TriniT: Relationship Queries on Extended Knowledge Graphs

Entity search over text corpora is not geared for relationship queries where answers are tuples of related entities and where a query often requires joining cues from multiple documents. With large knowledge graphs, structured querying on their relational facts is an alternative, but often suffers from poor recall because of mismatches between user queries and the knowledge graph or because of weakly populated relations.

TriniT is a search engine for querying and ranking on extended knowledge graphs that combine relational facts with textual web contents. Our query language is designed on the paradigm of SPO triple patterns, but is more expressive, supporting textual phrases for each of the SPO arguments. We present a model for automatic query relaxation to compensate for mismatches between the data and a user’s query. Query answers – tuples of entities – are ranked by a statistical language model. We present experiments with different benchmarks, including complex relationship queries, over a combination of the Yago knowledge graph and the entity-annotated ClueWeb’09 corpus.

The TriniT demo for the VLDB 2016 demo submission is undergoing maintenance.

• Mohamed Yahya, Denilson Barbosa, Klaus Berberich, Qiuyue Wang, and Gerhard Weikum. Relationship Queries on Extended Knowledge Graphs. WSDM 2016.

Detailed results from our WSDM 2016 submission "Relationship Queries on Extended Knowledge Graphs" can be found in this package.

TriniT combines facts using Open Information Extraction (OpenIE) with the Yago2s Knowledge Base to form its XKG. We run OpenIE on the FACC1 annotated ClueWeb09 corpus to produce a total of 392M extractions (65M facts). 

Additionally, to facilitate query relaxation, we extract paraphrases of KG and XKG relations. Our dataset contains a total of 172M scored relation paraphrase pair.

If you are interested the the data above, please contact us.

MPI: Mohamed Yahya, Klaus Berberich, Gerhard Weikum

External: Denilson Barbosa, Qiuyue Wang