sklearn-esa 0.0.13

An ESA implementation in python.

Get the required resources

scp -r [e-mail]:/home/weci2587/projects/args-topic-modeling/resources .

To run the ESA-script with all terms run:

For normal ESA:

./esa-all-terms.py  --similarity cos
                    --matrix-path <path_to_resources>/resources/esa-plain/<debatepedia|strategic-intelligence|wikipedia>.mat
                    --model-path <path_to_resources>/resources/esa-w2v/GoogleNews-vectors-negative300.bin
                    --model-vocab <path_to_resources>/resources/esa-w2v/w2v-vocab.p
                    --input-path <path_to_input_file>
                    --output-path <path_to_output_file>

For word2vec-ESA:

./esa-all-terms.py  --similarity max
                    --matrix-path <path_to_resources>/resources/esa-w2v/<debatepedia|strategic-intelligence|wikipedia>.mat
                    --model-path <path_to_resources>/resources/esa-w2v/GoogleNews-vectors-negative300.bin
                    --model-vocab <path_to_resources>/resources/esa-w2v/w2v-vocab.p
                    --input-path <path_to_input_file>
                    --output-path <path_to_output_file>

To run the word2vec-ESA with reduced terms run:

./esa-top-n-terms.py    -n <number_of_terms> 
                        --corpus-path <path_to_resources>/resources/corpora/<debatepedia|strategic-intelligence|wikipedia>.csv
                        --model-path <path_to_resources>/resources/esa-w2v/GoogleNews-vectors-negative300.bin
                        --model-vocab <path_to_resources>/resources/esa-w2v/w2v-vocab.p
                        --input-path <path_to_input_file>
                        --output-path <path_to_output_file>

The input document must be a csv file with "|" as the separator and must contain the column "document", which is used as the input text for the ESA.