/, where the vowel V is optional, (4)(5), which have the form /-jVk/, and (6-10), which probably had the form /-mVk/ assuming that m > n or n > m changes in some cases. These groups seem to match a natural grouping of the Minoan words into those that end with B, with L B and with i F B, respectively.
Old Hungarian contains two letters that denote the /k/ phoneme: and . According to some researchers one letter was used only within the words and the other was used only at the end of words. When carving the symbols into wood, a diamond is a convenient simplification of a circle, which may have denoted a human head [18]. Therefore, the shapes of the Minoan B symbol and the Old Hungarian symbol have a connection. Moreover, the Minoan symbol depicts the head of a man with prominent hair. The Mansi word for man is /kom/, while the ancient Greek word for hair was /komi/, which may have been borrowed from Minoan. This shows a /k/ or a /ko/ phonetic connection between the two symbols. Table 2 shows another pair of blocks that allows us to suspect that the symbol R is a conjunction symbol, meaning "and," a disjunction symbol, meaning "or," or it is some prefix. The shape of this symbol read from leftto-right suggests that it may denote two paths that merge together, that is, a conjunction.

Conjunction
When rotated ninety degrees, the symbol also reminds one of the Old Hungarian symbol, which denotes the /ʃ/ phoneme and occurs in the Hungarian words s and és that both mean "and." Table 3 shows the doubling of some symbols before the hypothetical suffixes.

Assimilation by consonant doubling
The doubling of consonants before suffixes is common in Hungarian. These doublings result from assimilation between the last consonant of the root and the beginning consonant of the suffix. Table 4 shows some examples.
Therefore, the Minoan a and X symbols behave similarly to the Hungarian doubled consonants and likely denote one of the consonants that is doubled in Hungarian except /ʃ/, which we already associated with R.

Assimilation by palatalization
Assimilation can occur without a doubling in case of some consonants. Table 5 gives some examples from Hungarian.   The palatalized sounds in Table 5 may not have been originally used in the Hungarian language within word roots, but they tend to occur naturally with the addition of suffixes that start with /j/. It is likely that in the Minoan language the palatalized sounds also first occurred as a result of assimilation. Table 6 shows that in block 22 a palatalization during assimilation can be suspected because the apparent assimilation yields a symbol that is rarely used. Moreover, it is never used at the beginning or the end of words, where palatalization is absent. It is also noticeable that it occurs only before LB, which we already associated with the /-juk, -jük/ suffix. Compare Phaistos Disk block 22 with the Arkalochori block 3, where there is no assimilation sound in a similar context before B which we associated with the /-Vk/ suffix.

Minoan and Hungarian Script Similarities
The grammatical comparisons in Section 3 already enable the identification of the phonetic values of some of the Phaistos Disk Minoan symbols as shown in the first two columns of Table 7. From Table 7, it is apparent that the Old Hungarian alphabet has a strong connection to the Minoan symbols. After such a realization, the logical step is the thorough comparison of all Minoan and Old Hungarian symbols to identify possible phonetic values of the Minoan symbols. The details of this step are described in Revesz [11].