### Tags: #AI #AI/OCR #OCR #RomanCoins #Digital_Humanities #DigitalProcess #Digital_Archives #Digital_Archaeology --- # What I was trying to do I am exploring the most efficient and accurate methods to perform OCR. I am exploring Google AI Studio and Gemini 2.5 Pro (05-06). The objective is to see how accurate the output is with 16th-19th century numismatic manuscripts written in Latin, French, and English. German and Greek will also be explored in the future. - link to the relevant github repo: like this: [Roman Coinage and Otho's Reign | Google AI Studio](https://aistudio.google.com/prompts/1BTRBqz1zFcF3ryAEK3eSGp8Q2nm7D6lF) ## what I did - step 1 - I first inputted a 'tone' for the AI to focus on. In the 'System Instructions' I inputted: "You are an expert in Optical Character Recognition. The image I will feed you is from a 17th century manuscript and is about Roman coins. " - step 2 - Next I uploaded the image that I wanted Gemini to perform OCR on into the chat prompt. The image was from Foy-Valliant Jean. 1743. *Numismata imperatorum Tomanorum praestantiora a Julio Caesare ad Postumum usque*. Romae: Sumtibus Caroli Barbiellini & Venantii Monaldini. Getty Research Institute. xxv. Accessed from the Internet Archive on May 13th, 2025. See [Numismata imperatorum Romanorum, Internet Archive](https://archive.org/details/gri_33125008668408/page/XXV/mode/1up) - step 3 - The Prompt used with the image was "Please perform OCR and the following image:" - This is the image used: [[Foy-Vaillant numismata imperatorum.png]] - The out put is below in callout box labelled 'Output' >[!Output] >Antiquarios paſſim proſtant; pauciſſimi etiam ærei exiſtunt Antiochiæ percuſſi, & Alexandriæ, quo ſerius Othonis ad imperium evecti nuncium perve- nerat. Adſcribi id nequit brevitati ejus imperii, quod ultra tres menſes non excurrit. Si enim intra id temporis ſpatium ſignari potuere nummi ex ar- gento, & auro, cur non ex ære, quod multo para- tius eſt, & magis obvium? An fortaſſe quod nuta- ret Senatus inter Othonem, & Vitellium magnis itineribus cum exercitu Germanico in Italiam con- tendentem, cuinam potius adhærendum foret, & po- pulo inter utrumque ſuſpenſo non exhibendus vide- retur vultus Othonis in moneta ærea, quæ vulgaris eſt, & inter populum ſpargitur, contra vero præ- bendus militi in denario argenteo, & aureo, cui pro Othone ſolvenda ſtipendia, & cum æmulo di- micandum? Vel fortaſſe quia eo deliciarum, & vanitatis deflexerat Otho nitidus, & calamiſtratus, & munditiarum pene muliebrium, ut inquit Suetonius in ejus vita, ut nollet ſuum vultum in ære vileſcere, & rubigine infici, ac fœdari, ſed dumtaxat in auro ſculpi, atque argento? Vel forſitan magna vis pecuniæ æreæ ſub Claudio, Nerone, & Galba per- cuſſæ in ærario ſupereſſet, & magna contra caritas aureæ, & argenteæ, & huic ſubveniendum, illa primum diſtrahenda? Vel quod Triumviri Mone- tales, Othone in belli curas diſtraċto, æs omne, ferirent adpoſita in media area vetere nota S. C. propriiſque nominibus ad oram: quorum æreorum - Step 4 - I then asked the AI to identify the manuscript the text is from. "tell me what manuscript this excerpt is from?" - The output was >[!Output] >This text is a classic example of early modern numismatic scholarship. > >Based on the style, content (detailed discussion of Roman imperial coinage, specifically Otho's coins, referencing Suetonius), and the typical printing of the 17th century, this excerpt is from: > >**Jean Foy-Vaillant's *Numismata Imperatorum Romanorum Praestantiora a Julio Caesare ad Postumum et Tyrannos***. > >This work was first published in Paris in 1674 (though there were later editions, like 1682, 1692, 1743). The specific passage you've shown discusses the reasons for the relative scarcity of Otho's bronze coins (aes) compared to his silver and gold, despite his short reign. Foy-Vaillant explores various hypotheses, including the Senate's hesitation between Otho and Vitellius, Otho's supposed vanity (not wanting his image on baser metal, citing Suetonius), or a surplus of bronze coins from previous emperors. > >The reference "ut inquit Suetonius in ejus vita" (as Suetonius says in his life [of Otho]) is a strong pointer, as is the detailed numismatic reasoning typical of Foy-Vaillant's comprehensive work. > >Specifically, this passage can be found in the section dedicated