[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-117792-en":3,"doc-seo-117792-105":30,"detail-sidebar-cat-0-en-105":83},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},117792,687197100911,"Himbo","https://ap-avatar.wpscdn.com/avatar/a000239b6f1da00475?x-image-process=image/resize,m_fixed,w_180,h_180&k=1785132997149421697",6,"Technology","Logion - Machine Learning for Greek Philology","Machine-learning methods are developed to solve multiple tasks in Greek philology using a BERT model trained on a large premodern Greek dataset. The system identifies and corrects scribal transmission errors that had not been detected previously. It also fills textual gaps arising from manuscript material deterioration and benchmarks outputs against a domain expert. Best results occur when expert users receive model suggestions, and analysis of interpretability links specific attention heads to grammatical features of premodern Greek.","Logion: Machine Learning for Greek Philology  \nCharlie Cowen-Breen 􀀃 Creston Brooks y Johannes Haubold z Barbara Graziosi z  \narXiv :2305 .01099v1 [ cs .CL] 1 May 2023  \nAbstract  \nThis paper presents machine-learning methods to address various problems in Greek philology. After training a BERT model on the largest premodern Greek dataset used for this  \npurpose to date, we identify and correct previously undetected errors made by scribes in the process of textual transmission, in what is, to our knowledge, the first successful identification of such errors via machine learning. Additionally, we demonstrate the model's capacity to fill gaps caused by material deterioration of premodern manuscripts and compare the model's performance to that of a domain expert. We find that best performance is achieved when the domain expert is provided with model suggestions for inspiration. With such human-computer collaborations in mind, we explore the model's interpretability and find that certain attention heads appear to encode select grammatical features of premodern Greek.  \n1 Introduction  \nPremodern Greek texts have been preserved in manuscripts which feature gaps caused by the deterioration of materials (e.g. papyrus, parchment, paper) and errors introduced by scribes. In the case of some premodern texts, scribes copied what survived from earlier exemplars, now lost, in a long process of textual transmission. Homer's Iliad, for example, dates to the 8th c. BCE and reaches us via a process of hand-copying, in relay, over many centuries: the first extant papyrus fragments date to the 3rd c. BCE and the first manuscripts of the whole poem to the 10th c. CE. To produce editions of premodern Greek texts, philologists identify scribal errors which accrued in the course of textual transmission, try to emend them, and fill gaps caused by material deterioration.  \n􀀃 Department of Pure Mathematics and Mathematical Statistics, University of Cambridge  \ny Department of Computer Science, Princeton University z Department of Classics, Princeton University  \nIn a related study, Assael et al. train a multitask transformer-based model to date, place, and fill gaps in ancient Greek inscriptions [1] . Inscriptions display the original text, whereas most of what survives from antiquity reaches us via a long tradition of hand-copying from earlier exemplars. For this reason, Assael et al. focus on gaps caused by physical damage but not on copying errors.  \nThe approaches outlined here are designed to serve philologists working on all Greek texts preserved via manuscript tradition. In particular, we introduce the intellectual work behind Logion, a framework for assisting philologists in their work.1 The name means \"oracle\" in Greek, and we chose it to emphasize the need to interpret machinegenerated results. In a proof-of-concept paper, we have already used Logion to find previously undetected errors in premodern Greek texts [2] . In this paper, we describe and study the approaches used to arrive at those results. More generally, we show that several tasks associated with philological research are suitable for contextual language models.  \n1.1 Structure of the paper  \nIn section 1, we outline the training procedure fora premodern Greek BERT model [3], with what we believe to be the largest dataset used for this purpose to date.  \nIn section 2, we state the problem of detecting scribal errors and outline the approach we used in [2] to discover previously undetected errors in the work of Michael Psellus, an 11th-century Byzantine author. To validate the approach, we randomly generate errors that simulate those made by scribes and study the effectiveness of the model at discovering them.  \nIn section 3, we randomly generate artificial gapsand compare the suggestions of a domain expert to those produced by the model for filling these  \n1[https://github.com/charliecb/Logion.git](https://github.com/charliecb/Logion.git).  \ngaps. We find that gaps are filled with the highest a","cbCaiiTtau4gW5Pf","https://ap.wps.com/l/cbCaiiTtau4gW5Pf","pdf",719093,1,14,"English","en",105,"# Introduction\n## Structure of the paper\n## Training procedure","[{\"question\":\"What is Logion and what does it help with in Greek philology?\",\"answer\":\"Interpretability analysis is performed by examining attention heads. Certain attention heads appear to encode select grammatical features of premodern Greek, connecting model internals to linguistically meaningful information.\"}]","Logion - Machine Learning for Greek Philology | PDF",1785679597,35,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":78,"head_meta":80,"extra_data":82,"updated_unix":28},"logion-machine-learning-for-greek-philology","",{"@graph":36,"@context":77},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/technology/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/logion-machine-learning-for-greek-philology/117792/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-02",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71],{"name":72,"@type":73,"acceptedAnswer":74},"What is Logion and what does it help with in Greek philology?","Question",{"text":75,"@type":76},"Interpretability analysis is performed by examining attention heads. Certain attention heads appear to encode select grammatical features of premodern Greek, connecting model internals to linguistically meaningful information.","Answer","https://schema.org",{"og:url":52,"og:type":79,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":81,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":84},[85,89,93,97,102,105,110,115,120,123,127],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":86,"show_sort_weight":87,"slug":88},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":90,"show_sort_weight":91,"slug":92},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Exam",70,"exam",{"id":98,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},5,"Comic",60,"comic",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":103,"slug":104},50,"technology",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},7,"Healthcare",40,"healthcare",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},8,"Research & Report",30,"research-report",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},9,"Religion & Spirituality",20,"religion-spirituality",{"id":118,"doc_module":4,"doc_module_name":46,"category_name":121,"show_sort_weight":118,"slug":122},"World Cup","world-cup",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":124,"slug":126},10,"Lifestyle","lifestyle",{"id":128,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":98,"slug":130},19,"General","general"]