EventParser and EpiTest A User Guide
Contents
1. EXPOSITION class and predicate OBJECT DISPOSITION class and predicate social isolated or ferryman oe social requested from 18 to character 177 from 178 to character 283 IBy the great river which was newly swollen with heavy In the middle of the night loud voices wakened him it a rain and overflowing the old ferryman weary from the seemed that travelers wanted to be ferried across toil of the day lay in his little hut and slept v Currently selected EYENT 1 Total number of EVENTS 1 Return to EventParser Fig 1 6 EVENT BASE window Clicking the RETURN TO EventParser button returns us to the main window where a range of additional menus and buttons can be used to invoke further functions In particular the KNOWLEDGE BASE menu allows us to examine the knowledge database that is incrementally expanded as successive predicates are declared The knowledge database is saved in a file with the sbf suffix sememe database file and is displayed in the SEMANTIC DICTIONARY window where semantic terms appear in the order in which they were entered or generated In the example screenshot in figure 1 7 the knowledge database contains two pairs of opposites two pairs of direct antonyms and two pairs of direct synonyms The antonyms and synonyms are direct because their elements belong to the same predicate class EventParser and EpiTest A User Guide 11 EventParser 4 36 Semantic Dic2. with company 450 615 social at home 2542 2836 episode elSO2 6 ferryman physical asleep 18 177 cognitive hearing 178 283 ferryman social reacting 290 486 social unaffected 2470 2576 episode elSO2 7 snake social cooperative 6607 6767 social unable 6974 7122 snake cognitive unacquainted 8575 8873 lzi X Fig 2 3 List of EPISODES generated by EpiTest In this case the original fairy1 esf file contains only ten EVENTS but even so the EPISODE list displays a total of forty two virtual EPISODE constructs Representing EpiTest output in a list format like this makes clear that a EventParser and EpiTest A User Guide 17 considerable number of EPISODES theoretically possible ways of combining EVENTS can be identified by using a computer to analyse reception records esf files produced in EventParser The output does not lend itself to qualitative analysis it is not a reading whose merits can be evaluated by critics and the methods behind it cannot be judged against the aesthetic standards of a hermeneutics of interpretation It is obviously meaningless to ask which of the forty six EPISODE constructs or thirty eight ACTION constructs in our data is more correct more satisfying or even better in the s
3. If we enter our ten basic events in the sample file fairy esf and analyse this file in EpiTest the program generates the same five EPISODES as those we found manually It also produces 246 METASYNONYMS which allow it to uncover hidden semantic connections and more than quintuple the size of its semantic database On this extended basis EpiTest generates forty one additional unpredicted EPISODE constructs each of which fully satisfies the criteria of one of our five categories EISO _3 and EANISO _ Now these results could have been worked out manually given a few sharp pencils plenty of paper and even more patience but not in the 0 33 seconds which our program needed to complete the task When it has finished processing a file EpiTest presents the display shown in figure 2 2 System specification 2 66 GHz Pentium IV processor with 512 MB RAM Windows XP EventParser and EpiTest A User Guide 15 EPITEST version 3 85 a zialki Combinatory analysis of EVENT base files for EPISODE and ACTION constructs Programming and development jan c meister uni hamburg de CONSULT C eventparser fainy1 est CONSTRUCT C eventparser fairy1 aniso con Min Max Wirt Event Integration EVENTS 10 EPISODES gt 5 Ms wees Direct Indirect Opposite i xEpi SYNONYMS as ep Sea ACTIONS a a Ae eis earan META SYNONYMS 246 20 8 SAVE AEN FIOR 235 294 Generate isochrone EPISODES Total EIS01 EIS02 EISO3 19 4 11 4 I Generate anisochrone EPI
4. broad formal terms cognitive task which humans perform so easily and find so deceptively simple when they read action EventParser and EpiTest A User Guide 21 It should be emphasized again here that our final quantification of the ACTION PRODUCT is in no way meant to be an objective measure of the coherence of a given narrative text This is obvious not least because in itself a number like 235 294 says absolutely nothing Is 235 294 a lot Is it a little The question can be answered only in differential terms never absolutely For the time being then this modest number is nothing more than the result of a cognitive formula which is evaluated when we form EPISODE and ACTION constructs during the comprehension of texts and the reading of ACTION It is a formula that involves many different variables and dynamic linking processes of which we can be sure that only a fraction are known to us Nevertheless it is a fact that we can obtain results perfectly compatible with human intuition by applying our theoretical arguments and the practical model which culminates in the ACTION product which correlates real empirically identified EVENT constructs and algorithmically derived virtual EPISODE and ACTION constructs The reader will see this for himself in the final part of our study and if that fails to convince him he can experiment further with EventParser and EpiTest on data of his own choosing He might for example follow the author an
5. illustrated in figure 1 4 we therefore provide the DISPOSITION with the requested predicate which belongs to the same class social as the expo sitional predicate 8 1 The EventParser Program EventParser 4 37 EVENT Parsing Window TextBaseFiles Phrase Definition EVENT Definition Knowledge Base C COMP ACT EVPARSER FAIRY1 TBF FOCUS on OBJECT _ EXPOSITION status of OBJECT DISPOSITION status of OBJECT Class Predicate Class Predicate Cancel EVENT Definition ferryman social isolated social requested Redo Redo Redo requested integrated isolated In the middle of the night loud voices wakened him it seemed that travelers wanted to be ferried across NEXT STEP w Show EVENT Base file PHRASE MARKING PHRASE MARKERS STRING SEARCH _ CURRENT STRING Enter or select from _ f lt gt gt gt Phrase Definition Please define ERER rom 178 menu to 283 Fig 1 4 EVENT PARSING window with predicate list As this involves the use of a descriptive term which the program has not encountered before we have to enter it manually in the list box on the right hand side of the main window which now also contains integrated and isolated the terms that we entered when defining the EXPOSITION As before we are prompted to define the new requested predicate in the PREDICATE DEFINITION window which now displays a semantic square in which three posi
