Jokes represent one of the most understudied sources about nineteenth century society. Due to their ephemeral nature they slipped from attention as soon as they were no longer funny or topical. Digitisation of newspapers and books has made them available again, but due to their short nature they are not easily accessible through current generic keyword-based newspaper search systems. In this paper we present the Old Jokes Archive, which aims to provide a digital archive focused solely on jokes. The archive will support the full process from initial text acquisition to search and nally re-use by both academic and general public users.
Jokes represent one of the most ephemeral spoken (or written) interactions [
While jokes are often treated in a disposable fashion, they have nevertheless
played important roles in many historical cultures. In nineteenth-century Britain,
for instance, the possession of a good new joke represented signi cant cultural
capital, for to be a true wit was a position of social distinction [
The availability of jokes in newspapers also means that, unlike longer,
humorous novels and stories, jokes reached a much wider audience spanning all
social classes. Thus a single joke could in one week reach as many readers as
a best-selling novel might throughout its author's life-time (see for example [
At the same time, while the second half of the nineteenth century saw the
introduction of copyright laws [
The jokes themselves are a potentially invaluable source of information about historical cultures, as they are typically built upon an assumption of shared knowledge; a belief that an audience will immediately understand how a stock character (such as a mother-in-law) behaves, how a familiar social situation should ordinarily play out, or what to respond with when somebody says `knockknock.' Historians can reverse-engineer these jokes in order to uncover the ideas and attitudes that joke-writers and their editors assumed were widely held at the time. The subjects of jokes, and the dynamics of laughter (who is laughing at/with who), can also reveal valuable new insights into the power relations at work in historical communities.
Even though they represent such a rich data-source, jokes are not widely used by most historians. Many Victorianists, for instance, rarely venture beyond the cartoons of Punch magazine when attempting to make sense of the period's comic culture. Millions of historical jokes have never been examined by historians and are therefore ripe for further exploration.
The chief obstacle to this research centres on the di culty of nding and accessing historical jokes. Many were never recorded and have now been lost to us, while even those that were written down and preserved in books and newspapers are tricky to uncover. The large-scale digitisation of newspapers, books, and other print archives presents new opportunities for solving this problem. However, even with the help of keyword searches it remains di cult for researchers to locate speci c jokes pertaining to their interests. For example, a keyword search for the word 'lawyer' in a typical digital newspaper archive will return a jumble of millions of news stories, adverts, editorials, letters, serialised stories, and poetry, within which we might also nd jokes about the legal profession. At present there is no straightforward way to search speci cally for historical jokes. 2
The Old Jokes Archive (OJA) aims to address this gap, by providing the rst, large-scale digital repository of historical humour, targeted both at academic researchers and the wider public. To support this the OJA provides a range of functionalities centred around two areas: the acquisition of the joke data and search and (re-)use functionality. 2.1
The joke acquisition starts based on the existing digitisation of newspapers and joke books, with the nal goal being annotated text versions that can then be used by researchers and the general public. Throughout the process a semiautomatic approach is used, combining automatic image and text processing with manual validation and post-processing using a crowdsourcing approach. In the OJA's precursor { the Victorian Jokes Archive { we have tested a number of crowdsourcing methods, in particular to improve the accuracy of crowdsourcing results, the experience of which will be integrated into the OJA. The joke acquisition process follows these ve steps: 1. Identi cation As the digitisation has focused on whole newspaper pages, the rst step is in the identi cation of those areas of the scanned pages that contain jokes and then the splitting of those areas into the individual jokes. We are investigating automated methods for this, but have initially adapted techniques from the Digital Playbills project. 2. Transcription OCR will be used to create an initial transcription. Newspaper paper and typeface can be relatively poor (see Figure 1), resulting in comparatively high error rates. To deal with this we have developed error classi cation heuristics, that allow us to determine the quality of the OCR output. Based on this crowdsourcing users can be o ered a choice to work on a transcription that has a low error rate { mostly typo-correction { or a high rate { essentially transcription from scratch. Using this both users who have signi cant time to invest or those who just want to do something quick and simple can be o ered appropriate crowdsourcing jobs. 3. Classi cation The resulting corrected transcription is then classi ed using automated heuristics developed previously. The classi cation works at a high level, distinguishing categories such as question-and-answer, dialog, or puns. The classi cation is then used by the following steps to apply categoryspeci c heuristics. 4. Segmentation The joke's text is segmented into chunks. The exact chunks depend on the joke category determined in the previous step, but for example for question-and-answer jokes this would be segmenting the question and the answer element, while for dialog jokes it includes identifying speakers, spoken text, and asides. Also more generally we have developed heuristics to identify joke titles and attribution. 5. Annotation The chunks are then, where appropriate, annotated with speci c meta-data extracted from the chunk. Among others we have developed heuristics to identify dialogue speaker gender and are currently working to automatically identify social class, age, and jobs of speakers.
After running through all ve steps, the joke texts will be publicly available through the online archive. Jokes that have been OCRed, but lack the error correction and further processing will also be available, if the user also wishes to see those. 2.2
In order to make the resulting archive available, the OJA will be available online
and provide a range of access points to explore, search, and use the jokes:
{ Search The OJA will provide a state-of-the-art faceted search system
enabling users to narrow their search for jokes via keywords and via the
categories and annotations created in the acquisition stage.
{ Exploration While search works well for users who know what they want,
for the general public a more open-ended, browsing based interface will be
developed. Based on previous work [
While initially the focus will be on English-language jokes, the project will be built with multi-lingual content in mind. Additionally content will be available through a permissive open-culture license to encourage use and re-use both academically and for private reasons.
Jokes represent a so far largely untapped resource for investigating a range of historical questions, from language use to the spread of ideas. The Old Jokes Archive will act as a central data-source point, starting with nineteenth century English-language jokes, but with an aim to expanding both temporally and linguistically. The tools, methods, and practices developed in the course of this project will be applicable to future archival projects that aim to 'remix' existing digitised content by extracting, organising, and re-presenting it in new ways.
While the initial focus will be on the joke's texts, the long-term aim is to also consider visual aspects in the source data, such as wehn jokes were accompanied by drawings.