bmeaut / python_nlp_2018_spring

MIT License
8 stars 10 forks source link

Viterbi backtracking parameter #10

Open balAdamT opened 6 years ago

balAdamT commented 6 years ago

In course_material/07_Tagging/07_Tagging_lecture.ipynb when updating the backtracking table, the following line is presented for the algorithm: B[i,j]=argmaxkπ[k,j−1]⋅Tk,iB[i,j]=arg⁡maxkπ[k,j−1]⋅Tk,i Shouldn't this be also include ⋅Ei,j?

Other sources seem to include it and I can't think of any good explanation why omitting it would be reasonable. Is this just a typo or was there some consideration behind this choice?

juditacs commented 6 years ago

You are correct, tt's indeed missing from the equation. The only difference between the two equations is max vs argmax.

balAdamT commented 6 years ago

Thanks