search

dkt-projekt / technical-discussion

This repository is used for technical discussions.
0 stars 0 forks source link

Document Analysis - Storytelling template #17

Open sraeuchle opened 7 years ago

sraeuchle commented 7 years ago

Hi there,

currently you working on the Semantic Storytelling Api to generate stories with a collection of entities. Is it possible to match documents against story templates? This would be the reverse funktionality. For example: input: text (letter|article...) output: [biography|....]

I think this would be helpful for evaluating the stories, maybe for deep learning and search algorithms.

Please let me know what you think about this feature?

jmschnei commented 7 years ago

Hi,

I think that this is an interesting point. For now, we have considered the approach where we extract/generate a template from a set of annotated documents.

I have some doubt about your example. You mention that you offer as input a text, and then in brackets you mention letter/article. This letter/article is just the format of the document or a template/structure defined for this text?

Then, the output of the system will be a biography structure with information filled from the text or just a TAG defining the possible output as Biography?

sraeuchle commented 7 years ago

Hi,

thanks for your response.

What i mean with letter/article in the brackets is, that the text input should be independend of the structure itself.

Let me explain it a bit more detailed: Let's say a text is a story. -> A letter is a short story and an article could be also a story. So i have a story in textual form and want to know if this story matches a storytemplate.

The input would be: "I am a sample biographical text about Angela Merkel. She is born in..." The output would be a tag: "biography" (like the storyType parameter of the story generation api)

What do you think?