6. in an unbroken chain what EPISODES P y can be linked to in an isochronous order The answer obviously depends on the connective potential of the initial construct a in each case In the case of our Successive new connections will obviously become possible as we move through the second third and subsequent positions in the list The corresponding ACTION lists are represented as actlist 1 2 3 n in the program syntax However EpiTest generates them with an algorithm that is partially deterministic rather than fully recursive it takes from the database only the first EPISODE construct which isochronally follows its predecessor on any occasion A fully recursive algorithm would generate a huge number of combinatorial variants and we are not yet in a position to place such a burden on reader or computer Practical tests have shown that it takes long enough to calculate deterministic ACTION constructs alone of the 0 39 seconds needed to process the Matrix esf file only about 0 05 seconds were required to assemble the EPISODE constructs the rest of the time was spent building ACTION lists Processing Unterhaltungen esf the complete mark up file for the Conversations of German Refugees our example text in part 3 took over four hours on a 266 MHz Pentium II processor Advances in computer technology since the time of writing will have reduced this figure considerably by now If nothing else these figures illustrate the mag
7. literary action actually is But it is certainly worthwhile asking how and why we believe that a certain amount of ACTION is present in a given narrative It is this question which the programs EventParser and Epitest can help us to answer 3 Installation Guide EventParser and EpiTest have been designed for PC DOS computers running Windows operating software EventParser 4 37 has very modest system requirements and will normally work on a 486 Hz cpu machine with 8 MB RAM and an operating system from Windows 3 11 upwards tested on 95 98 2000 and XP on an ME platform the program was found to be instable EpiTest 3 85 is decidedly more demanding The minimum configuration on which it tested positively was an Intel Pentium P I 166 MHz 32 MB RAM Windows 95 2 also tested successfully under Windows 98 NT 2000 und XP on PII IN and IV machines Depending on the hardware configuration certain restrictions apply in terms of the complexity of esf files handled and the length of episode and action lists available for inspection in the user interface Under Windows 98 an attempt to read in esf files of more than 50 kb i e more than approx 50 events or to inspect output files which may have been successfully generated by EpiTest but exceed the display box memory limitation can lead to a program crash However no loss of data is to be expected The EpiTest analyses discussed in part 3 of Computing Action A Narratological Approach were ini
8. segmented EventParser is started by executing the file named ep437 exe Known bugs 1 Activating the option Include pre defined terms in the KNOWLEDGE BASE menu prior to the first processing of a new tbf file leads to a conflict with the succeeding automatic creation of anew sbf file Solution create an empty sbf file prior to invoking EventParser using a simple text editor Notepad etc and save it in text only format into the same directory as the associated tbf file which you wish to process The sbf file root name must be identical to that of the bf e g myfile tbf and myfile sbf 26 3 Installation Guide If during the declaration process for a new EVENT only a partially new definition of semantic terms is done i e if the dispositional term is taken from an exiting term list and then combined with a newly introduced expositional term in the PREDICATE DEFINITION window that same term will reappear in the bottom right box when next opening the PREDICATE DEFINITION window However this is merely a display error the old term will not be saved in the course of the second term declaration Whenever any character is entered into the text box on top of the term lists the program will interpret this as the definition of a meaningful new semantic term Unless you manage to exit from the PREDICATE DEFINITION window the only way to identify nonsensical or accidental term entries is to attach an easily identif
9. to as the EVENT corresponds to an assertion by the narrator or an assumption by the reader that two causally or chronologically linked states represent a transformation in the properties of a fictive object Points a to d below will serve as a brief reminder of how EVENTS are defined and an outline of how they are encoded the construction of EVENTS is known as encoding in Event Parser terminology Note that for brevity s sake EventParser as well as the EpiTest program to be described in chapter 2 use shortened terms and refer to the two types of STATE OF AFFAIRS 1 e expositional vs dispositional states as two STATUSES to a fully defined expositional STATE OF AFFAIRS as EXPOSITION and to its counterpart as DISPOSITION a It must be possible to assign every predicate to a pre existent predicate class b A FOCUS must be present where a FOCUS is an identifiable narratively based perspective of perception of an object in the narrated world c Under this FOCUS the state of a fictional object of perception must be seen to be distinguished by predicate at point in fictional time Such a complex of object and predicate is referred to as a fictional STATE OF AFFAIRS The first STATE OF AFFAIRS in an EVENT construct is termed the expositional STATE OF AFFAIRS EXPOSITION d Under the same FOCUS or a different FOCUS which has an equivalent epistemological function i e which fulfils the same epistemological truth condit
10. to define the name of the results file con suffix Once the program has completed its calculations and displayed the 16 2 The EpiTest Program message Database analysis completed the results can then be displayed in text fields EPISODES opens a list of the EPISODE constructs that have been found ACTIONS shows the possible ways of linking these EPISODE constructs in the form of a PROLOG predicate action 1 2 3 n where each of the numbers inside the parentheses is the index of an EPISODE which makes up the ACTION SAVE allows the results of the various calculations to be stored in a separate daf data analysis file This file can be opened later in any text editor and contains a one page statistical summary of the results EXIT quits the program The panel on the bottom right hand side displays the number of EPISODES that were generated in each category and the total number of ACTION constructs they produced In the lower left hand comer we can see the time taken by EpiTest to complete its calculations When EpiTest has finished generating EPISODE constructs on the basis of the fairy1 esf file and stored them in the fairy aniso con results file the user can click on EP SODES to display a list of the EPISODE constructs which were found as shown in the screenshot in figure 2 3 EPITEST version 3 85 ioj x Combinatory analysis of EVENT base files for EPISODE and ACTION constructs Programming and development jan c m
11. A typical topic of study would be what are the differences between the ACTION PRODUCTS of a set of sub narratives embedded in a single overall narrative b One text and n readers Research of this type usually analyses statistical patterns in how different readers interpret a single text A possible research topic might be given a single text and two groups of readers how and why does the mean ACTION PRODUCT of the readers in one group one differ from that of the readers in the other c N texts and one reader In this case we have a single reader whose individual way of reading action provides a means of analysing historical developments in a genre or the works of a given author It is basically a variant of a a typical theme might be how do differences in ACTION PRODUCT relate to the early and late novels of a particular author d N texts and n readers This case involves the differential analysis of a text sample using multiple readers The approach might be most profitable in a study where action is just one aspect of a broader empirical analysis of the style of an individual author or movement 24 2 The EpiTest Program As can be seen the four different types share the use of differential analysis to explore empirical data In Computing Action A Narratological Approach I argue that it is impossible to find an essentialist definition of action be the latter practical or aesthetic There is no point in asking what action even
