Speaker
Description
Dictionary consultation in Ancient Greek presupposes the ability to reconstruct canonical lemmas from highly inflected word forms, a task that can be challenging for users at any level. The lemma-based access model was a material constraint in printed dictionaries, but it can be addressed in digital dictionaries by exploiting the possibilities offered by computational morphology. The aim of this prototype is to link inflected Ancient Greek word forms to their corresponding dictionary lemmas, allowing users to search directly from inflected forms rather than exclusively from canonical lemmas. Although presented as a prototype, the model is designed to be reusable and applicable to any digital Ancient Greek dictionary.
Tools enabling morphological analysis of Ancient Greek forms already exist, most notably the Greek Word Study Tool of Perseus Digital Library and the Morpho function on Logeion. However, in both Logeion and the Perseus Digital Library, dictionary consultation and morphological analysis remain functionally separate. In Logeion, the dictionary interface allows access only to forms attested in its database, while morphological analysis is provided through a separate tool (Morpho). Similarly, in the Perseus Digital Library, the Greek Word Study Tool provides automatic lemmatization and analysis, while dictionary resources such as the Liddell-Scott-Jones Greek-English Lexicon are accessed independently, without systematic interaction between the two components. Although the output of morphological analysers typically provides a lemma linked to a dictionary entry, dictionary access and morphological analysis are presented through separate interfaces and rely on independent datasets: dictionary access is based exclusively on the headword inventory, while the morphological analyser operates independently of this inventory.
The model addresses this limitation by enabling dictionary access through inflected forms rather than exclusively through canonical lemmas. When an inflected word form is entered, the system first searches it within a curated dictionary inventory, which includes both canonical lemmas and selected inflected forms. If the input is found, the system directly returns the corresponding lemma. Only if no match is found is the form processed through an automatic morphological analyser.
At the core of the prototype lies a curated dataset linking inflected forms to dictionary-approved lemmas and their morphological analysis. In its current implementation, the dataset does not aim to represent a full dictionary inventory, but focuses on Ancient Greek polystemic (multi-stem) verbs, a particularly challenging category due to stem alternation and irregular morphology. This restricted scope allows the prototype to address a well-defined morphological problem while testing the underlying lexicographic model.
From the user’s perspective, the prototype processes each query through a sequential workflow. When an inflected word form is entered, the system first attempts to match it against the curated dataset. If a match is found, the corresponding lemma and morphological analysis are returned directly. If no match is found, the input is passed to an automatic morphological analyser, which generates one or more candidate lemmas that can be linked to dictionary entries.
When the user enters a form, the prototype provides three possible types of results. A GOLD result is returned when the input form is attested in the curated dataset. An ACCEPTABLE result is returned when the form is not present in the dataset but the automatic analysis proposes a lemma that is included in the controlled inventory. An EXTERNAL result is returned when the automatic analysis proposes a lemma that is not attested in the dataset. These latter results are explicitly marked as automatically generated. This classification makes the origin of the result explicit, allowing users to distinguish between data derived from the curated dataset and data generated through automatic analysis, and thus to navigate dictionary access more transparently.
From a technical point of view, the prototype consists of two main components: a curated dataset of inflected forms linked to their lemmas and their morphological analysis, and an automatic morphological analyser, specifically Morpheus, used as a fallback. The dataset is stored as a structured table, while the analyser is queried only when no match is found in the dataset. The prototype has been implemented in Python, with Unicode normalization for polytonic Greek and a simple web-based interface developed in HTML.
This prototype demonstrates how a form-to-lemma access model can be applied to Ancient Greek, combining curated lexical data and automatic analysis to enable inflected-form search while preserving a clear distinction between different types of results. It is currently implemented as a local system and is not yet integrated into an online dictionary environment. Future work will focus on integrating the model into a fully functional digital dictionary and testing its application on a larger lexical inventory.