This repository contains a server used in the bodi-generator project to answer questions not implemented in a chatbot. This is done by running NLP tools in the server that can be accessed through server endpoints.
These are the current provided functionalities:
- Text-to-table: the input is a natural language query and the output is the provided tabular answer. This is done by TabularSemanticParsing. Note that this language model translates English to SQL statements. So, when the input is not in English, a translation to English must be performed. Currently, supported languages are Spanish, Catalan and English.
Before deploying the server, you may want to edit some variables for your purpose.
- checkpoint_path: the path to the model checkpoint
- csv_dir: the directory where the csv files (i.e. the database) are stored
- CONFIG_FILE_PATH: the path to a .properties file that contains properties used within the server. These
properties
are:
- xls.importer.xls: the name of the csv file used as a database
- csv.delimiter: the delimiter or separator of the tabular data file cells (e.g.
,
,\t
(tab)) - SERVER_URL: the URL where the server is deployed
- TEXT_TO_TABLE_ENDPOINT: the endpoint of the server used to make requests to query the database with natural language sentences
To install all the components:
./install_server.sh
And then, to deploy the server:
./run_server.sh