12. EPISODE and ACTION constructs which it contains Texts with a low level of EVENT integration but a high potential for ACTION synthesis are weak in action not because they are episodic like the above example but simply because they lack available EPISODES If a text has low levels of both EVENT integration and ACTION synthesis it will lack connectable EVENT constructs and therefore the raw material of a proper story in the first place The final case consists of texts which when read produce a high level of EVENT integration and a high level of virtual EPISODE integration the most extreme example of such a text is one in which all the constituent EVENTS are arranged in an uninterrupted succession which consists of and then connectors We are now in a position to abandon the interim terms action potential and episodicity Each of the four representative text types described above can be quantitatively represented as the product of the EVENT integration and EPISODE integration percentage values In the demonstration file this value which we shall refer to as the virtual ACTION PRODUCT is 54 117 4 347 235 294 The EpiTest window displays this figure under the final heading in the top right hand panel EPISODE constructs support an almost unlimited range of possibilities the ACTION PRODUCT is a measure of how much of this potential can be synthesized logically in a global ACTION construct the latter is as noted above defined in very
13. EventParser and EpiTest A User Guide JAN CHRISTOPH MEISTER This User Guide gives a brief introduction to the concepts and functionality of the markup tool EventParser currently version 4 37 and the analysis tool EpiTest currently version 3 85 The underlying narratological theory of narrated action is discussed in detail in my book Computing Action A Narratological Approach Berlin New York De Gruyter 2003 Please report any problems encountered with either program to mail jcmeister de You are welcome to copy or change any of the program code for non commercial purposes Any commercial use or exploitation of EventParser and EpiTest or part thereof requires my written authorization Contents 1 The EventParser Program 0 0cececeeceeceeeeeeeeeneeneeaes 2 2 The EpiTest Program cccececee cee eeeec eee eeeeneeeeenes 12 3 Installation Guide 2 nicseitscusaieiece akaa 24 4 Installed Files 2 2 0 0 0 c cece cece cece eens eee eneeaeeaenenseneeneenes 27 2 EventParser and EpiTest A User Guide 1 The EventParser Program The most important lesson to be drawn from our theoretical discussion of the concept of action in Computing Action A Narratological Approach 2003 is that a fictive happening even in its most basic form X and then Y never enters our consciousness in and of itself it can do so only in the form of an interpretive construct This basic construct of action logic which we refer
14. SODES Total EANISO1 EANISO2 27 12 15 A Generate ACTION CONSTRUCTS Total 38 Processing time in sec 0 391 EXIT Results not saved to daf file Fig 2 2 EpiTest user interface The various features of the program can be briefly described by considering the nine buttons in the screenshot reproduced in figure 2 2 The four buttons in the panel on the upper left hand side relate to the database file being evaluated in this case it is fairyl esf which was previously produced in EventParser CONSULT allows the user to open a new file EVENTS lists the current file in a text field SYNONYMS displays a similar list of categorially homogenous direct and categorially heterogenous indirect synonyms METASYNONYMS calls the algorithm for generating METASYNONYMS see above and lists them in a text field To the right of these last three buttons the program displays the number of events that have been read and the number of user defined and program generated semantic terms After the program has read and prepared a esf database the user should specify what kind s of combinatorial analysis are to be performed Depending on the checkboxes that have been selected EpiTest will generate isochronous EPISODE constructs anisochronous EPISODE constructs or ACTION constructs The buttons in the top right hand corner concern the actual generation saving and display of these constructs CONSTRUCT initiates the generation process and asks us
15. base dotted line ACTION constructs are generated by combinatorially connecting EPISODE constructs in Algorithm 3 Module 5 The first ACTION The flowchart illustrates how our software exploits the combinatorial capabilities of PROLOG EpiTest first builds a list of all the EPISODE constructs that fit the appropriate criteria and then initiates a second recursive process for building ACTION constructs However EpiTest is not really an intelligent program although it adds newly generated factual knowledge constructs and metasynonyms to its knowledge database it does not generate dynamic rule knowledge of any kind 14 2 The EpiTest Program construct is saved in the dynamic knowledge database before the program recursively checks whether further EPISODES can be added to the ACTION chain Complete ACTION constructs are saved in the con construct file 4 The constructs that have been generated are subjected to quantitative statistical analysis in Algorithm 4 The reader may well ask why we should spend a considerable amount of time and effort designing these computational algorithms in order to implement what is already a highly abstract formal system of narrative theory and action logic Some quantitative data may help illustrate the potential benefits In our illustration of the different types of EPISODE we analysed the combinatorial potential of ten EVENTS and identified five acceptable EPISODES which can be produced from them
16. d decide to test the programs by processing an extract from Lewis Caroll s Alice in Wonderland which was deliberately encoded to produce a narrative with a single uninterrupted strand This extract represents the opposite of the above extracts from the Fairy Tale Because of its typically episodic combination of high EVENT integration and low EPISODE integration the Fairy Tale example yielded an ACTION product of 235 294 Compare this with the results of processing the Alice esf file which yields an ACTION product that is almost ten times higher than that of Matrix esf Like the latter Alice esf contains just ten EVENT constructs but it permits the formation of an ACTION chain which contains a maximum of nineteen isochronally joined EPISODES Practical Analysis Using EpiTest In the preceding pages we have described a first attempt to put the theory developed in our book Computing Action A Narratological Approach into practice Part 3 of that book describes the full scale application of our software to a real text Goethe s Conversations of German Refugees However it is well to remember that our model and methods are not without their limits These weaknesses are particularly obvious when we come to deal with the anisochronous EVENT types Readers can process the semantics of u Alice esf is included with the downloadable files 22 2 The EpiTest Program anisochronous EVENT narratives effortlessly and intuitively because they use t
