This two days to be tortured,
I wrote a business of garbage should be sent to this kind of thing,
Need is to do a thing can identify the meaning of the text of user input, such as user said "open air conditioning I can identify" to "air conditioning" intentions,
Do preliminary logic about, such as white list and regular filter such as you, but the last step, according to the idea is that if regular make uncertain, is done using the vector machine model,
Which is the SVM,
But actually don't know how to use this at all,
Ok, if introduced, then the problem, consult ah ~ ~

Found a source on the Internet, probably saw the libsvm used to generate text is how to model, calculation of some CHI, TF, IDF, TF - such as UDF, themselves through the process to generate the corresponding files and model, but there is a problem in the use,
LibSVM data format is such
Label 1:2 value: the value... .
According to my understanding, the label corresponding to the intentions, the corresponding is which a 123 words (words and made corresponding relation), the value corresponds to the weight of the word is,

But according to my reading of the source code, find the value of the weighted value is Tfidf, is the value and the current word in the text and the sample frequency and so on, so if I use the old model to identify a new text, how do I get/this value to calculate?

CodePudding user response:

Poof, a person...

CodePudding user response:

This estimate involved in machine learning, artificial intelligence, may be Java version of the very few people know, suggest to artificial intelligence section ask there

CodePudding user response:

refer to the second floor qybao response:

this estimate involved in machine learning, artificial intelligence, may be Java version of the very few people know, suggest to artificial intelligence section ask there

. Slag Java development, into the line, what all don't understand, thank you very much