Status

• NLPBA corpus of 2000 abstracts with tag annotations of Protein, DNA, and RNA. No relations
• LLL05 corpuswith 85 sentences with relations of genic interaction (to anything)
• Stub implementation of Relna here.
• README for Relna follows below.

Goal

1. Expand NLPBA corpus with:
   • Protein to RNA relations (reason, more interesting biologically & we don't have to deal with the problem of distinguishing Protein vs Genes)
   • Protein to DNA elements (very specific cases... TODO)
   • [ ] Write as PubAnnotation format & anndoc (tagtog) format
2. Create method to given a text or PMID, recognize these kind of relations (meaning, entity names have to also be recognized)
   • Make method public
   • Well documented for totally new users (HOWTO)

Approach

1. Expand dictionary by bootstrapping method:
   1. Annotate seed corpus of 20 abstracts (with relations)
   2. Develop method for relationship extraction
   3. Predict on a bigger set of abstracts
   4. Review manually (gradually check only low confidence cases)
   5. Add to whole corpus and repeat process

RELNA HOWTO

INSTALLATION

1. Install gradle on your system.
2. Run install.sh.

ALTERNATE (Deprecated)

1. Install gradle on your system.
2. Unzip the genia.mod.tar.gz from resources/ and copy it to src/main/resources/gimli/resources/tools/gdep
3. Navigate to the parent directory (where build.gradle is placed) and run gradle assemble

RUNNING THE TOOL

Navigate to build/libs and run the tool with java -jar mthesis-ashish-*.jar with the following arguments:

• -i path-to-input-file
• -o path-to-output-file
• -w writer-format - anndoc or json
• -r reader-format - can be txt or iob2 (optional argument)
• -s string - text to tag given from command line (optional argument)

If reader format (-r) argument is not given, the program will guess based on the extension.

If string (-s) argument is given, the program will automatically read from the command line for an input.

Example

Input format : iob2, -r specified, writer format : anndoc java -jar build/libs/mthesis-ashish-*.jar -r iob2 -i sample/iob2/corpus1.iob2 -w anndoc -o sample/iob2/output1.ann.json

Input format : iob2, -r not specified, writer format : json java -jar build/libs/mthesis-ashish-*.jar -i sample/iob2/corpus1.iob2 -w json -o sample/iob2/output1.json

Input format : txt, -r specified, writer format : anndoc java -jar build/libs/mthesis-ashish-*.jar -i sample/txt/corpus2.txt -r txt -w anndoc -o sample/txt/output2.ann.json

Eventually, the jar file will be called relna.

If the file is in txt format, it must follow the pattern below.

<identifier>
<title>
<abstract>

If the file is in iob format, it must follow the pattern below.

<identifier>

<sentence begin>
.
.
<token>
.
.
<sentence end>

<sentence begin>
.
.
.
<sentence end>