ReMatch

K-Cap 2017 Project

Capturing Knowledge in Semantically-typed Relational Patterns to Enhance Relation Linking

Note: The evaluation results for K-Cap 2017 paper is in "Evaluation results" Folder.

For Installation and Running

Please read the entire README file before doing anything

Python Version

Python 2.7 👍

Required packages

• numpy
• glove_python
• sklearn
• practnlptool
• textrazor
• cPickle
• distance
• nltk
• SocketServer
• urllib
• json
• re

Required data files

• glove precomputed data files
  • download from (http://nlp.stanford.edu/data/glove.6B.zip) the default file used by the code is (glove.6B.50d.txt)
• patty data files (included )

Required Data

• text-razor API key

Code explanation:

PS. File names are self explanatory

1. Tagger: POS tagger
2. Splitter: split the question into combinations
3. Embedder: glove wrapper to convert question into vectors
4. Reader: PATTY data reader
5. Backend: the complete process of reading PATTY data and create embeddings, with the cosine similarity code
6. Frontend: the complete process of reading a question and processing it
7. Textrazor_Api: the API wrapper for the textrazor service
8. main: where the magic happens
9. api: for the web UI interface
10. webService: for calling the system as a web service locally

Running local web service

Running the service via ./webService.py [port]

and calling it is simple i.e. (http://localhost/question_url_encoded)

• Response will look like this:

{
 "results":[
    "result 1",
    "result 2", ...
 ],
 "parts":[
    "part 1",
    "part 2", ...
 ],
 "pos":[
    [
       "word",
       "pos tag"
    ], ...
 ],
 "relation 1": "dbpedia relation lable",
 "relation 2": "dbpedia relation lable",
 ...
 "relation N": "dbpedia relation lable",
 "gen_question":"generalized question here",
 "question":"the input question"
}

Fast run

please run the code once using the main file to create the *.dat files that will be just loaded other times which will reduce processing time because not extra processing is done.

running main file is straightforward ./main.py

Any other issues while running the code:

Please email 📧 to Yaser ([email protected]) or Kuldeep ([email protected]) if you face any problem.