prima

Personal Research Management for the IA

Installation

Note: This project uses gensim, nltk tokenizer, pandas, and pdfminer in addition to the internetarchive python library for some of the tools and runs with python 2.7 and c. You will need to download them by following the instructions on the websites.

1. Clone this repository to somewhere in your file system.

2. To install the tool, first download SQLite and place the files titled sqlite3.c and sqlite3.h in the prima/src directory. Then run

    $ sudo python setup.py install
   

3. To create the appropriate directories for your file system run the following commands where project_name is the desired file name for a collection to be saved in

    $ init_workspace.sh
    $ cd workspace
    $ init_project.sh project_name
    $ cd project_name
    $ init_collection.sh
   

4. To download a collection into the auto-generated source/ directory, run the following where collection_name is a valid id for a collection in archive.org (for example this collection would use collection_name=toronto)

    $ fetch_collection.sh collection_name 
   

5. After completing steps 1-5, run the following to get stats on your collection where tool is one of the options listed below with the appropriate parameters

    $ toolname params
   

6. For every new collection to be created, make sure you're in the workspace directory and repeat step 3 lines 3-5 and step 4 with the new collection name before using any tools.

Tools

Current available tools included in the prima and basic examples are:

1. BM25 (default k=10)

    $ bm25.sh [k] "sample query here"
   

2. K-means clustering (default k=3)

    $ k_means_clusterer.sh [k]
   

3. Latent Dirichlet allocation (default k=100)

    $ lda.sh [k]
   

4. Latent semantic indexing (default k=100)

    $ lsi.sh [k]
   

5. MinHash (default k=10)

    $ min_hash.sh
    $ min_hash_sim.sh source/folder/document [k]
   

6. tf-idf

    $ tfidf.sh
   

7. Word count

    $ word_count.sh
   

More detail on how exactly to use these can be found in the wiki.