17. dicate classes in the EVENT DEFINITION menu The class terms are prespecified in the current version they were chosen on a purely intuitive basis one of the shortcomings of our theory and methods as they stand In part 1 we decided that the isolated predicate should be assigned to the EXPOSITION selected in figures 1 1 and 1 2 If we click on the social option in the class list an empty list appears not shown in the screenshot We enter our new term isolated at the top of this empty list and confirm the new entry by pressing ENTER The PREDICATE DEFINITION window is then displayed as in figure 2 2 3 Predicate Definition Window EventParser 4 37 You have entered a new PREDICATE term please define it by entering its opposite in the New OPPOSITE term box EventParser will save these terms in the semantic dictionary associated with the current TextBaseFile 4 New exposition OPPOSITE New exposition term integrated Ka ae gt isolated z es implication implication contradiction L 9 iets l amp opposition Event STATUS Exposition Eveni PTAS aoci Fig 1 3 PREDICATE DEFINITION window expositional predicate In the PREDICATE DEFINITION window EventParser prompts the user to enter the opposite term which accompanies isolated The layout of the window shows the upper axis of a semantic square in the making We define integrated as the opposite term of isolated and by doing so giv
18. e text listing as is already the case with our ferryman in figure 1 2 Isolated errors can be corrected by using the REDO buttons next to each field Alternatively the user can reject the entire EVENT definition by clicking the CANCEL EVENT DEFINITION button each stage in the EVENT definition process must then be repeated from scratch EventParser 4 37 EVENT Parsing Window TextBaseFiles Phrase Definition EVENT Definition Knowledge Base C COMP ACT EVPARSER FAIRY1 TBF Derne FOCUS on OBJECT a ste Class ferryman Redo 1 relational before after 2 physical deadjalive 3 emotional repulsive attractive 4 cognitive unknown known 5 material poorrich 7 moral bad goo 8 aesthetic ugly beautiful 9 philosophical false true NEXT STEP Show EVENT Base file PHRASE MARKING PHRASE MARKERS STRING SEARCH CURRENT STRING Please define Search from 18 Ss to 177 Fig 1 2 EVENT DEFINITION menu in operation The screenshot in figure 1 2 shows the state of the program when the user has defined ferryman as an actant FOCUS confirmed the definition by pressing ENTER or clicking the MEXT STEP button and opened the list of nine expo Note however that predicate lists from previous sessions can be reused in new texts simply by copying the bf files 6 1 The EventParser Program sitional STATUS pre
19. e the program a first small indication of the normative and cognitive frame of reference which applies to the EVENT being defined After entering the new opposite term and clicking the BACK TO EVENTPARSER button we are returned to the main EVENT PARSING window The fields in the grey panel above the text listing now display the current state of the EVENT definition we can see that the FOCUS is ferryman and that the expositional STATUS of the object is described by the isolated predicate EventParser and EpiTest A User Guide 7 which is a member of the social predicates class The definition of the selected segment of text as an EXPOSITION is now complete Note that the numerical indices of the first and last characters in the selection 18 and 177 respectively are displayed in the lower right hand corner of the window under the heading CURRENT STRING The program s help panel now prompts us to repeat the process by selecting and defining the dispositional STATE OF AFFAIRS in a similar manner It is up to the individual interpreter to decide where this second STATE OF AFFAIRS is located and how wide the scope of its definition should be In chapter 1 4 of Computing Action we decided that the DISPOSITION is contained in the following segment In the middle of the night loud voices wakened him it seemed that travellers wanted to be ferried across Goethe 1989 70 We also decided to define a categorially homogenous dispositional STATUS As
20. ecution and should help readers who are not familiar with PROLOG to understand how the program and its various models operate Comments are sections of a source code listing inserted by the programmer to clarify his code to the reader or indeed sometimes to himself Comments are completely ignored during program execution in PROLOG syntax they are marked by a preposed or enclosed in A file with the source code is part of the installation see section 4 of this document EventParser and EpiTest A User Guide 13 INPUTS po ene wre a esf record file I T Vv Predicates Synonyms Antonyms Ger n of EPISODEs ACTION structs 4 Quantitative Analysis partially recursive OUTPUT con construct file Fig 2 1 EpiTest flowchart 7 However it will probably be easier to digest the flowchart shown in figure 2 1 With the help of the diagram we can describe a complete execution cycle of EpiTest as follows 0 Synonyms antonyms and EVENT constructs are read from the record file broken lines METASYNONYMS are generated in Algorithm 1 Module 6 in the program listing They are then added to the program s dynamic knowledge database solid line EPISODE constructs are generated by testing for combinatorial connections in Algorithm 2 Module 4 The constructs are saved in the con construct file and added to the dynamic knowledge data
21. edge comes into its own when the recipients need a EventParser and EpiTest A User Guide 23 heuristics with which to explain protensionally anisochronally arranged semantic EVENT terms In contrast to this ideal solution we have employed something of a stopgap measure in the practical application of the EventParser and EpiTest programs to our example texts the six individual narratives in Goethe s Conversations of German Refugees A maximum of ten pairs of opposite terms have been formulated in advance for each of EventParser s nine heuristic categories Taken together these terms are an intuitively plausible model albeit an admittedly crude one of the knowledge context which I believe to be of crucial importance in the identification of anisochronous EPISODES The results of the analysis are described in the third and final part of Computing Action after a discussion of the critical background against which our narratological study in literary computing makes sense In principle there are at least four different combinations of text and reader in which EventParser and EpiTest can be used to mark up and combinatorially analyse record files Together the four combinations represent a typology of action analysis each of them concerns a different methodological problem a One text and one reader In this relatively simple situation we are usually concerned with differential analysis of the components of a single more complex text