to Emperor Otho. For example, in the 1692 Paris edition (often found in two volumes), this would be in Volume 2, very early in the section on Otho. - Step 5 - Next I asked the AI to translate the text. "Translate the following into English :" Then copied and pasted the OCR output into the chat prompt. The output was the following. >[! Output] >Okay, here is a translation of the provided Latin text into English. This is a scholarly text discussing numismatics, so the language is somewhat formal. > >They are commonly found among antiquarians; very few bronze coins also exist, struck at Antioch and Alexandria, where the news of Otho's elevation to the empire had arrived later. This cannot be ascribed to the brevity of his reign, which did not extend beyond three months. For if, within that space of time, coins could be struck from silver and gold, why not from bronze, which is much more readily available and more common? > >Or perhaps [was it] because the Senate was wavering between Otho and Vitellius (who was hastening to Italy with the German army by forced marches), as to whom it should rather adhere, and, with the populace suspended between the two, the likeness of Otho should, it seemed, not be displayed on bronze coinage, which is common and is circulated among the people; but on the other hand, [it should be] provided to the soldiery on silver denarii and gold aurei, to whom stipends were to be paid on behalf of Otho, and with whom he had to contend against a rival? > >Or perhaps because Otho, sleek, with curled hair, and of almost womanly fastidiousness, as Suetonius says in his life [of Otho], had declined to such a degree of luxury and vanity that he would not want his likeness to be cheapened on bronze, and to be tainted with rust and defiled, but only to be engraved in gold and silver? > >Or perhaps [was it that] a great quantity of bronze money, struck under Claudius, Nero, and Galba, remained in the treasury, and, on the other hand, [there was] a great scarcity of gold and silver coinage, and this [scarcity] needed to be addressed, [meaning] that [existing bronze coinage] should first be distributed? > >Or [was it] that the Triumviri Monetales, with Otho distracted by the cares of war, were striking all the bronze with the old mark S.C. (Senatus Consulto) placed in the middle of the field, and with their own names on the edge: of which bronze coins... > >--- > **Notes on the translation and text:** >- **paſſim proſtant**: Literally "they stand forth everywhere." I've translated it as "They are commonly found among antiquarians," implying these coins are what antiquarians deal with or have available. >- **percuſſi**: "Struck" (as in coins being minted). >- **nuncium perve-nerat**: "news had arrived." >- **ſignari potuere nummi**: "coins were able to be marked/struck." >- **paratius eſt, & magis obvium**: "is much more readily available and more common." >- **nutaret Senatus**: "the Senate was wavering/hesitating." >- **magnis itineribus**: "by great journeys" or "by forced marches." >- **vultus Othonis**: "the face/likeness of Otho." >- **moneta ærea, quæ vulgaris eſt**: "bronze coinage, which is common." >- **denario argenteo, & aureo**: "silver denarius and gold aureus" (types of Roman coins). >- **cum æmulo di-micandum**: "[he] had to contend/fight with a rival." >- **nitidus, & calamiſtratus**: "sleek/polished, and with curled/crimped hair" (referring to Otho's well-known vanity). >- **munditiarum pene muliebrium**: "of almost womanly fastidiousness/elegance." >- **in ære vileſcere**: "to become cheap/debased on bronze." >- **rubigine infici, ac fœdari**: "to be infected/tainted with rust, and defiled." >- **pecuniæ æreæ**: "bronze money." >- **in ærario ſupereſſet**: "remained/was surplus in the treasury." >- **magna contra caritas aureæ, & argenteæ**: "on the other hand, a great scarcity/dearness of gold and silver [coinage]." >- **huic ſubveniendum, illa primum diſtrahenda**: "[the scarcity of gold/silver] needed to be remedied/helped, [while] that [existing bronze] should first be distributed/disposed of." >- **Triumviri Monetales**: The official board of three men responsible for minting coins. >- **Othone in belli curas diſtraċto**: "with Otho distracted by the cares of war" (ablative absolute). >- **æs omne**: "all the bronze." >- **vetere nota S. C.