PrIx user interface
for the Chancery collection

Introduction

This search interface provides textual access to the iconic "Trésor des Chartes" (or "Chancery") collection of manuscripts, which encompasses 83,000 page images reproducing the medieval registers produced by the French royal chancery. The registers date from 1302 to 1483 and are handwritten mainly in Latin, French and Spanish. While automatic transcription of medieval sources is notoriously difficult because of the greatly variable handwriting styles, this corpus is even more challenging because of its multilingual content and the large number of archaic and abbreviated words. In the HIMANIS project (2015-2017), this large image collection was made searchable through plain-text queries using Probabilisitic Indexing (PrIx) technology. The present search interface is a new and enhanced version of a prototype interface developped in HIMANIS for demonstration purposes.

If you are using this resource and find it useful, please include in your publications a reference to:

• Théodore Bluche, Sébastien Hamel, Christopher Kermorvant, Joan Puigcerver, Dominique Stutzmann, Alejandro H. Toselli, and Enrique Vidal: «Preparatory KWS Experiments for Large-Scale Indexing of a Vast Medieval Manuscript Collection in the HIMANIS Project». In 14th IAPR International Conference on Document Analysis and Recognition. ICDAR 2017 proc., pp.312-17, 2017. DOI: 10.1109/ICDAR.2017.59.

User interface guidelines

• Queries are typed on the query box, on the left of the "Search" button.
• Text typed in the query box is automatically case and diacritics folded. For example, "Paris" becomes "paris", "Sébastien" becomes "sebastien" and "Août" becomes "aout" (without the quotes).
• A confidence box and slider are provided to express a confidence threshold which determines the desired precision-recall tradeoff.
• A "Max. results" box is provided to indicate the maximum number of spots to be retrieved and shown. For example, if this number is set to 20 and the confidence is set to 40%, the system shows at most 20 spots with confidence higher than 40%; and if no (or 0) confidence is set, the system retrieves at most 20 spots with the highest confidence.


• Text images are organized according to this hierarchy: HOME, Collection, Volume and Page
• Click on the front page (HOME) "Chancery" banner with no query to enter the full Chancery collection. Then you can see the 199 Volumes indexed, each showing the number of images it contains (ranging from less then 100 to more than 1000).
• At each level of the hierarchy (except the lowest, page level), the elements are shown by means of thumbnail images or miniatures, along with some information; namely, a) an identifier of the element, b) a confidence bar representing the system's self-reliance that the query appears somewhere in this element and c) the number of elements in the lower level where the query may appear with a confidence above the threshold specified. By hovering the mouse over the confidence bar, the actual confidence value is shown.
• Queries can be formulated at any level: HOME, full Collection (chancery), specific Volume (JJ...), or individual Page.


• Pan and Zoom. When a (part of a) page image is shown, it can be explored by moving the mouse while holding its right button. In addition, the mouse wheel can be used to zoom in and out.
• Double-click. At any point of an image, the left mouse button can be double-clicked to show all the words or character sequences which are indexed at this point in the image PrIx.
• Tablets and smart phones. With minor limitations, this user interface is also operational for touch-screen tablets or smart phones.

Types of queries

• Abbreviations are not queried as they appear in the images; instead the corresponding expanded forms are used. For example, use "paris" to find "Paris" and "par.", or "johannes" to find "Johannes", "Joh^s", "Johẽs", etc.


• Approximate spelling can be used for any query word. The special symbol '~' can be appended to any word to allow searching for words which differ from the given one in at most one character. Larger dissimilarities can be specified by appending a number after the '~' symbol. For example, use 'Theodore~' to find 'Theodore', 'Teodore', 'Theodori', 'Theodoro' and 'Theodores', or 'Theodore~2' to additionally find 'Theodorum', 'Theodorici', 'Teodoro', etc.
  
• Wildcard spelling can be used for any query word. The symbol '*' is used as a wildcard representing any character string. For example, use 'Theodor*' to search for 'Theodoro', 'Theodore', 'Theodori', 'Theodorici', 'Theodorum', etc. Similarly, the symbol '?' can be used as a wildcard representing any single character. For example, use 'Johann?s' to search for 'Johannas', 'Johannes', or 'Johannis'.
  
