readme.txt

Brief introduction

DeepPavlov is an open-source conversational AI library built on TensorFlow, Keras and PyTorch, which
has a Knowledge Base Question Answering model that support perform knowledge base qa on both English and
Russian.
This project focus on reconstruct a KBQA system on Chinese by replace corresponding semantic parsing
components into Chinese version and finetuned on Chinese domain. Below are some details.


A Review on knowledge base question answer systems.
There is a point of view that distinguish different nlp question answer systems, that may be the
organization of answers. The question answer system is a system that transform (or map) natural language
to the different hierarchy of answer space. The answer space can be both probability space 、
distance space or discrete space. The former two space can be arrived by nlu classifier or distance
statistics. The last one is this project focus on and related topic as TableQA(natural language to SQL)
is a similar topic where this project focus on "translate" natural language to SPARQL query.


A Summary on natural language to SPARQL query.
Mainly two kinds of natural language to SPARQL query methods. One is a reconstruct method that try to
decide a suitable base form of SPARQL query and fill it by correspond semantic components extract from
natural language, The other is a transform method that make this process like a translation between
natural language and SPARQL query. The above discuss also compliant with TableQA and other natural language
to query topics.


Open Source Projects
Nowadays(2022 and before) ,In open source community, DeepPavlov[https://github.com/deeppavlov/DeepPavlov]
and Haystack[https://github.com/deepset-ai/haystack/] are two projects that implement KBQA use the above two
different methods. And this project focus on adjust the former into Chinese language domain.