22. efinition of the concept of action but things are different if EVENT EPISODE and ACTION are defined semiotically as dynamic constructs of reception Every STATE OF AFFAIRS that we encounter in the empirical or fictional world is not just projectively configured in anticipation of a possible change which it could undergo in the future it is also the base of a retrospective semantic protension which can make it the result of just such a change which has already taken place Between the theoretical possibility of this semantically preconfigured construct and its concrete instantiation there lie many worlds a marvellously comprehensive knowledge of the semantic relations which can accumulate through the symbolic representation of the phenomena of at least one possible world In its present state EpiTest has no access to the complex heuristic world knowledge of the semantic ordo naturalis which is constantly available to natural readers Until this shortcoming is rectified the rare occasions when EpiTest does combine anisochronous EVENTS into EPISODES and ACTIONS will be nothing more than flukes In order to overcome this obvious weakness in EpiTest and the theory behind it we would have to commit ourselves to using concepts from cognitive theory such as the script and the frame These concepts postulate a schematically represented mental knowledge of standardized situational and pragmatic patterns which is constantly accessible to recipients the knowl
23. eister uni hamburg de C eventparser fairy1 est C eventparser fairy1 aniso con o Min Max Virt Event Integration 5 30 46 54 1176 Direct Indirect Opposites Min Max Avg Episode Integration 19 4 20 2 2 4 4 34782 Ee Kis s Action Product META SYNONYMS 246 20 8 235 294 episode 20 episode elSO1 1 ferryman social isolated 18 177 social requested 178 283 ferryman social with company 450 615 social at home 2542 2836 J episode elS01 2 ferryman social isolated 18 177 social requested 178 283 ferryman social reacting 290 486 social unaffected 2470 2576 episode elSO1 3 ferryman social withdrawn 18 177 social cooperative 290 486 ferryman social with company 450 615 social at home 2542 2836 episode elS01 4 ferryman social withdrawn 18 177 social cooperative 290 486 ferryman social reacting 290 486 social unaffected 2470 2576 episode elSO2 5 ferryman physical asleep 18 177 cognitive hearing 178 283 ferryman social
24. ement involved in allocating descriptive terms It displays lists from which depending on the stage in the mark up process we can select focus descriptors or predicates that describe the expositional and dispositional statuses When the user loads a text for the first time the list of narrator and actant FOCUS descriptors is empty so as with all such lists the user must enter the appropriate term manually in Our practical demonstration is based on the opening segment of Goethe s Conversations of German Refugees 1795 which is discussed in detail in Computing Action A Narratological Approach EventParser and EpiTest A User Guide 5 the top field of the list The program automatically saves terms entered in this way and displays them in the list the next time it is displayed The STATUS options are slightly different for they require one of the nine semantic classes to be selected before the class specific lists can be viewed As these latter lists are initially empty we must enter the necessary predicate terms one by one when we start processing a new text The program does not restrict the type and scope of the user s predicates thus when using EventParser to encode a text especially for a series of experiments the researcher should ensure the comparability of the terms by establishing appropriate conventions in advance When a descriptive term is assigned it is displayed in the relevant field of the grey panel just above th
25. ense of being more faithful to the original text Qualitative evaluations where the original text of the literary narrative is paramount might therefore seem irreconcilable with the quantitative calculations on which our model is based But this is not the case by introducing the concept of virtual action potential we will we able to distinguish individual texts or even entire corpora from one another by studying the values of this new statistic The action potential has only one constant factor the literary text a concrete object which consists of a finite set of symbolic signs that are arranged so the empirical evidence tells us according to certain syntactic and grammatical rules The signs must be processed in a number of ways through reading interpretation recombination evaluation comparison and so on before we can even begin to speak of concepts let alone actions The signs are processed under the influence of many variables the reader s ability expectations intertextual and world knowledge and aesthetic norms to name but a few These variables differ not only from reader to reader i e from constructor to constructor but also inside individual reading subjects and receptive processes where closer observation shows that they can vary dynamically from text to text and even from sentence to sentence With each successive sentence the reader knows more discovers new associations and expects different things Reception
26. he framework of a prespecified comprehensive and well structured knowledge context This is very different from the incrementally assembled set of axioms which EventParser stores in its sbf files The case of an anisonchronous EPISODE can illustrate the difficulties which can result Assuming that narration and therefore reading is strictly sequential and does not employ embedding techniques the prehistory supplied by EVENT must be encountered after EVENT2 The order in which the acts of reception i e EVENT construction take place is irrelevant in our theory and the EpiTest algorithm with which we are concerned So far so good Now EVENT and EVENT must have explicitly connectable predicates if they are to be linked semantically by EpiTest The practical problem is that recipients tend to provide the EVENTS with such predicates only if they have mentally arranged the analeptically narrated EVENTS in the order of the fictive ordo naturalis beforehand The EANISO example encourages this to be sure but the bracketing and assistance it provides are the exception rather than the rule Narratologists typically treat the practical receptive necessity of returning to the ordo naturalis of the fabula as evidence that the ability to reconstruct a quasi natural chronological order at the diegetic level is a necessary prerequisite for identifying a consistent action order This may well be reasonable from the perspective of an intentional or causal d
27. iTest listings ep generator3 85 pl Uncompiled PROLOG file and epitestlisting doc source code listing EpiTest epitest epitest385 exe Program files epitest385 ovl epitest385 ini Initialisation file Alice esf Integrated protocol file created Fairytale by EventParser for subsequent Matrix processing with EpiTest Unterhaltungen metaterms msf Temporary file listing newly generated meta terms con daf Construct file listing all EPISODE und ACTION constructs generated by EpiTest Data analysis file containing the statistical results of the combinatory exploration of an esf file with EpiTest Please note when manipulating any sbf bf ebf or esf file make sure to insert a single line feed at the end of the file This is a crucial syntax requirement for list files ignoring it will result in a program crash