• A large dissimilarity degree and/or a small number of real characters in approximate and wildcard spelliling typicllly lead to huge, useless amounts of retrieved spots, more so if the query is issued at a high hierarchy level (e.g., at the root or HOME). To avoid these useless queries, the maximum dissimilartty and the minimum amount of real characters are limited depending on the hierarchy level where the query is issued.
  As a general recomendation, approximate and wildcard spelling should be used at higher hierarchy levels mainly for long words, and using only the smallest spelling dissimilarity (1), or a minimum number of wildcard symbols.
  Both approximate spelling and wildcards can entail large server computing demands and should be used with care!


• Compound queries. Individual words can be combined into "compound queries" in three ways: a) boolean (AND, OR, NOT) queries, b) sequence queries and c) macros.
• Boolean queries. AND, OR, NOT relations are interpreted at the full page-image level and the number of matches is the total number of word occurrences matching the query. The operators are: AND: "&&" (or just blank space), OR: "||", NOT: "-" (before the words which have to be negated). Parenthesis "(", ")", can be used for grouping: Examples: "paris && (france || flandres)": pages with at least one instance of "paris" AND at least one instance of either "france" OR "flandres", or both; "france - flandres": pages with at least one instance of "france" but with no instance of "flandres".
• Proximity AND queries. Are AND queries with a number between the two "&" symbols to specify how far apart the AND components are allowed to be. The number is a percentage of the whole image size. For example, 'Angleterre &10& France' retrieves images where 'Angleterre' is at most 10% of the page images size apart from 'France'.
• Sequence queries. Are sequences of words, expressed using the symbols "[" and "]". Sequences are not exact strings; they allow extra words to appear among the stated words. Examples: "[ludovico francorum]" (where "francorum" is an expanded abbreviation) would retrieve pages with strings such as "ludovico franc.", "Ludovico rege franc.", "ludovico dei gra. francorum", etc. The number of matches is computed at page-image level and one match corresponds to the complete sequence. Sequences can be seamlessly mixed with boolean operators.


• Macros are predefined complex queries generally used to retrieve information which is semantically homogeneous. In the current system version, macros can only be defined by the system administrator. Macros can be used by means of their predefined names preceded by the "@" symbol. The default system configuration provides six examples of macros: "@mois" which retrieves the names of the months of the year in french, "@mensis" for the month names in latin, "@moismensis", for all the month names in french and latin, and symilarly for weekday names: "@jour", "@dies" and "@jourdies". Macros can be seamlessly mixed with word sequences and boolean operators. The concrete definitions are as follows:
  
  @mois="(janvier || fevrier || mars || avril || mai || may || juin || juillet || aout || septembre || octobre || novembre || decembre)"
  
  @mensis="(januari* || februari* || marti* || aprilis || maio || maii || junio || junii || julio || julii || august* || septemb* || octob* || novemb* || decemb*)"
  
  @moismensis="(@mois || @mensis)"
  
  @jour="(lundi || mardi || mercredi || jeudi || vendredi || samedi || dimanche)"
  
  @dies="(lune || martis || mercuri || mercurii || iovis || veneris || sabbat* || [(die||diem||dies) dominic*])"
  
  @jourdies="(@jour || @dies)"


• The approximate locations of spotted queries in each page are marked with rectangles (called "bounding boxes") surrounding the corresponding words. The color of a rectangle expresses the degree of confidence the system has in the corresponding spot. Exact confidence values can be seen by hovering the mouse over the rectangles.
• If a word appears more than once in a text line, only the instance with greatest confidence is shown.
• The precision-recall tradeoff desired for each query is specified by mans of the Confidence threshold. The default value is 50%. A higher value results in more precision (little or none wrong spots) and less recall (some, or many existing instances of the query may not be retrieved); a lower value results in less precision (some, or many wrong spots) and more recall (all or most the existing instances can be retrieved).
• Queries are interpreted, and results are shown, hierarchically. If the query is issued at the top level (HOME), the system first shows the Volumes where the query may appear with a confidence above the threshold specified. Then if the same query is issued for a specific Volume, the system shows all the Pages where the query may appear with a confidence above the threshold. Finally if it is issued for a specific Page, the system shows locations where the words involved in the query may appear with a confidence above the threshold.

Query examples