see SCENARIO for a working tutorial

» see SCENARIO for a working tutorial

Ontologies are useful for their condense and structured format of a domain of knowledge. Specifically their organised terms and corresponding synonyms. Many NLP tools don’t utilize ontologies, Jabberwocky uses ontologies for synonym curation. Here we provide a full-depth explanation, informative scenarios, and working examples for the Jabberwocky toolkit - for installation instructions, see the Jabberwocky repository.

About the Commands

Below is an-indepth explanation of the commands which you can use with Jabberwocky.

bandersnatch

bandersnatch curates synonyms for a list of key terms / or words of interest from an ontology of your choice, you provide a list of ontology synonym tags. note: it is recommended your list of keywords are exactly the classes from your chosen ontology (all in lowercase).

Usage

$ jab-bandersnatch --help
Usage: jab-bandersnatch [OPTIONS]

Options:
  -o, --ontology TEXT     file of ontology.  [required]
  -s, --synonymtags TEXT  list of XML tags for synonym curation.  [required]
  -k, --keywords TEXT     list of class labels you want to use to search.
                          [required]
  --help                  Show this message and exit.

Running

$ jab-bandersnatch --ontology pocketmonsters.owl --synonymtags ontology_synonym_tags.txt --keywords words_of_interest.txt
$ jab-bandersnatch -o pocketmonsters.owl -s ontology_synonym_tags.txt -k words_of_interest.txt

Output

a .json file: output_ontology_label_synonyms.json of the classes and synonyms for your reference - this can be used for the catch command

catch

catch essentially “catches” key elements / sentences from textual data using a .json of key terms and their synonyms, if not using the output_ontology_label_synonyms.json from bandersnatch, then you can provide your own. The main element of catch is the textfile, which can be .txt or .json - if a .json is provided you need specify the parameter for the field that contains the textual data to process.

Usage

$ jab-catch --help
Usage: jab-catch [OPTIONS]

Options:
  -k, --keywords TEXT        list of terms and synonyms you want for grep, can
                             be from the ontology output.  [required]
  -t, --textfile TEXT        JSON or TXT file of text you want annotate.
                             [required]
  -p, --parameter TEXT       parameter of the the JSON text data.
  -i, --innerparameter TEXT  inner parameter of the the JSON text data if
                             expecting replies.
  --help                     Show this message and exit.

Running

$ jab-catch --keywords output_ontology_label_synonyms.json --textfile example_textfile.txt
$ jab-catch -k own_labels_synonyms.json -t example_tweets.json -p tweet-comment -i tweet-reply

Output

a .txt file: output_terms_match_raw.txt which includes all elements / sentences from the text file which includes a term of interest
a .json file: output_terms_match.json which includes the posts for each word of interest

bite

bite runs a tf-idf statistical analysis: searching for important terms in a text corpus. a user can use a list of key terms to remove from the text in order to avoid being in the statistical model - meaning other terms may be ranked higher. note: with the .json input you need specify the field inside the JSON that contains the textual data to process (same as catch).

Usage

$ jab-bite --help
Usage: jab-bite [OPTIONS]

Options:
  -k, --keywords TEXT        list of terms and synonyms you want to remove
                             from tf-idf analysis.
  -t, --textfile TEXT        JSON or TXT file of text you want annotate.
                             [required]
  -p, --parameter TEXT       parameter of the the JSON text data.
  -i, --innerparameter TEXT  inner parameter of the the JSON text data if
                             expecting replies.
  -g, --graph TEXT           make True if you want a plot of top 30 terms.
  -l, --limit TEXT           change if want a different plot limit.
  --help                     Show this message and exit.

Running

$ jab-bite --textfile facebook_posts.txt 
$ jab-bite -k output_ontology_label_synonyms.json -t example_tweets.json -p tweet-comment -i tweet-reply
$ jab-bite -k own_labels_synonyms.json -t facebook_posts.txt -g True

Output

a .tsv file: tfidf_results.tsv of all terms and their tf-idf score
a .pdf file: tfidf_plot.pdf the plot output which is requested if a user makes --graph True and presents the (default) 30-top scoring terms

arise

arise inserts synonyms in an ontology: you define these synonyms (e.g. “exact”, “broad”, “related”, or “narrow”) - these new synonyms may be based on the tf-idf statistical analysis from bite.

Usage

$ jab-arise --help
Usage: jab-arise [OPTIONS]

Options:
  -o, --ontology TEXT  file of ontology.  [required]
  -f, --tfidf TEXT     TSV file of the synonyms you want to add, can be based
                       from the tf-idf results.  [required]
  --help               Show this message and exit.

Running

$ jab-arise --ontology pocketmonsters.owl --tfidf tfidf_new_synonyms.tsv
$ jab-arise -o pocketmonsters.owl -f tfidf_new_synonyms.tsv

Output

a .owl file: updated_ontology.owl

jabberwocky

toolkit for those nonsensical ontologies