28. iable counter term x WOW allowing you to prune the esf file manually with a text editor at a later stage Under certain conditions the predicate definition routine can result in an underdefined semiotic square This can only be corrected manually by checking the resultant esf file with a text editor prior to processing the protocol file in EpiTest 3 3 Running EpiTest 3 85 EpiTest 3 85 is started by running the file epitest385 exe The program was written in LPA Prolog which conforms to the Edinburgh Standard For details on this implementation of Prolog see http www lpa co uk abt htm After installing EpiTest you will find an initialisation file epitest385 ini in the relevant diretory The command line in this file reads command h4096 12048 t1024 D1 01024 p4096 and contains so called program switches that define details of Prolog memory management and dynamic assertion and manipulation of clauses EventParser and EpiTest A User Guide 27 during recursion loops safe dynamic mode Please consult the LPA manual for details on how to manipulate these parameters Known bugs 1 When accidentally skipping the routine for generating metaterms which is normally executed by clicking the METASYNONYMS button the program may be unable to generate episode or action constructs This happens whenever the synonyms declared in EventParser lack explicit semantic connectivity Usually the higher the overall n
29. in general and the reception of ACTIONS in particular is fundamentally dynamic in nature and cannot be reconstructed in a supposedly ideal reading situation in order to facilitate the experiments of our computational approach However readers do not evaluate literary narratives on a purely subjective basis by considering the richness of their action logic the coherence and depth of their chains of fictive happenings and the originality and elegance with which they combine events Readers also take part in discussions with other readers Theoretically the semiotic process may continue into infinity as Eco has described but in practice we will terminate it on the basis of our pragmatic needs The same I believe applies to the discussion be it naive or critical or philosophical of the action potential of a narrative text Only if the reading receptive construction of action can create synthesis meaning does it make sense to try to define action potential in theory and measure it in practice The closer a receptively constructed ACTION clings to the referential 18 2 The EpiTest Program framework of fictively represented actors things situations and transformations the more it is reduced to nothing more than a denotational function but the more an ACTION distances itself from the fictive substance borne by a possible world and is condensed into a sequence of transformations which can be reconstructed as abstract formal propositions
30. ions the state of the identical object from c must be seen to be distinguished by predicate at point t in fictional time This is the dispositional STATE OF AFFAIRS DISPOSITION Program specific terms appearing on the user interface of EventParser and EpiTest in particular those defining windows and buttons are generally identified by small caps in italics e g EVENT PARSING window The use of EventParser has to date been confined to OBJECT EVENTS the following discussion therefore assumes the epistemological congruence of FOCUS and object However we must not ignore the theoretical distinction between the two and nor should we forget the presence of the DISCOURSE EVENT as a theoretical concept EventParser and EpiTest A User Guide 3 Like all symbolic systems narrating texts are characterised by a referential economy that suppresses information which is redundant or relatively easy for recipients to deduce from contextual evidence To enable our theory to cope with this fact we must formulate a supplementary rule which governs what kind of object can be involved in an EVENT construct e The fictional object in an EVENT construct must fall into one of two classes The first class consists of fictional objects which are referenced directly e g by personal names or pronouns or indirectly e g by iterative or durative phrases such as during or at once in a literal narrative assertion of their existence The second less
31. n and EVENT s FOCUS by overflowing the old ferryman weary from the toil of the day lay in his little Ai MERCIET In the middle of the night loud voices wakened him it ali kingithe EVENT seemed that travelers wanted to be ferried across DEFINITION menu Stepping outside he saw hovering over his moored boat two large and selecting from will o the wisps who insisted they were in a great hurry and wished they its lists were already across The old man pushed off without delay and rowed across the river with his usual skill while the strangers hissed at another in an unfamiliar very animated language and occasionally burst into loud laughter as they capered now about the sides and benches now upon the bottom of the boat The boat is rocking cried the old man And if you are so wild it might capsize Sit down you wisps They burst into laughter at the very idea mocked the old man and were wilder than ever He bore their mischief with patience and soon reached the other side Cancel EVENT Definition NEXT STEP v Show EVENT Base file Phrase Definition menu ex MARKING MARKERS Tan MEL o E Ca AEN tae SEARCH _ CURRENT STRING Please define s h from 18 HEN to 177 Fig 1 1 EVENT PARSING window The help area now asks us to use the EVENT DEFINITION menu to define the FOCUS The EVENT DEFINITION menu is the key user interface el
32. nitude of the 20 2 The EpiTest Program demonstration file we obtain a striking result despite the high EVENT integration of 54 percent the forty six virtual EPISODE constructs cannot produce an ACTION list with more than two elements Consider again the second row of figures in the top right hand panel of the EpiTest window in figure 2 3 On the right of the ACTION button we can see numerical values for the length of the shortest ACTION chain generated the length of the longest ACTION chain generated and the average length of all generated ACTION chains In the present example the value of each statistic is two Under the heading EPISODE INTEGRATION we can see the percentage of virtual EPISODE constructs which are combined in the longest generated ACTION chain In our example we have obtained a very small value 4 3 percent Although this result may seem somewhat disconcerting at first it is easily explained Because our example EPISODES reused the same segment of text several times it follows that the possibilities for building extended ACTION chains must be severely limited We are concerned with a narrative text that has been read as having ten EVENTS on which small basis a considerable number of possible EPISODES can be formed but the text remains episodic in the non technical sense of the word When we consider the opposite situation it becomes even more apparent that the episodicity of a text is related to the number of EVENT
