search

SvenElyes / Textanalytics

MIT License
1 stars 0 forks source link

Relations - Processing - Extract the relation #3

Open fmunzlin opened 3 years ago

fmunzlin commented 3 years ago

Any verse relevant to a chosen character has to be processed. If only one character has been assigned, we can not extract relation information. (but we may be able to enhance the character profile) If more than one character has been assigned, we may be able to extract an relation. This has to be done in this issue.

Any self variable of the relation class has to be updated in this task until then, they are None. Therefore, both are closely related

d3ithwen commented 3 years ago

paper: https://ieeexplore.ieee.org/document/8554714 for general steps of finding relations

aileen-reichelt commented 3 years ago

For the relations, we want to determine whether it is a positive or a negative relation. In order to do this, we need to perform a sentiment analysis on the relevant (= both characters mentioned) verse/text passage. Since we don't have labeled data, we should use a pre-trained sentiment classifier for this task.

aileen-reichelt commented 3 years ago

Additionally, we can simply count how often two characters appear together. Later on, we could represent this info in a graph with edges of different sizes.

fmunzlin commented 3 years ago

currently the relation gets extracted by (a) a bag of words (b) native baysian analyzer by TextBlob. Score can be calculated but will take approx. 20h for the hole bible. Has not been done because the Characters cell (e.g. ['Jesus', 'Maria', 'Joseph']) each verse is missing. can afterwards be calculated. May be enhanced by other classification approaches.