Ah, thanks for the vote of confidence :) I also discovered Apache OpenNLP 
Natural Language Processing (Thanks Apache!)

I guess now I can follow up on IBM's 2010 patent: "Method and system for 
automatic computation creativity and specifically for story generation" - 
http://www.freepatentsonline.com/7333967.pdf

They've got the theme idea right, but where they fail to realized:

A Beat of action is an event, a scene is an event, a sequence is an event, and 
a story is an event, are all examples of events. Each event goes through the 
same processes of thought

* Big_Opening: Responsible for extracting and demonstrated strong character 
characterization and a 
ranked list of answer-bearing documents, using a query 
formulated using information provided by the CharacterProfile + 
PersonalityBias. 

* Catalyst | Goal Evaluator: Responsible for encountering the 
question to determine the information needed (question type, 
answer type, key terms, etc.). 

* Big_Event | Question Analyzer:  Responsible for analyzing the 
solution space to determine if the information need is available (question 
type, 
answer type, key terms, etc.). 

* Pinch | Retrieval Strategist (RS):  Responsible for extracting a 
ranked list of answer-bearing documents, using a query 
formulated using information provided by the Question 
Analyzer + CharacterProfile + PersonalityBias and now dedicated to the outcome 
of the event. 

* Crisis | Information eXtractor (IX):  Responsible for extracting 
and scoring/ranking answer candidates from the answer 
bearing documents, based on expectation bias. 

* Climax | Answer Generator (AG):  Responsible for removing 
duplicates and selecting/filtering answers. Formulated using information 
provided by the Reversal + opposing CharacterProfile + opposing 
PersonalityBias. 

* Realization | Post Catalyst Analysis: A pilot-like wave is generated 
backwards and tries to close the Story loop (or Story event)

* Denouement: Character acts and improves based on what he's learned. (based on 
adjusted CharacterProfile + new PersonalityBias)

because these steps, detailing the event, is scale invariant, no state machine 
is needed.

The lead character should employ goals-based learning, in a way consistent with 
dramatic story design. As complications are introduced (search frictions), the 
lead character should conduct collision detections to determine if he is 
getting closer to, or further from, achieving his goal.

Since the story engine is a probabilistic system, it searches for the best 
match for this query, not just for 
strings that meet all the criteria.  For instance, a document that referenced a 
communication involving 
[Theme:good triumphs over evil when good has no fear of 
failure][Location:Woods], [Context:woods], [subtext:triump], 
[George:Exhausted,Max:Determined] is likely 
to be ranked highly, even if it does not fall in the data range specified.  
This allows users to enter highly 
detailed queries without the risk that important documents that do not exactly 
match 
the query will not be found.
 
---------------------------------------------------------------------------------


-Jericho West

Reply via email to