Below is available list of topic analysis methods that have been implemented so far.
In all the cases of the specified topic analysis methods, running each method would return a link which points to a page that would show the status of the analysis if its not finished yet or the results of the analysis if the analysis has completed.
Existing parameters are
- docs: a collection of documents, either supplied as an array of strings or a collection of .txt files. At least two documents should be given. Note that a single document can consist of a single sentence. In this way you can extract topics from a single text file by first tokenizing it at the sentence level and treating each sentence as a single "document".
- num_topics: The number of topics to extract. The minimum value is 2.
- topic_divider: This value is an integer value. If it has a value of zero, then num_topics would be used. If it has a positive value, then the number of topics would be
the
(number of documents)/topic_divider - maxiter: This value gives the maximum EM (expectation maximum algorithm) iteration that is allowed. Although a suitable value depends on the supplied documents and the value of beta, a good value can be 22.
- beta: This is a floating point number with range of (0,1]. It is used as a tempering parameter in the EM algorithm. A recommended values are values closer to 1 including 1 itself. Choosing the value of 1 means no tempering is used.
Here are sample json files to try with this algorithm, either from the snet client or the dApp: longer sample, shorter sample .