**: "the old mark S.C." (Senatus Consulto - by decree of the Senate), a common mark on Roman bronze coinage. >- **propriiſque nominibus ad oram**: "and with their own names on the edge." >- **quorum æreorum**: "of which bronze coins..." (The sentence is incomplete in the provided text, likely continuing on the next page of the manuscript). > >The author is exploring various reasons why bronze coins of Emperor Otho are rare, despite his short reign, especially compared to his gold and silver issues. The total token count for these prompts was 7,959. The temperature was set at 1 (neutral, 0 being no creativity, and 2 being very creative). The AI breakdown on how it translated the text is interesting. This will help with confirming the translation and its accuracy. - Step 6 - Lastly I asked the AI "What sources did you draw from to perform the translation?" Its output was: >[! Output] >My translation is based on my training as a large language model. The process involves: >1. **Lexical Knowledge:** My training data includes vast amounts of Latin text from various periods (Classical, Medieval, Renaissance, Early Modern) and their translations into English, as well as Latin dictionaries and glossaries. This allows me to recognize individual words like "Antiquarios" (antiquarians), "ærei" (of bronze/copper), "percuſſi" (struck, as in coins), "imperium" (empire/power), "Senatus" (Senate), "nummi" (coins), "denario argenteo" (silver denarius), "aureo" (gold coin/aureus), etc. >2. **Grammatical Understanding:** I've learned Latin grammar (declensions, conjugations, case functions, syntax, common constructions like the ablative absolute, subjunctive uses, etc.) by processing countless examples. This allows me to parse the relationships between words in a sentence. For example, understanding that "Othone in belli curas diſtraċto" is an ablative absolute construction meaning "with Otho distracted by the cares of war." >3. **Contextual Interpretation:** The surrounding text helps me disambiguate word meanings and choose the most appropriate English equivalents. For example: >- Knowing the text is about Roman coins helps interpret "æs" as bronze coinage rather than just the metal bronze. >- The phrase "magna contra caritas aureæ, & argenteæ" is understood in an economic context to mean "a great scarcity/dearness of gold and silver [coinage]," rather than "charity." >- Recognizing names like "Otho," "Vitellius," "Claudius," "Nero," "Galba," and "Suetonius" places the text historically and thematically. >4. **Specialized Vocabulary:** My training includes texts related to history, classics, and numismatics, so terms like "Triumviri Monetales" (the three men in charge of the mint) and the abbreviation "S.C." (Senatus Consulto - by decree of the Senate, a common mark on Roman bronze coins) are familiar. >5. **Pattern Recognition:** I identify common Latin idioms, phrases, and rhetorical structures and find their closest English counterparts. > >Essentially, I draw upon a massive, integrated knowledge base derived from the text and code I was trained on. It's not a case of looking up words in a single dictionary one by one, but rather a complex process of pattern matching, contextual analysis, and generation based on the learned relationships between words and concepts in both Latin and English. > >For a text like this, my internal process would be similar to how a human expert in historical Latin and numismatics would approach it, relying on years of accumulated knowledge and experience with similar texts. > ## challenges The only challenge at this moment is getting familiar with Google AI Studio as I have not used it before. The UI is somewhat daunting but there are many tutorials via YouTube. One example is [Google AI Studio In 26 Minutes - YouTube](https://www.youtube.com/watch?v=13EPujO40iE) ## thoughts on where to go next Next is to confirm the translation, though it appears accurate. The bigger concern at this moment is how the AI can handle larger files with multiple images of text, or if the AI can perform the same tasks with a pdf that can be up to 500 pages. Processing in batches does not appear to be an option at the moment unless I can figure out how to use Google Gemini API to access my files locally to perform batch OCR. ### Related Links [[Wicked Byzantine Problems]]