PDD (Poetic Devices Detector) Introduction Disclaimers WORKING WITH PDD Text Preparation Viewing Results Viewing Visualizations Resetting PDD
Introduction PDD is an Excel-based distant reading research tool created by Avraham Roos (
[email protected]) who is doing his PhD research (UvA) on the English (re)translations of the Hebrew Haggadah, the semi-liturgic booklet religious Jews read out-loud on Passover Eve in a family-setting. Roos is creating a digital corpus of more than 50 different English translations dating from the first (1770) to modern ones and is using digital tools to analyze and visualize these. More about his research HERE and HERE. The Haggadah text contains many poetic devices and as such is a real challenge for translators. Some disregard most and focus on accuracy of content while others attempt to imitate the textual structures as closely as possible. This tool helps to differentiate between these two groups by focusing on and visualization of three distinct poetic devices: End of line rhyme, end of line word repetition and initial letter abecedarian acroscicons. The initial hypothesis is that a high frequency of any of these (or a combination) in comparison to other translations in the corpus will point to a translation in which also other poetic devices are distinct. This can thus be used as a starting point for further close reading. Although in principle one could just read through all versions in the corpus to manually identify the poetic translations, this is not practical in larger corpora. In addition, the tool sometimes highlights places of interest easily missed with cursory reading. The tool in its present state is in the proof of concept state. Some haggadot translations have been entered together with several comparison texts (Othello Act 1 and Wikipedia Articles on Neville Chamberlain and Ogden Nash.) Upon downloading the tool it will reside on the local host, meaning results for new texts entered by the user will be saved in the user’s program only and will not be visible/ shareable with others. Disclaimers: The current version of PDD: ● will only work when macros are enabled ● will only work properly on English language based Excel versions. ● saves the final score allotted to any checked version and its breakdown comparing it to all earlier checked versions but does not save the texts and visualizations themselves. ● does not work for texts longer than 3000 lines. (Othello Act 1 = 935 lines)
WORKING WITH PDD Text Preparation In order to use PDD, one needs a text broken up into lines. All hyperlinks and non-alphanumeric characters need to be removed. All blank lines and empty spaces at beginning or end of lines should be removed. There are several tools that can do all this, I use a macro in Word to automatically format the text. PDD depends on IPA (International Phonetic Alphabet) transcription. In order to transcribe the text mentioned above, please enter it into the Lingarado website: http://lingorado.com/ipa/ and copy the transcription. If the source text is in British English, select this option for transcription, if in American English select that one. Running PDD PDD opens on the welcome screen:
1. Write version name in cell C5 2. Paste text in left column and IPA in right column. If IPA has hyperlinks, these should first be removed by copying into notepad and copying from notepad into PDD. 3. Press Run
Viewing Results Upon completion of Run, user is taken to final results sheet (sheet 4).
The last version run on PDD will be highlighted in yellow (column A) ranked according to its final score (column G) in the list of all priorly checked versions. PDD counts the number of times it finds end of line rhymes (= RH), end of line repetitions (=R) and evidence of abecedarian acroscicons (=ABC) This final score (=F) is calculated as follows: F=5*RH+R+5*ABC RH is acknowledged for AA, ABA and ABCA rhymes. ABC is acknowledged for any three or more successive letters This scoring system is based on trials and proved the most successful for our purposes.
The higher F, the more “poetic” the version is expected to be. For this proof of concept version text length is not taken into consideration as can be seen from the fact that Othello Act 1 scores much higher than a shortened version of the same text. However, this tool will primarily be used to compare Haggadot which are all approximately the same length. When comparing two excerpts from Wikipedia each of 1330 words long with the same amount of text from shortened Othello we see that PDD ranks Othello higher than the Wikipedia on Ogden Nash which contains several short poems and an abecedarian acrostic and places the regular text on Chamberlain last. The importance of choosing the correct IPA transcription (British or American English) can be seen from the following results screenshot where PDD checked the same haggadah twice, once with American IPA and once with British. The final scores for these two differ indicating that PDD recognized more rhymes when American IPA (pronunciation) was used. Not all words rhyme the same in the two Englishes:
Viewing Visualizations To view PDD visualizations, click on the link or go to Sheet 1
PDD highlights in red end of line repetitions:
PDD highlights in green end of line rhymes:
The visualizations of repetition and rhyme show three columns. The one on the left looks at a row X and compares it to X+1, the middle column looks at row X and compares it to X + 2 while the one on the right looks at X and compares it to X + 3. For rhymes this would correspond to finding the rhyme patterns AA, ABA and ABCA respectively. RH is defined by last vowel SOUND and following consonant sounds (This is why IPA is used) so that words like night and kite will come up as a rhyme: naɪt kaɪt. PDD does not take stress into consideration nor differentiates between two syllable and one syllable rhymes because the tool is meant as a very crude initial distant reading tool that will be a catch-all for any rhyme (male and female). Areas identified as poetic by PDD should be closer examined using existing poetry tools (such as Poemage, PoemViewer, RhymeDesign) which do identify and differentiate between male and female rhyme, stress and meter and are all defined as tools “in support of close reading”. The above shown example is especially interesting because PDD reveals here a TT (English) rhyming pattern in a part of the ST where the Hebrew original does NOT rhyme, thus highlighting a part of the text that might otherwise have been overlooked in a large corpus proving PDD’s value. In order to see visualization of ABC patterns, click on Sheet 3 (ideally all visualization should take place on the same sheet but for this proof of concept it was decided to stop programming).
The following example shows how PDD highlights in orange ABC patterns:
Only instances of three or more letter sequences are labeled by PDD as true. This is done to prevent accidental scoring of two consecutive letters.
Resetting PDD At the end of each run PDD can/ should be reset by clicking the “clear” button in any of the Sheets 1, 2, or 4. The tool is ready to be re-used but the list of checked versions and their score is saved and reordered each time a new version is analyzed. If this list is important, do not forget the save PDD before exiting it completely.
I would like to thank the forum members of www.MrExcel.com who have helped me with my Excel problems whenever I couldn’t figure out how to do something. Without them I would not have been able to build this!
A.Roos, 2016