33. obvious class is composed of fictional objects whose existence at point in time can be hypothetically postulated on the basis of points of narrative indeterminacy by back projecting it from 4 the dispositional STATE OF AFFAIRS It should be emphasized that rule e is not bidirectional it cannot be inverted so as to allow us to look into the future beyond an expositional STATE OF AFFAIRS In other words the rules of interpretation allow us to assume that if the text documents the present existence of an object the object must also have existed virtually in the fictional past We cannot however speculate about how if at all such an object will exist in the fictional future after the most recent description of the state of the fictional world in the narrative The only exception to this is when the narrative itself opens the fictional world to the future for example by making prophetic predictions We cannot test an assertion that epistemologically i e in terms of FOCUS transcends the inter preter s maximum possible stable knowledge state It follows that we cannot falsify ACTION constructs whose formation depends on such an assertion they are therefore beyond the scope of our approach We can now turn to the practical description of EventParser Apart from the welcome window which appears when it is opened the program contains a total of four windows These windows provide an interface with which the user can designate and ma
34. r the heading EVENT INTEGRATION The action potential of a text which basically measures the degree to which a given reading can be translated into overall ACTION constructs can now be evaluated in terms of the combinatorial affinities between atomic EVENT constructs and molecular EPISODE constructs However we shall not stop here our quantitative measure of action potential will also consider the scope for EPISODES themselves to be joined together in action logic This means that having considered local EPISODE forming EVENT synthesis above we must now consider global ACTION forming EVENT synthesis as well Ideally we would define a series of semantic criteria similar to those used in the EPISODE matrix this would allow our theory to represent our intuitive knowledge that the scope of a particular action narrative in the form of an ACTION construct depends primarily on the presence of a dominant semantic category or theme However such a refinement is beyond the remit of the present study and we must be content with a provisional pragmatic criterion instead EpiTest s combinatorial algorithm is therefore based on the following plausibly simple broad rather than narrow definition of the ACTION construct two EPISODES form an ACTION if and only if they are sequentially ordered in the semiotic continuum of the text This test therefore does no more than provide a solution to the following question assuming that EPISODE is the first
35. rk up the EVENTS in a text and make declarations to elucidate the world knowledge on which his definition of EVENTS depends 1 The EVENT PARSING window is the heart of the text processing interface 2 The PREDICATE DEFINITION window allows the user to define new predicates 3 The EVENT BASE FILE window allows the user to view and select encoded events 4 The SEMANTIC DICTIONARY window lists all the descriptive terms currently in use The list represents the program s knowledge database which expands dynamically as a text is processed 4 1 The EventParser Program Once the text to be processed has been loaded it is displayed in the EVENT PARSING window The user can then process the text by performing a prescribed series of definition operations The help panel on the right hand side of the window displays instructions which guide the user through each stage of the mark up process The first step in defining our first EVENT is to mark the text which contains its EXPOSITION We do this by clicking and dragging with the mouse to select the appropriate range of characters as illustrated in figure 1 1 EventParser 4 37 E ENT Parsing Window TextBaseFiles Phrase Definition EVENT Definition Knowledge Base C COMP ACT EVPARSER FAIRY1 TBF FOCUS on OBJECT j EXPOSITION status of OBJECT DISPOSITION status of OBJECT Class Predicate Class Predicate Tale Step 2 Define the By the great river which was newly swollen with heavy rai
36. t possible to save formally consistent mark up data in event database list database and sememe database files for subsequent differential and relational processing The program creates the list database and sememe database files behind the scenes they are then automatically combined to produce a master file with the suffix esf event sememe file in the format required for analysis in EpiTest Superficially EventParser is little more than a simple mark up tool but it also has a far more important function at every stage it forces the user to explicitly define the semantic terms being used and thus make clear the symbolically represented world knowledge which influences the encoding A number of auxiliary functions see the panel at the bottom of the EVENT PARSING window have been added to assist the user in dividing texts into EVENTS Searching for user defined characters or strings PHRASE MARKERS takes some of the effort out of selecting segments of text PHRASES A second search option is provided by the STRING SEARCH field in which the user can enter a string for which to search by scanning forward from the insertion point Unlike the PHRASE MARKERS field the STRING SEARCH option does not alter the selection its primary function is to locate specific names adjectives or similar features 12 1 The EventParser Program process In the next section of our discussion we shall turn to EpiTest and the question of how to design an a
37. the more it becomes a hollow empty shell of meaning the illusory meaning of a receptive happening that portrays itself as the only true action From this perspective the empirical reading and construction of ACTION can be seen as an attempt to find a way through the world of latent possibilities contained in a network of EVENTS Our definition of the EPISODE as the smallest such ACTION construct provides us with a unit with which to measure and quantify this world of possibilities We shall begin by considering the upper and lower theoretical values that can arise If only a single EVENT is present it is impossible to form an EPISODE at all Once two EVENTS the minimum quantity needed for an EPISODE are present there are two ways of combining them The definitions below assume that our two EVENTS are 1 Egg from chicken and 2 Chicken from egg a Hyperdetermination every EVENT capable of entering into a connection can enter into one and only one such connection Thus we must read our two EVENTS either as 1 2 or as 2 1 The formula required in this case is simple for n EVENTS the minimum number of EPISODE constructs Epmin is given by the formula Epmin ni2 b Hyperconnection every EVENT can connect with every other EVENT to form a pair but not with itself The ontic primacy of chicken over egg is just as conceivable as the primacy of egg over chicken and there are two possible EPISODES 2 1 and 1 2 respec
38. tially run on a P II configuarion 266 Mhz 32 MB RAM Windows 2000 and then re run on an AMD Thunderbird 800 MHz 128 MB RAM and aP IV 2 66 Ghz under Windows XP 3 1 The Installation File compact exe The installation file compact exe is a self extracting file available at http www jcmeister de downloads software comp action compact exe EventParser and EpiTest A User Guide 25 Clicking this file will start the automatic installation process note that no registry entries are made The default installation directory is c compact A detailed listing of all files extracted and stored during this process is found in section 3 3 of this installation guide Assuming that the default settings have been accepted the two programs will be extracted into the following directories e EventParser 4 37 saved as ep437 exe in c compact eventparser e EpiTest 3 85 saved as epitest385 exe in c compact epitest Desktop links and icons for both programs should be created manually Note that after installation all extracted files must have the read and write properties activated copying the files between a hard drive and a CD ROM can lead to these properties being disabled in which case neither program will work properly 3 2 Running EventParser 4 37 Having been written in a by now outdated version of Visual Basic 3 0 EventParser 4 37 can only handle tbf text files of up to 45 kb size Larger files will therefore have to be
