I have a classification problem. I have clusters called 'Experience', 'Education', 'Abilities' . The labelled data (72,000+ entries with all clusters together) with two columns looks like below.

year of education               Education
years education                 Education
years of educational            Education
two years of education          Education
years of education beyond       Education
education four year             Education
of proven sales experience      Ability
knowledge of and                Ability
experience or education high    Ability
assigned knowledge skills       Ability
accountable for driving         Ability
administrative and leadership skills    Experience
advanced negotiations skills            Experience
must have keyboarding skills            Experience
must have skills                        Experience
activities preferred skills             Experience
of clinical skill                       Experience

I have to give a string and find it whether it belongs to either Experience or Education or Ability based on the trained model. The examples of strings .

string1 = "There is a requirement of four year professional degree"
string2 = "Able to drive the teams to higher levels"
string3 = "Must have programming experience in C, C++"

when I test these string it should be able to classify the string in to either of the clusters.

What are the possible ways to train my model ? Refering to word2vec and doc2vec, will these models work? I couldn't find any relevant examples to train model on words and test on strings. Any ideas on how to work?

  • $\begingroup$ "I couldn't find any relevant examples to train model on words and test on strings." You didn't look very hard ;) This is text classification, you'll find plenty of examples. $\endgroup$ – Erwan Jun 13 '19 at 16:48

I am kind of confused by the question. But I am guessing you are asking what model to use and how to represent words.

This seems like a pretty introductory example so you may want to try simple Bag-of-words representation. Try pushing that into a Naive Bayes classifier as a first pass. If you are unhappy with the results, you can implement word2vec and more complex methods.

| improve this answer | |
  • $\begingroup$ if it works with word2vec, please update your answer or please provide any reference link with examples where I can read and understand. $\endgroup$ – Raady Jun 14 '19 at 8:14
  • $\begingroup$ @Raady: Why do you want to use word2vec? Start with simple methods, see e.g. pythonmachinelearning.pro/… (google text classification python for more examples) $\endgroup$ – Erwan Jun 14 '19 at 12:52
  • $\begingroup$ Literally the first thing on Google under "word2vec text classification": Link $\endgroup$ – Collin Cunningham Jun 18 '19 at 0:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.