Answer the question
In order to leave comments, you need to log in
How to automate project type detection?
Every day 60-80 projects fall for evaluation. Will fall soon about 200 and more.
There are 50 project types. Will be more detailed in the future.
There is a history of manual selection of projects. So far 1000 projects with assigned categories.
Is it possible to somehow write a program or script that will recommend choosing the appropriate category.
By keywords in the description and title of the project. If ambiguous - then offer options.
If it is unambiguous then assign categories. If the script is wrong, then the person will manually correct it.
How can this be done as cheaply and easily as possible?
Answer the question
In order to leave comments, you need to log in
We split the line with the description into words, each run through soundex . Then we count the words.
We bind indexes of already selected words to categories. And in conclusion, we are looking for categories by indexes. When adding (changing) a category, we add (remove) new indexes, if any.
In addition to the simple index value in the categories, the "weight" of the indexes must be specified for sorting the categories when matching "project - category". (That is, the higher the sum of the "weights" of all found words in the category, the higher the position at the exit. This also affects the "unambiguous" choice of automation. If it is above a certain threshold, then we select unambiguously, otherwise the choice or manual change)
In fact - this is the implementation of search engines before Google times.
Didn't find what you were looking for?
Ask your questionAsk a Question
731 491 924 answers to any question