39. tionary Window C COMP ACT EVPARSER FAIRY1_sbf EventParser saves any new EXPOSITION and or DISPOSITION predicates and opposites as defined by the user The program then automatically generates the respective antonyms and synonyms for these terms and captures them in its data base the sbfile This file cannot be edited within EventParser opposite isolated integrated dirantonym isolated requested dirantonym integrated ignored dirsynonym integrated requested dirsynonym isolated ignored opposite requested ignored Semantic terms are grouped into the following categories opposites good bad synonyms good nice antonyms good ugly A further distinction is made betwe ect and indirect antonyms and synonyms direct terms are those generated from predicates defined in identical indirect those generated from predicates defined under separate CLASSES CLASS membership of EXPOSITION and DISPOSITION predicates also differentiates between homocategorial projective and heterocategorial protensional EVENT constructs Return to EventParser Fig 1 7 SEMANTIC DICTIONARY window We have now reached the end of our discussion of the main components and principles of the EventParser software The program allows us to define EVENTS in accordance with the theoretical and methodological criteria set out above and makes i
40. tions have already been filled The first pair of opposite terms isolated and integrated has moved to the bottom of the square the new term requested occupies the top right hand corner By defining its opposite as ignored in the top left hand field see figure 1 5 we can complete our first semantic square In the newly created square the diagonal from isolated to requested represents the real event construct while the diagonal from ignored to integrated represents our first virtual event construct that is an implicitly suggested EVENT which is possible according to our world knowledge and compatible with the narrated world as we have read it EventParser and EpiTest A User Guide 9 EventParser 4 37 Predicate Definition Window L ignored requested fly fly i V Fig 1 5 PREDICATE DEFINITION window dispositional predicate When we return to the EVENT PARSING window EventParser gives us the opportunity to add the EVENT we have just defined to the EVENT database We do this by clicking the ADD TO EVENT BASE button after which we are presented with a list of all the EVENTS that have been defined see figure 1 6 Clicking on any EVENT in the list displays its associated segments of text 10 1 The EventParser Program EventParser 4 37 EYENT Base Window EventBase Files Lists C COMP ACT EVPARSER FAIRY1 ebf event ferryman social isolated 18 177 social requested 178 283
41. tively For n EVENTS the maximum theoretical number of EPISODE constructs EPmax is given by the formula Epmax n n 1 In a practical context these simple formulae can be applied to obtain numerical values which describe the minimum and maximum levels of episodicity in a concrete text They delineate the spectrum of EVENT combinations which are theoretically possible on the basis of a particular reading before syntagmatic and semantic criteria are considered Putting the formulae into practice using the demonstration data of the ten EVENTS discussed above we find that there is a theoretical minimum of 10 2 5 EPISODES and a theoretical maximum of 10 10 1 90 EPISODES The actual number of virtual EPISODE constructs which can be produced will lie somewhere inside the theoretical range of five to ninety in the present case it is forty six Expressed as a percentage this gives us an EVENT integration of EventParser and EpiTest A User Guide 19 just over 54 percent In less formal terms this means that a the EVENTS in the text can be combined into eighty five theoretically possible distinct EPISODE constructs and b of these eighty five theoretically possible constructs forty six i e 54 1 percent can be virtually instantiated on the basis of the semantic and formal predications with which the EVENTS have been encoded EpiTest calculates this value for us and displays it in the upper right hand corner of the window unde
42. umber of semantic terms declared the more likely they will possess explicit connectivity By contrary a low amount of terms may seem perfectly related to each other for the human reader who has access to metaphorical and world knowledge but not to the machine which relies on explicit definitions only This characteristic of EpiTest is intended On termination of the program an accidental error message can sometimes appear This can be ignored Within EpiTest the automatically generated con and daf output files can only be saved into the same directory as the esf input file 28 4 Installed files After a successful installation the following files should be available Program Directory File name Function EventParser eventparser ep437 exe Program files Cmdialog vbx Crystal vbx These files must all be in the Mscomm vbx same directory Some additional Msole2 vbx dll files will be saved here as Threed vbx well Alice Standard lbf Standard seme term files Fairytale Standard sbf Matrix These contain default semantic Unterhaltungen terms and may be changed or augmented tbf TextBaseFile Text files to be processed max size is 45 kb sbf EventParser files which are lbf automatically created ebf EventParser and EpiTest A User Guide 29 EpiTest esf Integrated protocol file which EventParser creates automatically by combining sbfand ebf files Ep
43. utomated tool for analysing the combinatorial potential of EVENTS What distinguishes a pair of EVENTS which can be linked to form an EPISODE What kind of relations exist between two such EVENTS And do interpreters prefer some ways of linking EVENTS to others Only when we have considered these questions in a theoretical context will we be able to devise generative algorithms which answer them The algorithms will be used in EpiTest to search for virtual EPISODE constructs and ACTION metaconstructs in EventParser record files In the process it will become clear that formally represented world knowledge a side effect of EventParser mark up is the cornerstone of combinatorial analysis in EpiTest 2 The EpiTest Program The EpiTest program was developed with a specific task in mind combinatorial analysis of esf files produced by EventParser The heart of the program consists of a set of algorithms which search saved databases of EVENTS for connections which correspond to the categories in our episode matrix EpiTest was written in PROLOG Programming in Logic a programming language which has long been regarded as ideal for developing artificial intelligence models and is also one of the standard tools of computational linguistics The code listing of EpiTest version 3 85 the version designed and used when writing the present study runs to about 2 240 lines including comments The comments document the route taken through the code during ex
Download Pdf Manuals
Related Search