search

taalbrecht / JournalVis

Journal article network visualization tool
0 stars 0 forks source link

Create "relevant thoughts" search that identifies most relevant sentence/sentences in articles #11

Open taalbrecht opened 8 years ago

taalbrecht commented 8 years ago

Create method to identify sentences most strongly associated with each topic. Perhaps include option to highlight parts (sentences/words) of an article corresponding to each topic to help with review. Include polarity analysis with this visualization feature.

taalbrecht commented 8 years ago

Topic calculation and polarity calculation per sentence are done. Add weighting by number of words to calculation for topic relevance of each sentence.

taalbrecht commented 8 years ago

Use shannon entropy relationship between number of words and amount of information (see DOI: 10.1002/j.1538-7305.1951.tb01366.x

Multiply total information:

Total sentence entropy×proportion of sentence made up of each topic

Use this metric to rank sentences from most relevant to least for each topic. Display 3 or 4 most relevant sentences when viewing each topic to help user understand what each topic is about

taalbrecht commented 5 years ago

Completed

taalbrecht commented 5 years ago

Use this metric to rank sentences from most relevant to least for each topic. Display 3 or 4 most relevant sentences when viewing each topic to help user understand what each topic is about